Sdea Ii PDF

Proceedings of the International Conference
Symmetries, Differential Equations and Applications
Courtesy: Google (Maxwells Fields)
Commemorating the sesquicentennial of the presentation by J C Maxwell
A Dynamical Theory of the Electromagnetic Field

to the Royal Society of London, 1864.
Sponsors
Higher Education Commission Pakistan (HEC)
Pakistan Academy of Sciences (PAS)
NUST University
International Mathematical Union (IMU)
International Center for Theoretical Physics (ICTP)
Pakistan Science Foundation (PSF)
National Center for Physics (NCP)
Editors
Sajid Ali (Pakistan)
Fazal M Mahomed (South Africa)
Asghar Qadir (Pakistan)
Proceedings of the International Conference
Symmetries, Differential Equations and Applications (SDEA-II)

Islamabad, Pakistan, January 27 30, 2014
The proceedings of the first International Conference Symmetries, Differential Equations and Applications
(SDEA-II) appeared as a special issue in the Journal of Applied Mathematics (JAM) in 2013. The first conference
in the series commemorated the bicentenary of the birth of variste Galois who introduced the notion of a group
and made fundamental contributions to the solution of polynomial equations. This conference was second in the
series and dedicated to the presentation of J. C. Maxwell of his theory of electromagnetism to the Royal Society of
London in 1864.
Journal: International Journal of Modern Physics Conference Series (IJMPCS)

Publisher: World Scientific
Publishing Year: 2015
Online ISSN: 2010-1945
Format: Open-Access
Sponsors
Preface
This issue is devoted to the Proceedings of the Second Conference on Symmetries, Differential Equations and
Applications. The first in the series, held at the University of the Witwatersrand, Johannesburg, South Africa
in January 2012 had been named (with exceptional foresight) the first Conference on Symmetries, Differential
Equations and Applications, SDEA-I. It had then been decided to hold a second (hence the ``First" in the
name) conference and to shift the venue to either Turkey or Pakistan. Since Turkey was already involved in
arranging a large number of international conferences in Mathematics, it devolved on Pakistan to hold the
second. The first one was held on the bicentennial of the birth of Evariste Galois and was dedicated to him.
The organizers were Fazal Mahomed, David Mason and Kevin Periera of Wits, South Africa with Mehmet
Pakdemirli of Turkey and Asghar Qadir of Pakistan added as the international members. The Conference was
of a very good standard and its proceedings, edited by F.M. Mahomed, M. Pakdemirli and A. Qadir, were
published in the Journal of Applied Mathematics.
The venue for SDEA-II was decided as the National University of Sciences and Technology, Islamabad. The
intention had been that the series be held roughly every second year. Since the convenient time for most
people, on account of winter (in the North) or summer (in the South) vacations being January, it was decided
to hold it precisely two years later, in January 2014. The first one having been dedicated to Galois, we needed
to see what possibilities were available for the year 2014. It turned out that this was the sesquicentennial of
Maxwell's theory of electromagnetism. Not that the paper was published in that year but that it was presented
to the Royal Society of London in 1864. Since David Mason (who is Scottish) and Asghar Qadir are both
great admirers of Maxwell and his work, the former was to present the dedication talk at the Conference, with
help from the latter. Unfortunately, he got ill and was unable to travel for the Conference, so the latter
presented the talk. Nevertheless, both have contributed to writing the paper for the Proceedings and it appears
at the start.
The original advertisement for the areas to be covered included, but were not limited to: Lie theory in
approximate and complex differential equations; Noether symmetries and their extensions; methods for first
integrals and conservation laws; linearization; stochastic differential equations; geometrical flows; moving
coframe formalism; classical mechanics, continuum mechanics and general relativity; equations of finance;
bi-Hamiltonian formalism. Of course, it was not expected that we get all of these covered, nor were they.
Apart from the dedication paper there were ten more, listed in the contents given next. Following the same
spirit of SDEA-I all conference contributions went under peer and double blind review process. We are most
grateful to all the authors for their work in preparing the manuscripts for these proceedings, to the reviewers
for their very thorough refereeing and back to the authors for further efforts to incorporate the modifications
required. We express our special thanks to the English Editor, Rabiya Asghar Qadir, for her work in checking
through the English.
We are also very grateful to the School of Natural Sciences (SNS) for its incredible support especially Prof.
Azad A Siddiqui (Principal) for his continuous efforts and encouragement. We would like to express our
gratitude to the NUST University for key infrastructural support, to the Commission for Science and
Technology (COMSTECH) of the Organization of Islamic Countries, to the Pakistan Academy of Sciences, to
the Higher Education Commission of Pakistan and to the Pakistan Science Foundation for generous support.
The National Center for Physics (NCP), Pakistan is also acknowledged for providing hospitality to the local
participants
Sajid Ali
(NUST University, Pakistan)
Fazal M Mahomed
(University of Witwatersrand, South Africa)
Asghar Qadir
(NUST University, Pakistan)
Contents
Dedication
Sesquicentennial of the Presentation by James Clerk Maxwell of his paper `À
Dynamical Theory of the Electromagnetic Field'' to the Royal Society of London
Asghar Qadir, David Mason
Technical Contributions
i)
Integrable systems with unitary invariance from non-streching geometric curve

flows in the Hermitian Symmetry Space Sp(n)/U(n)
Stephen C Anco, Esmaeel Asadi, Asieh Dogonchi
ii)
Noether gauge symmetries of geodesic motion in stationary and nonstatic Godeltype spacetimes
Ugur Camci
iii)
A review on symmetries for certain Aedes Aegypti Models

Igor Leite Freire, M. Torrisi
iv)
Symmetry Lie Algebras and properties of linear ordinary differential equations with
maximal dimension
M. Folly-Gbetoula, Abdul H Kara
v)
Lie symmetry analysis of some time fractional partial differential equations

E. H. El Kinani, A. Ouhadan
vi)
What symmetries can do for you

M.C. Nucci
vii)
Higher dimensional systems of differential equations obtainable by iterative use of

complex methods
Asghar Qadir, Fazal M Mahomed
viii) Geometrization of Lie and Noether symmetries and applications

Michael Tsamparlis
ix)
Stochastic symmetries of Wick type stochastic ordinary differential equations

Gazanfer Unal
x)
A finite differences approach to finding the exact solutions of differential equations

Servei Zuev
April 1, 2015
1:23
WSPC/INSTRUCTION FILE
Dedication
International Journal of Modern Physics: Conference Series

c World Scientific Publishing Company

Sesquicentennial of the Presentation by James Clerk Maxwell of his

paper A Dynamical Theory of the ElectromagneticField to the
Royal Society of London
Asghar Qadir
School of Natural Sciences, National University of Sciences and Technology, Campus H-12,
Islamabad 44000, Pakistan
asgharqadir46@gmail.com
D. P. Mason
DST-NRF Centre of Excellence in Mathematical and Physical Sciences, Differential Equations,
Continuum Mechanics and Applications, School of Computational and Applied Mathematics,
University of the Witwatersrand, Johannesburg, Private Bag 3, Wits 2050, South Africa
david.mason@wits.ac.za
Published: April, 2015
Let us work while it is day, for the night is coming and work by day
leads to rest by night
JC Maxwell
James Clerk Maxwell is generally regarded as the greatest contributor to the development
of Physics in the time between Newton and Einstein. His most important contributions
are the Kinetic Theory of Gases and Electromagnetism which is the unified theory of
Electricity and Magnetism. Although his major work on Electromagnetism was published
in 1865 it was read at a meeting of the Royal Society of London in 1864. The sesquicentennial of the theory correctly falls in 2014. In this article that event is celebrated.
Parts of his early and professional life are described. Aspects of his many contributions
are discussed but mainly we concentrate on his contributions through thermal and electromagnetic Physics.
1. Introduction
James Clerk Maxwell is generally regarded as the leading mathematical physicist
of the era between Newton and Einstein. He was an outstanding applied mathematician with powerful physical insight. He introduced the important concept of
the field into theoretical physics. He recognised the need to check his theoretical
predictions experimentally and performed the experiments himself.
On 8th December 1864 Maxwell read before the Royal Society of London his
paper A Dynamical Theory of the Electromagnetic Field. The year 2014 marks the
sesquicentennial of that presentation which is regarded as one of the most important
events of classical physics. In this article we celebrate that presentation. The paper
1
April 1, 2015
1:23
Dedication
Asghar Qadir & David Mason
was later published in the Philosophical Transactions of the Royal Society in 18651 .
We will describe parts of Maxwells life and work. Of his many contributions
to science, we will mention his work on colour and Saturns rings and the establishment of the Cavendish Laboratory at the University of Cambridge, but we will
mainly concentrate on his contributions to the Kinetic Theory of Gases for which he
introduced statistical methods into Physics and to electromagnetism and the electromagnetic theory of light. There are many books on Maxwell. The texts which
were consulted during the preparation of this paper are26 .
2. Life and work
James Clerk Maxwell was born on 13th June 1831 in a three storey house at 14 India
Street, Edinburgh. On his fathers side he was descended from the Clerks of Penicuik
in Midlothian, which was a distinguished Scottish family consisting of Parliamentarians and lawyers. Maxwells great-grandfather had married his cousin Dorothea
Maxwell and inherited the Maxwell estates provided he assume the Maxwell name,
which he did as Clerk Maxwell. It was through his great-grandfather that Maxwell
inherited the estate of Glenlair in Kirkcudbrightshire in south-west Scotland. His
mother was Frances Cay from Northumberland. The Cay family had achieved distinction in the legal and mathematical fields. A member of the Cay family was
Second in the Tripos at Cambridge University in 1752 and a Cay cousin was 6th
Wrangler in 1864.
Maxwell spent his childhood on the Glenlair estate. His father John had designed
and built the house on the estate and Maxwell in turn would add to the building.
As a child Maxwell showed a curiosity of the world around him and wanted to know
how it worked. If anything moved or shone he would ask the question, what is
the go of that? His mother looked after his early education until she died from
abdominal cancer in 1839 at the age of 47 years. For a time Maxwell was taught by
a private tutor before starting school at Edinburgh Academy in 1841.
Maxwell spent six years at Edinburgh Academy. He received a good grounding in Latin and Greek and a good background in Mathematics and some Physics.
Maxwell had a natural ability at geometry which he demonstrated from an early
age. While still at school, in 1846, he published his first paper which was on oval
curves7 . This geometrical vision contributed significantly to Maxwells physical insight by allowing him to visualise bodies and their motion. Maxwell made friends
with Peter Guthrie Tait who although the same age as Maxwell was a year behind
him at school. The two remained friends throughout Maxwells life. Tait was Senior
Wrangler and First Smiths Prizeman at Cambridge University in 1852 and later
became Professor of Natural Philosophya at Edinburgh University. In their correspondence Maxwell was dp/dt because of the equation dp/dt = JCM in Taits book
a At
the time of Newton, there was no separate Physics, it was called Natural Philosophy. This
practice has been carried on in the older Scotish Universities.
April 1, 2015
1:23
Dedication
Dedication to J. C. Maxwell
on thermodynamics. Tait was T0 and William Thomson, Lord Kelvin,who became

a friend of both Maxwell and Tait, was T.
Maxwell spent three years at Edinburgh University. He studied Logic, Physics,
Mathematics, Chemistry and some Geology. The Professor of Natural Philosophy
was James Forbes who had read Maxwells paper on Oval Curves to the Royal
Society of Edinburgh because Maxwell was considered too young to present the
paper himself. Forbes allowed Maxwell to use his laboratory to perform his own
experimental investigations. This was an important contribution to Maxwells early
scientific training in performing experiments. Forbes also helped Maxwell to develop
a literary style for his papers which had authority but was still fresh and informal6 .
Maxwell published two papers while at Edinburgh University. They were read to
the Royal Society of Edinburgh by the Professor of Mathematics, Phillip Kelland,
because Maxwell was still considered too young to present them himself. The first
paper was a geometrical paper on Rolling Curves8 which followed on from his work
on Oval Curves. The second paper was on the Equilibrium of Elastic Bodies9 . It
dealt with the deformation of solid bodies when subjected to stress. He developed the
mathematical theory of photoelasticity based on elastic strain functions and verified
the results with his own experiments which used polarised light. The field equations
of elasticity had been developed by GG Stokes and in this study Maxwell acquired
insight into field theory which he was to use later when setting out the equations of
Electromagnetism. Maxwell used one of the colourful patterns from photoelasticity
as a design for a woolen kettle holder which he hung from his fireplace at Glenlair3 .
He went to Cambridge University in October 1850. He was in residence at Peterhouse for the Michaelmas termb and then migrated in December to Trinity College.
He surprised the tutor (later Master of Trinity College) to whom he applied to migrate by producing a bundle of research papers and saying Perhaps these may show
you that I am not unfit to enter your College2 . His tutor was William Hopkins of
Peterhouse who had prepared many Senior Wranglers for the Mathematical Tripos.
Hopkins was methodical and brought order to Maxwells work although Maxwell
still took his own way in his studies. In 1853 Hopkins described Maxwell as the
most extraordinary man he had met within the whole range of his experience; it
appeared impossible for Maxwell to think incorrectly on physical subjects 10 . If
the subject allowed, Maxwell used diagrams although the other students may have
solved the problem more easily using analysis. He proceeded from one idea to another instead of relying on mathematical symbols and equations. He was a member
of the Apostle Club and the essays he prepared contributed to the high literacy
quality of his later papers. He attended regularly the lectures of George Gabriel
Stokes. Maxwell was already acquainted with some of the papers of Stokes from his
own work on elasticity. His friend Tait has observed that it may safely be said that
no high Wrangler of recent years ever entered the Senate-house more imperfectly
b The
Feast of Saint Michael is on 29th September, the beginning of the autumn (fall) term in a
number of English-speaking universities.
April 1, 2015
1:23
Dedication
trained to produce paying work than did Clerk Maxwell10 . Maxwell was Second Wrangler in the Mathematical Tripos. The Senior Wrangler was the Canadian
Routh. He was first equal with Routh in the Smiths prize which had a separate
examination.
In 1854 Maxwell started postgraduate research at Cambridge University. He
was elected a Fellow of Trinity College in October 1855. He worked in two main
areas, colour vision and Faradays lines of force although he also started research
in the kinetic theory of gases which became an interest for the rest of his life. In
March 1855 he read to the Royal Society of Edinburgh a paper on the mixture of
colours and colour blindness11 . It was mainly an experimental paper. By using multicoloured spinning discs invented by Newton and developed by Forbes at Edinburgh
University he showed how any colour could be represented by a suitable mixture
of three primary colours, red, green and blue. Thomas Young, fifty years earlier,
had taken the primary colours to be red, green and violet. Maxwell described the
rules of colour combination with the aid of an equilateral colour triangle with red,
green and blue at each vertex. White light is obtained from an equal mixture of
red, green and blue. With paints, an equal mixture of red, yellow and blue pigments
gives white but Maxwell found that no mixture of red, yellow and blue light gave
white light. There is an essential difference between mixing pigments as with paints
and mixing lights. The pigments extract colour so that the light that is seen is the
colour the paint was unable to absorb while mixing lights is combining colours6 . It
is relevant to remark that the Nobel Prize for Physics 2014 was awarded jointly to
Isamu Akasaki, Hiroshi Amano and Shuji Nakamura for the invention of efficient
blue light-emitting diodes which has enabled bright and energy-saving white light
sources. In 12 it is stated that: A little further on in the future three-colour
light-emitting diodes may replace the combination of blue light emitting diodes and
phosphor for efficient lighting. Maxwell tiled the floor of the entrance hall to his
house at Glenlair with coloured tiles as shown in Figure 1. Red, green, blue and
white tiles feature prominently in the floor pattern.
He read a paper13 on Faradays lines of force to the Cambridge Philosophical
Society in two parts, the first on 10 December 1855 and the second on 11 February
1856. It formulated mathematically Faradays idea of a line of force and replaced
the model of action-at-distance by the concept of a field. It was the second longest
of Maxwells research papers and marked the beginning of his work on Electromagnetism.
The Chair of Natural Philosophy (Physics) in Marischal College, Aberdeen, became vacant in 1855 and he applied. He would be closer to Glenlair and to his father
who was not in good health and he would also have a laboratory. He was appointed
to the Chair but his father died in April 1856. He resigned as a Fellow of Trinity
College, Cambridge but was later elected an Honorary Fellow. He started his duties
at Marischal College towards the end of 1856.He gave fifteen lectures a week and also
evening classes for artisans, a practice he started at Cambridge and was to continue
April 1, 2015
1:23
Dedication
Fig. 1.
Tiles laid by Maxwell on the floor of the entrance hall to Glenlair.
while at Kings College, London. He introduced the idea of students doing experiments themselves instead of only watching them being done. He was to do this also
at London University and at the Cavendish Laboratory in Cambridge. One of his
students was George Reith who was father of John Reith, later Lord Reith, the first
Director-General of the BBC. Maxwell signed the class certificate of George Reith
which Lord Reith hung in his office at the BBC. This linked Maxwell who discovered
electromagnetic waves with Reith who shaped their application in broadcasting4 . In
1858 Maxwell married Katherine Dewar, the daughter of the Principal of Marischal
College. Maxwell continued his research on colour vision at Aberdeen. A long paper
on the theory of compound colours was published by the Royal Society of London
which describes observations by himself and his wife Katherine to determine the
spectral curve on the colour triangle14 . In 1860 he was awarded the Rumford Medal
of the Royal Society of London for his work on colour vision. Maxwell made two
other significant research contributions at Aberdeen, his essay for the Adams Prize
and his work on the Kinetic Theory of Gases.
The Adams Prize Essay had been founded in 1848 by St Johns College, Cambridge, to commemorate the theoretical discovery of the planet Neptune by John
Couch Adams. In 1843 Adams was Senior Wrangler and also won the First Smiths
prize15 . The Prize was set every two years and open to Cambridge graduates. The
subject in 1855 was to investigate the structure of Saturns rings. Maxwell put a
lot of effort into this investigation. It was already known that a ring could not be
April 1, 2015
1:23
Dedication
a uniform solid because it would be unstable. Maxwell showed that a ring could
be a nonuniform solid body but the required distribution of mass did not agree
with observations. He also showed that unstable waves would be formed in a fluid
ring. He therefore concluded that a ring could only be composed of an indefinite
number of disconnected masses. The masses revolved round Saturn with different
velocities according to their distances from the planet. In December 1856 Maxwell
submitted the essay On the stability of motion of Saturns rings, which was later
published16 . It was over 80 pages in length and full of mathematical techniques
and physical insight. Due to the difficulty of the subject it was the only entry. The
Astronomer Royal, Sir George Airy, described it as one of the most remarkable applications of Mathematics to Physics that he had ever seen. Maxwell was awarded
the Adams Prize for 1855. His prediction was confirmed by the photographs sent
back to Earth by the Voyager 1 and 2 spacecraft in the early 1980s.
Maxwell was attracted to the Kinetic Theory of gases by the work of Rudolf
Clausius who was developing a molecular theory of gases. It has also been suggested
that his model of Saturns rings as made up of a very large number of masses with
a range of velocities depending on distance from the planet led him into Kinetic
Theory. His first contribution was the Maxwell distribution law giving the velocity
distribution of molecules in thermal equilibrium in a gas. He presented the law at the
meeting of the British Association for the Advancement of Science in Aberdeen in
September 1859. The derivation of the Maxwell distribution law made no reference
to collisions between molecules. It may be the first law of physics that depends on
probability. The probabilistic model of matter is now fundamental to physics. The
presentation was published as a paper in two parts in 1860? . The paper covered a
wide range of topics in Kinetic Theory and is regarded as one of Maxwells greatest
papers. It contained the counter intuitive prediction that the viscosity of a gas is
independent of its pressure. This was verified in 1865 by experiments which Maxwell
and his wife, Katherine, performed in London.
In 1860 the two colleges in Aberdeen, Marischal College and Kings College,
were fused together to form Aberdeen University. The new Chair of Natural Philosophy was filled by the former professor at Kings College who was senior to
Maxwell. Maxwells post ceased to exist and he had to find another position. Both
Maxwell and Tait applied for the Chair of Natural Philosophy at Edinburgh University which became vacant on the retirement of Forbes. Tait was appointed to
the Chair. Maxwell then applied for the vacant Chair of Physics and Astronomy at
Kings College, London and was accepted in July 1860. Before traveling to London,
Maxwell spent the summer at Glenlair. He contracted smallpox, possibly when he
went to buy a horse at a local fair. He was very ill but Katherine took great care of
him. In later life Maxwell said that she saved his life.
Maxwell was Professor at Kings College from 1860 to 1865. It was full of scientific activity and the most creative time of his career in which his most important
papers were published. They moved into a large house of four storeys at 8 (now
16) Palace Gardens Terrace, Kensington. It was 4 miles to Kings College which he
April 1, 2015
1:23
Dedication
could walk or take a horse-drawn bus ride. He was able to walk to meetings of the
Royal Society and at the Royal Institution. Kensington Gardens and Hyde Park
were close by where Maxwell and Katherine, who were enthusiastic horse riders,
frequently went for a ride in the afternoon. The house also had a large attic where
Maxwell did experiments with the assistance of Katherine. A lecturer was appointed
to assist Maxwell in his duties. The post was eventually held by Grylls Adams who
was the younger brother of John Adams in whose honour the Adams Prize was
founded. Adams succeeded Maxwell to the Professorship when he resigned in 1865.
In his inaugural lecture Maxwell stressed that he wanted above all to help people
learn to think for themselves.
While Maxwell was at Cambridge he had started corresponding with Faraday.
He now met him personally. He would often attend sessions at the Royal Institution
including Faradays Friday evening lectures. They got along well based on a strong
mutual respect but the relation was not close because Faraday was 40 years older
than Maxwell. Maxwell and Faraday had dinner together in May 1861 when Maxwell
gave a lecture at the Royal Institution on colour vision. At this lecture Maxwell
produced the first colour photograph which was of a tartan ribbon. He took black
and white photographs through red, green and blue filters and then projected and
superposed the results through the same filters3 . Three weeks after his lecture at
the Royal Institution he was elected a Fellow of the Royal Society of London in
recognition of his research on Saturns rings and colour vision. The following week
he celebrated his 30th birthday.
Maxwell, with the assistance of Katherine, performed experiments to test his
prediction from Kinetic Theory that the viscosity of a gas is independent of pressure. The gas used was air. He determined the viscosity by measuring the rate of
damping of the oscillation of a stack of three discs suspended by a thin wire. This
torsional pendulum was sealed inside a large glass vessel and the air pressure was
adjusted by a pump. He also tested experimentally a second prediction that the viscosity should vary as the square root of the absolute temperature. Maxwell analysed
the two sets of results at Glenlair during the summer break. Calculations then were
done using log tables. He forgot to take his log tables with him and he had to do all
the calculations by hand6 . The experimental results showed that the viscosity was
constant over a wide range of pressure which verified his prediction. The second set
of experimental results showed that the viscosity did not vary as the square root
of the absolute temperature but came closer to varying as the absolute temperature. The two laws were derived from Kinetic Theory under different assumptions.
The pressure dependence law held for any type of molecule while the temperature
law depended on the assumption that the molecules were perfectly elastic spheres
colliding with each other. Also Clausius had found errors in Maxwells derivation
of the law of equipartition of energy that the kinetic energy in a gas is divided
equally between linear and rotational energy. These and other considerations motivated Maxwell to undertake further investigations in Kinetic Theory which led to
his major paper On the dynamical theory of gases18 .
April 1, 2015
1:23
Dedication
Maxwell was a member of the committee of the British Association for the
Advancement of Science which was formed to investigate a system of units for
electricity and magnetism. Experiments were performed at Kings college. Maxwell
introduced the method of dimensional analysis in which all physical quantities are
defined in terms of mass, length and time. The recommendations of the Committee
made a significant contribution to the international system of units adopted in 1881.
Maxwells most important work at Kings College was on Electromagnetism. In
paper13 he introduced the displacement current and showed that light waves were
electromagnetic waves. Polarised light had been an interest of Maxwell since, as
a school boy at Edinburgh Academy, he had been taken to the workshop of the
optician William Nicol who had discovered how to polarise light using prisms cut
from Iceland spar. In paper1 which we are celebrating he presented the equations
of Electromagnetism, now known as Maxwells equations and was able to explain
polarization. His prediction of electromagnetic waves was eventually confirmed experimentally by Heinrich Hertz in 1888. The implications of this will be discussed
in more detail in Section 4. Maxwell resigned from Kings College in January 1865
and a few months later returned to Glenlair which was to be his main residence
until 1871. He wanted to have time to write-up his research and also to improve his
estate. He enlarged the house on the estate during 1867. A bridge was built across
the river Urr by his cousin William Cay who was a Civil Engineer. Figure 2 contains
a letter written by Maxwell a little later in 1876 which illustrates the commitment
he made to the running of his estate. He visited his tenants and took an interest
in the local school. Life in the country may have inspired the paper On hills and
dales which was an early exercise in topology19 . He maintained a correspondence
with his colleagues, especially Stokes, Tait and Thomson (Lord Kelvin) and received
journals and proofs of his manuscripts by mail. He had a special post box built in
a stone wall about 1/2 mile from the house and would walk there and back each
day to take and collect mail. He spent the winters in London and was an external
examiner for the Mathematical Tripos at Cambridge.
Maxwell continued his research on the dependence of the viscosity of a gas on
temperature. Instead of modelling the gas molecules as colliding perfectly elastic
spheres, he assumed that they repel each other by a force that is proportional to
rn where r is the separation distance. He found that the mathematics could be
greatly simplified by setting n = 5 and that this also gave a linear relation between
viscosity and absolute temperature. This agreed better with his experiments in
London but further experiments by other investigators showed that in general the
viscosity was not proportion to the temperature although for some gases n = 5.
In his derivation of relations for viscosity, diffusion and thermal conductivity he
introduced the concept of a relaxation time which is now widely used in science.
His paper On the dynamical theory of gases18 in which this work was published
gave strong support to the model that gases are composed of molecules.
Maxwell wrote the book The theory of heat while at Glenlair20 . It contained
for the first time Maxwells thermodynamic relations between pressure, volume,
April 1, 2015
1:23
Dedication
Fig. 2.
Letter written by Maxwell on improvement to his Glenlair estate.
April 1, 2015
10
1:23
Dedication
temperature and entropy. His contribution in this regard will be discussed later in
Section 3.
The main part of Maxwells greatest work, Treatise on electricity and magnetism was written at Glenlair between 1865 and 1871. A photograph of the bay
window of the room in which he wrote the Treatise is shown in Figure 3. It was published in 1873 after Maxwell had returned to Cambridge. The writing of the Treatise
gave him the opportunity to organise his own thoughts on the subject. It contained
almost everything then known on electricity and magnetism. It also contained new
results on the radiation pressure exerted by electromagnetic waves including light.
Maxwell used two representations for the equations, the cartesian component form
and the more concise quaternion form. He introduced the terminology, curl, gradient
and convergence to help with the physical interpretation. The terms curl and gradient are now used in vector calculus but convergence has been changed to its negative,
divergence. The 1000 pages of the Treatise contain the word aether only once3 .
Maxwell wanted to concentrate on the consequences of his theory rather than on
its foundations. The Treatise ranks after Newtons Principia as the most important
book in the history of physics. The Chair of Experimental Physics at Cambridge
Fig. 3. The bay window of the room at Glenlair in which Maxwell wrote his Treatise
on Electricity and Magnetism. The house was largely destroyed by fire in 1929.
University was established in 1870. After Thomson and Helmhotz had declined the
appointment it was offered to Maxwell who accepted on 8 March 1871. The Chan-
April 1, 2015
1:23
Dedication
11
cellor of the University The eighth Duke of Devonshire, William Cavendish, had
donated funds to build a Physics Laboratory and to purchase equipment. Maxwell
visited Thomsons Laboratory at Glasgow University to learn from their experience.
He designed the building with the architect as he had done when extending Glenlair.
He took great care to select the best equipment. He designed some of the apparatus
himself. The building had been known as the Devonshire Laboratory6 . The name
was changed to the Cavendish Laboratory at the suggestion of Maxwell to honour
both the Duke and his great uncle, the physicist Henry Cavendish. The Cavendish
Laboratory was completed in 1874.
In his inaugural lecture Maxwell pointed out how careful measurement has been
rewarded by the discovery of new fields of research and by the development of new
scientific ideas. As Director of the Cavendish Laboratory one of Maxwells duties
was to oversee the research in the Laboratory. He would suggest a research problem
if asked but otherwise would let the students choose their own topic. Maxwell was
an excellent supervisor of students in a laboratory and was greatly respected. Students were attracted to the Cavendish Laboratory including Strutt (Lord Rayleigh),
Poynting and Fleming who invented the thermionic valve. There was a revival of
Physical Sciences at Cambridge.
In 1874 Maxwell was asked by the Duke of Devonshire to edit for publication
the papers of Henry Cavendish, who was more interested in doing research than in
publishing. Maxwell was impressed by the originality of Cavendishs work. Many
of the experiments were on electricity which Maxwell performed again. The editing
took a lot of time but it was completed and a book was published in 187921 .
The most important paper Maxwell wrote while at the Cavendish Laboratory
was inspired by Boltzmanns work on the Kinetic Theory of gases22 . Maxwell introduced the idea of an ensemble average and stated the ergodic hypothesis. He gave a
new derivation of the distribution of velocities of a gas in equilibrium which led to
a proof of the principle of the equipartition of energy. He also showed that mixtures
of gases could be separated by a centrifuge. This paper is one of the foundations of
statistical mechanics.
Maxwells health had always been good until the beginning of 1877 when he
developed heartburn. By April 1879 he found it difficult to swallow. He spent the
summer at Glenlair. On 2 October he was told by his doctor that, like his mother, he
had abdominal cancer and had about one month to live. He returned to Cambridge
to have the assistance of Dr Paget and be with his friends and colleagues. He died
on 5 November 1879. He is buried in Parton Churchyard near his Glenlair estate.
The house at Glenlair was largely destroyed by fire in 1929.
Maxwell died at the height of his powers. The characteristics he demonstrated
when young remained with him all his life. He had a large capacity for sustained
work. He had great powers of concentration even when there were distractions, a
retentive memory, a high degree of originality, a deep physical insight and mathematical ability. Although reserved he had confidence in his own powers as a scientist.
He was kind and patient with a lively sense of humour which is clearly displayed in
April 1, 2015
12
1:23
Dedication
his poetry and correspondence. Maxwell was eccentric in his manners and he was
sometimes difficult to follow in conversation. At meals he was often lost in thought.
He prepared his lectures very conscientiously and thoroughly but they were not well
attended. Horace Lamb who attended his lectures at Cambridge said that he had
his full share of misfortunes at the blackboard. David Gill, a student of Marschal
College and future Director of the Royal Observtatory, Cape of Good Hope (in
what is now South Africa), commented that to those who could catch a few of the
sparks that flashed as he thought aloud at the blackboard in lecture, or when he
twinkled with wit or suggestion in after-lecture conversation, Maxwell was supreme
as an inspiration6 . He demonstrated social responsibility and compassion throughout his life. He leant books from the library at Glenlair to young farm workers.
When a youth he preferred to travel by third class in a railway journey. He said
that he preferred a hard seat. At Cambridge, Aberdeen and London he gave evening
courses at Working Mens colleges. As a research supervisor in the laboratory he
inspired his students with his generous advice and kindness. He liked country life
and was a good horseman. He was conscientious in his responsibilities to his tenants
on the Glenlair estate. He was of average height, strong and sturdy from his country
upbringing and walked with a spring in his step. At Cambridge he walked every day
from his house to the Cavendish Laboratory with his dog Tobi. We have this word
picture of Maxwell working in the Cavendish Laboratory3 :
In performing his private experiments at the laboratory Maxwell was very neathanded and expeditious. When working thus, or when thinking about a problem, he
had a habit of whistling, not loudly, but in a half-subdued manner. He could carry
the full strength of his mental facilities rapidly from one subject to another and could
pursue his studies under distractions which most students would find intolerable.
On these occasions he used in a manner to take his dog into his confidence and
would say softly, Tobi, Tobe, at intervals and after thinking and working for a
time, would at last say It must be so; Plato thou reasonest wellc .
3. The Kinetic Theory of Gases
Although Maxwells greatest achievement is regarded as his theory of Electromagnetism, his contribution through his attempts to explain thermal phenomena has
no less far-reaching consequences. It is worth while to review the understanding of
heat and the thermal properties of matter prevalent at the time.
Aristotle had believed that there are four terrestrial elements, viz. earth, water, air and fire; and a fifth celestial element, aether, on the basis of the five Platonic solids (whose faces are regular polygons). Chemical studies in the mid 1660s
did not fit in with this view of fire fully and the phlogiston theory of heat was
proposed. It regarded heat as a fifth terrestrial element, violating the Aristotelian
c The
verse is from Cato: A Tragedy (1713) by Joseph Addison (1672 - 1719).
April 1, 2015
1:23
Dedication
13
basis. Later chemical investigations demolished the five elements of Aristotle and
introduced many more. Now a fluid theory of heat was proposed which did not fare
much better. With the observation of the heating effect of mechanical work and
the use of heat to do mechanical work, a much better understanding was required.
This came with the First Law of Thermodynamics, which related heat to work and
essentially gives the Law of Conservation of Energy.
With the concept of energy there was the need to provide an internal energy
for any system. This energy could be due to mechanical work and would then be
the product of the pressure and volume of a gas, or it could be chemical energy as
when combustion occurred, or it could be due to sound, or it could be due to heat.
Sound was understood as compression waves in the medium but heat remained to
be understood. The intensity of heat being the temperature there was need for a
quantity corresponding to the volume that, when multiplied by the temperature,
would give the heat component of the internal energy. The problem was that there
was no such quantity seen in the laboratory. This postulated quantity was labeled
entropy. Investigations led to the statement of the Second Law of Thermodynamics
that concerned this entropy. It was a very different kind of law from the laws of
mechanics or the First Law of Thermodynamics; it was an inequality instead of
being an equality! What it said was that the entropy of a closed thermodynamic
system always increases. Since the efficiency of a heat engine reduces as entropy
increases, this direful law seemed to say that the Universe will wind down.
Before his major innovation, Maxwell had already made significant contributions
to thermodynamics by the use of partial differential calculus to relate rates of change
of different thermal quantities. Maxwells relations led to clear physical implications
of entropy that were observed. (They are still taught to Physics students, but the
contribution could hardly be regarded as revolutionary.)
He now addressed the two major problems in thermodynamics: the nature of
heat; and the concept of entropy and the Second Law of Thermodynamics. Mechanics, being regarded as a sound base, Clausius and others used Bernoullis view
of fluids as composed of molecules. They tried to explain heat as the kinetic energy
of the molecules of a gas. Clausius went to the extent of introducing a statistical
concept into the mechanical description of gases, namely the mean path traversed
by a molecule between two collisions.
Clausius work needs Statistics but did not use it. Maxwell made use of Statistics,
not merely averages. He used the distribution function to make the proposal of the
Kinetic Theory of gases quantitative and concrete. The distribution function gives
the probability that a molecule has a given momentum vector. He took a box of
volume d3 p about the momentum vector p in momentum space. The probability
density, f (p), is then the ratio of the number of molecules in the box divided by the
total number of molecules
in the system. Summing up over all such volumes in the
R
momentum space V f (p)d3 p must be unity where V is the volume of momentum
space. This will then automatically lead to the velocity and energy distributions
for the molecules. Now the average kinetic energy must be the thermal energy,
April 1, 2015
14
1:23
Dedication
which is a constant (called the Boltzmann constant, normally denoted by k) times

the temperature. This explains the temperature. Mathematically, the temperature
must be the weighted average of the kinetic energy of all the molecules with this
distribution function, over an infinite momentum space. Writing the momentum
space in spherical polar coordinates the average kinetic energy is then
Z
3
2
f (p)p4 dp := kT ,
< K.E. >=
(1)
m 0
2
having taken the average momentum of the molecules to be zero, where m is the
mass of the molecule. This gives the Maxwell momentum distribution
2
fp (p) = (2mkT )3/2 ep /2mkT .
(2)
One can replace the momentum by the velocity, v, to obtain the Maxwell velocity
distribution
m 3/2 2 mv2 /2kT
4
) v e
,
(3)
fv (v) = (
2kT
and thence the Maxwell energy distribution

r
E 1 3/2 E/kT
fE (E) = 2
(
)
e
.
kT
(4)
For a given momentum vector, p, pfp (p)d3 p gives the probability of the momentum being p, in the momentum space volume d3 p. Integrating over the volume
must give zero for the system to be at rest. Now, take the component of p along
the vector area elements on the surface and integrate their product, p.dS, over the
entire surface of the gas container. This gives the pressure, P of the gas, which turns
out to satisfy the ideal gas equation of state, P V = N kT , if we treat the molecules
as point particles that do not interact, where N is the number of molecules in the
system. (Allowing for the volume and interactions of the molecules gives the van
der Waals equation of state.)
This provides a demonstrable explanation of heat in terms of the mechanics of
Newton. The difference between the classical picture and the kinetic picture can be
tested by allowing a beam of hot gas to escape through an orifice in the container.
Kinetic Theory predicts that the particles will spread out due to gravity in the
up-down direction when they hit a screen, because there will be some with greater
momentum that will go nearly straight and others with less momentum that will
be pulled down more. The classical expectation would be that it should spread
uniformly about the geometrical shadow of the hole on the screen.
Does this theory explain the concept of entropy? Maxwell worried over this
matter and finally came up with an explanation. He considered a gas consisting
of two types of molecules in two joined boxes with a tiny hole in the partition
between the two boxes. He now placed a microscopic demon (a name given by
William Thomson) guarding the hole from one side of the hole. This demon would
separate the molecules that cross the hole, so that over time there would be two
April 1, 2015
1:23
Dedication
15
pure gases in the two boxes, which violates the Second Law of Thermodynamics.
Similarly, even for only one type of molecule, the demon could separate the fast
from the slow molecules so that the temperatures of the two connected boxes would
be different, again violating the Second Law. However, the work done by the demon
in blocking and sending on the molecules has to be accounted for and so the demon
does work to locally reduce the entropy, but increases its own by at least as much.
The moral of this story is that in a system without such a demon, the molecules are
stopped from separating into a hot and cold side, or into two types of molecules,
by the improbability of the separation occurring spontaneously. How unlikely is it?
Well, in a cubic centimetre of gas at normal temperature and pressure, the number
of molecules is given by Avogadros number, viz. 61023 . Consider air in a small
cube of one centimetre side. It consists of nearly one part of oxygen to four parts of
nitrogen, with some other gases mixed in. For our purposes, take it to be in precisely
this ratio. Thus there are about 1.2 1023 molecules of oxygen and 4.8 1023 of
nitrogen. The probability of the separation occurring spontaneously is, then about
51.8 1047 , which is as good as zero.
Maxwells great contribution to physical thinking here is the realization that
the Second Law of Thermodynamics is statistical in nature. In a letter to Tait he
said that he introduced the demon to show that the second law of thermodynamics
is a statistical law which applies to all molecules and not to individual molecules.
It is worth noting that the second law is arguably the most robust law in Nature.
Even the first law breaks down in a Universe that changes with time, according
to Einsteins theory of General Relativity, but the Second Law, being statistical, is
robust. The larger the system the more accurate the assumption of things running
down. Contrariwise, the smaller the system, the less definite is the prediction.
Another extremely important contribution is the use of the thought experiment,
which was a standard technique of Einstein and used extensively to this day. We
do not know of a prior use of a genuine thought experiment in the modern sense,
namely one that cannot be performed in a laboratory. Thinking about Maxwells
demon helped Leo Szilard develop information theory and the Szilard engine in
192923 . More recent developments of Maxwells demon can be found in2425 .
The statistical methods of Maxwell were developed further by Boltzmann, using
Hamiltons formulation of mechanics and forming the space constituted of both position and momentum, called phase space. In particular, he developed the MaxwellBoltzmann distribution that incorporates collisions into the theory. Though the
formal definition of entropy as the natural logarithm of the number of phase space
states available, relies on Boltzmanns formulation of Statistical Mechanics, its substance is already there in Maxwells Kinetic Theory of Gases. These developments
later led Max Planck to the postulate that the energy is absorbed and emitted by
molecules in discrete quanta, which explained the observed behaviour of radiant
energy. This led Einstein to postulate the existence of quanta of energy. Nowadays,
the statistical thinking of Maxwell has been developed much further and is arguably
the most crucial underlying paradigm.
April 1, 2015
16
1:23
Dedication
4. Electromagnetism
When Maxwell started his work on electricity and magnetism, much had already
been discovered. The concept of the electric and magnetic lines of force had been
provided by Faraday. It is particularly easy to see for magnetism by plotting the
direction of the needle of a small compass as one moves the compass along the
lines of force about a bar magnet. They are seen still more easily by putting a bar
magnet on a white sheet of paper and pouring iron filings on to the paper around
the magnet. Since the filings behave like tiny iron needles, they perform the same
job as the compass needle, but do so simultaneously, giving a picture of the field as
a whole. Ampere and Faraday had already stated their Laws giving the magnetic
effects of electric currents and the electric effects of moving magnets.
Maxwell formalized Faradays more physical idea of lines of force as the field
intensity, thereby providing the first clear concept of the electric and magnetic fields.
Since the lines of force emanate from some source, they must have a starting point.
Since we do find separate charges (electric monopoles) the electric lines of force
need have no ends. However, since we do not see magnetic charges i.e. monopoles
separately, the magnetic lines of force must have ends as well. Looking at the flux
of lines of force emerging from some region we can use Gauss divergence theorem
to conclude laws for the electric and magnetic field intensities, E and B,
.E = / , .B = 0 ,
(5)
where . is the divergence operator, is the electric charge density and is the
dielectric constant, which varies from material to material. It has a finite value for
the vacuum, usually denoted by 0 , which will be relevant shortly. These laws are
generally referred to as Gauss laws.
Amperes Law relates the magnetic flux across a loop of wire to the current
flowing through it:
I
B.dS = I ,
(6)
where dS is the vector area element for the surface containing the loop of wire,
is the magnetic permittivity which varies according to the material used for the
surface in which the wire is embedded. For the vacuum it is 0 and I is the electric
current. This can be written in differential (i.e. local) form instead of the integral
form by using Stokes theorem as
B = j ,
(7)
where j is the current per unit length in a piece of the loop of wire.
Faradays Law, that the electromotive force induced by a time-varying magnetic
field is given by the rate of change of that field, can be stated mathematically as
E=
B
.
t
(8)
April 1, 2015
1:23
Dedication
17
Equations (5), (7) (with a modification) and (8) are called the Maxwell equations. If all the laws were already in place and the magnetic effects of electric
currents and the electric effects of moving magnets were already known, was there
no genuine contribution by Maxwell? Did he simply compile all the results obtained
by others in his treatise and receive the credit due to the rest? The answer is an
emphatic NO! The clear formulation of the field concept is crucial to the modern
understanding of Electromagnetism and other fields and provides the paradigm for
modern Physics. Moreover, Maxwell immediately obtained many vital new insights
and discovered new facts on their basis. Let us examine the contributions in more
detail.
The idea of a field is there in embryonic form in Faradays lines of force. However, it does not make clear the ramifications of that concept and the dramatic
break it entails with the Newtonian view. While Newton had localized causes with
corresponding local effects Maxwell provided the nonlocal field. Thus, while Newton had a problem with the action at a distance that his law of gravity entailed,
Maxwell provided that action through the mediation of his all-pervasive field. Further, that field implied that all prior data everywhere would determine subsequent
events everywhere and there would be no localized causes for any effects. Instead
of action at a distance being odd, it is localized action that needs to be explained.
As if this contribution was not enough, the next one goes still deeper. Maxwell
noted that there is an underlying symmetry between the electric and magnetic fields
in the absence of sources. The Physics underlying it is that moving electric charges
lead to magnetic fields and moving electric fields lead to magnetic effects, even where
the charges are not present. This duality requires a fundamental unity in the two
fields. This observation provides the paradigm for the search for unification of forces
of Nature to this day. The further realization that where there is such a physical
unification there will be some associated symmetry that is somehow broken, is
equally significant and is the reason Maxwells contributions are being celebrated
in the SDEA-II at the sesquicentennial of his presentation of the theory to the Royal
Society. This, again, is an underlying paradigm for modern High Energy Physics
and the unification attempts. The physical realization that it is the effect of each
on the other that signals a unification is yet another vital insight that is used.
Maxwell immediately realized that Amperes Law, eq.(7), violated the required
symmetry. Because of his deep conviction that symmetry will rule, he set about
finding what was missing to complete the desired symmetry. The current density is
just the rate of change of the charge density. There must, then also be the rate of
change of the electric field which has to be inserted into that equation:
B = J +
E
.
t
(9)
He called the inserted term the displacement current and set about searching for
it experimentally by looking for a deflection in a needle due to the varying electric
field with no possibility for electric charges to flow. It is a tribute to his ingenuity
April 1, 2015
18
1:23
Dedication
and conviction that he found this term. Here was a case where he predicted that
the prevailing theory was wrong and demonstrated an effect that was not consistent
with that theory but fitted his theory. This, again, has become the paradigm for
modern developments through Theoretical Physics, and indeed the driving force for
doing laboratory experiments to-day.
For the next major contribution take the curl of eq. (8), commute with t
and use eqs. (9) and (5) and the vector operator identity that the curl of the curl
of a vector is the gradient of the divergence of the vector minus the Laplacian
acting on the vector to obtain the wave equation for the electric field with a source.
Similarly, taking the curl of eq. (9) and using the other three equations with the
same identities, gives the wave equation with a source for the magnetic field. Hence
there must be electromagnetic waves. In other words, these waves will be present
even where there are no sources. The wave equation for the electromagnetic field
obtained carries the speed of the wave in it. Identifying the speed by its square
appearing as the coefficient of the Laplacian part of the wave operator, it turns out
to be 1/ . Using the values for the vacuum one obtains a speed that is extremely
close to the speed of light. Hence, not only did Maxwell discover that there would
be electromagnetic waves but that light is an electromagnetic wave ! This discovery
confirmed the wave theory of light and provided a complete testable explanation
of such diverse phenomena as the dependence of the refractive index of a material
on its dielectric constant and its magnetic permittivity, as well as providing the
frequency of light in terms of its wavelength, of chromatic aberration and many
other similar phenomena.
Potentials had been introduced to deal with the gravitational force in terms of
the associated energy to simplify calculations by Lagrange and used by Laplace
to discuss the gravitational force outside a gravitating body and by Poisson to
extend it to the interior of gravitating bodies, yielding equations known by the
names of their originators. The gravitational potential is a scalar function whose
gradient is the negative of the gravitational force. Maxwell extended this to the
electric and magnetic forces to obtain corresponding scalar potentials, and U .
However, because the curl of a gradient is zero and so the divergence of a curl is zero,
on account of eq. (5), the magnetic field can be expressed as the curl of a vector
potential, A, while the electric field cannot. Taking this fact to be fundamental,
Maxwell used the vector potential for his further formulation.
Notice that the addition of the gradient of an arbitrary scalar function of space
and time to this vector potential would make no difference to the magnetic field.
Hence the magnetic field is invariant under this transformation. However, to accommodate the displacement current Maxwell found that he had to re-define the
electric potential, writing it as the sum of an electrostatic potential and the time
derivative of the magnetic vector potential, so that
E =
A
.
t
(10)
April 1, 2015
1:23
Dedication
19
Thus corresponding to the addition of the gradient to the magnetic vector potential
there must be a subtraction of its time derivative in the electric potential. Mathematically, the electromagnetic field is invariant under the gauge transformation
of the potentials,
= A + f (x, y, z, t) , = f (x, y, z, t)/t .
AA
(11)
To put the unified electromagnetic field in terms of potentials he used the four
component generalization of the complex numbers called quaternions proposed by
Hamilton. They are defined as q = r + ai + bj + ck, where a, b, c, r IR , i2 = j 2 =
k 2 = 1. He then noted that this new four component potential can be changed by
adding the gradient of an arbitrary scalar function of space and time without making
any difference to the physical electromagnetic field. This freedom of choice of the
potential is called gauge freedom. This was in some ways the most remarkable
and significant of his many discoveries. The quarternions have often been considered
for transcribing special relativity as the structure is remarkably similar to that of
Minkowski space and essentially gives the same group structure as well.
From here Maxwell could have taken two short steps and arrived at the Special
Theory of Relativity but was held back by his reverence for Newtons mechanics
that had provided so many insights. One step involved in fully accepting the electromagnetic field as itself providing the medium in which light (or electromagnetic
waves) move and thus dispensing with the old ether. Instead he spent much of his
dwindling energy on trying to construct a mechanical model of the ether, which
could support electromagnetic waves, to demonstrate that it could be done. The
other was to pursue the fact that his theory gives a constant speed of light for all
observers. Since the principle of special relativity was already taken for granted, no
more was needed. (It may be mentioned that his Kinetic Theory of gases had also
brought him to the verge of discovering the quantum of electromagnetic waves.)
Many physicists have placed Maxwell at par with Galilieo, Newton and Einstein
as among the greatest ever, with no means of deciding who was the greater. In the
case of the latter two, there was an annus mirabiles associated with each. For
Maxwell there was no single such year but a whole stream of nearly equally fundamental developments flowed from him from about 1860 to about 1870, anni
mirabilis or a decade mirabiles. (Of course, for Einstein there was much more
that he developed long after his miracle year, notably his theory of general relativity and his application of that theory to cosmology.) Among those who consider
Maxwell to be an unsung hero of Physics, was Abdus Salam (Nobel Laureate),
who said it to one of us (AQ) personally, and his mentor Paul Adrian Maurice Dirac
(Nobel Laureate) who he quoted. It is worth noting that Maxwell did not have the
advantage of the modern tensor notation and developed all his ideas on the basis of
his physical intuition and the cumbersome notation using separate symbols for each
separate component of the vector fields and potentials. Seeing them written out in
their full gory detail is an object lesson in the utility of modern tensor notation and
a reminder of his genius, that he saw the essential simplicity inherent in the mess
April 1, 2015
20
1:23
Dedication
of his equations. When one sees the equations in the language of differential forms
in four dimensions, with one definition of the field, one obvious set of identities and
one working equation, one loses a sense of what he had to look at. He nevertheless
set up the fundamental search for unification, with his theory of Electromagnetism,
that was followed by Einsteins unification of space and time, based on Maxwells
unification of Electromagnetism and leading on to the gauge unification of the electromagnetic and weak nuclear forces and set the stage for the current attempts at
the gauge unification of all forces of Nature.
Which of Maxwells developments is taken as more important, the kinetic theory
of gases or Electromagnetism, we leave to the readers to decide for themselves.
The current consensus seems to be that the field concept and production of the
model workable field theory is the more important. From the point of view of the
conference on symmetries that would certainly be endorsed. After all, the key feature
of Electromagnetism, the gauge freedom to choose the electromagnetic 4-vector
potential, provides the basis of using symmetries for physical purposes. It plays
a role through Noethers theorem and thence in the modern gauge theories of
quantum field theory; be they Abelian or nonabelian. However, the importance of
the second law of thermodynamics and its understanding provided by Maxwell may
lead to that being regarded as more fundamental in later ages.
5. Conclusion
The contribution of Maxwell that we celebrate in this paper relates to Electromagnetism, which is regarded as his greatest achievement, not only as the introduction
of the field concept but also as the simplest and most elegant example of a field
theory. While much of the technological use of electricity does not require his field
theory, there is so much that does, such as its use in wave guides and hence in
fibre optics, its use in the theory of refraction and its use for the propagation of
radio and television waves, that one wonders how much of the modern world would
have been possible without it. Although it is because of its use that its significance
may be appreciated by everybody, it is because of its development of the human
understanding of the world we live in that it may be considered to have had the
most dramatic impact. It provides the basic language of the Physics of today and
lies at the root of Albert Einsteins special theory of relativity. One may recall that
the original paper in relativity was entitled On the electrodynamics of moving bodies
26
. It has generally been assumed that despite having introduced the field concept
which obviates the need for a luminiferous aether, Maxwell continued to believe
in it. However, it has been pointed out that he may have given it as a sop to the
conservatism of his colleagues of the time, to be dispensed with at a future time.
Of course, we will now never know which was his true belief.
We have also mentioned numerous other contributions of Maxwell, while still
missing many others. One that we did dwell on was his development of the Kinetic
Theory of gases that lies at the base of the other major method of modern Physics,
April 1, 2015
1:23
Dedication
21
namely Statistical Mechanics. It is worth mentioning one aspect of the introduction

of his demon. To the best of our knowledge this is the first real thought experiment
used in Physics. Previous such experiments were merely idealizations of experiments
that could actually have been performed, such as Galileos smoothly sailing boat to
illustrate his principle of the relativity of uniform linear motion. Maxwells demon
is not such a thought experiment. It makes a point of principle that comes out loud
and clear by considering an impossible experiment. Such experiments have now
become the mainstay of theoretical physical reasoning.
Maxwell died very young at the age of 48. It is interesting to speculate how
much he might have achieved had he lived to a riper age with the good health that
so many later physicists have enjoyed. It is certainly not fanciful to speculate on
his discovering Special Relativity, which is a direct consequence of his theory that
we are celebrating. For that matter, Einstein formulated his General Relativity as a
tensor field theory; a direct generalization of the Maxwell vector field theory. Again,
it is certainly not pushing credulity to suppose that he would soon have developed
Statistical Mechanics. With his interest in the theory of radiation, he may even have
arrived at Plancks quantum of energy emitted or absorbed.
He seems to have been a twentieth century man trapped in the nineteenth century, as the following excerpts form his letters show. They enhance the scientist,
the teacher, the man with his knowledge, his beliefs and his wittiness.
Letter to Professor WG Adams, Natural Science Tripos, 3rd December 1873
10
I do not approve of the plan of a physical society considered as an instrument for

the improvement of natural knowledge. If it is to publish papers on physical subjects
which would not find their place in the transactions of existing societies, or in scientific journals, I think the progress towards dissolution will be very rapid. (omissis)
For the evolution of science by societies the main requisite is the perfect freedom of
communication between each member and any one of the others who may act as a
reagent.
Letter to Rev. CJ Ellicott, Lord Bishop of Gloucester and Bristol, 11 Scroope
Terrace, Cambridge, 22nd November 1876 10 :
But I should be very sorry if an interpretation founded on a most conjectural scientific hypotheses were to get fastened to the text in Genesis, even if by so doing
it got rid of the old statement of the commentators which has long ceased to be intelligible. The rate of change of scientific hypothesis is naturally much more rapid
than that of Biblical interpretations, so that if an interpretation is founded on such
an hypothesis, it may help to keep the hypothesis above ground long after it ought to
be buried and forgotten. At the same time I think that each individual man should
do all he can to impress his own mind with the extent, the order, and the unity of
the universe, and should carry these ideas with him as he reads such passages as
April 1, 2015
22
1:23
Dedication
the 1st Chap. of the Ep. to Colossians (see Lightfoot on Colossians, p. 182), just as
enlarged conceptions of the extent and unity of the world of life may be of service
to us in reading Psalm viii.: Heb. ii. 6, etc.
Letter to Mrs Maxwell, Trinity College, Cambridge, 3rd January 1870
10
There is a tradition in Trinity that when I was here I discovered a method of throwing
a cat so as not to light on its feet, and that I used to throw cats out of windows. I
had to explain that the proper object of research was to find how quick the cat would
turn round, and that the proper method was to let the cat drop on a table or bed
from about two inches, and that even then the cat lights on her feet.
Acknowledgements
We thank the referees for useful comments and suggestions. DPM thanks the National Research Foundation, Pretoria, South Africa, for financial support.
References
1. JC Maxwell, A dynamical theory of the electromagnetic field, Philos. Trans. Roy.
Soc. Lond. CLV (1865) 459 - 512.
2. WD Niven, The Scientific Papers of James Clerk Maxwell Vols. 1 & 2 (Cambridge
University Press, Cambridge, 1980).
3. I Tolstoy, James Clerk Maxwell (Canongate, Edinburgh, 1981).
4. James Clerk Maxwell Commemorative Booklet produced by the James Clerk Maxwell
Foundation on the occasion of the Fourth International Congress on Industrial and
Applied Mathematics coming to Edinburgh in July 1999.
5. I James, Remarkable Physicists: From Galileo to Yukawa (Cambridge University Press,
Cambridge, 2004).
6. B Mahon, The Man who Changed Everything: The Life of James Clerk Maxwell (John
Wiley and Sons, Chichester, 2004).
7. JC Maxwell, On the description of oval curves and those having a plurality of foci,
Proc. Roy. Soc. Edinb. II April 1851 (1844-1850) 89 - 91.
8. JC Maxwell, On the theory of rolling curves, Trans. Roy. Soc. Edinb. XVI (1849)
519 - 540.
9. JC Maxwell, On the equilibrium of elastic solids, Trans. Roy. Soc. Edinb. XX (1853)
87 - 120.
10. L Campbell and W Garnett. Life of James Clerk Maxwell, MacMillan and Co., London,
1882.
11. JC Maxwell, Experiments on colour, as perceived by the eye, with remarks on colour
blindness, Trans. Roy. Soc. Edinb. XXI (1857) 275 - 289.
12. Class for Physics of the Royal Swedish Academy of Sciences, Scientific Background on
the Nobel Prize in Physics 2014. Efficient blue light-emmitting dioes leading to bright
and energy-saving white hight sources. Nobelprizw.org.2014.
13. JC Maxwell, On Faradays lines of force, Trans. Camb. Philos. Soc. X (1864) 27 83.
14. JC Maxwell, On the theory of compound colours and the relations of the colours of
the spectrum, Philos. Trans. Roy. Soc. Lond. 150 (1861) 57 - 84.
April 1, 2015
1:23
Dedication
23
15. WM Smart. John Couch Adams and the Discovery of Neptune, Nature 158, 648 - 652,
1946.
16. JC Maxwell, On the stability of the motion of Saturns rings, Monthly Notices of
the Roy. Astronom. Soc. 10 (1859) 297 - 384.
17. JC Maxwell, Illustrations of the dynamical theory of gases I: On the motion and
collision of perfectly elastic spheres, Philos. Mag. XIX (1860) 19 - 32; II On the
process of diffusion of two or more kinds of moving particles among one another, 33
- 37.
18. JC Maxwell, On the dynamical theory of gases, Phil. Trans. Roy. Soc. Lond., CLVII
(1868) 49 - 88.
19. JC Maxwell, On hills and dales, Philos. Mag. XL (1870) 421 - 426.
20. JC Maxwell, Theory of heat (Text-books of Science, London, 1870).
21. JC Maxwell (ed.), The electrical researches of the Honourable Henry Cavendish FRS
written between 1771 and 1781, edited from the original manuscripts in the possession
of the Duke of Devonshire, KG (Cambridge, 1879).
22. JC Maxwell, On Boltzmanns theorem on the average distribution of energy in a
system of material points, Trans. Camb. Philos. Soc. XII (1879) 547 - 570.
23. M Klein, Maxwell, his Demon, and the Second Law of Thermodynamics, American
Scientist 58, 84-97, 1970.
24. EE Daub, Maxwells Demon, Studies in History and Philosophy of Science, Part A.
Volume 1, Issue 3, 213 - 227, 1970.
25. H Leff and AF Rex (Eds), Maxwells Demon 2: Entropy, Classical and Quantum
Information, Computing, Institute of Physics, London, 2003.
26. A Einstein, Zur elektrodynamik bewegter k
orper, Annalen der Physik 17 (1905) 891;
English translation On the electrodynamics of moving bodies by M. Saha University
of Calcutta 1920; and later by G.B. Jeffery and W. Perrett, Methuen and Co. Ltd.
1923.
April 1, 2015
1:14
Stephen*Anco


INTEGRABLE SYSTEMS WITH UNITARY INVARIANCE FROM

NON-STRETCHING GEOMETRIC CURVE FLOWS IN THE
HERMITIAN SYMMETRIC SPACE Sp(n)/U (n)
STEPHEN C. ANCO
Department of Mathematics and Statistics, Brock University
St. Catharines, Canada
sanco@brocku.ca
ESMAEEL ASADI
Department of Mathematics, Institute for Advanced Studies in Basic Sciences
Zanjan 45137-66731, Iran
easadi@iasbs.ac.ir
ASIEH DOGONCHI
Department of Mathematics, Institute for Advanced Studies in Basic Sciences
Zanjan 45137-66731, Iran
a.dogonchi@iasbs.ac.ir
A moving parallel frame method is applied to geometric non-stretching curve flows in the
Hermitian symmetric space Sp(n)/U (n) to derive new integrable systems with unitary
invariance. These systems consist of a bi-Hamiltonian modified Korteweg-de Vries equation and a Hamiltonian sine-Gordon (SG) equation, involving a scalar variable coupled to
a complex vector variable. The Hermitian structure of the symmetric space Sp(n)/U (n)
is used in a natural way from the beginning in formulating a complex matrix representation of the tangent space sp(n)/u(n) and its bracket relations within the symmetric Lie
algebra (u(n), sp(n)).
Keywords: geometric curve flow, Hermitian symmetric space, integrable system, parallel
frame, unitary invariance
1. Introduction
In the theory of integrable systems, the modified Korteweg-de Vries (mKdV) equation and the sine-Gordon (SG) equation are two integrable equations of basic importance. They have an elegant geometric origin that arises from the differential
invariant of a curve that undergoes certain geometric non-stretching flows (which
locally preserve the arclength of the curve) in the Euclidean plane and the sphere,
respectively. For generalizing this geometric picture to higher dimensional spaces, it
becomes natural to transform from a Frenet frame along the curve which determines
the differential invariants of the curve to a parallel frame 1 which determines differ1
April 1, 2015
1:14
Stephen*Anco
Stephen C. Anco & Esmaeel Asadi & Asieh Dogonchi
ential covariants of the curve. This approach has been used in numerous geometric
spaces (see Refs. 2, 3, 4, 5, 7, 6, 8, 9, 10, 11, 12, 13, 14).
There is a general construction of a parallel frame 15 for non-stretching curve
flows in Riemannian symmetric spaces M = G/H. The Cartan structure equations
of this frame have the property that they explicitly yield a pair of Hamiltonian and
symplectic operators in all symmetric spaces. These operators can be used to derive
a hierarchy of group-invariant bi-Hamiltonian mKdV equations and SG equations
whose variables are given by the components of the Cartan connection matrix along
a curve undergoing particular geometric non-stretching flows 15 .
In this paper, we apply this general method 15 to the Hermitian symmetric
space Sp(n)/U (n) and get U (n 1)-invariant integrable scalar-vector mKdV and
SG equations. These equations and their bi-Hamiltonian integrability structure naturally involve both the Hermitian inner product and the complex structure J on
the symmetric Lie algebra (u(n), sp(n)/u(n)).
Using this Hermitian structure, we show how to introduce a natural complex
matrix representation for the symmetric Lie algebra (u(n), sp(n)/u(n)) as well as the
Lie bracket relations between u(n) and sp(n)/u(n)). In this way, the flow equation
can be naturally written in terms of a scalar variable and a complex vector variable
that transform properly under the unitary gauge group of the underlying parallel
frame. If instead a standard real matrix representation were to be used, there would
be a problem of how to translate the flow equation with real variables into a unitarily
invariant form. Our approach fully resolves this issue. (See also the examples in
Refs. 10, 13, 14.)
More generally, depending on whether the symmetric space is real, Hermitian,
or quaternionic, one may use such extra geometric/algebraic structures from the
beginning on the tangent space of M as well as on the associated symmetric Lie
algebra (h, g/h) and its Lie bracket relations.
The outline of the paper is as follows. In section 2, we review the definition of
Riemannian symmetric spaces and also Hermitian symmetric spaces using symmetric Lie algebras. For the Hermitian symmetric space Sp(n)/U (n), we decompose
the subalgebra u(n) and the quotient space sp(n)/u(n) into parallel and perp subspaces relative to a regular element in the Cartan subspace in sp(n)/u(n). We give
a complex matrix representation of these subspaces and also the bracket relations
between them, which will be essential later for setting up a parallel frame for curves
in Sp(n)/U (n).
In section 3, we first review the standard frame field and connection formulation
of the Riemannian structure common to all symmetric spaces, and then we review
the definition of a moving parallel frame for non-stretching curve flows in symmetric
spaces. In section 4, we consider non-stretching curve flows in the Hermitian space
M = Sp(n)/U (n). By utilizing a moving parallel frame and then taking the pull
back of the torsion and curvature 2-forms from M to the surface swept out by a
curve flow, we get Cartan structure equations which encode a pair of Hamiltonian
and symplectic operators. These results will be just stated without proof. See Ref. 15
April 1, 2015
1:14
Stephen*Anco
Integrable systems in the Hermitian symmetric space Sp(n)/U (n)
for details and an explicit proof for curve flows in general symmetric spaces. Then
we derive a hierarchy of U (n 1)-invariant bi-Hamiltonian equations, starting with
a U (n 1)-invariant mKdV equation. All of these equations are integrable systems
for an imaginary scalar variable coupled to a complex vector variable. We next
derive a U (n 1)-invariant SG equation for the same variables by using the kernel
of the symplectic operator and doing an algebraic reduction of the parallel part of
the flow. There is a general method behind this reduction which will be explored in
a subsequent paper25 .
2. Algebraic structure of Sp(n)/U (n)
A general Riemannian symmetric space M = G/H is defined by 16 a simple Lie
group G and an involutive compact Lie subgroup H in G. Any linear frame on M
provides a soldering identification 17 between the tangent space Tx M at points x
and the vector space m = g/h. There is an orthogonal decomposition g = h m
with respect to the Cartan-Killing form with the Lie bracket relations
[h, h] h,
[h, m] m,
[m, m] h.
(1)
A Hermitian symmetric space is a Riemannian symmetric space that possesses a

covariantly-constant (1, 1) tensor J which acts on each tangent space as a complex
structure, J 2 = id.
Cartan subspaces of m are defined as a maximal abelian subspace a m with
the property that it is the centralizer of its elements, a = m c(a). It is well-known
16
that any two Cartan subspaces are isomorphic to one another under some linear
transformation in Ad(H) and that the action of the linear transformation group
Ad(H) on any Cartan subspace a generates m. The dimension of a as a vector space
is equal to the rank of M .
To define a moving parallel frame on non-stretching curve flows, which will be
presented in the next section, we need to choose an element e in the Cartan subspace
a. This element will be identified with the frame components of the tangent vector
of a curve. For any choice of e, the corresponding linear operator ad(e) induces a
direct sum decomposition of the vector spaces m = sp(n)/u(n) and h = u(n) into
centralizer spaces mk and hk and their orthogonal complements (perp spaces) m
and h with respect to the Cartan-Killing form. The Lie bracket relations on mk ,
m , hk , h coming from the structure of g as a symmetric Lie algebra (1) are given
by
[mk , mk ] hk ,
[mk , hk ] mk ,
[hk , hk ] hk ,
(2)
[hk , m ] m ,
[hk , h ] h ,
(3)
[mk , m ] h ,
[mk , h ] m ,
(4)
while the remaining Lie brackets

[m , m ],
[h , h ],
[m , h ]
(5)
April 1, 2015
1:14
Stephen*Anco
obey the general relations (1).

Through the Lie bracket relations (2)(4), the operator ad(e) maps h into m ,
and vice versa. Hence ad(e)2 is well-defined as a linear mapping of each subspace
h and m into itself.
2.1. Symmetric Lie algebra (u(n), sp(n)/u(n))
Hermitian symmetric spaces have been classified by Cartan and others. See Ref. 16
for full details. For the specific Hermitian symmetric space Sp(n)/U (n), we will
provide all the algebraic details needed for later use.
The symplectic Lie algebra sp(n) consists of all matrices in gl(2n, C) satisfying
the condition

0 I
g + g t = 0, g t = g, =
.
(6)
I 0
An involution on gl(2n, C) which leaves sp(n) invariant is defined by (g) = g. The
set of matrices in sp(n) that is invariant under is isomorphic to the unitary Lie
algebra u(n). Thus the Lie algebra sp(n) as a symmetric Lie algebra orthogonally
decomposes as sp(n) = h m into the eigenspaces of :
h :=u(n) g,
(h) = h,
m :=sp(n)/u(n) g,
(m) = m.
The matrix representation of g = sp(n) is given as

A B
sp(n), At = A, B t = B
B A
(7)
(8)
(9)
in which A, B gl(n, C). The matrix representation of the vector spaces m =

sp(n)/u(n) and h = u(n) is given as

A B
(A, B) := i
m, A, B gl(n, R), At = A, B t = B,
(10)
B A

C D
(C, D) :=
h, C, D gl(n, R), C t = C, Dt = D.
(11)
D C
The complex structure J of the tangent space To M = m is represented by
J = 12 iI u(n), so then the action of ad(J) on m is simply given by multiplication
by i. Therefore one can identify (A, B) m as a complex symmetric matrix A+iB
s(n, C). The elements (C, D) in h can be naturally written as a complex matrix
C + iD u(n).
With this complex representation, the bracket relation (1) is not just a matrix
commutator due to the role of complex structure J.
Lemma 2.1.
April 1, 2015
1:14
Stephen*Anco
(1) The Hermitian matrix representation of the vector space m = sp(n)/u(n) and
Lie subalgebra h = u(n) is given as
(A1 ) m,
(C1 ) h,
A1 gl(n, C),
C1 gl(n, C),
At1 A1 = 0,
t
C1
+ C1 = 0.
(12)
(13)
(2) The Lie bracket relations (1) have the following matrix representation in which
(A1 ), (A2 ) m and (C1 ), (C2 ) h:
[A1 , A2 ] = A2 A1 A1 A2 h,
(14)
[A1 , C1 ] = A1 C 1 C1 A1 m,
(15)
[C1 , C2 ] = C1 C2 C2 C1 h.
(16)
(3) The restriction of Cartan-Killing form on sp(n) to m = sp(n)/u(n) is a

negative-definite inner product
hA1 , A2 i = 8(n + 1)Re(tr(A1 A2 )).
(17)
(4) The vector space m = sp(n)/u(n) is of dimension n2 + n and of rank n. The

n-dimensional vector subspace a m = sp(n)/u(n) generated by real, diagonal
matrices (E) m is a Cartan subspace.
An element e of the Cartan subspace a is called a regular element 16,18 if its
centralizer subspace c(e) in g = sp(n) is of maximal dimension. Any real, diagonal
matrix Eii gl(n, R) whose only non-zero component is a 1 in its ith row and ith
column (with 1 i n) is regular element (Eii ) a. We will choose
1
e := (E11 ) a
(18)
where R is a normalization constant. We choose this constant so that e has unit

norm,
1 = he, ei = 8(n + 1)/
which determines
= 8(n + 1).
(19)
In the following lemma, we give an explicit matrix representation of the perp

and parallel subspaces of m and h determined by e. These subspaces are defined by
the properties
ad(e)mk = 0,
hm , mk i = 0
ad(e)hk = 0,
hh , hk i = 0.
and
Lemma 2.2.
April 1, 2015
1:14
Stephen*Anco
(1) The matrix representation of mk and m in m = sp(n)/u(n) is given as

ak 0
a a
(ak , Ak ) =
mk , (a , a ) :=
m
0 Ak
at 0
in which ak R, a iR, a Cn1 , and Ak s(n 1, C).
(2) The matrix representation of hk and h is given as

0 0
c c
(Ck ) =
hk , (c , c ) =
h
ct 0
0 Ck
(20)
(21)
in which c iR, c Cn1 , and Ck u(n 1).

(3) The dimension of perp and parallel subspaces is given as
dim(mk ) = n2 n + 1,
dim(hk ) = n2 2n + 1,
(22)
dim(m ) = dim(h ) = 2n 1.
(4) The regular element (18) in the Cartan subspace a is represented as
1
e = (1, 0) mk .
(23)
In particular the linear operator ad(e) gives an isomorphism of m and h :

1
ad(e)(a , a ) = (2a , a ) h ,
1
ad(e)(c , c ) = (2c , c ) m .
(24)
To write out the explicit Lie bracket relations on m = mk m and h = hk h ,

we will use the following inner products and outer products. For a, b Cn1 , we
note
abt + bat = 2Reha, bi R,
abt bat = i2Imha, bi iR,
(25)
where
ha, bi = abt
(26)
is the Hermitian inner product. Also, we define

at b bt a := a b so(n 1, C),
at b + bt a := a b s(n 1, C),
(27)
and
t
at b b at := ab u(n 1),
Proposition 2.1.
(28)
April 1, 2015
1:14
Stephen*Anco
(1) The bracket relations (2)(4) between the perp and parallel subspaces of sp(n)
are given as:
[mk , mk ] =[(ak1 , Ak1 ), (ak2 , Ak2 )] = (Ak2 Ak1 Ak1 Ak2 ) hk
(29a)
[mk , hk ] =[(ak , Ak ), (Ck )] = (0, Ak Ck Ck Ak ) mk
(29b)
[hk , hk ] =[(Ck1 ), (Ck2 )] = (Ck1 Ck2 Ck2 Ck1 ) hk
(29c)
[mk , m ] =[(ak , Ak ), (a , a )] = (2ak a , a Ak ak a ) h
(29d)
[mk , h ] =[(ak , Ak ), (c , c )] = (2ak c , ak c c Ak ) m
(29e)
[m , hk ] =[(a , a ), (Ck )] = (0, a Ck ) m
(29f)
[hk , h ] =[(Ck ), (c , c )] = (0, c Ck ) h
(29g)
(2) The remaining bracket relations (5) between perp spaces are given as:
[m , m ]hk =[(a1 , a1 ), (a2 , a2 )]hk
=(at2 a1 ) hk
(30a)
[m , m ]h =[(a1 , a1 ), (a2 , a2 )]h

=(2iImha1 , a2 i, a1 a2 a2 a1 ) h
(30b)
[m , h ]mk =[(a , a ), (c , c )]mk

=(2Reha , c i 2a c , a c ) mk
(30c)
[m , h ]m =[(a , a ), (c , c )]m
=(2iImha , c i, a c c a ) m
(30d)
[h , h ]hk =[(c1 , c1 ), (c2 , c2 )]hk

=(c2 c1 ) hk
(30e)
[h , h ]h =[(c1 , c1 ), (c2 , c2 )]h

=(2iImhc1 , c2 i, c1 c2 c2 c1 ) h
(30f)
(3) The Cartan-Killing form on m is given as

h(a1 , a1 ), (a2 , a2 )i = 8(n + 1)Re(a1 a2 + 2ha1 , a2 i).
(31)
The adjoint action of the Lie subalgebra hk h = u(n) on g = sp(n) generates

the linear transformation group Hk H = Ad(H), leaving invariant the element e
in the Cartan subspace a m = sp(n)/u(n). The group H is given by the unitary
group U (n 1) U (n) which has the matrix representation

1 0
U (n 1) ' Hk , Ck U (n 1).
(32)
0 Ck
This unitary group U (n 1) acts on the subspace m as
Ad(Ck )(a , a ) = (a , a Ckt ) m .
The action of
Hk
(33)
' U (n 1) on mk is given by
Ad(Ck )(ak , Ak ) = (ak , Ck Ak Ckt ) mk .
(34)
April 1, 2015
1:14
Stephen*Anco
Notice that these actions leave the scalar component unchanged. This observation
will help us to give a geometric interpretation of the algebraic reduction used later
to derive the SG flow.
Proposition 2.2. The linear map ad(e)2 on m , h ' iR Cn1 is given by
ad(e)2 (a , a ) = 1 (4a , a ) m ,
ad(e)2 (c , c ) = 1 (4c , c ) h .
The irreducible subspaces of m and h in this representation are (a , 0) iR and
(0, a ) Cn1 on which ad(e)2 has respective eigenvalues 4/ and 1/.
3. Moving parallel frames for non-stretching curve flows in
Riemannian symmetric spaces
The Riemannian structure of the space M = G/H is most naturally described 17,19
in terms of a m-valued linear coframe e and a h-valued linear connection whose
torsion and curvature
R := d + 21 [, ]
T := de + [, e],
(35)
are 2-forms with respective values in m and h, given by the following Cartan structure equations
T = 0,
R = 12 [e, e].
(36)
Here the brackets denote the wedge product combined with the Lie bracket.
The underlying Riemannian metric on the space M = G/H is given by
g = he, ei
(37)
in terms of the Cartan-Killing form restricted to m.

Now consider any smooth flow (t, x) of a curve in M . The flow is called nonstretching if it preserves the G-invariant arclength ds = |x |g dx, with |x |2g =
g(x , x ), in which case we can put |x |g = 1 without loss of generality (whereby x
is the arclength parameter).
For flows that are transverse to the curve (such that x and t are linearly
independent), (t, x) will describe a smooth two-dimensional surface in M . The
pullback of the torsion and curvature equations (36) to this surface yields
Dx et Dt ex + [x , et ] [t , ex ] = 0,
(38)
Dx t Dt x + [x , t ] = [ex , et ],
(39)
with
ex := ecx ,
et := ect ,
x := cx ,
t := ct ,
where Dx , Dt denote total derivatives with respect to x, t.
(40)
April 1, 2015
1:14
Stephen*Anco
For any non-stretching curve flow, these structure equations (38)(40) encode
an explicit pair of Hamiltonian and symplectic operators once a specific choice of
frame along (t, x) is made. For the case of curve flows in two-dimensional and
three-dimensional Riemannian spaces, see Refs. 3, 6. A proof for the general case of
curve flows in a general Riemannian symmetric space M = G/H is given in Ref. 15.
See Refs. 20, 21, 22, 23 for a related, more abstract formulation.
As shown in Ref. 15, a natural moving parallel frame can be defined by the
following two properties which are a direct algebraic generalization of a moving
parallel frame in Euclidean geometry1 :
(i) ex is a constant unit-norm element e belonging to a Cartan subspace a m,
i.e. Dx ex = Dt ex = 0, hex , ex i = 1.
(ii) x belongs to the perp space h of the Lie subalgebra hk h of the linear
isotropy group Hk H = Ad(H) that preserves ex , i.e. hhk , x i = 0.
A moving frame satisfying properties (i) and (ii) is called H-parallel and its
existence can be established by constructing 15 a suitable gauge transformation on
an arbitrary frame at each point x along the curve.
Through property (i), the set of curve flows (t, x) in M = G/H can be divided
into algebraic equivalence classes defined by the orbit of the element ex = e in a m
under the action of the gauge group H = Ad(H).
4. Bi-Hamiltonian structure and a hierarchy of U (n 1)-invariant
mKdV and SG flows
In a general symmetric space M = G/H, the Cartan structure equations (38)
(40) yield a flow equation on the components of the Cartan connection along the
curve, where the flow is specified by the perp component of t . To write down the
Hamiltonian and symplectic operators appearing in the flow equation, we use the
following notation:
ex = e a mk ,
(41)
et = hk + h mk m ,
(42)
t = $k + $ hk h ,
(43)
x = u h .
(44)
Also we write
h = ad(ex )h h .
(45)
Then we have the following theorem from Ref. 15.

Theorem 4.1. The Cartan structure equations (38)(39) for any H-parallel linear
coframe e and linear connection pulled back to the two-dimensional surface (t, x)
in M = G/H yield the flow equation
ut = H($ ) + h ,
$ = J (h )
(46)
April 1, 2015
1:14
10
Stephen*Anco
where
H = K|h ,
J = ad(e)1 K|m ad(e)1
(47)
are compatible Hamiltonian and symplectic operators that act on h -valued functions and that are invariant under Hk , as defined in terms of the linear operator
K := Dx + [u, ] [u, Dx1 [u, ]k ].
(48)
These operators arise directly from projecting the Cartan structure equations
(38)(40) into the parallel and perp subspaces of h and m, which yields
Dx hk + [u, h ]k = 0,
(49)
Dx h + [u, hk ] + [u, h ] [$ , e] = 0,
(50)
and
Dx $k + [u, $ ]k = 0,
(51)
Dx $ ut + [u, $ ] + [u, $ ] + h = 0.
(52)
We will essentially use these equations (49)(52) throughout the paper.

As shown in Ref. 15, there are two natural flows that each give rise to a groupinvariant integrable system. One flow is generated by the x-translation symmetry
of the Hamiltonian and symplectic operators, which yields a group-invariant mKdV
equation. The other flow is defined by the kernel of symplectic operator, which
produces a group-invariant SG equation. These two equations are at the bottom of
a hierarchy of higher-order integrable systems.
Theorem 4.2. Composition of the operators H and J yields a recursion operator
R = HJ that produces a hierarchy of Hk -invariant flows (46) on u starting from
the flow
h = ux
(53)
which gives an integrable group-invariant mKdV equation. The kernel of the recursion operator R yields a further Hk -invariant flow (46) on u defined by
J (h ) = 0.
(54)
This flow gives an integrable group-invariant SG equation after an algebraic reduction is made.
The integrability structure of these two flows is shown in detail in Ref. 15.
5. Bi-Hamiltonian flow equations in Sp(n)/U (n)
We will now derive the U (n 1)-invariant mKdV flow and U (n 1)-invariant SG
flow in the Hermitian symmetric space M = Sp(n)/U (n). Employing the notation
April 1, 2015
1:14
Stephen*Anco
11
and preliminaries in Sec. 2.1, we consider a non-stretching curve flow (t, x) that
has a U (n)-parallel framing along given as
1
e = (1, 0) R s(n 1, C) ' mk ,
= 8(n + 1)
u = (u, u) iR Cn1 ' h ,
(55)
(56)
and
hk = (hk , Hk ) R s(n 1, C) ' mk ,
h = (h , h ) iR C
n1
' m ,
(58)
$k = (W ) u(n 1) ' hk ,
$ = (w , w ) iR C
n1
(57)
(59)
' h ,
(60)
as well as
1
h = (h , h ) = ad(e)h = (2h , h ) iR Cn1 ' h
(61)
using the matrix identifications (20)(21), where u, h , w , h iR are imaginary

(complex) scalar variables, hk R is a real scalar variable, u, h , w , h Cn1
Cn1 are complex vector variables, Hk s(n 1, C) is complex symmetric matrix
variable.
We remark that, since the rank of the space M = Sp(n)/U (n) is n, then for n 2
the element (55) belonging to the Cartan subspace of m = sp(n)/u(n) determines
one particular algebraic equivalence class of non-stretching curve flows in which the
tangent vector x of the curve is identified with the orbit of this element under the
action of the gauge group Hk = Ad(H) of the U (n)-parallel frame.
In terms of the variables (55)(61), the Cartan structure equations (49)(50)
and (51)(52) are respectively given by
hkx + 2h u + h ut + h ut = 0,
Hkx
ht u
(62)
u h = 0,
t
(63)
t
hx + 2uhk + h u h u 2w = 0,
(64)
(65)
hx + uHk hk u + u h h u + w = 0,
and
Wxk + wt u ut w = 0,
(66)
wx
(67)
ut uw
+ w u + 2h = 0,
wx ut + uWk + uw w u h = 0.
(68)
Writing these equations (62)(68) in the operator form (46), we obtain the flow
equation

u
w
h
w
h
=H
+
,
=J
,
(69)
u t
w
h
w
h
April 1, 2015
12
1:14
Stephen*Anco
in terms of the Hamiltonian operator

Dx
i2Imhu, i
H=
u
Dx + u + uDx1 u
and the symplectic operator
1
D uDx1 u
J = 41x
2 u uDx1 u

iImhu, i + 2uDx1 Rehu, i
.
Dx u + 2uDx1 Rehu, i + uDx1 u
(70)
(71)
5.1. mKdV flow

The mKdV flow is produced by the x-translation generator

1
h
2h
ux
.
=
=
h
ux
h
(72)
Substitution of this expression into the flow equation (69) yields an integrable mKdV
system for the variables (u, u):
ut 1 ux = 41 uxxx 32 u2 ux + i3Imhu, uxx i,
ut 1 ux = uxxx + 32 (|u|2 12 u2 )x u + 3(|u|2 21 u2 )ux 23 ux ux 34 uxx u,
(73)
where we have rescaled t t/, for convenience.
This system (73) has a bi-Hamiltonian structure and exhibits invariance under
the unitary group U (n1) acting on u and u by the transformations Ad(R)(u, u) =
(u, uR1 ) for all matrices R Hk = U (n 1).
5.2. SG flow
The SG flow is defined by

0=
w
w

=J

h
.
h
This yields the flow equation

1
ut
h
2h
=
=
ut
h
h
(74)
(75)
with (h , h ) satisfying
hx + 2uhk + h ut h ut = 0,
(76)
hx + uHk hk u + u h h u = 0,
(77)
where hk and Hk are determined by equations (62)(63).

Similarly to the method 9,10,13,14,24 used to derive SG flows in other symmetric
spaces, we seek a local expression for hk and Hk through an algebraic reduction of
the form
Hk = ht h s(n 1, C),
(78)
April 1, 2015
1:14
Stephen*Anco
13
with some coefficient (hk , h ) C. Notice that, under gauge transformations (34),
it is precisely the components Hk and h that change while the components hk and
h are invariant. This observation provides a geometrical motivation for the form
of the reduction (78).
To find the coefficient , we substitute expression (78) into equation (63) and
use equations (76) and (77) to eliminate x derivatives of h , h . Then since the
matrix Hk is symmetric, we expand equation (63) as a linear combination of the
symmetric matrices ht h and ut h + ht u. Putting their coefficients to zero, we
obtain
x 22 (h ut ) 2u = 0,
(79)
hk + h 1 = 0.
(80)
From equation (80) we obtain

=
hk h
1
= 2
C.
hk + h
hk + |h |2
(81)
The remaining equation (79) is then just a consistency condition between equation
(81) for and the equations (76)(77), which holds identically.
Hence we have found an expression for Hk as a function of hk , h h :
Hk =
hk h t
1
h h .
ht h = 2
hk + h
hk + |h |2
(82)
Next we find hk as a function of h , h . To do that we use the conservation law

Dx (h2k + |Hk |2 h2 + 2|h |2 ) = 0
(83)
which is admitted by the system of equations (76)(77) and (62)(63), where

|Hk |2 := tr(Hk Hk ) = ||2 |h |4 =
||2 = =
h2k
|h |4
,
+ |h |2
1
1
= 2
.
|hk + h |2
hk + |h |2
(84)
(85)
This conservation law is also given by Dx (hh, hi) = 0 in which h = hk + h is given

by equations (57) and (58).
Substitution of expressions (84) and (85) into the conservation law (83) gives
||2 + ||2 |h |4 + 2|h |2 = 1.
(86)
Solving this quadratic equation for ||2 , we obtain

p
||2 = (|h |2 21 ) 12 1 4|h |2 .
(87)
Then from equations (81) and (85) we derive the expression

p
hk = ||2 |h |2 .
(88)
April 1, 2015
14
1:14
Stephen*Anco
Finally, we take the x derivative of the flow equation (75) and substitute hk
and Hk given by the equations (88) and (82). Thus we get the following hyperbolic
system for the variables (u, u):
utx = 4uA + i4Imhut , ui,
utx = |u|2
A 21 ut
ut + Au uut 12 ut u,
A2 + 14 |ut |2
(89)
in which
A := hk = 12
q
p
1 2|ut |2 + 14 u2t 1 4|ut |2 .
(90)
This system (89) is invariant under the unitary group U (n1), which acts on u and u
by the transformations Ad(R)(u, u) = (u, uR1 ) for all matrices R Hk = U (n1).
Conclusion
The Hermitian symmetric space Sp(n)/U (n) is one of few such spaces listed in the
classification of symmetric spaces 16 . By adapting the moving parallel frame method
developed in Ref. 15 to derive group-invariant bi-Hamiltonian integrable systems
from non-stretching geometric curve flows in Riemannian symmetric spaces, we
have obtained new integrable mKdV and SG systems for coupled scalar-complex
vector variables. These systems are invariant under the unitary group U (n 1) and
have a bi-Hamiltonian structure.
Our derivation makes use of the complex structure of the symmetric space
Sp(n)/U (n) in an essential way to formulate a natural complex matrix representation for the spaces sp(n) and u(n) as well as for their Lie bracket relations. This
approach resolves the problem of how to express the usual real matrix representation for sp(n) and u(n) in a coupled complex form that transforms properly under
the unitary gauge group U (n). The same method can be applied to quaternionic
symmetric spaces as well.
Another problem that we have addressed is how to carry out systematically the
algebraic reduction needed to get a SG system from the underlying non-stretching
curve flow 24,9,10,13,14 . We show that this reduction has a very simple geometrical
formulation by using the projection of the parallel part of the curve flow into the
complement of tangent direction along the curve. This method is motivated by our
observation that the components given by this projection are invariant under gauge
transformations on the parallel frame along the curve. We are currently extending
this geometrical reduction for SG curve flows to general symmetric spaces 25 .
In a different direction, we plan to look at how to extend the work in Ref. 15
to derive nonlinear Schrodinger systems (complex and quaternionic) from nonstretching curve flows in symmetric spaces with Hermitian or quaternionic structures. Work is underway using examples of low-dimensional symmetric spaces 26 .
April 1, 2015
1:14
Stephen*Anco
15
Acknowledgments
EA acknowledges support from the Department of Mathematics and Statistics at
Brock University during a research visit in which this paper was finalized. SCA is
supported by an NSERC research grant.
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
R.L. Bishop, Amer. Math. Monthly 82 (1975), 246251.

R. Hasimoto, J. Fluid Mechanics 51 (1972), 477485.
A. Doliwa, P.M. Santini, Phys. Lett. A 185 (1994), 373384.
J. Langer and R. Perline, Phys. Lett. A 239 (1998), 3640.
J. Langer and R. Perline, Pac. J. Math. 195 (2000), 157178.
G. Mar Beffa, J.A. Sanders, and J.-P. Wang, J. Nonlinear Sci. 12 (2002), 143-167.
J. Sanders and J.-P. Wang, Moscow Mathematical Journal 3 (2003), 13691393.
S.C. Anco, J. Phys. A: Math. Gen. 39 (2006), 20432072.
S.C. Anco, SIGMA 2 (2006), 044 (18 pages).
S.C. Anco, in Symmetries and overdetermined systems of partial differential equations,
223250, IMA Volumes in Math. and Appl. 144 (Springer, New York, 2008).
E. Asadi and J. Sanders, in SPT 2007Symmetry and perturbation theory, 2128
(World Sci., Hackensack, NJ, 2008).
E. Asadi, PhD Thesis, Vrije Universiteit, Amsterdam 2008.
S.C. Anco and E. Asadi, J. Phys. A: Math. Theor. 42 (2009), 485201 (25 pages).
S.C. Anco and E. Asadi, J. Phys. A: Math. Theor. 45 (2012), 475207 (37 pages).
S.C. Anco, J. Geom. Phys. 58 (2008), 137.
S. Helgason, Differential Geometry, Lie Groups, and Symmetric Spaces (Amer. Math.
Soc., Providence, 2001).
S. Kobayashi, K. Nomizu, Foundations of Differential Geometry Vol. I and II, (Wiley
1969).
J. Lepowsky and G.W. McCollum, Trans. Amer. Math. Soc. 216 (1976), 217-228.
R.W. Sharpe. Differential Geometry, Graduate Texts in Mathematics Vol. 166
(Springer-Verlag, New York, 1997).
C.-L. Terng and G. Thorbergsson, Results Math. 40 (2001), 286-309.
G. Mari Beffa, SIGMA 4 (2008), 034 (23 pages).
G. Mari Beffa, Physica D 238 (2009), 100115.
G. Mari Beffa, Pacific J. Math. 247 (2010), 163-188.
S. Anco, T. Wolf, J. Nonlinear Math. Phys. 12 (2005), 1331. ibid. J. Nonlinear Math.
Phys. 12 (2005), 607608.
S.C. Anco and E. Asadi, In preparation.
A.M.G. Ahmed, S.C. Anco, E. Asadi, In preparation.
April 1, 2015
1:15
Ugur*Camci


Noether gauge symmetries of geodesic motion in stationary and

nonstatic G
odel-type spacetimes
Ugur Camci
Department of Physics, Akdeniz University, 07058 Antalya, Turkey
ucamci@akdeniz.edu.tr
In this study, we obtain Noether gauge symmetries of geodesic motion for geodesic Lagrangian of stationary and nonstatic G
odel-type spacetimes, and find the first integrals
of corresponding spacetimes to derive a complete characterization of the geodesic motion. Using the obtained expressions for x a of each spacetimes, we explicitly integrate the
geodesic equations of motion for the corresponding stationary and nonstatic G
odel-type
spacetimes.
Keywords: G
odel-type spacetime, geodesic equation, Noether gauge symmetry
1. Introduction
G
odels stationary solution1 for the Einstein field equations gives an exact cosmological solution of Einsteins field equations. This solution contains closed timelike
curves (CTCs) that is a particular property of Godels universe and it is geodesically
complete2 . Furthermore, it has no singularity and horizon. Firstly, the motion of free
particles in this cosmological model was analyzed by Kundt3 and Chandrasekhar
and Wright4 . Later the geodesic equations of motion in Godel-type spacetimes have
been analyzed by several authors. A physical model for the geodesic motion using
the effective potential was given by Novello et al.5 Then, Calvao et al.6 followed by
Novello et al.5 give a complete discussion of timelike geodesics and also treat null
geodesics for stationary G
odel-type spacetimes. The geodesic equations for some
special classes of stationary G
odel-type spacetimes with seven isometries were integrated by Reboucas and Teixeira7 . The analytical solution of the geodesic equations
of the original stationary G
odels universe for both massive and massless (lightlike)
particles has been derived by Grave et. al. 8 in a special set of coordinates. They have
generalized the work of Kajari et. al. 9 on the solution of lightlike geodesic equations. Dautcourt10 considered only the lightlike case and studied the lightcone of
the stationary G
odel-type metrics. It is indicated by astronomical observations that
our universe is rotating. The concept of a rotating universe with vanishing expansion and shear appeared with G
odels universe model, and the rotational symmetry
of G
odels metric comes from the existence of CTCs. But it is well known that
the universe is also expanding. Therefore, it would be important to consider and
1
April 1, 2015
1:15
Ugur*Camci
U. Camci
investigate of the nonstatic rotating universe model of Godel. As far as we know, a

study of geodesic motion for the nonstatic Godel-type spacetime is not exist yet.
Finding a simple expression for a constant of motion, i.e. a first integral, it will
be useful to solve the geodesic equations of motion. The Lagrangian formalism is
used to derive the equations of motion. Differential equations to posses a Lagrangian
are associated with the Noether symmetries that are describing physical features of
differential equations in terms of conservation laws admitted by them. The strict
Noether symmetry approach without a gauge term (see Refs.1114 ) is a kind of
symmetry in which the Lie derivative of Lagrangian L arising from the metric of
interest dragging along a vector field Y vanishes, i.e. Y L = 0. The Noether gauge
symmetry (NGS) approach which includes a gauge term1521 is a generalization of
the former strict Noether symmetry approach, and it will be discussed in section
4. Recently, we have obtained the NGSs of geodesic Lagrangian L and explicitly
integrated the geodesic equations of motion for the corresponding stationary Godeltype spacetimes22 .
Now let us recall the spacetime symmetries. Let X be any global smooth vector
field and gab the metric tensor field of any type on manifold M , then the natural concept of a symmetry is geometrically given as a mapping, and reduced to a
differential relation between X and gab as23
X gab = 2gab
(1)
where X is the Lie derivative operator along the vector field X, and = (xa )
is a conformal factor. The group of conformal motions generated by a conformal
Killing vector (CKV) field X is defined by Eq. (1). For ;ab 6= 0, the CKV field
X is said to be proper, otherwise it is a special conformal Killing vector (SCKV)
field (;ab = 0). The vector field X is a homothetic vector (HV) for ,a = 0, and
it is an isometry or a Killing vector (KV) field for = 0. The set of all CKV
(respectively SCKV, HKV and KV) form a finite-dimensional Lie algebra denoted
by C (respectively S, H and G). The connection between the KVs and NGSs of
spacetimes is examined by several works. For the spaces of different curvatures such
as maximally symmetric spacetimes and Bertotti-Robinson like spacetime, the existence of new conserved quantities has been discussed and conjectured15,16,17 . The
NGSs of FRW and class A Bianchi type homogeneous spacetimes have been studied
by Tsamparlis and Paliathanasis18,19 . Recently, a classification of plane symmetric
static spacetimes according to their geodesic Lagrangian considering NGS approach
has been provided by Ali and Feroze20 .
This study is organized as follows. In the following section, we give a short review about stationary and nonstatic Godel-type spacetimes and their properties. In
section 3, we present the equations of geodesic motion for stationary and nonstatic
G
odel-type spacetimes and their geodesic Lagrangian and Hamiltonian structure. In
section 4, we discuss the Noether gauge symmetry approach of geodesic Lagrangian
for G
odel-type spacetimes. In sections 5 and 6, we give exact solutions of Noether
gauge symmetry equations for stationary and nonstatic Godel-type spacetimes, re-
April 1, 2015
1:15
Ugur*Camci
Noether gauge symmetries in G

spectively. In final section, our conclusion with a brief summary of findings are
presented.
2. Stationary and Non-static G
odel-type Spacetimes
In cylindrical coordinates, xa = (t, r, , z), a = 0, 1, 2, 3, the line element for stationary G
odel-type spacetimes can be written2,24
2
ds2 = [dt + H(r)d] dr2 D2 (r)d2 dz 2 .
(2)
The necessary and sufficient conditions for a Godel-type manifold to be spacetime

homogeneous (STH, hereafter) are found as25,26,27,28
D00
= const m2 ,
(3)
D
0
H
= const 2
(4)
D
where prime denotes derivative with respect to the radial coordinate r. It is proved
that the necessary and sufficient conditions for a four-dimensional Riemannian
G
odel-type manifold to be locally homogeneous are those given by Eqs. (3) and
(4). Also, the four-dimensional homogeneous Riemannian Godel-type manifolds are
locally characterized by two independent parameters m2 and : the pair of (m2 , )
identically specify locally equivalent manifolds24 .
All STH Riemannian manifolds endowed with a stationary Godel-type spacetime
(2) from the general solution of Eqs. (3) and (4) are obtained as follows:
Class I : m2 > 0, 6= 0. For this case, the metric coefficients can be written by
H(r) =
2
[1 cosh(mr)]
m2
and
D(r) =
1
sinh(mr).
m
(5)
Class II : m2 = 0, 6= 0. The metric coefficients are

H(r) = r2
and
D(r) = r,
(6)
where only the essential parameter appears.

Class III : m2 2 < 0, 6= 0. Similarly, in this class the metric coefficients
leads to
2
1
H(r) = 2 [cos(r) 1] and D(r) = sin(r).
(7)
Class IV : m2 6= 0, = 0. We refer to the manifolds of this class as degenerated

G
odel-type manifolds, since the cross term in the line element, related to the rotation
in the G
odel model, vanishes. By a trivial coordinate transformation, one can
make H = 0 with D(r) given, respectively, by Eqs. (5) or (7) depending on whether
m2 > 0 or m2 2 < 0. Throughout this paper we have used the following
property
0 0
D
D2
= 1,
(8)
D
April 1, 2015
1:15
Ugur*Camci
U. Camci
which is valid for STH static G

odel-type spacetimes only. The form of the scalar
curvature R for the static G
odel-type spacetime is R = 2(w2 m2 ).
In standard coordinates, xa = (t, x, y, z), the nonstatic Godel-type spacetime
has the form
1
2
ds2 = [dt + Kemx dy] K 2 e2mx dy 2 dx2 dz 2 ,
(9)
2
where K is a function of time only, and m is a constant that does not appear in
G
odels original paper. For the nonstatic Godel-type spacetime, the scalar curvature
is given by R = m2 2K/K.
If K is a constant, the above metric is reduced to
G
odels the original stationary rotating universe.
The G
odel metrics are mainly interesting for their high degree of
symmetry25,27,28,29 . All classes of stationary Godel-type spacetimes admit at least a
G5 group of motions. In a special case m2 = 4w2 , it has been shown that the group
of motions is G7 , a maximal symmetry group of stationary Godel-type spacetimes27 ,
which are given in appendix A. It has been pointed out that not only the original
G
odel metric but also the STH stationary Godel-type spacetimes does not admit
HVs 30,31 . The proper CKVs and complete conformal algebra of a stationary Godeltype spacetime have been computed by Tsamparlis et al.32 The Ricci collineations
(RCs) and contracted RCs of STH stationary Godel-type spacetimes are studied
by Melfo et al.31 The matter collineations (MCs) of that spacetimes are obtained
by Camci and Sharif33 . In this work, we aim to give a complete classification for
stationary and nonstatic G
odel-type spacetimes according to the Noether gauge
symmetries of their geodesic Lagrangian.
3. The Equations of Geodesic Motion
For any spacetime metric, the motion of massive or massless particles is described
by geodesic Lagrangian as
1
gab x a x b V (xc ).
2
which gives rise to the equations of motion
L=
(10)
x
a + abc x b x c = g ab V,b
(11)
with the following constraint to be fulfilled

gab x a x b = ,
(12)
where the functions abc (xe ) are the connection coefficients of the metric. Here,
the dot over a symbol denotes a derivative with respect to proper time (for
massive particle motion) or with regard to an affine parameter (for masslesslightlike-geodesics) along the solution curve, and V (xe ) is the time independent
potential function. Furthermore, the conservative force field F a (xe ) can written as
F a (xe ) = g ab V,b . We have = 1, 0, +1 for spacelike, lightlike (or null) and timelike
geodesics, respectively.
April 1, 2015
1:15
Ugur*Camci

Using the stationary G

odel-type spacetime (2), the geodesic Lagrangian density
takes the following form
i
1 h 2
(13)
L=
t r 2 z 2 + (H 2 (r) D2 (r)) 2 + H(r)t V (t, r, , z).
2
One may obtain the Euler-Lagrange equations of motion by varying of the geodesic
Lagrangian (13) with respect to the coordinates t, r, and z as given by
t + H + H 0 r + V,t = 0,
r + (HH 0 DD0 ) 2 + H 0 r V,r = 0,
H t + (H 2 D2 ) + 2(HH 0 DD0 )r + H 0 r t V, = 0,
(14)
z V,z = 0.
(17)
(15)
(16)
The energy functional or Hamiltonian of the dynamical system, EL , associated with

the Lagrangian (13) is found as
L
L
L
L
+ r
+
+ z
L
EL = t
r
z
t
i
1 h 2
t r 2 z 2 + (H 2 D2 ) 2 + H t + V (t, r, , z).
(18)
=
2
pr = r,
Introducing the momenta pa = gab x b = L/ x a we have pt = t + H ,
2
2
p = H t + (H D ), pz = z.
Then EL becomes

1 2
1
2
EL =
pt p2r p2z 2 (p Hpt ) + V (t, r, , z).
(19)
2
D
For the nonstatic G
odel-type spacetime (9), the geodesic Lagrangian (10) has
the form

1 2
1
L=
t x 2 z 2 + K 2 (t)e2mx y 2 + K(t)emx ty V (t, x, y, z).
(20)
2
2
Varying this Lagrangian with respect to t, x, y and z one can find the Euler-Lagrange
equations of motion as follows:
1
t + Kemx y + mKemx x y KK,t e2mx y 2 + V,t = 0,
2
m
x
+ K 2 e2mx y 2 + mKemx ty V,x = 0,
2
1
K,t 2 emx
t + Kemx y + mtx + K,t emx ty + mKemx x y +
t +
V,y = 0,
2
K
K
z V,z = 0.
(21)
(22)
(23)
(24)
The energy functional associated with the Lagrangian (20) is

L
L
L
L
+ x
+ y
+ z
L
x
y
z
t

1 2
1
=
t x 2 z 2 + K 2 e2mx y 2 + Kemx ty + V (t, x, y, z).
2
2
EL = t
(25)
April 1, 2015
1:15
Ugur*Camci
U. Camci
In terms of the momenta

pa , i.e., pt = t + Kemx y,
px = x,
py =

mx
mx
t + Ke y/2
and pz = z,
the EL becomes
Ke
"

2 #
1 2
emx
2
2
EL =
p px pz 2 pt
py
+ V (t, x, y, z).
(26)
2 t
K
4. The Noether Gauge Symmetry Approach
The Noether gauge symmetry (NGS) is defined as follows. Let us consider a vector
field
Y=
+ a a
(27)
where and a are depend on and xa . Here, is an independent variable and

xa ( ) are dependent variables, i.e. the configuration space of the Lagrangian (10) is
Q = (xa ), whose tangent space is T Q = (xa , x a ). The first extension of the above
vector field is given by
Y[1] = Y + a
,
xa
(28)
in which a = D a x a D , where D is the operator of total differentiation with

respect to , i.e., D = / + x a /xa .
The vector field Y is a NGS of a Lagrangian L(, xa , x a ) if there exists a gauge
function f (, xa ) such that 34
Y[1] L + L (D ) = D f
(29)
which takes the alternative form18,19

,a = 0,
b
gab ,
= f,a
gab = , gab ,
V = , V f,
(30)
where is the Lie derivative operator along = a xa . It is noted here that the set
of all NGSs form a finite dimensional Lie algebra denoted by N . The significance
of NGS is clearly comes from the fact that if Y is the Noether gauge symmetry
corresponding to the Lagrangian L(, xa , x a ), then
I = L + ( a x a )
L
f
x a
(31)
is a first integral or a conserved quantity associated with NGS vector field Y. Then
it follows from this relation for the geodesic Lagrangian (10) that
I = EL + a
L
f,
x a
(32)
where EL = x a L/ x a L. Now we seek the condition in order that the Lagrangian

(13) would admit NGS.
Recently the Noether gauge symmetries of geodesic Lagrangian for some
spacetimes have been calculated, and classified according to their symmetry
April 1, 2015
1:15
Ugur*Camci

generators1519,22 . Here, we investigate the NGSs of then stationary and nonstatic

G
odel-type spacetimes.
For the stationary G
odel-type spacetimes (2), the Noether gauge symmetry
equations (29) or (30) yield 19 partial differential equations:
,t = 0,
0
,
+
3
, +
0 1
,r = 0,
2
H,
, = 0,
f,t = 0,
,z = 0,
1
,
0
H,
(33)
2
2
),
+ f,r = 0,
+ (H D
f, =
0
2
0
2
1
f,z = 0, 2,t + 2H,t , = 0,
,r + H,r ,t = 0,

0
0
2
2
2 2
H + , + H ,t + , + (H D ),t H, = 0,
0
2
3
1
,z
+ H,z
,t
= 0,
2,r
, = 0,
0
1
2
2 2
1
3
H,r , + (H D ),r = 0,
,z
+ ,r
= 0,
0
0 1
2
2 2
0
2(HH DD ) + 2(H D ), + 2H, (H 2 D2 ), = 0,
0
2
3
3
H,z
+ (H 2 D2 ),z
,
= 0,
2,z
, = 0,
0
1
2
3
V,t + V,r + V, + V,z + , V + f, = 0.
0,
(34)
(35)
(36)
(37)
(38)
(39)
(40)
(41)
The general solution to the above NGS equations has been found in Ref. [22],and
reviewed in the next section for each classes I-IV for the Godel-type spacetimes.
For the nonstatic G
odel-type spacetimes (9), the NGS equations (29) are found
as follows:
,t = 0,
,x = 0,
,y = 0,
,z = 0,
(42)
1
2
0
2
1
0
f,y = 0, (43)
,
+ Kemx ,
f,t = 0, ,
+ f,x = 0, Kemx ,
+ K 2 e2mx ,
2
3
0
2
0
2
1
,
+ f,z = 0, 2,t
+ 2Kemx ,t
, = 0,
,x
+ Kemx ,x
,t
= 0,
(44)
K,t 0
emx 0
1
0
2
2
+ m 1 +
+ ,t
+ ,y
+ Kemx ,t
, = 0,
K
K ,y
2
0
2
3
1
,z
+ Kemx ,z
,t
= 0,
2,x
, = 0,
1
2
1
3
1
0
= 0,
,z
+ ,x
= 0,
H,y
Kemx ,x
K 2 e2mx ,x
2
K,t 0
2
1
2
0
+ m 1 + ,y
+ emx ,y
, = 0,
K
K
2
1 2 2mx 2
3
mx 0
3
,y Ke ,z K e
,z = 0,
2,z
, = 0,
2
V,t 0 + V,x 1 + V,y 2 + V,z 3 + , V + f, = 0.
(45)
(46)
(47)
(48)
(49)
(50)
We will give the solution of above NGS equations in section 6.

5. The Solution of Noether Gouge Symmetry Equations for
Stationary G
In this section we review our earlier work on geodesics of stationary Godel-type
spacetimes22 . After some algebra, we have calculated the general solution to Eqs.
(33)-(41) in order to get NGSs for each of the classes I, II, III and IV when the
potential function V (xe ) vanishes.
April 1, 2015
1:15
Ugur*Camci
U. Camci
Class I : m2 > 0, 6= 0.
The NGS vector field components and the gauge function for class I with the
condition m2 6= w2 are found as
= c1 , 0 =
2 =
2w
H
(c2 cos + c3 sin ) + c4
+ c5 , 1 = c2 sin + c3 cos ,
D
m
D0
(c2 cos + c3 sin ) c4 m,
D
3 = c6 + c7 ,
f = c6 z,
(51)
which means that we have seven NGSs, i.e. the five KVs X1 , ..., X5 given by (A.2)
in appendix A, and two non-Killing NGSs
Y1 = ,
Y2 = z
with gauge term f = z.
(52)
For class I, the first integrals (32) associated with X1 , ..., X5 , Y1 and Y2 are
found as
h
i
2w
I1 m H t + (H 2 D2 ) ,
(53)
I1 = t + H ,
I2 = z,
I3 =
m
n
o

sin
I4 =
H(1 + D0 )t + H 2 + (H 2 D2 )D0 cos r,
(54)
D
n
o

cos
I5 =
H(1 + D0 )t + H 2 + (H 2 D2 )D0 + sin r,
(55)
D
I6 = EL ,
I7 = z + z,
(56)
where the EL is the Hamiltonian (19) of the dynamical system and yields
(
)

2
1
1
2w
I3
2
2
2
EL =
I1 I2 2 HI1
I1
r .
2
D
m2
m
(57)
L
Hence the Hamiltonian EL is conserved ( EL = dE
d = 0). Therefore, the vector
field Y1 is the trivial NGS.
r and z from Eqs. (53)-(56). Then the integration for the
,
It can be solved t,
coordinate z easily give z = pz + I7 , and we obtain all x a and a constraint
equation from the first integrals given above as

1
t = 2 p H + pt (D2 H 2 ) ,
D
1
= 2 [pt H p ] ,
D
z = pz ,
2
p2t
(58)
(59)
(60)
V (r),
(61)
r = I4 cos + I5 sin ,
(62)
(63)
r =
pt H + p D + (I4 sin + I5 cos )D = 0,

where we have defined the effective potential
V (r) :=
1
2
(pt H p ) + p2z + .
D2
(64)
April 1, 2015
1:15
Ugur*Camci

and the constants of motion pt I1 , p 2wI1 /m2 I3 /m and pz I2 represent

the conservation of energy, angular momentum and z component of momentum,
respectively. Furthermore, we can see from the Eq.(12) that = 2I6 , where =
1, 0+1 for spacelike, null and timelike geodesics, respectively. The Eq. (61) with the
effective potential in (64) is the generalization of radial equation given in reference
[6]. Furthermore, we have seen that there exist a new radial equation, the Eq. (62),
which depends not on r, but only on .
Introducing a new variable u = m2 H/4w which is equivalent to sinh2 (mr/2) for
the class I (see the Eq. (42) of ref. 6 ), the Eq. (61) gives
2
u =
m2 p2t

m2 2
2
2
,
u + (1 + 2w)u
4
(65)
where , 2 and are defined as

= 2 1 +
4w2
m2
2 =
p2z +
,
p2t
p
.
pt
(66)
Using the above definitions, the effective potential takes the form of
V (r) =
2
p2t
H 2 + 2 p2t .
2
D
(67)
Therefore, from the radial equation (61), one can accomplish a complete characterization of the motion which depends essentially on the parameters , , m and w.
This characterization separates the motion into three distinct cases > 0, = 0,
and < 0. For the trajectories of physical particles, it follows from the Eq. (61)
that 0 2 1 (see also Refs. [5] and [6]).
The general solution of (65) is given by
i
p
1 h
1 2 + 2w + (1 2 + 2w)2 m2 2 sin(mpt ( 0 )) ,
2
(68)
where 6= 0 and (1 2 + 2w)2 m2 2 0. Using the new variable u = m2 H/4w
in Eqs. (58) and (59) for t( ) and ( ) we find
u( ) =

2
p
(1
+
w)
4w
u
t
t =
+ pt 1 2
,
u
m
1+u
wpt
pt m2 /4
=
,
1+u
u(1 + u)
(69)
(70)
Now we shall use the solution (68) of the radial equation to solve the above equations
for t( ) and ( ). For 6= 0, after some algebra, we have obtained the following
April 1, 2015
10
1:15
Ugur*Camci
U. Camci
general solutions of the Eqs.(69) and (70)

"
#
(1 + p) tan(mpt ( 0 )/2) + q
2w( + 4w/m2 )
p
arctan
t( ) = p
m (1 + p)2 q 2
(1 + p)2 q 2

4w2
+pt 1 2 + t0 ,
(71)
m
"
#
(1 + p) tan(mpt ( 0 )/2) + q
m( + 4w/m2 )
p
( ) = p
arctan
2 (1 + p)2 q 2
(1 + p)2 q 2

2
(72)
arctan
{p tan(mpt ( 0 )/2) + q} + 0 ,
m
where (1 + p)2 > q 2 , t0 = t(0) and 0 = (0) are integration constants. Here we
have introduced the parameters p and q as
s
m2 2
1 2 + 2w
,
q := p2
,
(73)
p :=
2
4
where p2 m2 2 /4. For = 0, the general solution of the Eq. (65) yields
u( ) =
( 2 1)
[1 + sin(mpt ( 0 ))] ,
2
(74)
and using this solution in Eqs. (69) and (70) it follows that

(1 + p) tan(mpt ( 0 )/2) p
8w2
p
arctan
t( ) =
1 + 2p
m3 (1 + 2p)

2
4w
(75)
+pt 1 2 + t0 ,
m

(1 + p) tan(mpt ( 0 )/2) p
2w
( ) = p
arctan
+ 0 , (76)
1 + 2p
m (1 + p)
where p = (1 2 )/2.
In the special class I case, where m2 = 4 2 , i.e. m = +2w, we find the NGS
vector field components and the gauge function as
= c1 ,
2w H
H
(c2 cos + c3 sin ) + c4
[c8 cos(mt + ) + c9 sin(mt + )] + c5 ,

D
m
D
1 = c2 sin + c3 cos + c8 sin(mt + ) c9 cos(mt + ),
(77)
0
D
1
2 = (c2 cos + c3 sin ) + [c8 cos(mt + ) + c9 sin(mt + )] c4 m,
D
D
3 = c6 + c7 , f = c6 z.
0 =
Then one finds that there are nine NGSs which are seven KVs X1 , ..., X7 given by
(A.3) in appendix A and two non-Killing NGSs given by (52), where m = +2.
April 1, 2015
1:15
Ugur*Camci

11
The first integrals (53)-(56) for X1 , ..., X5 , Y1 and Y2 together with m = +2w
are same ones for this special class I. The remaining first integrals associated with
X6 and X7 given in (A.3) are

I3
2w
cos(mt + )
H 2 I1 +
sin(mt + ) r,
(78)
I8 =
D
m
m

sin(mt + )
I3
2w
I9 =
H 2 I1 +
+ cos(mt + ) r.
(79)
D
m
m
which yields
r = I8 sin(mt + ) + I9 cos(mt + ),
(80)
pt H p + [I8 cos(mt + ) + I9 sin(mt + )] D = 0.
(81)
Class II : m2 = 0, 6= 0.
For the class II, where H(r) = wr2 and D(r) = r, the NGS vector field
components and the gauge function are obtained as
= c1 ,
0 = wr(c2 cos + c3 sin ) + c4 ,
1
2 = (c2 cos + c3 sin ) + c5 ,
r
1 = c2 sin + c3 cos , (82)
3 = c6 + c7 ,
f = c6 ,
(83)
which give seven NGSs, i.e. the five KVs X1 , ..., X5 given in appendix, and two nonKilling NGSs same as (52). Hence, the first integrals associated with those vector
fields are
I1 = t wr2 ,
I2 = z,
I3 = wr2 t + r2 (w2 r2 1),

1
I4 = sin wrI1 I3 r cos ,
r

1
I5 = cos wrI1 I3 + r sin ,
r
i
1 h 2
I6 =
t + r 2 + r2 (1 w2 r2 ) 2 + z 2 + wr2 t,
2
I7 = z + z,
(84)
where I6 = EL and = 2I6 which takes the values 1, 0 and 1 for spacelike, null
and timelike geodesics, respectively. Using the obtained first integrals given above,
it follows that the coordinate z is z( ) = pz + I7 and

t = pt 1 w2 r2 p w,
(85)
p
= wpt 2 ,
(86)
r
r 2 = p2t V (r),
(87)
(88)
p
I4 sin + I5 cos = wpt r ,
r
(89)
April 1, 2015
12
1:15
Ugur*Camci
U. Camci
where pt = I1 , pz = I2 and p = I3 are constants of motion related with energy,

z component of momentum and angular momentum, respectively, and the effective
potential is defined as
h
p i2
+ p2z + .
V (r) = wpt r +
(90)
r
It is easily seen that one can derive the radial equation (87) from Eqs. (86), (88)
and (89). Introducing the new variable v = r2 the radial Eq. (87) becomes
v 2 = 4w2 p2t v 2 + 4v 4p2 ,
with defined by = p2t 2wpt p p2z . The general solution of (91) is
q
+ 2 4w2 p2t p2 sin [2wpt ( 0 )] ,

v( ) r2 =
2
2
2w pt
(91)
(92)
where 2wpt p . Then after substitution (92) into (85) the general solution of
the resulting differential equation for t( ) gives

wq 2
4w2 p2t p2 cos [2wpt ( 0 )] + t0 .
+
(93)
t( ) = pt wp
2pt
2
Finally, using (92) in (86) the behavior of the coordinate is given by
(
)
2wpt p
tan [wpt ( 0 )] + 2w2 p2t
1
p
+ 0 ,
( ) = wpt + p
tanh
4 2 w4 p4t 2
4 2 w4 p4t 2
(94)
q
where = 2 4w2 p2t p2 and 4 2 w4 p4t > 2 4w2 p2t p2 . For a special case such
as p = I3 = 0 the first integrals simplify considerably. In the latter special case the
geodesic equations can be integrated completely and the solution reads

1 2
t( ) = pt
(I4 + I52 )
2pt

1 2
(I4 I52 ) sin(2wpt ) 4I4 I5 cos2 (wpt ) + t0 ,
(95)
+
2
4wpt
1
r( ) =
[I5 cos(wpt ) I4 sin(wpt )] ,
(96)
wpt
( ) = wpt + 0 ,
(97)
z( ) = pz + I7 ,
(98)

where p2t I22 + I42 + I52 = , and 0 is an integration constant.
Class III : m2 2 < 0, 6= 0.
In class III, where m2 2 < 0, 2 > 0 and w 6= 0, it follows that the KVs
X1 , X2 , X4 , X5 and the non-Killing NGSs Y1 , Y2 are same form as given the above,
but only X3 has the form as (2w/)t + .
Class IV : m2 6= 0, = 0.
In this class, where m2 6= 0, w = 0, the metric functions are taken as H(r) = 0
1
and D(r) = m
sinh(mr) for m2 > 0, or D(r) = 1 sin(r) for 2 = m2 > 0.
April 1, 2015
1:15
Ugur*Camci

13
The NGS vector fields and the gauge function for this class are
= c1 , 0 = c2 z + c3 s + c4 , 1 = c5 cos + c6 sin ,
(99)
0
D
2 =
(c5 sin + c6 cos ) + c7 , 3 = c2 t + c8 + c9 , f = c3 t c8 . (100)
D
This result yields nine NGSs, i.e., the six KVs X1 , ..., X6 given in appendix and two
non-Killing NGSs same as (52), and extra one non-Killing NGS
Y3 = t
with gauge term f = t.
(101)
If m2 = = 0, then the line element (2) is clearly Minkowskian. Therefore, this

particular case has not been included in this study.
The first integrals of this class are
I1 = t,
I2 = z,
I3 = z t tz,
I4 = r cos + DD
sin ,
0
I5 = r sin + DD
cos ,
i
1 h 2
t + r 2 + D2 2 + z 2 ,
I6 = D2 ,
I7 =
2
I8 = z + z, I9 = t t,
(102)
where I7 = EL . From the Eq.(12), we get that = 2I7 , and = 1, 0, 1 for

spacelike, lightlike and timelike geodesics, respectively. The above constants of mo and z.
r,
tion can be solved for t,
,
Then, we find from (102) that t = I1 I9
and z = I2 + I8 , which means that along any geodesic the time coordinate t and
the axial coordinate z vary uniformly with respect to its affine parameter . It also
follows from (102) that
I6
= 2 ,
D
(103)
I62
,
D2
D0
I4 sin + I5 cos +
I6 = 0,
D
I1 I8 I2 I9 = 0.
r 2 = I12 I22 + 2I7
(104)
(105)
(106)
(107)
Here, the constants of motion representing conservation of energy, angular momentum and z component of the momentum are pt = I1 , p = I6 and pz = I2 ,
respectively. Thus, we can write the Eq. (104) as
r 2 = p2t V (r),
(108)
where the effective potential is defined by

V (r) =
p2
+ p2z + .
D2
(109)
April 1, 2015
14
1:15
Ugur*Camci
U. Camci
Now, we introduce a new variable = m2 D2 = sinh2 (mr) for m2 > 0. Thus the
Eq. (108) in terms of the new variable yields

2 = 4m2 p2t 2 ( + m2 2 ) m2 2 ,
(110)
where = (p2z + )/p2t 1 and = p /pt . The general solution of this equation is
( ) =

1
( + m2 2 ) + ( m2 2 ) sin (2mpt ( 0 )) .
2
(111)
Considering this new variable in the Eq. (103) for ( ) we get

2
m p
=
,
(112)
which has the general solution as

"
#
m2 2 ( + m2 2 ) tan mpt ( 0 )
( ) = arctan
+ 0 .
2m
(113)
When the new variable has the form = 2 D2 = sin2 (r) for 2 = m2 > 0,
the Eq. (108) give

2 = 42 p2t 2 ( 2 2 ) 2 2 ,
(114)
and the general solution of this equation is
i
1 h
( ) = 3/2 4 ( 2 2 ) + (2 2 )e2pt ( 0 ) + 4e2pt ( 0 ) .
8
(115)
Then the Eq. (103) for ( ) has the general solution
#
" s
(2 2 ) 2pt ( 0 )
1

e
2 + 0 ,
(116)
( ) = 1 arctan
2 (1 + 2 2 )
where 1 =
(2 2 )(1+2 2 )
, > 0, 2 2 > and (1 + 2 2 ) > 0.
6. The Solution of Noether Gouge Symmetry Equations for

Nonstatic G
In this section, after some algebra, we have calculated the general solution to the
NGS Eqs. (42)-(50) for nonstatic Godel-type Spacetimes in order to get NGSs when
the potential function V (xe ) is a constant, say V0 .
If K is a constant, say k, then we find the NGS vector field components and the
gauge function as
2 = c2
emx
+ c3 , 1 = c2 y + c4 ,
mk
my 2
c4 my + c5 , 3 = c6 + c7 ,
2
0 = 2c2
= c1 ,

e2mx
mk 2
f = c6 z. (117)
April 1, 2015
1:15
Ugur*Camci

15
Thus we have seven NGSs, that is, the five KVs X1 , ..., X5
X3 = x myy , X4 = t ,

2 mx
m 2 e2mx
y ,
X5 =
e
t yx +
y
mk
2
mk 2
X1 = y ,
X2 = z ,
(118)
and two non-Killing NGSs

Y1 = ,
Y2 = z
with gauge term f = z.
(119)
The above KVs (118) are linearly independent KVs of Godels stationary
spacetime25 . For this case the first integrals (32) associated with X1 , ..., X5 , Y1
and Y2 are found as
I1 = ke
mx

t + k emx y ,
2
I2 = z,
I3 = x myI1 ,
I4 = t + kemx y,

m 2 e2mx
2I4 mx
I5 = y x + I1
y
+
e
,
2
mk 2
mk
I6 = EL ,
I7 = z + z,
(120)
(121)
(122)
(123)
where the EL is the Hamiltonian (19) of the dynamical system and yields

1 2
k 2 2mx 2
2
2
y + V0 .
EL =
I I2 x e
2 4
2
(124)
Then the constraint equation (12) to be fulfilled is

2
k2
t + kemx y x 2 e2mx y 2 z 2 = ,
2
(125)
which give
x 2 +
k 2 2mx 2
e
y = I42 I22
2
(126)

2
t + kemx y = I42 .
(127)
and
We point out that the Eqs.(126) and (127) are same equations as in Eqs. (11) and
(12) of Chandrasekhar and Wright4 to integrate explicitly geodesic equations in
G
odels universe, respectively. From the first integrals (120)-(123) the integration
x,
for the coordinate z easily gives z = pz + I7 , and we can find t,
y,
z and the
April 1, 2015
16
1:15
Ugur*Camci
U. Camci
constraint equation as follows:

2py mx
t =
e
pt ,
k

py
2
y = emx pt emx ,
k
k
z = pz ,
(128)
(129)
(130)
x = (mpy )y I3 ,
2
(131)
p2t
x = V (x),
mpy 2
2pt mx
py 2mx
e
+ I5 = 0,
y + I3 y
e
+
2
mk
mk 2
(132)
(133)
where V (x) is defined by

V (x) =
e2mx
pt emx
p y 2
+ p2z + ,
k
(134)
and py I1 , pz I2 and pt I4 represent constants of motion. Also it follows from

(12) that I6 = (/2 + V0 ). Now introducing a new variable u = emx in Eq. (132)
we get

2 2
4
2
2 2
2
u = m pt u +
u 2 ,
(135)
k
k
where = py /pt , = 2 + 1 and 2 = (p2z + )/p2t . The general solution of this
differential equation is given by
u( ) =
2 h
1 2( 2) sin (pt m ( 0 )) ,
k
where > 0. Thus using this new variable in Eq. (128) we find

t = pt 2 1
ku
(136)
(137)
which has the general solution of

"
#
2
2( 2) 1
t( ) =
arctanh
+ tan (pt m ( 0 )) pt ( 0 )+t0 , (138)
mp
p
p
p
where t0 is an integration constant, p = 2 2 8 + 7 and 6= 2 2/2. To find

the explicit form of y( ) we consider Eq.(131) instead of Eq.(129) and find that
"
#
pt ( 2) 2 cos pt m ( 0 )
1
+ I3 + y0 ,
y( ) =
(139)
mpy 1 + 2( 2) sin pt m ( 0 )
where y0 is a constant of integration.
April 1, 2015
1:15
Ugur*Camci

17
If K = k eat , where k and a are nonzero constants, then the NGSs and gauge
function are found as follows:

c2
2 (at+mx)
0
= c1 , =
ay + e
+ c3 , 1 = c2 y + c4 ,
m
k

2
c2 1 2(at+mx)
2
2
2 y
e
=
(a + m )
c3 ay c4 my + c5 ,
(140)
m k2
2
3 = c6 + c7 ,
f = c6 z,
which yields again seven NGSs, five of them are KVs X1 , ..., X5
X1 = y , X2 = z , X3 = x myy , X4 = t ayy ,

2
2 (at+mx)
1
1 2(at+mx)
1
2
2 y
ay + e
t yx +
(a + m ) 2 e
y (141)
X5 =
m
k
m
2
k
and two non-Killing NGSs same as (119). The first integrals for X1 , X2 , X3 , Y1 and
Y2 are same as the Eqs. (120) and (123). The remaining first integrals associated
with X4 and X5 are
I4 = t + keat+mx y ayI1 ,
(142)

2(at+mx)
2
e
y
I4
2
I1
(a2 + m2 )
+
ay + e(at+mx) . (143)
I5 = y x +
m
2
k2
m
k
Using the definitions for constants of motion pt and py (= I1 ) then the first integral
I4 becomes
I4 = pt (apy )y,
(144)
which means that y( ) is a constant such as

y=
1
(pt I4 ),
apy
(145)
where apy 6= 0. Therefore, the remaining first integrals give

t( ) = pt + t0 ,
a
x( ) = pt + x0 ,
m
z( ) = pz + I7 ,
a
m
I3 = pt + (I4 pt ),
m
a
p2z +
m2 a2
=
, py = keat0 +mx0 pt ,
p2t
m2
(146)
(147)
(148)
(149)
(150)
where t0 and x0 are integration constants, and m, a 6= 0.

For any other form of K(t), we have the following NGS vector field components
and gauge function from the NGS Eqs. (42)-(50) of nonstatic Godel-type Spacetimes
= c1 , 0 = 0, 1 = c2 , 2 = c2 my + c3 , 3 = c4 + c5 , f = c4 z, (151)
April 1, 2015
18
1:15
Ugur*Camci
U. Camci
which gives five NGSs, i.e., three KVs X1 , ..., X3

X1 = y ,
X2 = z ,
X3 = x myy ,
(152)
and two non-Killing NGSs as given in (119). The first integrals associated with the
above NGSs are same as (120) and (123). Using the appeared first integrals, one
can find that

K mx
mx
I1 = Ke
t + e y , I2 = z,
I3 = x myI1 ,
(153)
2
I4 = EL ,
I5 = z + z,
(154)
where EL is given by

K 2 2mx 2
1 2
2
2
e
y x I2 + V0 .
t + 2I1 y
EL =
2
2
(155)
We point out here that pt is not a constant of motion for this case. The constants
of motions for arbitrary K(t) except that it is a constant or exponential function
of t are py I1 and pz I2 . Thus, it follows from (153)-(155) that z = pz + I5 ,
i.e., the axial coordinate z vary uniformly with respect to the parameter along
the geodesic and
x = (mpy )y I3 ,
(156)
2 mx py mx
e
t ,
(157)
y = e
K
K
2py
x 2 + t2 = 2 e2mx p2z
(158)
K
To get the solution of the above system of differential equations we need to introduce
first the form of K(t), and later try to solve that system. For example, if py vanishes,
then we find that
x( ) = I3 + x0 ,
(159)
Z
dt
y( ) = 2em(I3 x0 )
+ y0 ,
K(t)
q
t( ) = (I32 + p2z + ) + t0 ,
(160)
(161)
where x0 , y0 and t0 are integration constants. Here the obtained solution is physically acceptable if = 1 (spacelike geodesics) and I32 + p2z < 1.
7. Conclusions
In this paper, we have obtained the NGSs of geodesic Lagrangian L for stationary
and nonstatic G
odel-type spacetimes. The stationary Godel-type spacetimes corresponding to the classes I, II, III and IV admit the algebra N7 G5 . In special class I
(where m2 = 4w2 ) and class IV, we have found 9 NGS generators. The NGS algebra
admitted by the special class I is N9 G7 . The stationary Godel-type spacetime in
class IV admits the algebra N9 G6 . The nonstatic Godel-type spacetimes admit
April 1, 2015
1:15
Ugur*Camci

19
the algebra N5 G3 . In special case where K(t) = keat , we have found 7 NGS
generators, and the NGS algebra is N7 G5 .
We obtained the first integrals admitted by geodesic Lagrangians for both stationary and nonstatic G
odel-type spacetimes that are due to the existence of NGS
vector fields including the KVs. Using the obtained first integrals in almost all classes
of G
odel-type spacetimes, we have derived the analytical solutions of geodesic equations which represent the usefulness of the NGSs.
Acknowledgements
This work was supported by The Scientific Research Projects Coordination Unit
of Akdeniz University (BAP). Project Number: 2013.01.115.003. The author would
like to thank The Abdus Salam International Centre for Theoretical Physics (ICTP)
for the financial grant of the successful conference Symmetries, Differential Equations and Applications (SDEA-II), National University of Sciences and Technology,
School of Natural Sciences, Islamabad, Pakistan held in January 27-30, 2014.
Appendix A. KVs of Stationary G
The KV fields as well as corresponding Lie algebra of the classes I-IV of STH Godeltype spacetimes (2) have been determined by Reboucas et al. 29 , and the results are
stated in the following theorem:
Theorem : The four-dimensional homogeneous Riemannian Godel-type manifolds admit a group of isometry Gr with
(i) r = 5 in classes I (where m2 6= 4w2 ) , II and III;
(ii) r = 6 in class IV;
(iii) r = 7 in the special case of class I, where m2 = 4 2 .
For all STH Riemannian manifolds endowed with a static Godel-type spacetime
(2) the KVs are obtained as follows:
Class I : m2 > 0, 6= 0. For this case, the general solution of Eqs. (3) and (4) can
be written by
2
[1 cosh(mr)]
m2
and the five KVs X1 , ..., X5
H(r) =
and
D(r) =
1
sinh(mr).
m
2
t m ,
m
H
D0
X4 = sin t + cos r
sin ,
D
D
0
H
D
X5 = cos t sin r
cos ,
D
D
The Lie algebra has the following non-vanishing commutators:
X1 = t ,
[X3 , X4 ] = mX5 ,
X2 = z ,
(A.1)
X3 =
[X4 , X5 ] = mX3 ,
[X5 , X3 ] = mX4 .
(A.2)
April 1, 2015
20
1:15
Ugur*Camci
U. Camci
It should be noticed that the expressions for all KVs are time-independent.
In the special class I case, where m2 = 4 2 , i.e. m = +2w, the seven KVs
X1 , ..., X7 are given by
X1 = t ,
H
X4 =
D
H
X5 =
D
H
X6 =
D
H
X7 =
D
X3 = t m ,
D0
sin t + cos r
sin ,
D
D0
cos t sin r
cos ,
D
X2 = z ,
(A.3)
1
cos(mt + ) ,
D
1
sin(mt + )t cos(mt + )r + sin(mt + ) ,
D
cos(mt + )t + sin(mt + )r +
The corresponding Lie algebra has the following non-vanishing commutators:

[X3 , X4 ] = mX5 ,
[X4 , X5 ] = mX3 ,
[X5 , X3 ] = mX4 ,
[X1 , X6 ] = mX7 ,
[X6 , X7 ] = mX1 ,
[X1 , X7 ] = mX6 .
Class II : m2 = 0, =
6 0.
For this class, where H(r) = wr2 and D(r) = r, the five KVs X1 , ..., X5 are
given by
X1 = t ,
X2 = z ,
X3 = ,
1
X4 = r sin t cos r + sin ,
r
1
X5 = r cos t + sin r + cos
r
(A.4)
The Lie algebra will have the following non-vanishing commutators:

[X3 , X4 ] = X5 ,
[X3 , X5 ] = X4 ,
[X4 , X5 ] = 2X1 .
Class III : m2 2 < 0, 6= 0. Similarly, the integration of the conditions

for homogenity Eqs. (3) and (4) leads to
H(r) =
2
[cos(r) 1]
2
and
D(r) =
1
sin(r).
(A.5)
In class III, where m2 2 < 0, 2 > 0 and w 6= 0, it follows that the KVs
X1 , X2 , X4 , X5 are same form as given above for the class I, but only X3 has the
form as (2w/)t + .
Class IV : m2 6= 0, = 0. We refer to the manifolds of this class as degenerated
G
odel-type manifolds, since the cross term in the line element, related to the rotation
in the G
odel model, vanishes. By a trivial coordinate transformation, one can
make H = 0 with D(r) given, respectively, by Eqs. (5) or (7) depending on whether
April 1, 2015
1:15
Ugur*Camci

21
m2 > 0 or m2 2 < 0. The six KVs X1 , ..., X6 are given by

X1 = t ,
X2 = z , X3 = z t + tz ,
D0
X4 = cos r
sin ,
D
D0
cos , X6 = ,
X5 = sin r
D
The non-vanishing commutators of NGSs are
[X1 , X3 ] = X2 ,
(A.6)
[X2 , X3 ] = X1 ,
2
[X4 , X5 ] = m K6 ,
[X5 , X6 ] = X4 ,
[X6 , X4 ] = X5 .
References
1. K. G
odel, An example of a new type of cosmological solution of Einsteins field equations of gravitation, Rev. Mod. Phys. 21, 447 (1949).
2. H. Stephani, D. Kramer, M. A. H. MacCallum, C. Hoenselaers and E. Herlt, Exact Solutions of Einstein Field Equations, (Cambridge University Press, Cambridge,
2003).
3. W. Kundt, Tr
agheitsbahnen in einem von G
odel angegebenen Modell, Z. Phys. 145,
661 (1956).
4. S. Chandrasekhar and J.P. Wright, The geodesics in G
odels universe, Proc. Natl.
Acad. Sci. 47, 341 (1961).
5. M. Novello, I. D. Soares and J. Tiomno, Geodesic motion and confinement in G
odel
universe, Phys. Rev. D 27, 779 (1983).
6. M. O. Calv
ao, I.D. Soares and J. Tiomno, Geodesics in G
odel-type space-times, Gen.
Rel. Grav. 22, 683 (1990).
7. M. J. Reboucas and A. F. F. Teixeira, Features of a relativistic space-time with seven
isometries, Phys. Rev. D 34, 2985 (1986).
8. F. Grave, M. Bauser, T. M
uller, G. Wunner and W.P. Schleich, The G
odel universe:Exact geometrical optics and analytical investigation on motion, Phys. Rev. D
80, 103002 (2009).
9. E. Kjari, R. Walseri W. P. Schleich and A. Delgado, Sagnac effect of G
odels universe, Gen. Rel. Grav. 36, 2289 (2004).
10. G. Dautcourt, The lightcone of G
odel-like spacetimes, Class. Quant. Grav.27, 225024
(2010).
11. S. Capozziello and A. de Felice, f(R) cosmology from Noethers symmetry, JCAP 08,
016 (2008).
12. U. Camci and Y. Kucukakca, Noether symmetries of Bianchi I, Bianchi III and
Kantowski-Sachs spacetimes in scalar-coupled gravity theories, Phys. Rev. D 76,
084023 (2007).
13. Y. Kucukakca, U. Camci and I. Semiz, LRS Bianchi type-I universes exhibiting Noether
symmetry in the scalar-tensor Brans-Dicke theory, Gen. Rel. Grav. 44, 1893 (2012).
14. M. Sharif and S. Waheed, Noether symmetries of some homogeneous universe models
in curvature corrected scalar-tensor gravity, JCAP 02, 043 (2013).
15. T. Feroze, F.M. Mahomed and A. Qadir, The connection between isometries and symmetries of geodesic equations of underlying spaces, Nonlinear Dynam. 45, 65 (2006).
16. T. Feroze, New conserved quantities for the spaces of different curvatures, Modern
Phys. Lett. A25, 1107 (2010).
April 1, 2015
22
1:15
Ugur*Camci
U. Camci
17. T. Feroze and I. Hussain, Noether symmetries and conserved quantities for spaces with
a section of zero curvature, J. Geom. Phys. 61, 658 (2011).
18. M. Tsamparlis and A. Paliathanasis, Lie and Noether symmetries of geodesic equations
and collineations, Gen. Rel. Grav. 42, 2957 (2010).
19. M. Tsamparlis and A. Paliathanasis, The geometric nature of Lie and Noether symmetries, Gen. Rel. Grav. 43, 1861 (2011).
20. F. Ali and T. Feroze, Classification of plane symmetric static space-times according
to their Noether symmetries, Int. J. Theor. Phys. 52, 3329 (2013).
21. Y. Kucukakca and U. Camci, Noether gauge symmetry for f(R) gravity in Palatini
formalism, Astrophys. Space Sci. 338, 211 (2011).
22. U. Camci, Symmetries of geodesic motion in G
odel-type spacetimes, JCAP 07, 002
(2014).
23. G. H. Katzin, J. Levine and W. R. Davis, Curvature collineations: A fundamental
symmetry property of space-times of general relativity defined by the vanishing Lie
derivative of the Riemann curvature tensor, J. Math. Phys. 10, 617 (1969).
24. A. K. Raychaudhuri and S. N. Thakurta, Homogeneous space-times of G
odel type,
Phys. Rev. D 22, 802 (1980).
25. M. J. Reboucas and J. Tiomno, Homogeneity of Riemannian space-times of G
odel
type, Phys. Rev. D 28, 1251 (1983).
26. M. O. Calv
ao, M. J. Reboucas, A.F.F Teixeira and W.M. Silva, Notes on a class of
homogeneous space-times, J. Math. Phys. 29, 683 (1988).
27. A. F. F. Teixeira, M. J. Reboucas and J. E.
Aman, Isometries of homogeneous G
odel
type spacetimes, Phys. Rev. D 32, 3309 (1985).
28. M. J. Reboucas and J. Tiomno, A class of inhomogeneous G
odel-type models, Nuova
Cimento B 90, 204 (1985).
29. M. J. Reboucas and J. E.
Aman, Computer-aided study of a class of Riemannian
space-times, J. Math. Phys. 28, 888 (1987).
30. G. S. Hall and J. da Costa, Affine collineations in space-times, J. Math. Phys. 29,
2645 (1988).
31. A. Melfo, L. A. Nunez, U. Percoco and V. M. Villalba, Collineations of G
odel-type
space-times, J. Math. Phys. 33, 2258 (1992).
32. M. Tsamparlis, D. Nikolopoulos and P. S. Apostolopoulos, Computation of the conformal algebra of 1+3 decomposable spacetimes, Class. Quant. Grav. 15, 2909 (1998).
33. U. Camci and M. Sharif, matter collineations of spacetime homogeneous G
odel-type
metrics, Class. Quant. Grav. 20, 2169 (2003).
34. E. Noether, Invariante Variationsprobleme, G
ottingen Math. Phys. Kl. 2, 235 (1918);
English translation by M.A. Tavel,Invariant Variation Problems, Transport Theory
and Statistical Physics 1(3), 186 (1971).
April 1, 2015
1:15
Mariano*Torrisi
A REVIEW ON SYMMETRIES FOR CERTAIN AEDES AEGYPTI

MODELS
IGOR LEITE FREIRE

Centro de Matem
atica, Computaca
o e Cognic
ao
Universidade Federal do ABC - UFABC
Rua Santa Ad
elia, 166, Bairro Bangu, 09.210 170
Santo Andr
e, SP - Brasil;
E-mail: igor.freire@ufabc.edu.br/igor.leite.freire@gmail.com
MARIANO TORRISI
Dipartimento di Matematica e Informatica,
Universit`
a Degli Studi di Catania,
Viale Andrea Doria, 6, 95125 Catania, Italia;
E-mail: torrisi@dmi.unict.it/m.torrisi12@gmail.com
We summarize our results related with mathematical modeling of Aedes aegypti and
its Lie symmetries. Moreover, some explicit, group-invariant solutions are also shown.
Weak equivalence transformations of more general reaction diffusion systems are also
considered. New classes of solutions are obtained.
Keywords: Aedes aegypti mosquitoes, Reaction-diffusion equation, Model for population
dynamics , Equivalence transformations, Symmetries, Invariant solutions
1. Introduction
In this paper we revisit our previous results1,2,3,4 using group methods for mathematical modeling of Aedes aegypti mosquitoes. This species is the main vector of
dengue, a disease characterized, among other symptoms, by high fever and severe
pain in the body, which can be fatal in many cases.
Such a disease mainly occurs in countries located in the tropical zone of the
globe, particularly those with a hot and humid climate. Although for years it was
essentially restricted to the mentioned zones, it may be possible that in the next
decades places and countries that until now have never been affected by this plague
will be as a result of global warming. Most interested in the investigation about
dengue and its vector are those who are currently affected. Climate change around
the world can lead to modifications of the weather and such modifications are sufficient to attract the attention of those living in places that in the next decades can
develop hot and humid weather becoming eventually fruitful places for the reproduction of Aedes aegypty and, consequently, be potentially affected by the dissemination
of dengue.
1
April 1, 2015
1:15
Mariano*Torrisi
I.L. Freire & M. Torrisi
Our research about Aedes aegypti was first concerned with the dispersal dynamics of the mosquito. We introduced as a new mathematical model, a partial
differential equation system and carried out a complete group classification with
respect to some constitutive parameters characterizing the model1,2 . Then we investigated the nonlinear case admitting the largest symmetry Lie algebra3 . From
the biological point of view, the last one corresponds to an infestation in a place
where wind currents can be neglected.
Motivated by our previous results1,2,3 , we considered a wide class of nonlinear
reaction-diffusion systems that enclose our models4 . Those classes possessed three
arbitrary constitutive functions, some of them depending on more than one variable.
Therefore the classical Lie group analysis is not appropriate for an investigation
concerned with invariance properties5,6 . For this reason we looked at such classes
from the point of view of equivalence transformations8,9,10 .
The plan of the paper is as the follows: in section 2 we recall some models for Aedes aegypti, following those that inspired us to introduce our systems
considered1,2 . Next, in section 3, we show the Lie point symmetry generators found
in 1,2 . From these Lie point symmetry generators, we get some explicit solutions for
some of the models obtained during the group classification. In particular we study
the nonlinear case, which biologically corresponds to an infection in region with
weak wind currents3 . Looking for further generalizations of our models, we consider
weak equivalence transformations4 in section 5. Since we are revisiting our results,
and in some sense, summarizing them, in some parts the presentation might closely
follow the previous works1,2,3,4 .
2. New models
Although the Aedes aegypti diffuses dengue among humans, we consider only
mosquitoes populations1,2,3 .
The first PDE model of the Aedes aegypti population was introduced by Tahahashi et al.11 . In that model the mosquito population was divided into two subpopulations: the winged u and the aquatic v. Their corresponding balance equations
in an adimensional form are written as the following class of nonlinear reaction
diffusion systems
ut = uxx ux + k v(1 u) 1 u,
(1)
vt = k(1 v)u (2 + )v,

where k, , 1 , 2 are parameters depending on the environment and is related
to the wind flux, which is responsible for the advection motion of the winged population; k being the ratio between the carrying capacities related to the amount of
findable nutrients, k1 , and the occupation of the available breeders k2 ; the specific
rate of maturation of the aquatic form into winged female mosquitoes ; 1 and 2
are the mortality of the winged population and the mortality of aquatic population,
April 1, 2015
1:15
Mariano*Torrisi
Symmetries for models of Aedes Aegypti
respectively. Apart from , all of the remaining parameters are strictly positive.
Further details are given in Takahashi et. al.11 . Additionally, the reader interested
in the dimensional version of system (1) is guided to that reference, where system
(1) is deduced point by point from its dimensionalised partner.
Generalisations of the model (1) can be obtained by considering nonlinear diffusion and nonlinear wind advection. Then, in this case we can write
p
q
ut = (u ux )x 2u ux + k v(1 u) 1 u,
(2)
vt = k(1 v)u (2 + )v,
where p, q R.
We can consider the interaction between both winged and aquatic populations
to be weak, since they are located in two different environments. In this way, the
logistic terms k v(1 u) and k(1 v)u could be replaced. We removed those terms
and based on the modeling of Proteus Mirabilis12 they were substituted, respectively,
by (v + u) and k(u + v), that is, we introduced the system

k
p
q
ut = (u ux )x 2u ux + k v + ( k 1 )u,
(3)
vt = ku + (k 2 )v.
System (3) can be considered as a Malthusian model13 (that is, without saturation of mosquitoess growth). Therefore, from the biomathematical point of view,
it is well suited for short time scales, that biologically may correspond to a fast
infestation due to consecutive rainy days.
For further developments in this paper we remark that all of the systems shown
in this section belong to the following class of nonlinear advection-reaction-diffusion
equations
ut = (f (u)ux )x + g(u, v, ux ),
(4)
vt = h(u, v).
Such a system can describe the dynamics of two interacting population densities,
u and v. The first balance equation takes into account some advection effects ux ,
together with the diffusion of the winged population u and the reaction effects. In
our cases the advection effects are due to external stimuli of wind effects on the
individuals of the population u. In the balance equation for density v only the socalled reaction terms are taken into account. This equation is concerned with the
aquatic population.
3. Symmetries
In our previous researches symmetries of Aedes aegypti models1,3 , we focused our
attention in the system (3) because the system (1) only admits translations in space
April 1, 2015
1:15
Mariano*Torrisi
and time. In order to show some results obtained in1 we recall here a few elements
about symmetries.
Definition 3.1. A symmetry for the system (3) is a non-degenerate point transformation of independent and depend variables

x = x (x, t, u, v),

t = t (x, t, u, v),
T :=
(5)
u = u (x, t, u, v),

v = v (x, t, u, v),
which leaves the system invariant.
Definition 3.2. A continuous one-parameter (local) Lie group of transformations
is a family G of transformations

x = x (x, t, u, v, ),

t = t (x, t, u, v, ),
(6)
T :=
u = u (x, t, u, v, ),

v = v (x, t, u, v, ),
which is locally a C -diffeomorphism in a subset S R2+2 with coordinates
(x, t, u, v), depending analytically on the parameter in a neighborhood D R
of = 0 and reduces to the identity transformation when = 0.
The linear form of (6)
x = x + 1 (x, t, u, v) + O(2 ),
t = t + 2 (x, t, u, v) + O(2 ),
u = u + 1 (x, t, u, v) + O(2 ),
v = v + 2 (x, t, u, v) + O(2 ),
(7)
where
1 (x, t, u, v) :=
x
|=0 ,

2 (x, t, u, v) :=
t
|=0 ,

u
v
|=0 ,
2 (x, t, u, v) :=
|=0 ,

allow us to introduce the vector field
X = 1 (x, t, u, v)
+ 2 (x, t, u, v) + 1 (x, t, u, v)
+ 2 (x, t, u, v) . (8)
x
t
u
v
This operator is usually called an infinitesimal generator of the transformation or
infinitesimal generator of the Lie point symmetry. Then, given a transformation (7),
it is possible to obtain the corresponding generator (8). Vice-versa, given a generator
of the type (8), it is possible to obtain its transformation using the exponential map,
that is, the transformation is given by (x, t, u, v) 7 (eX x, eX y, eX u, eX v). The
set of all infinitesimal generators leaving invariant a given equation endowed with
the commutator [X, Y ] = XY Y X, necessarily form a Lie algebra.
1 (x, t, u, v) :=
April 1, 2015
1:15
Mariano*Torrisi
With respect to (3), its Lie point symmetry generators were obtained1 by using the Lie infinitesimal criterion and the system (3) was completely classified with
respect to the constitutive parameters p, q, . The result obtained there are summarized in the following Table 1:
0
0
0
6 0
=
0
43
6= 43 , 0
2q
6 0
=
Generators
x , t
t)v
x , t , uu + vv , B(x, t)u + B(x,
2
x , t , x x 3xuu 3xvv , 2xx 3uu 3vv
x , t , pxx + 2uu + 2vv
x , t , qxx + uu + vv
appearing in the second row of the Table 1 are solutions

The functions (B, B)
of the following linear system

+ ( k 1 )B,
Bt = Bxx + k B
(9)
Bt = kB + (k 2 )B.
Remark 3.1. Although system (3) has the parameters , 1 , 2 , k, as well as p, q
and , only the last three affect the group classification in the nonlinear cases. When
the nonlinearities are removed, the remaining parameters influence the symmetry
group in the sense that the solution of (9) depend on the parameters , 1 , 2 , k
t)v has some dependence on these
and, consequently, the generator B(x, t)u + B(x,
constitutive parameters.
4. Invariant Solutions
In this section we give a short review of some reductions of the system (3) derived
in references1,3 .
4.1. Case 6= 0
In this case we have an advection due to wind currents, acting on the winged density
population u.
4.1.1. The case p = 2q 6= 0
In this subcase, the symmetries are translations in space and time, and a dilation
(x, t, u, v) 7 (q x, t, u, v), > 0. Then, the linear combination of the Lie point
symmetry generators is given by
X = c3 t + (qx + c2 )x + uu + vv .
April 1, 2015
1:15
Mariano*Torrisi
The invariant surface conditions imply

q
e c3
qx+c2 ,
u = e c3 U (),
v = e c3 V ().
(10)
Then system (3), after having puting p = 2q, gives the reduced ODE system:
1
c3 (U
1
c3 (V
+ qU 0 ) = 2q 3 4 U 2q1 (U 0 )2 + 2q 2 3 U 2q U 0 + q 2 4 U 2q U 00 +
+2q 2 U q U 0 + k V + ( k 1 )U,
(11)
+ qV 0 ) = kU + (k 2 )V.
In order to bring this system in an autonomous form we observe that the system
(11) is invariant under the following associated 14 dilation transformation group
1
= a1 ? , U = a q U ? , V = a q V ? .
We can do the following change of variables15,16,17,18 ,
= ln , V1 () =
Putting Vi, :=
form
d Vi
d
q1
, V2 () =
V
1
(12)
it is possible to rewrite system (11) in the following autonomous
V1, = 2q 3 V12q1 (V1, 1q V1 )2 + 2q 2 V12q (V1, 1q V1 )+

i
h
+V12q q 2 V1, 1q V1, 1q + 1 (V1, 1q V1 ) +

+2qV1q V1, 1q V1 + k V2 + k 1 V1 ,
(13)
V2, = kV1 + (k 2 )V2 .

which can be written as a 3 3 dynamical system.
4.1.2. The case p 6= 2q
In this subcase the only admitted symmetries are the temporal translations and the
spatial translations. Therefore we can look for invariant solutions of the form u =
U ()and v = V () where = x t where is an arbitrary constant representing
the velocity of wave propagation. In this case the reduced system assumes the
following form:

U 0 = (U p U 0 )0 2U q U 0 + k V + k 2 U,
(14)
V 0 = kU + (k 2 )V.
As in the previous system, we can rewrite (14) like a 3 3 first order system to
perform a qualitative analysis. If we are interested in a formal solution, it is possible
to verify its existence, at least for some special values of p and q.
For instance if we take p = 0 and q = 1 it is easy to verify that
U = A1 + A0 ,
V = B1 + B0 ,
April 1, 2015
1:15
Mariano*Torrisi
are solutions of (14) provided that A1 , A0 , B1 , B0 are solutions of the following

nonlinear algebraic system
A1 = 2A0 A1 k B0 ( k 2 )A0 ,
2A21 ( k 2 )A1 k B1 = 0,
B1 = kA0 + (k 2 )B0 ,
kA1 + (k 2 )B1 = 0.
without requiring restrictions on the constitutive parameters.
4.2. Case = 0
This case corresponds to a situation where the wind currents can be neglected. For
instance, it might correspond to an infestation in a small region, between builds, in
which wind currents are very weak. Using Table 1 we can distinguish two possibilities
4.2.1. p = 4/3
In this subcase the system (3) reads
ut = (u 3 ux )x + k v + ( k 1 )u,
(15)
vt = ku + (k 2 )v,
whose Lie algebra, taking the results obtained in
into account, is spanned by
X1 = x , X2 = t , Xp = pxx + 2uu + 2vv ,
(16)
with p = 4/3 in Xp , and by the additional generator

X3 = x2 x 3xuu 3xvv .
(17)
From this generator, we can look for an invariant solution of the form
u(x, t) =
A(t)
B(t)
, v(x, t) = 3 .
x3
x
Substituting these functions into (3) and taking into account that = 0 and p =
4/3, we obtain the following linear system of ordinary differential equations

A0 = k B + k 1 A,
(18)
B 0 = kA + (k 2 )B,
April 1, 2015
1:15
Mariano*Torrisi
whose solutions are

bt
A(t) = a1 e 2 cosh
B(t) = a1 e
bt
2
t
2
sinh
bt
+ a2 e 2 sinh
+a2 e
t
2
i
b cosh 2t +
h
i

cosh 2t b sinh 2t
t
2
bt
2
(19)
where a1 , a2 are arbitrary constants and

h
i2

= 4 + 1
+ (k 2 ) > 0,
k

(20)
b = 1 + 2 + k , c = +
1 (k 2 ).
k
k
A particular solution is given by

"
!#
!
t
bt
cosh
2
e 2
t
t
bt
u(x, t) = e 2
, v(x, t) =
b cosh
.
sinh
x3
2x3
2
2
Recalling that (x, t, u, v) 7 (x + x0 , t, u, v) and (x, t, u, v) 7 (x, t + t0 , u, v) are
symmetries, we can obtain a new solution from the previous one, given by

(t+t0 )
cosh
b(t+t0 )
2
,
u
(x, t) = e 2
(x + x0 )3
"
b(t+t0 )
e 2
v(x, t) =
sinh
3
2(x + x0 )
(t + t0 )
2
!
b cosh
(t + t0 )
2
!#
.
A more general solution was obtained3 . Consider the linear combination

X = (c1 + 2xc4 + c3 x2 )x + c2 t (3xuc3 + 3uc4 )u (3xvc3 + 3vc4 )v
(21)
of the basis of the Lie symmetry algebra for the system (15). Then, by assuming
c3 = 1, from the invariant surface conditions, we get the following invariant solutions
u=
(x2
A()
B()
, v= 2
,
3/2
+ 2c4 x + c1 )
(x + 2c4 x + c1 )3/2
(22)
where
Z
= t c2
t+
c2
x+c4 ,
if c24 c1 = 0,

dx
c
x+c
2
4
= t
arctan
, if c24 c1 < 0,
x2 + 2c4 x + c1
c1 c24
c1 c24
x+c4 c24 c1
, if c24 c1 > 0.
t c22
ln
2 c4 c1
x+c4 + c24 c1
(23)
April 1, 2015
1:15
Mariano*Torrisi
If we neglect time translations, that is, c2 0 in (21) and, consequently, in (23),

the reduced system becomes

1
A0 = 3A 3 (c1 c24 ) + k B + k 1 A,
(24)
0
B = kA + (k 2 )B,
in the unknown functions A(t) and B(t).
The solutions of system (24) are not easy to obtain due to the nonlinearity in the
first equation. However, when c1 = c24 , it can be reduced to a single linear ordinary
differential equation. In fact it is the system (18). For further details, see 3 .
4.2.2. p
/ {4/3, 0}
In this subcase the system (3) reads
p
ut = (u ux )x + k v +

1 u,
(25)
vt = ku + (k 2 )v.
Its 3-dimensional Lie algebra is spanned by (16). We consider the following general
infinitesimal operator
X = pc1 X1 + c2 X2 + Xp
(26)
where c1 and c2 are arbitrary real constants.

By applying the invariant surface conditions we conclude that the invariant
solutions, with respect to (26), have the form:
2
u(x, t) = A()(x + c1 ) p , v(x, t) = B()(x + c1 ) p ,

where
=
et
c2
(c1 + x) p
and the functions A() and B() satisfy the following general reduced system of
ODEs

0
A = (2A A0 )2 + k B + k 1 A
h

i
00 2
+ Ap p42 p2 A + A0 p (1 c2 ) 1 + p2 + c2 Ap ,
(27)
0
B = kA + (k 2 )B.
As in the previous case, we neglect translations in time. Then (27) can be written
as

0
4+2p
p+1
+ k B(t) + k 1 A(t),
A (t) = p2 A(t)
(28)
0
B (t) = kA(t) + (k 2 )B(t).
April 1, 2015
10
1:15
Mariano*Torrisi
It is easy to ascertain by simple calculations that it is possible to reduce the

search for solutions of (28) to the search for solutions of the following second order
ODE:

(p + 1)(4 + 2p) p
A
(t)
+
A00 (t)
+
k
A0 (t) +
1
2
p2
k

4 + 2p
p
+ (k 2 )
A(t)
+
(k
A(t) = 0, (29)
2
1
p2
k
that is a family of Lienard equations. After having put
(p+1)(4+2p)
,
p2
d = (k 2 ) 4+2p
p2 ,
(30)
A00 (t) (a Ap b) A0 + (d Ap + c) A = 0,
(31)
a=
we are able to write (29) as
where b and c are given by (20). Some classes of solutions of this family were
obtained3 .
5. Weak equivalence transformations and symmetries for the
system (4)
We consider, now, the following class of nonlinear advection-reaction-diffusion equations
gux 6= 0,
ut = (f (u)ux )x + g(u, v, ux ),
(32)
vt = h(u, v)
from the point of view of weak equivalence transformations4 . After reviewing this
work we apply our results for other models that are potentially interesting from the
biomathematical point of view.
5.1. Equivalence transformations
Following Ibragimov et. al.7 , we consider an equivalence transformation a non degenerated change of independent and dependent variables t, x, u, v into t, x
, u
, v,
such that
x = x(
x, t, u
, v),
t = t(
x, t, u
, v),
u = u(
x, t, u
, v),
v = v(
x, t, u
, v),
(33)
that maps a system of the class (32) in another one preserving the differential
u, v) 6=
structure but, in general, with f(
u) 6= f (u), g(
u, v, u
x ) 6= g(u, v, ux ), h(
h(u, v). In the special case in which f(
u) = f (u), g(
u, v, u
x ) = g(u, v, ux ) and
h(
u, v) = h(u, v) an equivalence transformation is a point symmetry.
Following Ovsiannikov5 , equivalence transformations of systems of the type (32)
can be considered as transformations acting on the basic augmented space A
{t, x, u, v, ux , f, g, h}.
April 1, 2015
1:15
Mariano*Torrisi
11
Then we can consider a family of one-parameter equivalence transformations as

a group of transformations, acting on the basic augmented space A, of the type

x = x(
x, t, u
, v, ) t = t(
x, t, u
, v, ), u = u(
x, t, u
, v, ), v = v(
x, t, u
, v, ),
f = f (
x, t, u
, v, ), g = g(
x, t, u
, v, u
x , ), h = h(
x, t, u
, v, )
(34)
which is locally a C -diffeomorphism, depending analytically on the parameter
in a neighbourhood of = 0 and reduces to the identity transformation for = 0.
Following8,5,19,9,10,7,6,20 we consider the infinitesimal generator of the equivalence transformations of the system (32) that reads:
Y = 1 x + 2 t + 1 u + 2 v + 1 f + 2 g + 3 h
(35)
where all infinitesimal components 1 , 2 , 1 , 2 and i (i = 1, 2, 3) are sought, at

least in principle, depending on x, t, u, v, ux , f , g and h. In order to obtain the
determining system, which allows us to get the infinitesimal coordinates i , i and j
(i = 1, 2 and j = 1, 2, 3), we apply the Lie-Ovsiannikovs infinitesimal criterion. In
this case, according to our results4 , the generator of weak (see, e.g9,10,12 ) equivalence
transformations of (32) is given by the operator
Y = (x)x + (t)t + (t, u)u + (x, t, v)v + (20 0 )f f +
(36)
+ (t + (u t )g + (00 uu ux )f ) g + ((v 0 )h + t ) h .
Here we recall that a weak equivalence transformation is still a transformation of the
type (33) but it can map the arbitrary functions f , g, and h into another functions
that not only have different form but also have additional functional dependencies.
Their generators are obtained by simplifying the Lie-Ovsiannikov algorithm. Further
details concerned with the weak equivalence of system (32) and similar systems can
be found in references4,12 .
5.2. Symmetries for the system (32)
We recall the following theorem, proved by us4 . For related material see references
21,22,23,24,25
.
Theorem 5.1. Let (36) be an infinitesimal equivalence generator for the systems
(32), then the operator
X = (x)x + (t)t + (t, u)u + (x, t, v)v ,
(37)
which corresponds to the projection of Y on the space (x, t, u, v), is an infinitesimal

symmetry generator of the system (32) if and only if the constitutive equations,
specifying the forms of f, h and g, are invariant with respect to Y .
April 1, 2015
12
1:15
Mariano*Torrisi
For the system under consideration, in general, the constitutive equations whose
invariance must be requested are:
f = D(u),
g = G(u, v, ux ),
h = F (u, v).
(38)
The invariance condition reads

Y (f D(u))|(38) = 0, Y (1) (g G(u, v, ux ))
= 0, Y (h F (u, v))|(38) = 0,
(38)
(39)
where
Y (1) = Y + j uj + aA hA
a
is the first extension of (36) and
e k
j = Dje u
k Dj ,
ae A hA
e b
aA = D
b Da .
e k
ij
= Dje j u
ik Dj ,
For further details, see 4,19,9,10 .

From the invariance condition (39), we obtain the following equations
1 1 Du = 0,
2 11 Gux 1 Gu 2 Gv = 0,
3 1 Fu 2 Fv = 0 (40)
restricted to (38).
Substituting
11 = (u1 1 11 )u11 = (u 0 )ux
(41)
and taking into account (38) we can rewrite the equations (40) in the following form
(20 0 )D(u) (t, u)Du = 0,
t + (u )G
(uu u2x
00
(42)
0
ux )D (u )ux Gux +
Gu (t, x, v)Gv = 0,
((v )F + t ) Fu (t, x, v)Fv = 0.
(43)
(44)
It is easy to see that the translation generators

X0 = t ,
X1 = x
are Lie point symmetries of (32) for any functions f (u), g(u, v, ux ) and h(u, v).
In particular, they are the only members of the principal Lie algebra LP of the
equivalence algebra. In 4 we proved
Corollary 5.1. An equivalence operator for the system (32) belongs to Principal
Lie Algebra LP iff i = 0, j = 0, i = 1, 2, j = 1, 2, 3.
April 1, 2015
1:15
Mariano*Torrisi
13
5.3. Some extensions of the principal Lie algebra

In order to show how to get some possible extensions of the Principal Algebra from
(42)- (44) we assume that
G = ur ux + 1 u2 + 2 v,
(45)
where the parameters , r, 1 , 2 are constitutive parameters of the considered

phenomena. Here, for the sake of simplicity, we assume , 1 , 2 , Du 6= 0. We
would like to emphasize that in this section the form of the function G, as well as
certain values of the parameters and the restrictions considered, although similar
to some forms previously considered, were mainly chosen following a criterion of
mathematical simplicity. From (42) we conclude that = 1 x + 0 , where 1 , 0
are arbitrary constants. Then we can write (42) as
(21 0 )D(u) (t, u)Du = 0
(46)
while, after having taken (45) into account, (43) reads

t + (u t )(ur ux + 1 u2 + 2 v) Duu u2x (u 1 )ur ux +
(rur1 ux + 21 u) (t, x, v)2 = 0,
which can itself be rewritten in the form

t +(u 0 )(1 u2 +2 v)21 u2 +ux (0 0 )ur rur1 Duu u2x = 0.
From the last equation, we can easily conclude that = 1 (t)u + 0 (t) from the
coefficient of u2x . From the free term and the coefficient of ux , we obtain
= (1 r)a1 x + a0 , = a1 u, = a1 t + a2 , = 2a1 v,
where a0 , a1 and a2 are arbitrary constants. Then, as a consequence of Theorem 1,
in addition to the translations in x and t, we have the following dilation symmetry
generator
X = (1 r)x
+t u
2v ,
x
t
u
v
(47)
provided that D and F are solutions of the following differential equations

uDu = (2r 1) D,
uFu + 2vFv = 3F.
(48)
5.4. Reduction
3
Let us take r = 1. Then, a solution to (48) is given by D = u and F = au3 + bv 2 ,

where a and b are constants.
Using the generator (47), we have following system of ODEs:
dx
dt
du
dv
=
=
=
,
0
t
u
2v
April 1, 2015
14
1:15
Mariano*Torrisi
which gives us
f (x)
g(x)
(49)
,
v= 2 .
t
t
Substituting such function into the system (32), with D = u, G = uux +1 u2 +
3
2 v and F = au3 + bv 2 , we have
00
0
2
f = f f + f f + 1 f + 2 g,
(50)
3
g = af 3 + bg 2 ,
u=
The second equation of (50) can be brought to a third degree algebraic equation
for g whose solution must be substituted in the first equation. Of course we only
considered the positive root since u and v are densities and the negative value would
guide us to a negative function, which does not have meaning in this case. Once
solved the ODE for f and g, we can easily obtain a solution for the system using
(49).
6. Conclusions
In this paper we revisited the main results of our previous papers1,2,3,4 concerned
with the population dynamics of the Aedes aegypti, the main vector of dengue,
obtained using the Lie group analysis techniques.
We observe that the theory of continuous symmetries enables us to obtain, from
the system (3) five different models, corresponding to different cases with respect
to the parameters p, q and , as it was summarized on Table 1.
The parameters p and q are related, respectively, with nonlinearities in the diffusion and in the advection. On the other hand, the term corresponds to the wind
influence on the Aedes aegypti models. It is interesting to observe that from the Lie
point symmetry classification we must consider as special cases some that can have
biological significance or meaning, such as the case = 0, which was extensively
considered in 3 .
Inspired by models considered previously, in section 4 we show some applications
of equivalence transformation techniques concerned with class (32) in order to get
symmetries from new forms of the constitutive functions.
We have shown how it is easily possible, by applying the projection theorem
(5.1), to get an idea of the form of the constitutive functions and determine the
extensions of the principal Lie Algebra for the selected form of the constitutive
functions4 .
Acknowledgements
The authors are grateful to FAPESP (project no. 2014/05024-8) for financial
support. M.T. acknowledges the support from the GNFM (Gruppo Nazionale
per Fisica-Matematica) of INdAM. I. L. Freire also thanks CNPq (process no.
308941/2013-6) for financial support.
April 1, 2015
1:15
Mariano*Torrisi
15
References
1. I. L. Freire and M. Torrisi, Nonlin. Anal.: Real World Appl., 14, 13001307, (2013),
http://dx.doi.org/10.1016/j.nonrwa.2012.09.018.
2. I. L. Freire and M. Torrisi, On a Model for Aedes Aegypti Numerical analysis
and applied mathematics (ICNAAM 2012), Vols A and B, edited by: Simos, TE; Psihoyios, G; Tsitouras, C; et al. AIP Conference Proceedings 1479, American Institute
of Physics, New York, 2012, pp. 1373-1376.
3. I. L. Freire and M. Torrisi, Commun. Nonlinear Sci. Numer. Simulat., 19, 872879,
(2014) http://dx.doi.org/10.1016/ j.cnsns.2013.08.006.
4. I. L. Freire and M. Torrisi, Abstr. Appl. Analysis, Art. Number: 546083, 2014,
(2014), http://dx.doi.org/10.1155/2014/546083.
5. L. V. Ovsiannikov, Group Analysis of Differential Equations. Academic Press, New
York, (1982).
6. N. H. Ibragimov, CRC Handbook of Lie Group Analysis of Differential Equations,
Boca Raton, Florida: CRC Press, (1996).
7. N. H. Ibragimov, M. Torrisi and A. Valenti, Preliminary group classification of equation vtt = f (x, vx )vxx + g(x, vx ), J. Math. Phys., 32, 2988-2995, (1991).
8. I. Lisle, Equivalence transformations for classes of Differential Equations (University
of British Columbia, Ph. D. Thesis ), (1992).
9. M. Torrisi and R. Tracin`
a, Equivalence transformations and symmetries for a heat
conduction model, Int. J. Non-Linear Mechanics, 33, 473-487, (1998).
10. V. Romano and M. Torrisi, Application of Weak Equivalence Transformations to a
Group Analysis of a Drift-Difffusion Model, J. Phys. A: Math Gen., 32, 79537963,
(1999).
11. L. T. Takahashi, N. A. Maidana, W. C. Ferreira Jr., P. Pulino and H. M. Yang,
Mathematical models for the Aedes aegypti dispersal dynamics: traveling waves by
wing and wind, Bull. Math. Biol., 67, 509528, (2005), an references therein.
a , Exact solutions of a reaction-difusion system for Proteus
Mirabilis bacterial colonies, Nonlin. Anal.: Real World Appl., 12, 18651874, (2011).
13. J. D. Murray, Mathematical Biology, Springer, Berlin, (2002).
14. L. Dresner, Similarity solutions of nonlinear partial differential equations, Pitman
Advanced Pub. Program, (1983).
15. A. Donato, Invariant solutions and nonlinear wave propagation, Int. J. Non-Linear
Mechanics , 17, 307314, (1987).
16. W. F. Ames and A. Donato, On the evolution of weak discontinuities in a state characterized by invariant solutions, Int. J. Non-Linear Mechanics, 26, 167174, (1991).
17. M. Torrisi and M. V. Lalicata, Some similarity solutions for a hyper elastic half-space
with finite deformations, Int. J. Non-Linear Mechanics , 26, 15-23, (1994).
18. M.V. Lalicata and M. Torrisi, Group-analysis approach for a binary reacting mixture,
Int. J. Non-Linear Mechanics , 29, 279-288, (1994).
19. I. Sh. Akhatov, R. K. Gazizov and N. H. Ibragimov, Nonlocal symmetries: A heuristic
approach, J. Soviet Math., 55, 14011450, (1991).
20. C. M. Khalique, F. M. Mahomed and B. P. Ntsime, Group classification of the generalized Emden-Fowler-type equation, Nonlin. Anal.: Real World Applic., 10, 33873395,
(2009).
21. N. H. Ibragimov and M. Torrisi, A simple method for group analysis and its application
to a model of detonation, J. Math. Phys., 33, 39313937, (1992).
a, Equivalence transformations for Systems of First Order
Quasilinear Partial Differential Equations, Modern Group Analysis VI: Developments
in Theory, Computation and Application, (Edited by N. H. Ibragimov et F.M. Ma-
April 1, 2015
16
1:15
Mariano*Torrisi
homed) New Age International Publishers, (1996).

23. M. Torrisi, R. Tracin`
a and A. Valenti, Group analysis approach for a non linear differential system arising in diffusion phenomena, J. Math. Phys. 37, 4758-4767, (1996).
24. N. H. Ibragimov and N. S
afstr
om, The equivalence group and invariant solutions of a
tumour growth model, Commun. Nonlin. Sci. Num. Simulat., 9, 6169, (2004).
a, On a class of reaction diffusion systems: Equivalence transformations and symmetries, in Asymptotic Methods in Nonlinear Wave Phenomena,
Eds.: T. Ruggeri, M. Sammartino, 207-216, (2007), DOI: 10.1142/97898127089080018.
April 1, 2015
1:16
A*H*Kara


Symmetry Lie Algebras and properties of linear ordinary differential

equations with maximal dimension
Mensah Folly-Gbetoula
School of Mathematics, University of the Witwaresrand, 2050
Johannesburg, Gauteng, South Africa
Mensah.Folly-Gbetoula@wits.ac.za
A.H. Kara
School of Mathematics, University of the Witwaresrand, 2050
Johannesburg, Gauteng, South Africa
Abdul.Kara@wits.ac.za

Solutions of linear iterative equations and expressions for these solutions in terms of the
parameters of the first-order source equation are obtained. Based on certain properties
of iterative equations, finding the solutions is reduced to finding solutions of the secondorder source equation. We have therefore found classes of solutions to the source equations
by letting the parameters of the source equation be functions of a specific type such as
monomials, functions of exponential and logarithmic type.
Keywords: symmetries; iterative equation; equivalence transformation
1. Introduction
Consider nth order linear ordinary differential equations (LODEs) n (y) = 0, where
y = y(x). It has been proved that for n = 1 and 2, all equations can be written in
the canonical form y 0 = 0 and y 00 = 0, respectively. Lie showed that a linear ordinary
differential equation of a general order n > 2 is equivalent to the equation y (n) = 0 if
and only if its symmetry algebra has the maximal dimension n + 4.1 It has also been
shown in Ref. 1 that a LODE of order n > 2 has a symmetry algebra of maximal
dimension if and only if it is iterative. In Ref. 2 a list of three linear equations
(in their normal form) of order n = 3, 4 and 5 that are iterative is provided. In
this paper, we review the expression of the symmetry generator in terms of the
solutions u and v of the second-order source equation, y 00 + A22 (x)y = 0, obtained
by Krause and Michel.3 Given that these solutions can be expressed in terms of the
parameters r and s of the first-order source equation, r(x)y 0 + s(x)y = 0, we wish
to see how these generators can be expressed in terms of the parameters r and s.
Principal
corresponding author
1
April 1, 2015
1:16
A*H*Kara
M. Folly-Gbetoula & A.H. Kara
Some properties of a linear iterative equation (see Refs. 1, 2, 4, 5) are used in order
to generate the vector field that spans the Lie algebra in terms of the mentioned
parameters for some special cases of r and A22 .
2. Preliminaries and Definitions
2.1. Iterative equations
Linear iterative equations are the iterations n y = 0 of the linear first order equation
of the form
y r(x)y 0 + s(x)y = 0
n
(n1)
[y] ,
(1)
n N.
(2)
Linear iterative equations are known as equations reducible to the canonical form
y (n) = 0 by point transformations. We assume without loss of generality that a
linear iterative equation of a general order is of the form
n y Kn0 y (n) + Kn1 y (n1) + Kn2 y (n2) + + Knn1 y 0 + Knn y = 0
(3)
(E) : y (n) + A2n y (n2) + + Ajn y (nj) + + Ann y = 0
(4)
and let
Kji
be the normal reduced form of Eq. (3). All the coefficients

can be expressed in
terms of the parameters r and s and the form of the general coefficient is given by
the following theorem.
Theorem 2.1. (Ref. 4) In terms of the parameters r and s of the source equation,
the general coefficient Knj of the iterative equation Eq. (3) has the form
Knj =
Mj
X
Mj1
kj =j kj1 =j1
M2 X
M1
X

rkj j rkj1 j1 . . . rk1 1 rj . . . , (5)
k2 =2 k1 =1
for n 1 and 1 j n, and where the expressions for Ml , ij and j are given by

l
X
j
i
j
ij =
Ku , ku Z, Ml = n+
ij , j = n+
0j = M0 . (6)
2
2
2
u=i+1
2.2. Symmetry analysis of the linear iterative equation

In this section, the application of the algorithm for finding a symmetry of a linear
iterative equation is considered. It has to be noted that the infinitesimal generator
v = (x, y)x + (x, y)y , is well-known. Ref. 2 showed that its nth prolongation
has the form
" k1

n
X
X k n1
(i) (k+1i)
(i+1) (ki)
(n)
y f
y
f
pr v = v +
i
2
k=1 i=0
#

n1 0
(k)
(k)
+
f + y +h
y(k) .
(7)
2
April 1, 2015
1:16
A*H*Kara
Symmetry Lie Algebras and properties of linear ordinary differential equations with maximal dimension
Imposing the invariance criterion pr(n) v(E) = 0 ( whenever (E) = 0 holds) and
applying the algorithm of finding symmetry in Ref. 2 it was proven that

v = f (x)x +

n1 0
f (x) + c y + h(x) y ,
2
(8)
where
n2
X
1 0
(n + 1)! (3)
(i)
Ani
= 0.
f + A2n f 0 + A2n f = 0, h(n) +
n h
(n 2)!4!
2
i=0
(9)
In Ref. 5 it was proven that condition on f does not depend on the order of the
linear iterative equation and it was shown that condition (9) can be reduced to
f 000 + 4A22 f 0 + 2A22 f = 0, h(n) +
n2
X
(i)
Ani
= 0.
n h
(10)
i=0
Note. n linearly independent solutions of Eq. (4) are given by
yk = un(k+1) v k
0 k n 1,
(11)
u and v are the two linearly independent solutions of
y 00 + A22 y = 0.
(12)
Remark 2.1. It follows from the work in papers Ref. 2 (equations 2.17 and 3.2)
and Ref. 5 (equations 17 and 48) that all the coefficients of iterative equations of
order n can be expressed in terms of the coefficient A22 and its derivatives only. We
April 1, 2015
1:16
A*H*Kara
have proved that

n+1
2
An =
A22 ,
(13a)
3

n + 1 h 2 (1) i
A3n =
2A2
,
(13b)
4

1
n+1
22
4
2 (2)
(13c)
An =
3A2 + (5n + 7)A2 ,
5
3

i
n + 1 h 2 (3)
(1)
A5n =
4A2 + 2(5n + 7)A22 A22
,
(13d)
6

5
n+1
(1) 2
(4)
(2)
6
An =
5A22 + (21n + 29)A22 A22 + (7n + 10)A22
7
2

1
3
(13e)
+ (35n2 + 112n + 93)A22 ,
9

4
n+1
(5)
2
(1)
(1)
(2)
A7n =
6A22 + (35n2 + 112n + 93)A22 A22 + 12(7n + 10)A22 A22
8
3

8
2 2 (3)
+ (14n + 19)A2 A2
,
(13f)
3

63
n+1
(2) 2
(6)
(4)
+
A8n =
7A22 + 20(3n + 4)A22 A22 + (9n + 13)A22
9
5
1
(1)
(3)
4
14(12n + 17)A22 A22 + (175n3 + 945n2 + 1769n + 1143)A22
15

2
2
2
2 2 2 (2)
2
2 2 (1)
+ (315n + 996n + 817)A2 A2 + 2(105n + 339n + 284)A2 A2
, (13g)
5

112
n+1
(7)
(1)
(3)
2
(2)
(5n + 7)A22 A22 + 2A22
2A22 A22
(3n + 13)
A9n =
8 A22
10
3

280
70
2
(2)
(1)
(3)
2
(1)
(5n + 7)A22 + 9A22

8A22 A22 A22
(n + 3) +
48A22 A22
9
3

(1)
(2)
(3)
(5)
34A22 A22 20A22 A22 + A22
(3n + 5)A22 +

2 (1)
20
(1)
(2)
2 35n2 + 112n + 93 A22 A22 + 18 (7 n + 10) A22 A22 +
9

40
(3)
(5)
3
(1) 2
4 (14n + 19) A22 A22 + 9 A22
(n + 15) A22 +
16 A22 27 A22
3

3
10
(2)
(4)
(1)
40 A22 A22 + 3 A22
(5 n + 11)A22 +
2 35 n2 + 112 n + 93 A22
9

2
(1)
(2)
(4)
(1)
+45 (7 n + 10)A22
+ 18 (21 n + 29)A22 A22 + 90 A22
(n + 7)A22 +
1440 A22
(1) 3
720 A22
(1)
+ 6464 A22 A22
A22
(4)
(1)
A22
(2)
+ 1792 A22 A22
(3)

896 2 2 (5)
3
(1)
A2 A2 2048A22 A22
,
3
3248 2 (2) 2 (3)

A2 A2
3
(13h)
April 1, 2015
1:16
A*H*Kara
A10
n
"

22
2
(2)
2 (8)
22
2 (2)
=
9 A2 +
(5 n + 7)A2 + 9 A2
8 2 A22 A22
A22
5

2
(1)
2 (1)
2 2 (2)
2 (4)
27 A2
32 A2 A2 + 3 A2
(5 n + 13) 154 (5 n + 7)A22 A22 +

3
22
2 (3)
2 2 (1)
2 (3)
2 35 n2 + 112 n + 93 A22
2 A2
8 A2 A2 A2
(2 n + 7)
5

2
2
(2)
2 (1)
2 2 (2)
2 (4)
+45 (7 n + 10)A2
+ 18 (21 n + 29)A2 A2 + 90 A2
2 A22 A22
(n

n+1
11

11
3
(1) 2
(2)
(4)
(1)
4 16A22 27A22
40A22 A22 + 3A22
A22 65 8A22 A22
3

(2)
(3) 2 (1)
2
(2)
(1)
(3)
(4)
A22
A2 144 2 A22 A22
A22 + 165 A22 A22 + 100 A22 A22
!
4
11
2 (6)
175 n3 + 945 n2 + 1769 n + 1143 A22
4 A2
(7 n + 11)A22 +
45

2 (2)
(1) 2
+30 105 n2 + 339 n + 284 A22 A22
+ 6 315 n2 + 996 n + 817 A22 A22
+5)
+189 (9 n + 13)A22
!
(2) 2
(1)
(3)
(4)
+ 210 (12 n + 17)A22 A22 + 300 (3 n + 4)A22 A22

2 (1)
33
2 (6)
2
2 35 n2 + 112 n + 93 A22 A22 + 18 (7 n
+105 A2
(n + 17)A2 +
2

(1)
2 (1) 2 (2)
2 2 (3)
2 (5)
+10)A2 A2 + 4 (14 n + 19)A2 A2 + 9 A2
(n + 8)A22

(1)
99
(1)
(3)
2
(2)
(1)
(2)
20 8 A22 A22 A22
A22 + 88 2 A22 A22
+
A22 150 A22 A22
2

1071
2
(1)
(1)
(2)
2 2 (3)
2 (5)
2 (1)
48 A22 A22 34 A22 A22
120 A2 A2 + 7 A2
(n + 2)A2
5

1152
3
(1) 2
(2)
2 2 (3)
2 (5)
2 (1)
20A2 A2 + A2
A2
16A22 27A22
40A22 A22
5

(3) 7056
(4)
(2)
(1)
(3)
2
(2)
2 (4)
+3A2
A22 + 1260 8A22 A22 A22
2A22 A22
A22
A22 +
5

4914 2 (1) 2 (5) 2016 2 2 (6)
9 2
2
(2)
3
A2 A2
A2 A2
A2 9216 A22 A22 128 16 A22
5
5
40

(1)
(1) 2
(2)
(4)
(1)
(3)
27 A22
40 A22 A22 + 3 A22
A22 + 2080 8 A22 A22 A22
A22
!#
4608 A22
(2) 2
5280 A22
(1)
A22
(3)
3200 A22 A22
(4)
+ 128 A22
(6)
(13i)
Remark 2.2. In paper Ref. 5, they let n , n 3, be the linear operator corresponding to the linear iterative equation of order n with source equation y 00 +ay = 0
April 1, 2015
1:16
A*H*Kara
and they proved for n = 3, 4, 5, 6 that its general expression is
n
n1
Y
v 0
2n1k kn
n (yk ) =
= 0,
(k j) u
v
u
j=0
(14)
for 0 k n1. We have verified that Eq. (14) is also true for n = 7, 8, 9 and 10. It
worthwhile to mention that the validity of Eq. (14) for an arbitrary n is unknown.
Therefore, the symmetry algebra of Eq. (4) is spanned by
n1 0
(u v + uv 0 )yy ,
2
Vk = un1k v k y , k = 4, . . . , n + 3.
(15)
V0 = yy , V1 = u2 x + (n 1)uu0 yy , V2 = uvx +
V3 = v 2 x + (n 1)vv 0 yy ,
3. Symmetry Generator of nth-order LODEs in Terms of the

Parameters r and s of the Source Equation
We can notice that finding symmetries and solutions of linear iterative equations is
reduced to the case of the second-order source equation in the normal form y 00 +
A22 y = 0. Despite the low order of the source equation, its solutions u and v are
generally not available. In this section, we are interested in finding explicit solutions
of the equation in terms of the parameter r, and for r as large (i.e. arbitrary) as
possible. Recall that 4
A22 =
r02 2rr00
4r2
provided that
s=
n1 0
r.
2
(16)
3.1. Solutions of the source equation and symmetry for given

values of r
We have noted that the solutions of the source equation y 00 +ay = 0 are not available
for A22 arbitrary. So we are interested in the same problem from a different angle.
For given values of the parameter r of the source equation, find the solutions u and
v in terms of r. In view of Eq. (15), this amounts to solving
r02 2rr00
= A22
(17)
4r2
for r. So this problem is not easier than that of finding u and v directly, but here
one is only interested in equations determined by the source parameter r, and to
find u and v in terms of r. Note also that in virtue of Eq. (15) it is not necessary
to find an expression of the symmetries in terms of r ( it only suffices to express u
and v in terms of r).
3.1.1. Case r = constant
Values of s, A22 , u and v: Invoking equations Eq. (16) we have s = 0 and A22 = 0
so that the corresponding source equation is y 00 = 0. And then, the two linearly
April 1, 2015
1:16
A*H*Kara
independent solutions are

u = 1,
v = x.
(18)
These solutions can not be expressed in terms of the parameters r and s. However,
we can write the symmetries as functions of the variables x and y, in this somewhat
trivial case.
Symmetry generators: We use Eq. (15) and Eq. (18) to generate the vectors that
span the Lie algebra:
n1
V0 = yy , V1 = x , V2 = xx +
yy , V3 = x2 x + (n 1)xyy
2
Vk = xk y , k = 4, . . . , n + 3.
(19)
3.1.2. Case r(x) = c1 x + c2 , c1 6= 0
Values of s, A22 , u and v: For such value of r , we have s = (n + 1)c1 /2 and
2
A22 (x) = (c1 /(2c1 x + 2c2 )) . Equation Eq. (12) becomes

2
c1
y 00 +
y=0
(20)
2c1 x + 2c2
and this equation is of Eulers type. Two linearly independent solutions of equation
Eq. (20) are given by
v = r ln r.
(21)
u = r,
Symmetry generators in terms of the parameters r and s: Similarly, invoking Eq. (15)
, the n + 4 vectors that span the Lie algebra in this special case are given by
V0 = yy ,
V1 = rx syy ,
V2 = r ln rx (1 + ln r)syy
V3 = r ln2 rx ln2 r + 2 ln r syy , Vk = r

n1
2
lnk ry , k = 4, . . . , n + 3. (22)
3.1.3. Case r(x) = (c1 x + c2 )m , c1 6= 0, m 6= 0, 1, 2

Corresponding values of s, A22 and the linearly independent
solutions: The corre2
sponding value of A22 is thus given by A22 (x) = 2m m2 (c1 /(2c1 x + 2c2 )) and
equation Eq. (12) becomes

2

c1
00
2
y = 0.
(23)
y(x) + 2m m
2c1 x + 2c2
2t
Using the change of variable c1 x + c2 = exp 2mm
we get
2
d2 y
2
dy
+ y = 0,
dt2
2m m2 dt
(24)
where we assume that 2m m2 > 0 in order to avoid complex numbers. Therefore,

the solutions of Eq. (23), in terms or the parameter r, are given as follows
1
u = r2,
v=r
2m
2m
(25)
April 1, 2015
1:16
A*H*Kara
Symmetry generators in terms of the parameters r and s: We can now express the
symmetries in terms of the parameters r and s since we were able to express the two
solutions in terms of these parameters. Hence, the n + 4 symmetries are as follows
1m
1
2
V0 = yy , V1 = rx syy , V2 = r m x
r m syy
m(n 1)

2m
22m
n12k
k
2m
V3 = r m x
r m syy , Vk = r m + 2 y , k = 4, . . . , n + 3.(26)
m
3.1.4. Case r(x) = exp (mx), m 6= 0

Corresponding value of A22 and the linearly independent solutions: We get s = (n
2
1)mr/2 and A22 (x) = m2 /4. Equation Eq. (12) becomes y 00 (m/2) y = 0 and
1
1
has u = r 2 and v = r 2 as solutions.

Symmetry generators in terms of the parameters r and s: The knowledge of u, v
allows us to find the symmetries as follows
s
1
V0 = yy , V1 = rx syy , V2 = x , V3 = x + 2 yy
(27)
r
r
n12k
Vk = r 2 y , k = 4, . . . , n + 3.
(28)
3.1.5. Case r(x) = ln x
In this case, the expression for A22 is given by
A22 =
1 + 2 ln x
.
4x2 ln2 x
(29)
Solving
00
y +
1 + 2 ln x
4x2 ln2 x

y=0
(30)
is not easy and we have tried to find the solutions using Mathematica which gives
us

3
r
r
2
u = G2,0
,
v
=
,
(31)
1,2
2 12 , 12
2
where Gm,n
p,q is the Meijer G-function defined by
Qn
Qm

Z

1
j=1 (bj s)
j=1 (1 aj + s)
m,n a1 , . . . , ap
Qq
Q
z s ds,
Gp,q
z =
b1 , . . . , b q
2i L j=m+1 (1 bj + s) pj=n+1 (aj s)
(32)
3.2. Solutions of the source equation for given values of A22
As already noted, the solutions of a general linear iterative equation are easily
obtained using the simple formula Eq. (11) from the linearly independent solutions
u and v of the second-order source equation. In the previous section, we obtained
April 1, 2015
1:16
A*H*Kara
expressions of u and v in terms of r, for various large values of r. In this present

section, to find u and v, instead of choosing a value for r, we rather specify a value
for A22 . Our aim is to solve the source equation for A22 as large as possible. Despite
its simplicity, the general solutions of the second-order source equation y 00 +A22 y = 0
is not known, for A22 arbitrary. Only values of A22 that have not been obtained in
the previous section by a choice of r will be considered.
3.2.1. Case A22 = C, for an arbitrary constant C 6= 0
Let
y 00 + Cy = 0
(33)
be the second-order source equation (in normal form) generating a family of iterative
equations. First we note that since C is a constant, and since each equation in the
corresponding family of iterative equations (in normal form) has coefficients which
are only polynomials in C and its derivatives, it follows that each iterated equation
has constant coefficients. The converse is also clear from the relation

n3 n 2
n(n 1)(n + 1) 2
n+1
2
A2 =
A2 .
(34)
An =
A22 =
3
6
6
Consequently each iterated equation has constant coefficient if and only if A22 is a
constant. In Ref. 2 they proved that indeed, all coefficients in each iterated equation
depends solely on A2n (thus on A22 ) and its derivatives. Now, let
n (y) = 0
(35)
be the iterative equation of order n generated by Eq. (33), and suppose that its
labelling coefficient A2n is a constant, B say. To find the parameter r of the source
equation that generates Eq. (35), one needs to solve the equation

n3 n
n3 n r02 2rr00
B=
C, i.e. B =
(36)
6
6
4r2
for r. This has solution
"
r(x) = c2 cos
r
6
#
B
(2c1 + x) ,
n + n3
and the corresponding value of s(x) = (1/2)(n 1)r0 (x) has expression
"
#
r
r
r
3
B
B
s(x) =
c2 (1 + n)
sin 2 6
(2c1 + x) .
2
n + n3
n + n3
(37)
(38)
However, using the usual operator n and r alone, one can easily generate iterative
equations of any order n with A2n = B. For instance, using Eq. 13, we obtain the
April 1, 2015
10
1:16
A*H*Kara
list of canonical normal forms of iterative equations of order 3 to 10 with constant

coefficient:
y (2) + A22 y = 0,
y (3) + 4A22 y 0 = 0,
y (4) + 10A22 y 00 + 9 A22
y
(5)
20A22 y (3)
+ 64
2
y = 0,
2
A22
y (6) + 35A22 y (4) + 259 A22
y 0 = 0,
2
2
A22
y 00 + 225 A22
3
y = 0,
3
A22
y 0 = 0,
y + 2304
2
3
4
y (8) + +84A22 y (6) + 1974 A22 y (4) + 12916 A22 y 00 + 11025 A22 y = 0,
2
3
4
y (9) + 120A22 y (7) + 4368 A22 y (5) + 52480 A22 y (3) + 147456 A22 y 0 = 0,
2
3
4
y (10) + 165A22 y (8) + 8778 A22 y (6) + 172810 A22 y (4) + 1057221 A22 y 00
5
+893025 A22 y = 0.
(7)
56A22 y (5)
+ 784
(3)
Expressing the solutions of Eq. (35) in terms of r, one may proceed as follows: If
we let
(39)
u = cos[ Cx], and v = sin[ Cx]
be the linear iterative solutions to y 00 + Cy = 0. For instance, in terms of r, we have
s
0 2 2rr 00
0 2 2rr 00
r
r
, v = sin x
,
(40)
u = cos x
4r2
4r2
where r is given by Eq. (37). This gives u and v in terms of C = A22 . We know that
the linearly independent solutions to Eq. (35) are given by uk v n1k , 0 k n 1.
Solutions Eq. (39) correspond to the case where C 0 and it has to be noted that
the general solutions are given by
y = A1 exp{x} + A2 exp{
x},
(41)
for some constants A1 and A2 , and

are roots of x2 + C = 0 (C real ).
3.2.2. Case A22 = xm ,
y 00 + xm y = 0
In terms of special functions, Mathematica gives

m +1

2x 2
1
u = x J m+2
m+2
m +1

2x 2
1
,
v = x Y m+2
m+2
2 m1
A
where J and Y are Bessel functions and x = 2
.
(42)
(43)
April 1, 2015
1:16
A*H*Kara
3.2.3. Case A22 = ebx , b 6= 0

Equation Eq. (12) becomes
y 00 + ebx y = 0.
(44)
Therefore the two solutions in this case are given by

2 bx
2 bx
e
, v=Y
e
,
u=J
b
b
(45)
with A22 = ebx .

4. Parameters of the Transformed Equations
Consider the linear iterative equation in the standard form
n y Kn0 y (n) + Kn1 y (n1) + Kn2 y (n2) + + Knn1 y 0 + Knn y = 0,
(46)
where r(x)y 0 + s(x)y (y) is the original source equation. We assume that
n (w) Zn0 w(n) + Zn1 w(n1) + Zn2 w(n2) + + Znn1 w0 + Znn w = 0
(47)
is the transformed equation obtained from Eq. (46) by the equivalence transformations
x = f (z),
y = g(z)w,
(48)
where f and g are arbitrary functions.2 It is known that under an equivalence

transformation, the symmetry algebras of two given equivalent equations are isomorphic which means that Eq. (47) is an iterative equation with source equation
R(z)w0 + S(z)w = (w).
It has been proved that if the linear iterative equation is in its normal reduced
form the parameter, r, of the source equation and the parameter, R, of the transformed equation are such that

1 r0 (f )2 2r(f )r00 (f ) 04
R02 2RR00
002
0 (3)
=
f
3f
+
2f
f
,
(49)
R2
f 02
r(f )2
where f = f (z).5 Note that by assuming the equation to be in its normal form
the computations become easier since we are looking only at how the parameter
r is changing. We aim to establish a similar relationship for linear iterative equation in its standard form and to retrieve Eq. (49) using the well known group of
transformations,
x = f (z),
y = [f 0 (z)]
n1
2
w,
(50)
of the general linear equation in normal reduced form (see 8 ). The idea consists of
iterating with the operator n times and transforming the resulting equation with
Eq. (48) and call it D(z, w, n). On the other hand, use the operator to iterate
directly n times.
11
April 1, 2015
12
1:16
A*H*Kara
4.1. n = 2
By definition
(2) y r2 y 00 + (rr0 + 2rs)y 0 + (rs0 + s2 )y.
(51)
From Eq. (48) we have

g
dy
g0
= 0 w + 0 w0
dx
f
f
(52)
dy 0
1
= 03
[f 0 gw00 + (2f 0 g 0 f 00 g)w0 + (f 0 g 00 f 00 g 0 )w] .
dx
f (z)
(53)
y0 =
and
y 00 =
According to Eq. (52) and Eq. (53), one can rewrite Eq. (51) as follows

(2f 0 g 0 f 00 g)
g
1 g
0
+
K
(2) y = D(z, w, 2) = K20 02 w00 + K20
2 0 w
f
f 03
f

0
(f 0 g 00 f 00 g 0 )
1g
2
+
K
+
K
g
w.
+ K20
2 0
2
f 03
f
(54)
Setting n = 2 in Eq. (47), the transformed equation becomes

Z20 w00 + Z21 w0 + Z22 w = 0
Z20 w00
with Z21 w0
0 0
00
(55)
Z22 w
and the substitution of
in Eq. (54) leads to

0 1 g
0 0 (2f g f g)
1 0 g
K2 Z2 02 + K2 Z2
+ K2 Z2 0 w0
f
f 03
f

0 00
00 0
0
0 0 (f g f g )
1 0g
2 0
0 2 g
+ K2 Z2 0 + K2 Z2 g w = 0.
+ K2 Z2 02 + K2 Z2
f
f 03
f
(56)
n
The functions Km
, Zop , f and g do not depend on w and w0 therefore we can equate
the coefficients of the latter to zero to yield
Z12 = f 0 K12 +
2f 0 g 0 f 00 g
,
f 0g
Z22 = f 02 K22 +
f 0 g 0 1 f 0 g 00 f 00 g 0
K2 +
,
g
f 0g
(57)
where
Kjn =
Knj
,
Kn0
Zjn =
Znj
.
Zn0
(58)
4.2. n = 3
According to the definition of iterative equation,
(3) y = K30 y (3) + K31 y (2) + K32 y (1) + K33 y = 0.
(59)
From Eq. (53) we have

6f 00 g 0
g
3f 00 g 3g 0
3f 002 g 3g 00
(3)
(3)
(2)
y =
w
+
+
w
+
+
+ 03
f 03
f 04
f 03
f 04
f 05
f

3f 00 g 00
3f 002 g 0
g 000
f 000 g 0
f 000 g 0
04 w + 04 +
+ 03 04
w.
(60)
f
f
f 05
f
f
April 1, 2015
1:16
A*H*Kara
We substitute Eq. (53) and Eq. (60) in Eq. (59) and we replace Z30 w(3) by
(Z31 w(2) + Z32 w(1) + Z33 w).) Then, letting all the coefficients w, w0 and w00 equal
to zero in the resulting equation leads to the relationship that we are looking for:
3g 0
3f 00
+
+ K13 f 0
f0
g
3g 00
3f 002
f 000
2f 0 g 0 f 00 g
6f 00 g 0
0 + K13 (
) + K23 f 02
Z23 = 0 + 02 +
fg
f
g
f
g
0 00

3f 00 g 00
3f 002 g 0
g 000
f 000 g 0
f g f 00 g 0
Z33 = 0
+ 02 +
0 + K13
fg
f g
g
fg
g
02 0
f g
+K23
+ K33 f 03 .
g
Z13 =
(61)
(62)
(63)
4.3. n = 4
The expressions of 4 y and 4 w are given by
K40 y (4) + K41 y (3) + K42 y (2) + K43 y (1) + K44 y = 0,
(64)
Z40 w(4)
(65)
Z41 w(3)
Z42 w(2)
Z43 w(1)
Z44 w
= 0,
respectively. Using the same idea we get
6f 00
4g 0
0 ,
g
f
0 0
3f
g
6g 00
18f 00 g 0
4f (3)
15f 002
Z24 = f 02 K24 + (
3f 00 )K14 +
+
,
0
0
g
g
fg
f
f 02
6f 00 g 0
3f 002
3f 0 g 00
2f 02 g 0
f 00 f 0 )K24 + (
+
+
f (3) )K14
Z34 = f 03 K34 + (
g
g
f0
g
8f (3) g 0
18f 00 g 00
30f 002 g 0
10f (3) f 00
15f 003
4g (3)
f (4)
+
+
+
0 ,
0
0
02
02
03
fg
fg
f g
f
f
g
f
03 0
02 00
00 0 0
00 00
002 0
f
g
f
g
f
f
g
3f
g
3f
g
Z44 = f 04 K44 +
K34 + (
)K24 + (
+
0
g
g
g
g
fg
f 0 g (3)
f (3) g 0 1 g (4)
6f (2) g (3)
f (4) g 0
4f (3) g (2)
+
)K4 +
0
g
g
g
f 0g
fg
f 0g
15f 002 g 00
15f 003 g 0
10f (3) f 00 g 0
+
.
+
02
02
f g
f g
f 03 g
Z14 = f 0 K14 +
(66)
(67)
(68)
(69)
13
April 1, 2015
14
1:16
A*H*Kara
4.4. Case n = 5
A straightforward but tedious computations for n = 5 lead to
5g 0
10f (2)
,
(70)
g
f0
0 0

2
4f g
45f (2)
40f (2) g 0
10f (3)
10g (2)
Z25 =f 02 K25 +
+
, (71)
6f (2) K15 +
g
g
f 0g
f0
f 02
02 0

2
3f g
6f 0 g (2)
18f (2) g 0
15f (2)
Z35 =f 03 K35 +
+
3f (2) f 0 K25 +
+
0
g
g
f
g

(3) 0
(2) (2)
(2) 2 0
(3) (2)
30f g
60f g
135f
g
60f f
(72)
4f (3) K15
+
+
0
0
02
fg
fg
f g
f 02
Z15 =f 0 K15 +
105f (2)
10g (3)
5f (4)
+
,
f 03
g
f0

03 0
2
3f 02 g (2)
2f g
0 2 (2)
3
4
04 4
f f
K5 + 3f (2) f 0 f (3) +
Z5 =f K5 +
g
g

(2) 0 0
(2) (3)
(2) 3
6f f g
10f f
15f
4f 0 g (3)
K25 + f (4) +
+
2
g
f0
g
f0

2
18f (2) g (2)
30f (2) g 0
f (5)
15f (2) f (4)
8f (3) g 0
5g (4)
+
0 +
K15 +
0
g
g
fg
g
f
f 02
(73)
10f (4) g 0
10f (3)
105f (3) f (2)
105f (2)
210f (2) g 0
+
2
0
03
04
fg
f
f
f 03 g
f0
2
30f (3) g (2)

120f (3) f (2) g 0
135f (2) g (2)
40f (2) g (3)
+
+
,
0
0
02
fg
fg
f g
f 02 g
!
(1) 2 (3)
4
2
f (1) g (1) 5
f 03 g (2)
f 0 f (2) g 0
f
g
5
05 5
3
Z5 =f K5 +
K4 +
K5 +
g
g
g
g
0 (4)
2
f 0 f (3) g 0
3f 0 f (2) g (2)
3f (2) g 0
f (4) g (1)
fg
K25 +
g
g
g
g
g
3
6f 0 f (2) g (3)
4f (3) g (2)
10f (2) f (3) g 0
15f (2) g (2)
15f (2) g 0
+
+
K15
g
g
f 0g
f 0g
f 02 g
3
(74)
g (5)
f (5) g 0
10f (2) g (4)
5f (4) g (2)
105f (2) g (2)
+
0
g
fg
f 0g
f 0g
f 03g
4
105f (2) g 0
10f (3) g (3)
15f (2) f (3) g (1)
10f (3) g 0
+
+
+
+
f 04 g
f 0g
f 02 g
f 02 g
2
45f (2) g (3)

60f (2) f (3) g (2)
105f (2) f (3)
+
.
02
02
f g
f
f 03 g
In Eq. (46) we have y = y(x), Kni = Kni (x), i = 0, . . . , n and similarly in Eq. (47)
w = w(z), and Zni = Zni (z) , for j = 0, . . . , n. Now we have established a general
April 1, 2015
1:16
A*H*Kara
relationship between the Ajn and Bnj , i.e

Ajn = Kjn |K1n =0
and
Bnj = Zjn |Z1n =0 .
(75)
It has to be noted that for equations in normal form Eq. (4), i.e.,
y (n) + A2n y (n2) + + Ajn y (nj) + + Ann y = 0,
(76)
it is always possible to express the coefficients Ain for 2 i n in terms of the

coefficient A22 and its derivatives. Therefore, it suffices to make use of the coefficients
A22 and B22 . Such relationship between parameters of two given equivalent equations
in their standard form is not known. However, we have found the relationship for
equations of order n = 2, 3, 4 and n = 5.
For the sake of clarification, let us consider the equation of second-order in its
standard form and use the above results.
Example 4.1. For n = 2, the transformation Eq. (50) implies that g = f 0 .

Therefore, expression Eq. (57) becomes
Z12 = f 0 K12
1
1
Z22 = f 02 K22 + f 00 K12 02
2
f
(77)

3 002 1 0 (3)
f ff
.
4
2
(78)
Letting K12 = 0 and using the notation Eq. (75) we obtain Z12 = 0, as expected, and

1
3 002 1 0 (3)
2 04
2
B2 = 02 A2 f f + f f
.
(79)
f
4
2
Recall that
r02 2rr00
R02 2RR00
2
,
A
=
.
2
4R2
4r2
Then, from Eq. (79) we get the relation

R02 2RR00
1 r0 (f )2 2r(f )r00 (f ) 04
002
0 (3)
=
f
3f
+
2f
f
R2
f 02
r(f )2
B22 =
obtained in
(80)
(81)
using the reduced form.
5. Conclusion
We have expressed the infinitesimal generator in terms of the parameters r and s of
the first-order source equation for a linear iterative equation of order n, n 3. By
letting n be the linear operator corresponding to the linear iterative equation of
order n with source equation y 00 + ay = 0, we showed that n , 7 n 10, satisfy
equation Eq. (14). A list of linear iterative equations of order up to 10, in normal
form that have constant coefficient was given and parameters of the transformed
linear iterative equations (in standard form) of order n = 2, 3, 4 and 5 were also
obtained.
15
April 1, 2015
16
1:16
A*H*Kara
References
1.
2.
3.
4.
5.
6.
S. Lie, Math. Ann. 22, 213-253 (1888).

F.M. Mahomed and P.G. Leach, J. Math. Anal. Appl. 151, 80-107 (1990).
J. Krause and L. Michel,C.R. Acad. Sci. Paris 307, 905-910 (1988).
J.C Ndogmo and F.M. Mahomed, Cent. Eur. J. Math. 12(4), 648-657 (2014).
M. Folly-Gbetoula and A. H. Kara, J. Math. Anal. Appl. 415, 135-147 (2014).
P.J. OHara, R. Osteen and R.S. Rodriguez, Internat. J. Math. Math. Sci. 15(1), 183188 (1992).
7. F. M. Mahomed, Math. Meth. Appl. Sci. 20, 1995-2012 (2007).
8. F. Neuman, Aequationes Mathematicae 33, 123-149 (1987).
April 1, 2015
1:17
El*Kinani


LIE SYMMETRY ANALYSIS OF SOME TIME FRACTIONAL

PARTIAL DIFFERENTIAL EQUATIONS
E. H. El Kinani
A.A Group, Mathematical Department Moulay Ismal University, Faculty of Sciences and
Technics Errachidia, BP 509, Morocco.
Universit
e Moulay Ismal Ecole Nationale Sup
erieure des Arts et M
etiers (ENSAM), Mekn`
es,
B.P. 15290 Al Mansour, Maroc
h.elkinani@ensam.umi.ac.ma
A. Ouhadan
Centre R
egional des M
etiers de lEducation et de la Formation
Mekn`
es,BP 255, Morocco
aouhadan@gmail.com
This paper uses Lie symmetry analysis to reduce the number of independent variables
of time fractional partial differential equations. Then symmetry properties have been
employed to construct some exact solutions.
Keywords: Lie symmetry method, fractional derivative, time fractional differential equation, Similarity variables.
1. Introduction
In recent years, fractional differential equations attracted more attention as they
are widely used to describe several phenomena in many fields of sciences such as
fluid flow, biology, chemistry, control theory and other areas1 . Consequently, many
powerful and efficient methods for seeking exact and numerical solutions have been
developed including the Adomians decomposition method2 , the variational iteration method3,4 , transformation methods5 , the finite difference method6 , the first
integral method7 , and so on.
Lie symmetry analysis is an important method in studying differential equations
with integer order8,9,10 . This method has several applications including construction of new solutions from trivial ones, possibility of some non linear equations to
be linearized, construction of integrator factor, reduction of order, reduction of the
number of independent variables and so on. Concerning symmetry analysis of fractional partial differential equations, the recognized results, in the literature, are very
few and the method is still not sufficiently developed. Rigorous study of symmetries,
admitted by such fractional equations, was started by Gazizov et al. in11 where they
have used two basic fractional derivatives: Riemann-Liouville and Caputo.
1
April 1, 2015
1:17
El*Kinani
A. Ouhadan & E. H. El Kinani
We use Lie symmetries with the prolongation formula given by Gazizov et al.11
for the time fractional partial differential equation given by
u
2u
u
=
x
+ f (x) ,
2
t
x
x
(1)
where tu is the fractional derivative of order with 0 < < 1. The above
equation has been extensively studied in the case of = 1 by Craddock12 with
different expressions of the drift function f which satisfies a family of Riccati type
differential equations. In this paper, we are interested in two cases
for the drift
1+3x
.
function, the first is f (x) = R and the second is f (x) = 2(1+
x)
There is no unique definition of fractional derivatives13,14 . Here we use the
Riemann-Liouville version, defined by
( n
u
u
n , = n N
R t u(,x)
D u(t, x) = = t 1
n
t
(n) tn 0 (t)+1n d, n 1 < < n, n N ,
where is the well-known gamma function, and D satisfies
( + 1)
t
, > 0, > 1, t > 0 ,
( + 1 )
t
D 1 =
, 0, t > 0 .
(1 )
D t =
This paper is organized as follows. In the second section, a background of symmetry

analysis is given and the system of determining equations for symmetry generators
corresponding to studied equation is obtained. In the third section, the obtained
system of determining equations is solved and infinitesimal generators of each case
are given. In the fourth section, some special exact solutions of Eq. (1) are pointed
out. Finally, a conclusion is given.
2. Symmetry Analysis of Fractional Partial Differential Equations
Eq. (1) is a time fractional partial differential equation of the form
u
= F (t, x, u, ux , uxx , ...),
t
0 < < 1,
where subscripts denote partial derivatives. Equation (1) is invariant under a one
parameter continuous transformations 8,9 :
e
t = t + (t, x, u) + o(),
x
e = x + (t, x, u) + o(),
u
e = u + (t, x, u) + o(),
with corresponding infinitesimal generator given by
X = (t, x, u)
+ (t, x, u)
+ (t, x, u) ,
t
x
u
(2)
April 1, 2015
1:17
El*Kinani
Lie Symmetry Analysis of Some Time Fractional Partial Differential Equations
where , and are infinitesimals to be determined later. To apply Lies algorithm,

we need to extend the infinitesimal generator X to X of the form
X = X + x
+ xx
+ ,
ux
uxx
u
such that u = tu and x , xx , are extended infinitesimals of order 1,2 and

respectively. x and xx have the form
x = x + (u x )ux x ut u u2x u ux ut ,
xx
= xx + (2xu xx )ux xx ut + (uu
uu u3x
uu u2x ut
(3)
2xu )u2x
2xu ux ut
+ (u 2x )uxx 2x uxt 3u ux uxx
u uxx ut 2u ux uxt .
(4)
Nevertheless, the -th extended infinitesimal related to Riemann-Liouville fractional time derivative has the following form
= Dt () + Dt (ux ) Dt (ux ) + Dt (Dt ( )u) Dt+1 ( u) + Dt+1 (u),
where Dt denotes the total time fractional derivative. The -th extended infinitesimal can be rewritten as11,15,16
u
u
+ (u Dt ( )) u +
t
t
t

+ n
X
u
n+1
+
D
(
)
Dtn (u)
t
n
n
n
+
1
t
n=1
+
X
Dtn ()Dtn (ux ).
n
n=1
(5)
where
=
m k1
+ X
n X
X
X n k 1
n=2 m=2 k=2 r=0
n
r k!
m
nm+k
t
[u]r m (ukr ) nm k .
(n + 1 )
t
t
u
(6)
In our case, the term vanishes as a consequence of uu = 0. Equation (1) is invariant under transformations (2) if and only if the invariance condition is satisfies,i.e
(uxx + f 0 (x)ux ) f (x)x xxx = 0,
whenever
u = xuxx + f (x)ux , (7)
where f is the drift function. To obtain general form of infinitesimals , and

, we need to substitute expressions (3),(4) and (5) into invariance condition (7).
The equation depends on variables ux , uxx , ut , uxt , ... and Dtn u, Dtn ux for n =
1, 2, .... Those variables are considered to be independent variables.
April 1, 2015
1:17
El*Kinani
Splitting the defining Eq. (7) with respect to independents variables mentioned
above leads to the system of infinitely equations
t = u = x = u = uu = 0,
(8)
xu xt x(u 2x ) = 0,
(9)
f (x)t f 0 (x) + f (x)x x(2xu xx ) = 0,

u
u
f (x)x xxx = 0,
t n t

u
Dtn+1 ,
n = 0, 1, 2, ...
n tn
n+1
(10)
(11)
(12)
The above system is known as the system of determining symmetries and solving
it leads to infinitesimal generators of symmetries admitted by Eq (1).
3. Admitted Symmetries of Studied Equation
Examining equations (8) and (9) in the above system, we readily obtain
(x) = ax + m x,
and
(t) = at + b,
with a, b and m are arbitrary constants. In virtue of uu = 0, must be linear in
u. Thus
= g(t, x)u + h(t, x),
for some functions g(t, x) and h(t, x). On the other hand, Eq. (12) requires that
u
= 0,
then
g = g(x).
t
We substitute these in Eq. (11) to derive that the function h is an arbitrary solution
of the original fractional differential Eq. (1). Substituting of and their expressions
into Eq. (10) leads to the equation
1
x
1
m
(13)
g 0 (x) = mf (x) (a +
m)f 0 (x) .
2
x
4x x
8x x
Differentiation of the Eq. (11) over u leads to
xg 00 (x) + f (x)g 0 (x) = 0.
(14)
Differentiation of Eq. (13) with respect the independent variable x and substituting
this into Eq. (14), leads
1
a d
d
x Lf + 161 mx 2 [3 + 8Lf 8x Lf ] = 0.
2 dx
dx
with Lf = xf 0 f +
f2
2 .
(15)
Finally, the choice of a function y(x) leads to a drift

2
function f as a solution of the Ricatti type equation xf 0 f + f2 = y(x) and hence

Eq. (15) determines m and a.
April 1, 2015
1:17
El*Kinani
Cas 1: In this first case, we consider the fractional differential equation

2u
u
u
=
x
+ ,
(16)
2
t
x
x
where is an arbitrary constant. In this case we have xf 00 + f f 0 = 0. Using Eqs.
(15) and (13) we obtain
Subcase 1. = 12 . In this case a basis for Lie algebra of symmetries admitted
by Eq. (16) is

X1 = x
+ t , X2 = x
x
t
x
X3 = , X4 = u , Xh = h(t, x) ,
t
u
u
with h an arbitrary solution of Eq. (16).
Subcase 2. = 32 , for this choice of , the symmetry Lie algebra is spanned
by X1 , X3 , X4 , Xh and the infinitesimal generator X5 given by

1
X5 = x
u .
x 2 x u
Subcase 3. Therefore, if the drift does not belongs to { 21 , 32 } the only possible
symmetries are X1 , X3 , X4 and Xh .
1+3x
. In this case the drift function f satisfies
2(1+ x)
1 2
3
2 f = 8 and the equation arising from f is
Cas 2: f (x) =
0
xf f +
the Ricatti equation
u
u
1 + 3 x u
.
=x 2 +
t
x
2(1 + x) x
(17)
Thus from Eq. (15) we obtain that a and m are arbitrary. Finally, the symmetry algebra admitted by Eq. (17) is spanned by X3 , X4 , Xh and the infinitesimal
generators X6 and X7 given by
u ,
X6 = x
+t +
x
t 2(1 + x) u

1
u .
X7 = x
x 2(1 + x) u
4. Exact Solutions
In this section we use the obtained symmetries to construct exact solutions.
4.1. Reduction in the case f (x) =
1
2
with X2 + X4
x
Invariants of the operator X2 +X4 are z = t, and I = ue2
. The corresponding
2 x
invariant solution in this case has the form u(t, x) = (t)e
, and the reduced
equation is
Dt ((t)) = (t).
April 1, 2015
1:17
El*Kinani
Let us consider that the above equation is a fractional order differential equation
of order 2q with q a positive integer greater than 2. Then a solution of the above
reduced equation is given by 14
(t) =
q1

1 X
Et (k, 1) (1)qk1 Et (k, (1)q ) ,
2
k=0
where the function Et is defined by

Et (, a) = t
X
k=0
(at)k
.
( + 1 + k)
Consequently, the exact solution of original equation in this case is given by

u(t, x) =
q1
o
1 2x X n
Et (k , 1) (1)qk1 Et (k , (1)q ) .
e
2
2
2
k=0
4.2. Reduction in the case f (x) =
3
2
with X5
The expression of this operator is

1
x
u .
x 2 x u
X5 =
The corresponding invariants of X5 are

z = t,
and
I = ux 2 .
with invariant solution of the form

1
u(t, x) = x 2 (t),
and the reduced equation is
Dt ((z)) = 0.
Thus the invariant solution becomes
1
u(t, x) = cx 2 t1 ,
with c an arbitrary constant.
1+3 x
with X7
4.3. Reduction in the case f (x) = 2(1+
x)

This operator has the form X7 = x x 2(1+1x) u u

. Corresponding invariants
of X8 are z = t,
form
and I = u 1+1x = (z). Invariant solution in this case has the

u(t, x) =
1
(t),
1+ x
April 1, 2015
1:17
El*Kinani
and the reduced equation of (17) is

Dt (t) = 0.
The exact solution constructed in this case is given by
c
t1 ,
u(t, x) =
1+ x
where c is an arbitrary constant.
Remark 4.1. To obtain the group transformation generated by infinitesimal generator X6 , we solve the system of first order ordinary differential equations,
de
t
= (e
t, x
e, u
e),
d
de
x
= (e
t, x
e, u
e),
d
de
u
= (e
t, x
e, u
e),
d
subject to the initial conditions
e
t(0) = t,
x
e(0) = x,
u
e(0) = u.
The one-parameter groups G generated by X6 is given as follows 8 . The entries gives

the transformed point exp(X6 )(t, x, u) = (e
t, x
e, u
e):
1+ x
G : (xe , te ,
u).
x + e 2
As G is a symmetry group, so if u(t, x) is a solution of equation (1), then is the
transformed functions .u(t, x) which is given by
x+e2
u(e t, e x),
.u(t, x) =
x+1
where is sufficiently small real number. Using this invariance property, we can
derive new exact solutions of equation (1) from the known one.
5. Conclusion
Using Lie point symmetry analysis we see that the studied time fractional partial
differential equation can be transformed into fractional differential equations with
a single independent variable. Consequently, some exact solutions are found and an
example of a one parameter group of transformations is constructed.
Acknowledgments
The first author E. H. El Kinani would like to express his thanks to the organizers of
SDEA-II Jan 27th -Jan 30th (2014) in Islamabad Pakistan for their warm hospitality
during the conference. Special thanks to Dr. Sajid Ali for his cooperation. Our
thanks also to the anonymous referees for their careful reading of our manuscript.
April 1, 2015
1:17
El*Kinani
References
1. L. Debnath, Recent applications of fractional calculus to science and engineering, International Journal of Mathematics and Mathematical Sciences, Vol. 54, 3413-3442,
2003.
2. A. M. Wazwaz, S.M. El-Sayed, A new modification of the Adomian decomposition
method for linear and nonlinear operators, Appl. Math. Comput. Vol. 122(2001)No.3,
pp.393-405.
3. C. C. Wu, A fractional variational iteration method for solving fractional nonlinear
differential equations, Comput. Math. Appl., vol.61, pp. 2186-2190, 2011.
4. J. Ji, J. B. Zhang and Y. J. Dong, The fractional variational iteration method improved
with the Adomian series, Appl. Math. Lett., vol. 25, pp. 2223-2226, 2012.
5. B. Lu., B
acklund transformation of fractional Riccati equation and its applications to
nonlinear fractional partial differentialequations, Phys. Lett. A, vol. 376, pp. 2045-2048,
2012.
6. M.M. Meerschaert, H.P. Scheffler, C. Tadjeran., Finite difference methods for twodimensional fractional dispersion equation, J. Comput. Phys. 211(2006) 249261.
7. M. Eslami, B. Fathi Vajargah, M. Mirzazadeh and Anjan Biswas, Application of first
integral method to fractional partial differential equations, Indian Journal of Physics.
Volume 88, Issue 2, 177-184, (2014).
8. P.J. Olver, Application of Lie groups to Differential Equation. New York: Springer;
1986.
9. G. Bluman, S. Kumei, Symmetries and Differential Equations. Applied Mathematical
Sciences Series. 81 (Second ed.). New York: Springer-Verlag; 1989.
10. G.W. Bluman and J.D. Cole, The general similarity solutions of the heat equation,
Journal of Mathematics and Mechanics, vol. 18(1969), pp. 1025-1042.
11. R.K. Gazizov, A.A. Kasatkin, S.Yu. Lukashchuk, Continuous transformation groups
of fractional differential equations, Vestnik, USATU 9 (2007),pp. 125-135.
12. M. Craddock, E. Platen, Symmetry group methods for fundamental solutions, J. Differential Equations 207 (2004),pp. 285-302.
13. K.B. Oldham, J. Spanier, The fractional calculus.-Academic Press, 1974.-234p.
14. K.S. Miller, B. Ross, An introduction to the fractional calculus and fractional differential equations.-John Wiley and Sons, Inc; 1921.
15. R.K. Gazizov, A.A. Kasatkin, S.Yu. Lukashchuk, Symmetry properties of fractional
diffusion equations, Phys. Scr. T136 (2009), 014-016.
16. E. Buckwar, Y. Luchko, Invariance of a partial differential equation of fractional order
under the Lie group of scaling transformations, J. Math. Anal.Appl. 227 (1998). pp.
81-97.
17. S. Momani, Z. Odibat, Analytical solution of a time fractional NavierStokes equation
by Adomian decomposition method, Appl. Math. Comput. 177(2006),pp. 488-494.
18. R. Hilfer, Applications of Fractional Calculus in Physics, Singapore: World Scientific;
2000.
19. I. Podlubny, Fractional differential equations: an introduction to fractional derivatives, fractional differential equations, to methods of their solution and some of their
applications. New York: Academic Press; 1999.
April 1, 2015
1:18
Maria*Nucci


WHAT SYMMETRIES CAN DO FOR YOU
M.C. NUCCI
Dipartimento di Matematica e Informatica,Universit`
a degli Studi di Perugia,
& INFN Sezione di Perugia,
06123 Perugia, Italy
nucci@unipg.it
Several applications of Lie symmetries and its generalisation are presented: from turning butterflies into tornados, to its applications in epidemics, population dynamics, and
ultimately converting classical problems into the quantum realm. Applications of nonclassical symmetries are also illustrated.
Keywords: Lie symmetry; Population dynamics; Lagrangian; Jacobi last multiplier;
Noether symmetry; Classical quantization; Nonclassical symmetries.
1. Introduction
In his inaugural address at Kings College London in 1860, James Clerk Maxwell
said1 : We, while following out the discoveries of the teachers of science, must experience in some degree the same desire to know and the same joy in arriving at
knowledge which encouraged and animated them.
This paper is a review of some of the authors work in the area of symmetries,
those symmetries initiated by a great teacher of science, Sophus Lie, who based his
idea on the work of another sublime teacher of science, Karl Gustav Jacob Jacobia .
Which symmetries are we dealing with? In the Introduction of his book3 , Olver
tersely stated: The applications of Lies continuous symmetry groups include such
diverse fields as algebraic topology, differential geometry, invariant theory, bifurcation theory, special functions, numerical analysis, control theory, classical mechanics
quantum mechanics, relativity, continuum mechanics and so on. It is impossible to
overestimate the importance of Lies contribution to modern science and mathematics. Nevertheless, anyone who is already familiar with one of these modern manifestations of Lie group theory is perhaps surprise to learn that its original inspirational
source was the field of differential equations.
a Hawkins has established in Ref.2 the nature and extent of Jacobis influence upon Lie. As Hawkins
clearly stated2 , given the fact that the Jacobi Identity is fundamental to the theory of Lie groups,
Jacobis influence upon Lie will come as no surprise. But the bald fact that he inherited the Identity
from Jacobi fails to convey fully or accurately the historical dimension of the impact of Jacobis
work on partial differential equations.
April 1, 2015
1:18
Maria*Nucci
M.C. Nucci
Nearly thirty years later, and after hundreds of papers and many books, e.g.
Refs. 4-13 just to cite a few, have been published on this subject, a young theoretical
mathematician still states: The only useful symmetry is dilation (they can be used
to determine embeddings into Sobelev or Lebesgue spaces) and all other ones are
useless, while a senior theoretical physicist asks: What are these symmetries?
We shall show by means of some examples what symmetries can dob . Of course, we
do not claim to cover every aspect of the applications of Lie symmetries in the field
of differential equations. We hope to inspire some curiosity in both young and senior
mathematical physicists, in such a way that they wish to know what Lie symmetries
can do for them.
2. From a butterfly to a tornado with symmetries
The motion of a heavy rigid body about a fixed point is one of the most famous
problems of classical mechanics15 . In 1857 Maxwell himself wrote16 To those who
study the progress of exact science, the common spinning-top is a symbol of the
labours and the perplexities of men who had successfully threaded the mazes of the
planetary motions. The mathematicians of the last age, searching through nature
for problems worthy of their analysis, found in this toy of their youth, ample occupation for their highest mathematical powers. In 1750 it was Euler17 who derived
the equations of motion, which now bear his name, and described what is nowadays
known as the Euler-Poinsot case because of the geometrical description given by
Poinsotc about a hundred years later18 . It was Jacobi19 who integrated this case by
using the elliptic functions which he had developed (along with Legendre, Abel and
Gauss20 ) and mastered21 we have translated this fundamental text into Italian
and commented extensively22 .
More than 200 years later, in 1963, a paper was published23 in which was presented a system of three ordinary differential equations. The author considered a
hydrodynamical system developed by Rayleigh24 and reduced it by applying a double Fourier series as in Ref. 25. Thus he obtained what nowadays is the famous
Lorenz system26 . Three parameters are part of the Lorenz system. For particular values of those parameters the Lorenz system can be integrated in closed form
by means of Jacobi elliptic functions27 . We call this system the Lorenz integrable
system.
In Ref. 28 we applied Lie group analysis to a third-order differential equation,
which is equivalent to the Lorenz integrable system, and obtained a two-dimensional
Lie symmetry algebra, which we then used to integrate the Lorenz integrable system
in terms of Jacobi elliptic functions. In Ref. 29 we showed that the same Lie symb Since
the task of finding symmetries can be quite cumbersome, we have used our interactive
REDUCE programs14 throughout.
c Again in Ref. 16 Maxwell stated M. Poins
ot has brought the subject under the power of a more
searching analysis than that of the calculus, in which ideas take the place of symbols, and intelligible propositions supersede equations.
April 1, 2015
1:18
Maria*Nucci
metry algebra is admitted by a third-order differential equation which is equivalent

to the Euler equations of a torque-free rigid body moving about a fixed point. Then
a transformation was easily derived by which the Lorenz integrable system becomes
the Euler equations of a torque-free rigid body moving about a fixed point. Thus,
it can be stated that the Lorenz integrable system moves à la Poinsot.
In Ref. 29 the same transformation was applied to the Lorenz system with any
value of parameters, and consequently the Euler equations of a rigid body moving
about a fixed point and subjected to a torsion depending on time and angular velocity was obtained. The numerical solution of this system yields a three-dimensional
picture which resembles a tornado. Thus Lorenzs butterfly was transformed into a
tornadod .
Consider the Lorenz system23 :
x0 =
(y x),
0
y = xz + rx y,
z 0 = xy bz,
(1)
(2)
(3)
where
, b and r are parameters (a prime denotes differentiation with respect to ).
This system can be reduced to a single third-order ordinary differential equation30
for x, which admits a two-dimensional Lie symmetry algebra if
= 1/2, b = 1 and
r = 0. System (1-3) becomes
(y x)
,
2
0
y = xz y,
x0 =
0
z = xy z.
(4)
(5)
(6)
The corresponding third-order equation is:

2xx000 2x0 x00 + 5xx00 3x02 + 2x3 x0 + 3xx0 + x4 + x2 = 0,
and admits a two-dimensional Lie symmetry algebra L2 with basis:

1
/2
X1 = ,
X2 = e
xx .
2
A basis of its differential invariants of order 2 is given by:

3 0 x
x 2
0
00
x ,
= x + x +
x3 .
= x +
2
2
2
(7)
(8)
(9)
Equation (7) is reduced to the following first-order equation:

22
d In
d
= 2 ,
d
(10)
1972, Lorenz gave a talk at the December meeting of the American Association for the Advancement of Science in Washington, entitled Predictability: Does the Flap of a Butterflys Wings
in Brazil set off a Tornado in Texas?
April 1, 2015
1:18
Maria*Nucci
M.C. Nucci
which can be easily integrated31 to give:

1 + 4 42
= c1 ,
(1 + 2)2
(11)
where c1 is an arbitrary constant. Substitution of x and its derivatives into (11)

yields a second-order ordinary differential equation

2
1 + 4 x00 + 23 x0 + 21 x x3 4 x0 + 12 x x4
= c1 ,
(12)
2
(x3 + 2x00 + 3x0 + x) x6
which admits the Lie symmetry algebra L2 . Lies classification of two-dimensional
algebras into four canonical types32 allows us to integrate (12) by quadrature if we
introduce the canonical variables:
v = 2e /2 ,
u=
e /2
,
x
which transform equation (12) into

2
du
d2 u
4u 2
1+4
dv
dv
"
#2 = c1 ,

2
d2 u
du
2u 2 4
1
dv
dv
(13)
(14)
and operators (8) into

1 = v ,
X
2 = vv + uu .
X
Then the general solution of (14) can be easily derived32 to be:

Z
1/2
v
c1 2c2 u2 c22 u4
du = + c3 ,
2 c1
(15)
(16)
with c2 and c3 arbitrary constants. This solution which involves an elliptic integral
has already been obtained by Sen and Tabor30 by means of a lengthier analysis.
The Euler equations describing the motion of a heavy rigid body about a fixed
point with no torsion are
(B C)
qr,
(17)
A
(C A)
q =
pr,
(18)
B
(A B)
r =
pq,
(19)
C
with A, B and C being the principal moments of inertia, and p(t), q(t) and r(t) the
components of the angular velocity (a dot denotes differentiation with respect to
t). This system can be reduced to a single third-order ordinary differential equation
for, say, p, viz
p =
d3 p dp d2 p 4(C A)(A B) 3 dp
p
= 0,
dt3
dt dt2
BC
dt
(20)
April 1, 2015
1:18
Maria*Nucci
which admits a two-dimensional Lie symmetry algebra L2 with basis:

1 = t ,
2 = tt pp .
(21)
The two Lie symmetry algebras L2 and L2 that we have found are actually the same,
i.e. Type IV in Lies classification32 . Therefore they are linked by a transformation
which takes (, x) into (t, p). Prolongation to the second-order of the two equivalent
Lie symmetry algebras yields a transformation which takes the system (17)-(19)
into the system (4)-(6) as

4
= log 2 ,
t
pt
x=
,
2
C B 2
y=
qrt ,
2A

C B (C A) 2 (A B) 2 2
z=
r +
q t ,
(22)
2A
B
C
with the following condition on the moments of inertia
(A B)(A C)
1
= .
BC
4
(23)
A slightly more general condition could have been considered if one replaces 1/4 with
k/4 (k an arbitrary parameter). If one derives B from (23) by assuming A C > 0
and 4A 3C > 0, i.e.
B=
4A(A C)
,
4A 3C
then the transformation (22) turns into the following

4
= log 2 ,
t
pt
x=
,
2
(2A C)(2A 3C) 2
qrt ,
y=
2A(4A 3C)

(2A C)(2A 3C) 4A2 q 2 (4A 3C)2 r2 2
z=
t ,
8A2 (4A 3C)2
(24)
(25)
and the system (17)-(19) assumes the form

(2A C)(2A 3C)
qr,
A(4A 3C)
3C 4A
q =
pr,
4A
A
r =
pq.
4A 3C
p =
(26)
(27)
(28)
April 1, 2015
1:18
Maria*Nucci
M.C. Nucci
We also derived the inverse transformation, i.e.

t = 2e /2 ,
p = xe /2 ,
(4A 3C)ye /2
,
q= q
p
2 (2A C)(2A 3C)( y 2 + z 2 + z)
qp
Ae /2
y2 + z2 + z
,
r= p
(2A C)(2A 3C)
(29)
which takes the system (4)-(6) into the system (17)-(19) after substituting B as in
(24).
If one applies the transformation (29) to the general Lorenz system (1)-(3), then
the following equations are obtained
2(2A C)(2A 3C)
p
qr + (2
1) ,
A(4A 3C)
t
2 2
2 2
4A
q
(4A
3C)
r q
3C 4A
pr + (b 1) 2 2
q =
2
4A
4A q + (4A 3C) r2 t
2(4A 3C)3 A
+
r
(2A C)(2A 3C) [4A2 q 2 + (4A 3C)2 r2 ]
4A2 q 2 (4A 3C)2 r2 r
A
pq (b 1) 2 2
r =
4A 3C
4A q + (4A 3C)2 r2 t
8(4A 3C)A3
+
r
(2A C)(2A 3C) [4A2 q 2 + (4A 3C)2 r2 ]
p =
(30)
pr
,
t2
(31)
pq
.
t2
(32)
They can be interpreted as the Euler equations of a rigid body moving about a
fixed point and subjected to a torsion which depends on time t and angular velocity
(p, q, r) in the body-frame reference. Also the moments of inertia are related by
means of (24).
We used MAPLE in order to draw a three-dimensional plot of system (30)-(32),
and a tornado was indeed obtained29 .
3. Epidemics: periodic solutions with symmetries
In Ref. 33 we showed that for a certain relationship among the involved parameters
Lie group analysis, when applied to a SIRI disease transmission model formulated
by Derrick and van den Driessche34 , leads to a periodic general solution in apparent
contrast to the qualitative analysis performed in Ref. 34.
Derrick and van den Driessche formulated a model of disease transmission in
a nonconstant population of size N divided into three classes: susceptibles (S),
infectives (I) and recovereds (R).
Individuals move from one compartment to the next according to the following
flow diagram:
April 1, 2015
1:18
Maria*Nucci
bN
?
S
I(S, I, N)
I
I

I(R, I, N )
(d + )I
dS
?
(d + )R
?
The parameter b is per capita birth-rate, d per capita disease free death rate, excess
per capita death rate of infectives, excess per capita death rate of recovereds,
per capita recovery rate of infectives, and per capita loss of immunity rate of
recovereds. The incidence of disease in the susceptible class is given by the function
I(S, I, N ), while I(R, I, N ) is the transfer rate of the recovered class into the
infective class. The above hypotheses lead to the following differential equationse :
S 0 = bN dS + %R I(S, I, N ),
(33)
(34)
(35)
I = I[(S, I, N ) + (R, I, N ) (d + + )],

R = I (d + + %)R I(R, I, N ).
The analysis in Ref. 34 was mainly dedicated to show existence (or nonexistence)
of periodic solutions for the SIRS model (33)-(35) when proportions of individuals
in the three epidemiological classes are considered, i.e.
s = S/N,
i = I/N,
r = R/N .
(36)
With these variables system (33)-(35) becomes

s0 = b(1 s) + %r + si + sr i(s, i),
0
i = (b + + )i + i + ir + i(s, i) + i(r, i),

0
r = i (b + % + )r + ri + r i(r, i),
(37)
(38)
(39)
where (s, i) = (s, i, 1) = (S/N, I/N, N/N ) = (S, I, N ) and (r, i) = (r, i, 1)
= (R/N, I/N, N/N ) = (R, I, N ).
In Ref. 34 a theorem was presented and proved in order to establish under what
conditions system (37)-(39) does not possess periodic solutions in the feasibility
region
D = {s 0, i 0, r 0 : s + i + r = 1}.
(40)
An example of the nonexistence of periodic solutions was then introduced, namely

a special SIRI case of the general model (37)-(39) with % = = 0, (s, i) = s, and
e Here 0
denotes a derivative by t.
April 1, 2015
1:18
Maria*Nucci
M.C. Nucci
(r, i) = r. Since s + i + r = 1 it is possible to eliminate r and finally obtain the

following system:
s0 = b(1 s) ( )si,
(41)
i = i[( )s + ( )i ( + b + )].
(42)
In Ref. 35 Lie group analysis was applied to system (41)-(42), namely to either the
second-order equation in the unknown s that one obtains by deriving i from (41)
or the second-order equation in the unknown i that one obtains by deriving s from
(42). Several cases were found, even instances of hidden linearity. In particular, it
was found33 that when
b = 0,
= 2 ,
=
then a two-dimensional Lie symmetry algebra is admitted by equation

i00 = ((i2 i0 )i0 + i3 + b2 i2 + (i 1)2 i3 + ((i 1)i + i0 )i2
(i + 2i0 + ( + )(i 1)i)i2 + (i2 + i0 + i
+(i 1)i (2i 1)i)bi)/i,
(43)
which is obtained from system (41)-(42) by deriving s from equation (42), i.e.
s=
[b + ( )(i 1)]i + i0
,
( )i
(44)
and substituting it into equation (41). The Lie symmetry algebra is generated by
the operators
1 = tt ii ,
2 = t .
(45)
This means that equation (43) can be easily integrated by quadrature. Its general
solution is
a1

,
(46)
i=
a1 a2 a1 t
sin
(2 2 + 2 )

and from (44) one obtains:

a1 a2 a1 t
a1 cos
1
1

.
s=
a1 a2 a1 t
2
(2 2 + 2 )
sin

(47)
This general solution of system (41)-(42) is clearly periodic in apparent contrast

with the findings in Ref. 34. Note that the functions (46)-(47) are neither bounded
nor positive nor continuous, and do not belong to the feasibility region (40). In fact
b must be positive for nonexistence of periodic solutions. However in Ref. 34 the
condition b = 0 was allowed in order to show that system (37)-(39) has periodic
solutions if (s, i) = si, and (r, i) = 0.
April 1, 2015
1:18
Maria*Nucci
4. Determining Lagrangians from symmetries

The inverse problem of calculus of variation has attracted a lot of interest since in the
second half of the 18th century Euler36 and then Lagrange37 introduced the direct
problem, namely the idea of linking the solution of a differential equation to the
maximum/minimum of a functional, the celebrated problem of the brachistochrone
being indeed the most famous classical example. It will take hundreds of pages to
cite all the papers and books that have been published since. Most authors mark
the birthdate of the inverse problemf with the 1887-papers by either Helmholtz38 or
Volterra39 . Some other especially among the Russian speaking researchers pushes
the date slightly back to the 1886-paper by Sonin40 . Very few recognize the seminal
work by Jacobi, namely his 1845-paper41 and his 1842-1843 Dynamics Lectures
published posthumously in 188442 , available in English43 since 2009, where he links
his last multiplier to the Lagrangian for any even-order ordinary differential equation
(ODE). Actually both Volterrag and Sonin recognize the contribution of Jacobi last
multiplier in their papers, Sonin more explicitly than Volterra since he showed that
his own method involves the Jacobi last multiplier (p.10 in Ref. 40).
It was shown in Ref. 45 that Jacobi Last Multiplier yields the Lagrangian for
any equation of even orderh
u(2n) = F (x, u, u0 , u00 , . . . , u(2n1) ),
(48)
since it can be derived from the following formula

M 1/n =
2L
,
(u(n) )2
(49)
where M is the Jacobi Last Multiplier of equation (48) and L is its Lagrangian.
This formula was given by Jacobi himself in Ref. 41 p. 364.
We recall that Fels has proved46 that the Lagrangian is unique in the case of
fourth-order equation, provided it exists. In the case of equations of sixth and eight
order the uniqueness was proved by Juras in Ref. 47.
The method of Jacobi last multiplier was enhanced when Lie determined the
link with his symmetries48 , a link very easy to implement that allows us to derive
many multipliers and therefore Lagrangians.
As pointed out by Tonti49 many authors have dealt with the inverse problem of
calculus of variations by either using a formal approach or an operatorial approach
following on the steps of either Helmholtz or Volterra: for example Refs. 50-55 and
many others.
We have not underestimated the research of these very distinguished authors.
Yet when possible we prefer to follow Jacobi since his Last Multiplier has a direct
f Roughly
speaking, the problem of finding a Lagrangian if it exists.

his 1906 address at the Congress of Italian Naturalists44 Volterra wrote One of the most
celebrated discovery by the mathematician Jacobi, that of the principle of the last multiplier.
h We use a prime to indicate the derivative with respect to x.
g In
April 1, 2015
10
1:18
Maria*Nucci
M.C. Nucci
link to conservation laws and symmetries that are the essential elements that in
our opinion make the difference between a mathematical abstraction and physical
concreteness56 .
We would like also to mention that many systems do not admit a Lagrangian.
Nevertheless they may admit a Lagrangian if put in a different form as suggested
by Bateman57 , namely finding a set of equations equal in number to a given set,
compatible with it and derivable from a variational principle. In Ref. 58 it was
demonstrated how to construct many different Lagrangians for two famous examples
which were deemed by Douglas59 not to have a Lagrangian. Following Batemans
dictat different sets of equations compatible with those by Douglas and derivable
from a variational principle were found in Ref. 58.
In Ref. 56 it was shown that a method presented by Trubatch and Franco60
and later by Paine61 for finding Lagrangians of classic models in biology, is actually
based on finding the Jacobi Last Multiplier of such models. Using known properties
of Jacobi Last Multiplier it was shown how to obtain linear Lagrangians of systems
of two first-order ordinary differential equations and nonlinear Lagrangian of the
corresponding single second-order equation that can be derived from them, even in
the case where those authors failed such as the host-parasite model. Also it was
shown that the Lagrangians of certain second-order ordinary differential equations
derived by Volterra in Ref. 62 are particular cases of the Lagrangians that can
be obtained by means of the Jacobi Last Multiplier and consequently more than
one Lagrangian for those Volterras equations were derived. Here after a survey on
Jacobi Last Multiplier and its properties, we show two examples from Ref. 56.
4.1. Jacobi Last Multiplier

The method of the Jacobi Last Multiplier provides a means to determine all the
solutions of the partial differential equation
Af =
n
X
ai (x1 , . . . , xn )
i=1
f
=0
xi
(50)
or its equivalent associated Lagranges system

dx1
dx2
dxn
=
= ... =
.
a1
a2
an
(51)
In fact, if one knows the Jacobi Last Multiplier and all but one of the solutions,
namely n 2 solutions, then the last solution can be obtained by a quadrature. The
Jacobi Last Multiplier M is given by
(f, 1 , 2 , . . . , n1 )
= M Af,
(x1 , x2 , . . . , xn )
(52)
April 1, 2015
1:18
Maria*Nucci
11
where
f
f
x1 xn

1
1
(f, 1 , 2 , . . . , n1 )
x
x
1
n
= det
..
..
(x1 , x2 , . . . , xn )
.
.
n1
n1
x1
xn
=0
(53)
and 1 , . . . , n1 are n 1 solutions of (50) or, equivalently, first integrals of

(51) independent of each other. This means that M is a function of the variables
(x1 , . . . , xn ) and depends on the chosen n 1 solutions, in the sense that it varies
as they vary. The essential properties of the Jacobi Last Multiplier are:
(a) If one selects a different set of n 1 independent solutions 1 , . . . , n1 of
equation (50), then the corresponding last multiplier N is linked to M by
the relationship:
N =M
(1 , . . . , n1 )
.
(1 , . . . , n1 )
(b) Given a non-singular transformation of variables

:
(x1 , x2 , . . . , xn ) (x01 , x02 , . . . , x0n ),
then the last multiplier M 0 of A0 F = 0 is given by:

M0 = M
(x1 , x2 , . . . , xn )
,
(x01 , x02 , . . . , x0n )
where M obviously comes from the n 1 solutions of AF = 0 which correspond to those chosen for A0 F = 0 through the inverse transformation
1 .
(c) One can prove that each multiplier M is a solution of the following linear
partial differential equation:
n
X
(M ai )
i=1
xi
= 0;
(54)
viceversa every solution M of this equation is a Jacobi Last Multiplier.

(d) If one knows two Jacobi Last Multipliers M1 and M2 of equation (50), then
their ratio is a solution of (50), or, equivalently, a first integral of (51).
Naturally the ratio may be quite trivial, namely a constant. Viceversa the
product of a multiplier M1 times any solution yields another last multiplier M2 = M1 .
Since the existence of a solution/first integral is consequent upon the existence
of symmetry, an alternative formulation in terms of symmetries was provided by
Lie48,63 . A clear treatment of the formulation in terms of solutions/first integrals
April 1, 2015
12
1:18
Maria*Nucci
M.C. Nucci
and symmetries is given by Bianchi64 . If we know n 1 symmetries of (50)/(51),

say
i =
n
X
ij (x1 , . . . , xn )xj ,
i = 1, n 1,
(55)
j=1
a Jacobi last multiplier is given by M = 1 , provided that 6= 0, where
a1 an
1,1
1,n
= det .
.. .
..
.
(56)
n1,1 n1,n
There is an obvious corollary to the results of Jacobi mentioned above. In the case
that there exists a constant multiplier, the determinant is a first integral. This result
is potentially very useful in the search for first integrals of systems of ordinary
differential equations. In particular, if each component of the vector field of the
equation of motion is missing the variable associated with that component, i.e.,
ai /xi = 0, the last multiplier is a constant, and any other Jacobi last multiplier
is a first integral.
Another property of the Jacobi Last Multiplier is its relationship with the Lagrangian, namely Jacobis formula (49), that for any second-order equation
x
= (t, x, x)
(57)
becomes (see also Ref. 65)

M=
2L
x 2
(58)
where M = M (t, x, x)
satisfies the following equation
d
(log M ) +
= 0.
dt
x
Then equation (57) becomes the Euler-Lagrangian equation:

L
d L
+
= 0.
dt x
x
(59)
(60)
The proof65 is given by taking the derivative of (60) by x and showing that this
yields (59). If one knows a Jacobi Last Multiplier, then L can be obtained by a
double integration, i.e.:

Z Z
L=
M dx dx + `1 (t, x)x + `2 (t, x),
(61)
where `1 and `2 are functions of t and x which have to satisfy a single partial
differential equation related to (57), as it was shown in Ref. 66. However in Ref. 67,
April 1, 2015
1:18
Maria*Nucci
13
`1 , `2 were related to the gauge function F = F (t, x). In fact, we may assume
F
x
F
`2 =
+ `3 (t, x)
t
`1 =
(62)
where `3 has to satisfy the mentioned partial differential equation and F is obviously
arbitrary.
In Ref. 60 it was shown that a system of two first-order ordinary differential
equations
u 1 = 1 (t, u1 , u2 )
u 2 = 2 (t, u1 , u2 )
(63)
always admits a linear Lagrangian of the form

L = U1 (t, u1 , u2 )u 1 + U2 (t, u1 , u2 )u 2 V (t, u1 , u2 ).
(64)
The key is a function W such that

W =
U1
U2
=
u2
u1
(65)
and
d
1
2
(log W ) +
+
= 0.
dt
u1
u2
(66)
It is obvious that equation (66) is the equation (54) of the Jacobi Last Multiplier for
system (63). Therefore once a Jacobi Last Multiplier M (t, u1 , u2 ) has been found,
then a Lagrangian of system (63) can be obtained by two integrations, i.e.:
Z

Z

d
(67)
L=
M du1 u 2
M du2 u 1 + g(t, u1 , u2 ) + G(t, u1 , u2 ),
dt
where g(t, u1 , u2 ) satisfies two linear differential equations of first order that can
be always integrated, and G(t, u1 , u2 ) is the arbitrary gauge function that should
be taken into consideration in order to correctly apply Noethers theorem68 . If a
Noethers symmetry
= (t, u1 , u2 )t + 1 (t, u1 , u2 )u1 + 2 (t, u1 , u2 )u2
(68)
exists for the Lagrangian L in (67) then a first integral of system (63) is
L
L
L
(1 u 1 )
(2 u 2 ) + G(t, u1 , u2 ).
u 1
u 2
(69)
April 1, 2015
14
1:18
Maria*Nucci
M.C. Nucci
4.2. Volterra-Lotkas model

The Volterra-Lotkas model considered in Ref. 60 is the following:
w 1 = w1 (a + bw2 )
w 2 = w2 (A + Bw1 ).
(70)
In order to simplify system (70) we follow Ref. 60 and introduce the change of
variables
w1 = exp(r1 ),
w2 = exp(r2 )
(71)
and then system (70) becomes

r1 = b exp(r2 ) + a
r2 = B exp(r1 ) + A.
(72)
An obvious Jacobi Last Multiplier of this system is a constant, say 1, and consequently by means of (67) a linear Lagrangian of system (72) is
L[r] = r1 r2 r2 r1 + 2(B exp(r1 ) + b exp(r2 ) Ar1 + ar2 ) +
d
G(t, r1 , r2 ) (73)
dt
which (minus the gauge function G) was found in Ref. 60. Moreover we can derive a
Jacobi Last Multiplier for the Volterra-Lotka system (70) by using property (b). In
fact we have to calculate the Jacobian of the transformation (71) between (w1 , w2 )
and (r1 , r2 ) and this yields a Jacobi Last Multiplier of system (70), i.e.

1

0

(r1 , r2 )
w1

= 1 .
M[w] = M[r]
=
(74)
1 w1 w2
(w1 , w2 )
0 w
2
Finally, formula (67) yields a linear Lagrangian of system (70)
L[w] = log(w1 )
w 2
w 1
log(w2 )
+ 2(A log(w1 ) + a log(w2 ) Bw1 + bw2 ). (75)
w2
w1
This Lagrangian was not obtained in Ref. 60. We note that (70) is autonomous
and therefore invariant under time translation, namely t . It is easy to show that
the Lagrangian L[w] in (75) yields a time-invariant first integral through Noethers
theorem68 , i.e.:
L[w] + w 1
L[w]
L[w]
+ w 2
= A log(w1 ) a log(w2 ) + Bw1 bw2 = const. (76)
w 1
w 2
Following Ref. 60 we can transform system (72) into an equivalent second-order ordinary differential equation by eliminating, say, r1 . In fact from the second equation
in (72) one gets

r2 A
r1 = log
,
(77)
B
April 1, 2015
1:18
Maria*Nucci
and the equivalent second-order equation in r2 is the following

r2 = b exp(r2 ) + a (A r2 ).
15
(78)
A Jacobi Last Multiplier for this equation has to satisfy equation (59), i.e.:
d
(log M ) + b exp(r2 ) + a = 0
dt
(79)
d
(log M ) + r1 = 0,
dt
(80)
namely
by taking into account the first equation in (72), and consequently we get the
following Jacobi Last Multiplier for equation (78):
M1 = exp(r1 ) =
B
,
r2 A
(81)
the last equality thanks to (77). Then a Lagrangian can be obtained by a double
integration as in (61), i.e.

d
L1 = B (r2 A) log(A r2 ) r2 + b exp(r2 ) + ar2 + F (t, r2 ).
(82)
dt
The same Lagrangian (minus the gauge function F ) was obtained in Ref. 60. We
note that (78) is autonomous and therefore invariant under time translation. It is
easy to show that the Lagrangian L1 in (82) yields a time-invariant first integral,
through Noethers theorem68 , i.e.:
I1 = ar2 + r2 + A log(A r2 ) b exp(r2 ) = const.
(83)
4.3. Volterra-Verhulst-Pearl equation

In 1939 Volterra wrote62 : I have been able to show that the equations of the struggle
for existence depend on a question of Calculus of Variations (omissis). In order to
obtain this result, I have replaced the notion of population by that of quantity of life
[69]. In this manner I have also obtained some results by which dynamics is brought
into relation to problems of the struggle for existence. The quantity of life X and
the population N of a species are connected by the relation
N=
dX
.
dt
(84)
It is immediately obvious that this idea of raising the order of each equation is totally
different from that by Trubach and Franco who provided a method for finding a
linear Lagrangian for systems of first-order equations. Also Volterras method is
different from that of deriving a single second-order equation from a system of two
first-order equations: indeed Volterra takes a system of first-order equations and
transform it into a system of second-order equations.
April 1, 2015
16
1:18
Maria*Nucci
M.C. Nucci
One of the two equations considered by Volterra is the Verhulst-Pearl equation

dN
= N ( N )
dt
(85)
that through (84) becomes

dX
d2 X
=
2
dt
dt
dX
dt

.
(86)
At glance it admits a two-dimensional Lie symmetry algebra generated by the operators t and X . Then a Jacobi Last multiplier for (86) can be obtained by means
of (56), i.e.

dX
dX
dX
1 dt
dt dt
1
= M(86) = 1 =
(87)
(86) = det 1 0
0
dX
dX
(86)
dt dt
0 1
0
and consequently from Jacobis formula (58) we obtain the Lagrangian

1 dX
dX
1
dX
dX
L(86) =
log
+
log
+X
dt
dt
dt
dt
(88)
which is indeed the Volterras Lagrangian in Ref. 62.

Actually equation (86) admits an eight-dimensional Lie symmetry algebra generated by the following operators:

1 = exp(X t)t , 2 = exp(X) t + X , 3 = exp(X + t)X ,

4 = exp(X)X , 5 = exp(t) t + X , 6 = X ,
7 = exp(t)t ,
8 = t .
(89)
Therefore equation (86) is linearizable by means of a point transformation. In order

to find the linearizing transformation we have to look for a two-dimensional abelian
intransitive subalgebra, and, following Lies classification of two-dimensional algebras in the real plane63 , we have to transform it into the canonical form
u ,
yu
with u and y the new dependent and independent variables, respectively. We found
that one such subalgebra is that generated by X3 and X4 . Then it is easy to derive
that
1
y = exp(t),
u = exp(X t)
(90)
and equation (86) becomes

d2 u
= 0.
dy 2
(91)
As we have shown above the Volterras Lagrangian (88) of the equation (86) comes
from the Jacobi Last Multiplier that can be obtained by means of (56) with the two
April 1, 2015
1:18
Maria*Nucci
17
symmetries 8 and 6 in (89). This Lagrangian (88) admits two Noether symmetries and therefore two first integrals of equation (86) can be derived by Noethers
theorem68 , i.e.

dX
dX
6 = In6 = log
log
+ t
dt
dt

dX
8 = In8 = log
+ X.
(92)
dt
Other nine Jacobi Last Multipliers and therefore Lagrangians can be obtained by
means of (56) and any other combination of two symmetries in (89). The nine Jacobi
Last Multipliers are:
exp(t)
exp(X)
exp(2t X)
,
, JLM15 = dX
, JLM17 =

dX 2
dX 2
dX 3
dt
dt
dt
dt
JLM14 =
JLM18 =
exp(t X)

dX 2
dt
JLM34 =
dX
dt
, JLM23 =
exp(t X)
exp(t)
,
, JLM25 =
3
dX 2
dt
dX
dt
exp(t 2X)
exp(t + X)
exp(X)
, JLM36 =
, JLM37 =
,(93)
dX
dX
dt
dt
and the indices indicate which two of the symmetries in (89) have been used. Consequently the nine Lagrangians are:

dX
1
log
+X ,
Lag14 = exp(t)
dt

1 dX
dX
dX
dX
1
1
Lag15 = exp(X)
log
+
+ log
,
dt
dt
dt
dt
1
Lag17 = dX exp(2t X),
2 dt

1
dX
dX
dX
log
log
,
Lag18 = 2 exp(t X)
dt
dt
dt

1
dX
Lag23 = exp(X) log
+ X ,
dt
exp(t X)
,
Lag25 =
2 t dX
dt

2
dX
1
Lag34 = exp(t + 2X)
,
2
dt

1
dX
dX
dX
Lag36 = 2 exp(t + X)
log
dt
dt
dt

1
dX
dX
dX
Lag37 = exp(X)
log
+
.
(94)
dt
dt
dt
These Lagrangians admit a different number of Noether symmetries. The Lagrangians Lag17 , Lag25 , Lag34 admit five Noether symmetries, the possible higher
April 1, 2015
18
1:18
Maria*Nucci
M.C. Nucci
number. For example the Lagrangian Lag34 in (94) yields the following five Noether
symmetries and corresponding first integrals of equation (86)

dX
,
3 = Int3 = exp(X) +
dt
dX
4 = Int4 = exp(t + X)
,
dt

2
dX
5 = Int5 = exp(2X)
,
dt

dX
dX
,
6 + 2 8 = Int6 = exp(t + 2X)
dt
dt

2
dX
.
(95)
7 = Int7 = exp(2t + 2X)
dt
5. Quantizing with Noether symmetries
Different methods exist to describe the dynamics of quantum systems, namely
(i) the time-dependent Schr
odinger equation with the corresponding continuity
equation for the probability density;
(ii) the description using a time propagator, also called Feynman kernel or space
representation of the Green function;
(iii) the time-dependent Wigner function.
All three methods were shown to be linked via a dynamical invariant, the so-called
Ermakov invariant, in Ref. 70. Therefore we pursue the quantization of classical
problems by searching for a time-dependent Schrodinger equation.
In Ref. 71 it was inferred that Lie symmetries should be preserved if a consistent
quantization is desired. In Ref. 72 [ex. 18, p. 433] an alternative Hamiltonian for
the simple harmonic oscillator was presented. It is obtained by applying a nonlinear
canonical transformation to the classical Hamiltonian of the harmonic oscillator.
That alternative Hamiltonian was used in Ref. 73 to demonstrate what nonsense
the usual quantization schemesi produce. In Ref. 77 a quantization scheme that preserves the Noether symmetries was proposed and applied to Goldsteins example in
order to derive the correct Schr
odinger equation. In Ref. 78 the same quantization
scheme was applied in order to quantize the second-order Riccati equation, while in
Ref. 79 the quantization of the dynamics of a charged particle in a uniform magnetic
field in the plane and Calogeros goldfish system were achieved. In Ref. 80 the same
method yielded the Schr
odinger equation of an equation related to a Calogeros
goldfish, and Ref. 81 that of two nonlinear equations somewhat related to the Riemann problem82 . In Ref. 83, and Ref. 84 it was shown that the preservation of the
i Such
as normal-ordering74,75 and Weyl quantization76 .
April 1, 2015
1:18
Maria*Nucci
19
Noether symmetries straightforwardly yields the Schrodinger equation of a Lienard

I nonlinear oscillator in the momentum space 85 , and that of a family of Lienard II
nonlinear oscillators 86 , respectively.
If a system of second-order equations is considered, i.e.
x IRN ,
(t) = F(t, x, x),
(96)
that comes from a variational principle with a Lagrangian of first order, then the
quantization method that was first proposed in Ref. 77 consists of the following
steps:
Step I. Find the Lie symmetries of the Lagrange equations
= W (t, x)t +
N
X
Wk (t, x)xk .
k=1
Step II. Among them find the Noether symmetries

= V (t, x)t +
N
X
Vk (t, x)xk .
k=1
This may require searching for the Lagrangian yielding the maximum possible number of Noether symmetries66,87,88,89 .
Step III. Construct the Schr
odinger equationj admitting these Noether symmetries as Lie symmetries, namely
2it +
N
X
fkj (x)xj xk +
k,j=1
N
X
hk (x)xk + f0 (x) = 0
(97)
k=1
admitting the Lie symmetries

= V (t, x)t +
N
X
Vk (t, x)xk + G(t, x, ) ,
k=1
without adding any other symmetries apart from the two symmetries that
are present in any linear homogeneous partial differential equation, namely
,
(t, x) ,
where = (t, x) is any solution of the Schrodinger equation (97).

Here we present the quantization with Noether symmetries of a charged particle
in a uniform magnetic field in the plane as derived in Ref. 79. The corresponding
classical Lagrangian is
L=
j We
1 2
(x + y 2 ) + (y x xy)
assume ~ = 1 without loss of generality.
(98)
April 1, 2015
20
1:18
Maria*Nucci
M.C. Nucci
and consequently the Lagrangian equations are

x
= y,
y = x.
(99)
In Ref. 90 the following Schr

odinger equation was determined
2 2
(x + y 2 ) = 0.
(100)
4
The Lie symmetry algebra admitted by the linear system (99) has dimension
fifteen91 , and the classical Lagrangian (98) admits eight Noether symmetries generated by the following operators:
2it + xx + yy i(yx xy )
1
1
(sin(t)x + cos(t)y) x + (cos(t)x sin(t)y) y ,
2
2
1
1
X2 = sin(t)t (cos(t)x sin(t)y) x (sin(t)x + cos(t)y)y ,
2
2
X3 = t ,
X1 = cos(t)t
X4 = yx + xy ,
X5 = sin(t)x + cos(t)y ,
X6 = cos(t)x sin(t)y ,
X7 = y ,
X8 = x .
(101)
The Schr
odinger equation (100) admits an infinite Lie symmetry algebra generated
by the operator (t, x, y) , where is any solution of the equation itself, and also
a finite dimensional Lie symmetry algebra generated by the following operators:

1
2 sin(t) i cos(t) 2 (x2 + y 2 ) ,
Y1 = X1 +
4

1
Y2 = X2 +
2 cos(t) + i sin(t) 2 (x2 + y 2 ) ,
4
Y3 = X3 ,
Y4 = X4 ,
1
Y5 = X5 (x cos(t) + y sin(t)) ,
2
1
Y6 = X6 + (x sin(t) y cos(t)) ,
2
i
Y7 = X7 + x ,
2
i
Y8 = X8 y ,
2
Y9 = .
(102)
This known example supports the method introduced here, namely that the
Schr
odinger equation admits a finite Lie symmetry algebra that corresponds to
the Noether symmetries admitted by the classical Lagrangian plus the symmetry
Y9 admitted by any homogeneous linear partial differential equation.
April 1, 2015
1:18
Maria*Nucci
21
6. Nonclassical symmetries and heir-equations

Nonclassical symmetries were introduced in 1969 in a seminal paper by Bluman and
Cole92 . They obtained new exact solutions of the linear heat equation, i.e. solutions
not deducible from the classical symmetry method. (Some authors call them Qconditional symmetriesk of the second type, e.g. Ref. 94, while others call them
reduction operators, e.g. Ref. 95.)
The nonclassical symmetry method can be viewed as a particular instance of the
more general differential constraint method that, as stated by Kruglikov96 , dates
back at least to the time of Lagrange... and was introduced into practice by Yanenko,
see Ref. 97. The method was set forth in details in Yanenkos monograph98 that
was not published until after his death99 . A more recent account and generalization
of Yanenkos work can be found in Ref. 100.
The nonclassical symmetry method consists of adding the invariant surface condition to the given equation, and then apply the classical symmetry method. The
main difficulty of this approach is that the determining equations are no longer linear. On the other hand, the nonclassical symmetry method may give more solutions
than the classical symmetry method.
Galaktionov101 and King102 had found exact solutions of certain evolution equations which are not derived by either the classical or nonclassical symmetry method.
However, in Ref. 103 it was shown that these solutions can be obtained by iterating
the nonclassical symmetry method. Successive iterations generate new equations,
that were named heir-equations because, although more complex than the original
equations, they inherit the same Lie symmetry algebra. Invariant solutions of these
heir-equations are just the solutions found by Galaktionov and by King.
We recall that the heir-equations are just some of the many possible n-extended
equations as defined by Guthrie in Ref. 104.
In Ref. 103 it was also shown that this iterating method yields both partial
symmetries as given by Vorobev in Ref. 105, and differential constraints as given
by Olver106 .
Fokas and Liu107 and Zhdanov108 independently introduced the method of generalised conditional symmetries, i.e., conditional Lie-Backlund symmetries. In Ref. 109
it was shown that the heir-equations can retrieve all the conditional Lie-Backlund
symmetries found by Zhdanov. In Ref. 110 Goard has shown that Nuccis method
of constructing heir equations by iterating the nonclassical symmetries method is
equivalent to the generalised conditional symmetries method.
Moreover in Ref. 111 it was shown that one can find the nonclassical symmetries of any evolution equations of any order by using a suitable heir-equation and
searching for a given particular solution among all its solutions, thus avoiding any
complicated calculations.
k In
Ref. 93 this name was introduced for the first time.
April 1, 2015
22
1:18
Maria*Nucci
M.C. Nucci
In Ref. 112 a class of reaction-diffusion equations, i.e.

ut = uxx + cux + R(u, x),
(103)
with nonlinear source R(u, x), was introduced as a model that incorporates climate
shift, population dynamics, and migration for a population of individuals u(t, x)
that reproduce, disperse, and die within a patch of favorable habitat surrounded by
unfavorable habitat. It is assumed that due to a shifting climate, the patch moves
with a fixed speed c > 0 in a one-dimensional universe. In Ref. 113 nonclassical
symmetries of (103) were searched for. Several cases were obtained by using suitable
solutions of the heir-equations. Here after recalling the concept of heir-equations103
and their link to nonclassical symmetries111 , we outline its application to (103) and
present one of the cases determined in Ref. 113, where one can find more details.
6.1. Heir-equations and nonclassical symmetries
For the sake of simplicity let us consider an evolution equation in two independent
variables and one dependent variable of second order:
ut = H(t, x, u, ux , uxx ).
(104)
= V1 (t, x, u)t + V2 (t, x, u)x F (t, x, u)u
(105)
If
is a generator of a Lie point symmetry of equation (104) then the invariant surface
condition is given by:
V1 (t, x, u)ut + V2 (t, x, u)ux = F (t, x, u).
(106)
Let us take the case with V1 = 0 and V2 = 1, so that (106) becomesl :

ux = G(t, x, u)
(107)
Then, an equation for G is easily obtained. We call this the G-equation114 . Its
invariant surface condition is given by:
1 (t, x, u, G)Gt + 2 (t, x, u, G)Gx + 3 (t, x, u, G)Gu = (t, x, u, G).
(108)
Let us consider the case 1 = 0, 2 = 1, and 3 = G, so that (108) becomes:

Gx + GGu = (t, x, u, G),
(109)
which yields equation the -equation. Clearly:

Gx + GGu uxx .
(110)
l We have replaced F (t, x, u) with G(t, x, u) in order to avoid any ambiguity in the following discussion.
April 1, 2015
1:18
Maria*Nucci
23
We could keep iterating to obtain the -equation, which corresponds to:

x + Gu + G uxxx (t, x, u, G, ),
(111)
the -equation, which corresponds to:

x + Gu + G + uxxxx (t, x, u, G, , )
(112)
and so on. Each of these equations inherits the symmetry algebra of the original
equation, with the right prolongation: first prolongation for the G-equation, second
prolongation for the -equation, and so on. Therefore, these equations were named
heir-equations in Ref. 103. This implies that even in the case of few Lie point
symmetries many more Lie symmetry reductions can be performed by using the
invariant symmetry solution of any of the possible heir-equations, as it was shown
in Refs. 103, 115, 116.
6.2. Nonclassical symmetries of equation (103)
The Gequation of (103) is:
Gt + RGu Gxx 2GGxu G2 Guu cGx Ru G Rx = 0.
(113)
and the equation is

t + Ru + Ru GG xx 2Gxu 2xG G2 uu 2GuG
2 GG cx Ruu G2 Ru 2GRxu + Rx G Rxx = 0.
(114)
The particular solution of the -equation that we are looking for is

(t, x, u, G) = R(u, x) cG + F (t, x, u) V2 (t, x, u)G,
(115)
that replaced into (114) yields an overdetermined system in the unknowns F , V2

and R(u, x). We obtain a polynomial of third degree in G which four coefficients we
call di , (i = 0, 1, 2, 3) where i stands for the corresponding power of G. We impose
all of them to be zero. From d3 , we obtain
V2 (t, x, u) = ss1 (t, x)u + ss2 (t, x),
(116)
while d2 yields
1
1
F (t, x, u) = ss21 u3 +
3
2

ss1
2css1 2ss1 ss2 + ss3 (t, x)u + ss4 (t, x), (117)
x
with ssj (t, x), (j = 1, 2, 3, 4) arbitrary functions of t and x. Then after differentiating d1 four times with respect to u we obtain
4 R(u, x)
= 0,
(118)
u4
which implies that R(u, x) must be a polynomial in u of third degree at most, i.e.
R(u, x) =
a23 (x) 3 a2 (x) 2

u +
u + a1 (x)u + a0 (x),
6
2
(119)
April 1, 2015
24
1:18
Maria*Nucci
M.C. Nucci
where ai (x), (i = 0, 1, 2, 3) are arbitrary functions of x. Since none of the remaining

arbitrary functions depends on u, and d1 has now become a polynomial of degree
3 in u, we have to annihilate all the four coefficients, i.e. d1,i , (i = 0, 1, 2, 3). From
d1,3 we have that ss1 (t, x) must be a constant, and two cases raise:
3
a3 (x),
Case 1. ss1 =
2
Case 2. ss1 = 0.
In Case 1, from coefficients d1,2 , d1,1 , d1,0 we obtain ss2 , ss3 , and ss4 , respectively.
All of them are functions of x only, e.g.

1
4a03 (x) + 3a2 (x) + 2ca3 (x) ,
(120)
ss2 =
2a3 (x)
where 0 denotes differentiation with respect to x. Now the only remaining coefficient
is d0 which has become a linear polynomial in u. Therefore we are left with two
expressions to annihilate, namely an underdetermined system of two equations that
contain the derivative of a3 (x) up to fifth order, and fourth order, respectively, and
lower derivatives
of the other three functions a2 (x), a1 (x), and a0 (x). If we assume
a3 (x) = 3x, and a2 (x), a1 (x), a0 (x) to be constants then we obtain that
1
1
R(u, x) = x2 u3 + 3u2 + c2 u,
2
2
(121)
and
3x
1 + cx
2 + 3cx
, ss2 (t, x) =
, ss3 (t, x) = c
, ss4 (t, x) = 0. (122)
2
x
4x
Thus, (115) becomes
ss1 (t, x) =
x3 u3 + 2cu c2 xu + 6x2 uG 4G
,
4x
(123)
namely
x3 u3 + 2cu c2 xu + 6x2 uux 4ux
,
4x
that can be solved in closed form, i.e.
uxx =
u(t, x) =
c2 R2 (t)e
cx
2
c2 (1 + cx)e
R1 (t) + (cx 2)R2 (t)e
cx
2
(124)
cx
2
+ (10 + 5cx + c2 x2 )e
cx
2
(125)
with Rk (t), k = 1, 2 arbitrary functions of t. Substituting (125) into (103) yields the
following nonclassical symmetry solution
c2 c1 ec
u(t, x) =
c2 e
c2 t
4
t+ cx
2
+ c1 (cx 2)ec
c2 (1 + cx)e
2 t+ cx
2
cx
2
+ (10 + 5cx + c2 x2 )e
cx
2
(126)
with ck , k = 1, 2 arbitrary constants. We observe that

lim u(t, x) =
c2
,
cx 2
lim u(t, x) = 0
(127)
April 1, 2015
1:18
Maria*Nucci
25
and that u(t, x) < 0 for t > 0, x < 0. This means that the solution (126) is not
defined at x = 2/c and is positivem if x 0.
More cases can be found in Ref. 113.
7. What can symmetries do for you?
Symmetries can
transform a nonlinear problem into a linear one;

determine exact solutions, and/or conservation laws;
transform a butterfly into a tornado;
determine periodic solution for non-periodic problems;
determine Lagrangians for (nearly) all problems;
quantize a classical problem;
and much more since we do not claim to have exhausted all the possibilities.
Last but not least, symmetries may also inspire beauty and peace in you
https://www.flickr.com/photos/17667265@N07/3192922037/in/photostream/
Acknowledgments
The author acknowledges the support of the Italian Ministry of University and
Scientific Research through PRIN 2010-2011, Prot. 2010JJ4KPA 004.
References
1. P. M. Harman (ed.), The Scientific Letters and Papers of James Clerk Maxwell Vol. I
(Cambridge University Press, Cambridge, 1990).
2. T. Hawkins, Arch. Hist. Exact Sciences 42, 187 (1991).
3. P. J. Olver, Applications of Lie Groups to Differential Equations, (Springer-Verlag,
New York, 1986).
4. L. V. Ovsjannikov, Group Analysis of Differential Equations, (Academic Press, New
York, 1982).
5. W.F. Ames, Nonlinear Partial Differential Equations in Engineering, Vol. 2, (Academic Press, New York, 1972).
6. G. W. Bluman, and J. Cole, Similarity Methods for Differential Equations, (SpringerVerlag, New York, 1974).
7. G. W. Bluman and S. Kumei, Symmetries and Differential Equations, (SpringerVerlag, Berlin, 1989).
8. H. Stephani, Differential Equations. Their Solution Using Symmetries, (Cambridge
University Press, Cambridge, 1989).
9. J.M. Hill, Differential Equations and Group Methods for Scientists and Engineers,
(CRC Press, Boca Raton, FL, 1992).
10. N.H. Ibragimov (ed.), CRC Handbook of Lie Group Analysis of Differential Equations,
Vol. I: Symmetries, Exact Solutions, and Conservation laws, (CRC Press, Boca Raton,
FL, 1994).
m It
depends also on the values given to the arbitrary constants.
April 1, 2015
26
1:18
Maria*Nucci
M.C. Nucci
11. N. H. Ibragimov, Elementary Lie Group Analysis and Ordinary Differential Equations,
(Wiley, New York, 1999).
12. P. E. Hydon, Symmetry Methods for Differential Equations: A Beginners Guide,
(Cambridge University Press, 2000).
13. G. W. Bluman and S. C. Anco, Symmetry and Integration Methods for Differential
Equations, (Springer-Verlag, Berlin, 2002).
14. M. C. Nucci, Interactive REDUCE programs for calculating Lie point, non-classical,
Lie-B
acklund, and approximate symmetries of differential equations: manual and
floppy disk. In CRC Handbook of Lie Group Analysis of Differential Equations. Vol. 3:
New Trends in Theoretical Developments and Computational Methods, N.H. Ibragimov
ed., (CRC Press, Boca Raton, 1996) p. 415.
15. V. V. Golubev, Lectures on Integration of the Equations of Motions of a Rigid Body
about a Fixed Point (State Publishing House of Theoretical Technical Literature,
Moscow, 1953) English translation published by the Israel Program for Scientific Literature in 1960.
16. J. C. Maxwell, Transactions of the Royal Society of Edinburgh XXI (IV), 559 (1857).
17. L. Euler, Theoria motus corporum solidorum seu rigidorum ex primis nostrae cognitionis principiis stabilita et ad omnes motus, qui in huiusmodi corpora cadere possunt,
accommodata (Litteris et Impensis A. F. R
ose, Rostochii et Gryphiswaldiae, 1765).
18. L. Poinsot, J. Math. Pures Appl. 16, 9 (Première et deuxième Parties) 289 (Troisième
et dernière Parties) (1851).
19. K. G. J. Jacobi, J. Reine Angew. Math. 39, 293 (1850).
20. A. Natucci, Atti Accad. Ligure 9, 40 (1953).
21. K. G. J. Jacobi, Fundamenta Nova Theoriae Functionum Ellipticarum (Fratrum Borntraeger, Regiomonti, 1829).
22. L. Rosati, M. C. Nucci, and F. Mezzanotte, Edizione italiana e rilettura dei FUNDAMENTA NOVA di K. G. J. Jacobi a 150 anni dalla morte (18 Febbraio 1851 - 18
Febbraio 2001), Preprint RT n. 2001-4, Dipartimento di Matematica e Informatica,
Universit`
a di Perugia (2001).
23. E. N. Lorenz, J. Atmos. Sci. 20, 130 (1963).
24. Lord Rayleigh, Phil. Mag. 32, 529 (1916).
25. B. Saltzman, J. Atmos. Sci. 19, 329 (1962).
26. A. D. Dalmedico, Arch. His. Exact Sciences 55, 395 (2001).
27. H. Segur, Solitons and the Inverse Scattering Transform, Lect. Int. School Phys. Enrico Fermi, Varenna, Italy, 7-19 July (1980) Published in: Topics in Ocean-Physics,
edited by A. R. Osborne and P. Malanotte Rizzoli (North-Holland, Amsterdam, 1982),
p. 235.
28. M. C. Nucci, Mathl. Comput. Modelling 25, 181 (1997).
29. M. C. Nucci, J. Math. Phys. 44, 4107 (2003).
30. T. Sen and M. Tabor, Physica D 44, 313 (1990).
31. G. M. Murphy, Ordinary Differential Equations and Their Solutions (Van Nostrand,
Princeton, 1960).
32. S. Lie, Vorlesungen u
ber Differentialgleichungen mit Bekannten Infinitesimalen Transformationen (Teubner, Leipzig, 1912).
33. M.C. Nucci, Using Lie symmetries in epidemiology, In Conference on Diff. Eqns. and
Appl. in Math. Biology, Nanaimo, BC, Canada. Electron. J. Diff. Eqns., Conference
12 (2004) p. 87.
34. W. R. Derrick and P. van den Driessche, J. Math. Bio. 31, 495 (1993).
35. M. Rellini, Risoluzione di un modello deterministico non lineare che descrive
lepidemiologia dellinfluenza, Mathematics thesis supervised by M. C. Nucci (Uni-
April 1, 2015
1:18
Maria*Nucci
27
versity of Perugia, Italy, 2000).

36. L. Euler, Methodus inveniendi lineas curvas maximi minimive proprietate gaudentes,
sive solutio problematis isoperimetrici latissimo sensu accepti (Bousquet, Lausanne et
Geneva, 1744).
37. J. L. Lagrange, Miscellanea physico-mathematica societatis Taurinensia 2, 407 (1760).
38. H. Helmholtz, J. Reine Angew. Math. 100, 137 (1887).
39. V. Volterra, Rend. Acc. Lincei ser.IV III, 274 (1887).
40. N. Ya. Sonin, Warsaw Univ. Izvestiya 1-2, 1 (1886).
41. K. G. J. Jacobi, J. Reine Angew. Math. 29, 213 and 333 (1845).
42. K. G. J. Jacobi, Vorlesungen u
ber Dynamik. Nebst f
unf hinterlassenen Abhandlungen
desselben herausgegeben von A. Clebsch (Druck und Verlag von Georg Reimer, Berlin,
1884).
43. K. G. J. Jacobi, Jacobis Lectures on Dynamics tr. by K. Balagangadharan, ed. by
Biswarup Banerjee, Industan Book Agency, (American Mathematical Society, Providence, 2009).
44. V. Volterra, Proposta di una Associazione Italiana per il Progresso delle Scienze, in
Atti del Congresso dei Naturalisti Italiani, Milano, 1906, (Zanichelli, Bologna 1920)
p. 146.
45. M. C. Nucci and A. M. Arthurs, Proc. R. Soc. A 466, 2309 (2010).
46. M. E. Fels, Trans. Amer. Math. Soc. 348, 5007 (1996).
47. M. Jur
as, Acta Applicandae Mathematicae 66, 25 (2001).
48. S. Lie, Forhandlinger I Videnskabs-Selskabet I Christiania, 255 (1874).
49. E. Tonti, Inverse Problem: Its General Solution, in Differential Geometry, Calculus
of Variations and Their Applications T. M. Rassias e G. M. Rassias editor (Marcel
Dekker, New York, 1985) p. 497.
50. E. Tonti, Bull. Acad. Roy. Belg. 55, 137 and 262 (1969).
51. F. Magri, Int. J. Engng. Sci. 12, 537 (1974).
52. W. M. Tulczyjew, The Euler-Lagrange resolution, In Lecture Notes in Mathematics
No. 836 (Springer-Verlag, New York, 1980) p. 22.
53. V. Sarlet, J. Phys. A: Math. Gen. 15, 1503 (1982).
54. A. M. Vinogradov, J. Math. Anal. Appl. 100, 1 (1984).
55. I. M. Anderson and G. Thompson, The Inverse Problem of the Calculus of Variations for Ordinary Differential Equations (American Mathematical Society, Providence, 1992).
56. M.C. Nucci and K.M. Tamizhmani, J. Nonlinear Math. Phys. 19, 1250021 (2012).
57. H. Bateman, Phys. Rev. 38, 815 (1931).
58. M. C. Nucci and P. G. L. Leach, Phys. Scripta 83, 035007 (2011).
59. J. Douglas, Trans. Amer. Math. Soc. 50, 71 (1941).
60. S. L. Trubatch and A. Franco, J. Theor. Biol. 48, 299 (1974).
61. G. H. Paine, Bull. Math. Biol. 44, 749 (1982).
62. V. Volterra, Human Biology 11, 173 (1939).
63. S. Lie, Vorlesungen u
ber Differentialgleichungen mit bekannten infinitesimalen Transformationen (Teubner, Leipzig, 1912).
64. L. Bianchi, Lezioni sulla teoria dei gruppi continui finiti di trasformazioni (Enrico
Spoerri, Pisa, 1918).
65. E. T. Whittaker, A Treatise on the Analytical Dynamics of Particles and Rigid Bodies
(Cambridge University Press, Cambridge, 1999).
66. M.C. Nucci and P. G. L. Leach, J. Math. Phys. 48, 123510 (2007).
67. M. C. Nucci and P. G. L. Leach, Il Nuovo Cimento B 123, 93 (2008).
68. E. Noether, Nachr. d. K
onig. Gesellsch. d. Wiss. zu G
ottingen Math-phys. Klasse, 235
April 1, 2015
28
1:18
Maria*Nucci
M.C. Nucci
(1918).
69. V. Volterra V, Acta Biotheoretica 3, 1 (1937).
70. D. Schuch and M. Moshinsky, Phys. Rev. A 73, 062111 (2006).
71. M. C. Nucci, P. G. L. Leach, and K. Andriopoulos, J. Math. Anal. Appl. 319, 357
(2006).
72. H. Goldstein, Classical Mechanics, 2nd edn. (Addison-Wesley, Reading, 1980).
73. M. C. Nucci and P. G. L. Leach, J. Math. Anal. Appl. 406, 219 (2013).
74. D. Bjorken and S. D. Drell, Relativistic Quantum Mechanics (McGraw-Hill Book Co.,
New York, 1964).
75. W. H. Louisell, Quantum Statistical Properties of Radiation (John Wiley & Sons, New
York, 1990).
76. H. Weyl, Zeitschrift f
ur Physik 46, 1 (1927).
77. M. C. Nucci, Theor. Math. Phys. 168, 997 (2011).
78. M. C. Nucci, J. Phys.: Conf. Ser. 380, 012008 (2012).
79. M. C. Nucci, J. Nonlinear Math. Phys. 20, 451 (2013).
80. M. C. Nucci, Math. Notes Miskolc 14, 461 (2013).
81. M. C. Nucci, J. Phys.: Conf. Ser. 482, 012032 (2014).
82. G. Sierra, J. Phys. A: Math. Theor. 45, 055209 (2012).
83. G. Gubbiotti and M. C. Nucci, J. Nonlinear Math. Phys. 21, 248 (2014).
84. G. Gubbiotti and M. C. Nucci, Quantization of quadratic Lienard-type equations by
preserving symmetries, to appear in J. Math. Anal. Appl. (2015).
85. V. Chithiika Ruby, M. Senthilvelan, and M. Lakshmanan, J. Phys. A: Math. Theor.
45, 382002 (2012).
86. A. G. Choudhury and P. Guha, J. Phys. A: Math. Theor. 46, 165202 (2013).
87. M. C. Nucci and P. G. L. Leach, J. Math. Phys. 49, 073517 (2008).
88. M. C. Nucci and K. M. Tamizhmani, Il Nuovo Cimento B 125, 255 (2010).
89. M. C. Nucci and K. M. Tamizhmani, J. Nonlinear Math. Phys. 17, 167 (2010).
90. C. G. Darwin, Proc. R. Soc. Lond. A 117, 258 (1927).
91. A. Gonz
alez-L
opez, J. Math. Phys. 29, 1097 (1988).
92. G. W. Bluman and J. D. Cole, J. Math. Mech. 18, 1025 (1969).
93. W. I. Fushchych, W. M. Shtelen and M. I. Serov, Symmetry Analysis and Exact
Solutions of Equations of Nonlinear Mathematical Physics (Kluwer, Dordrecht, 1993).
94. R. Cherniha, J. Math. Anal. Appl. 326, 783 (2007).
95. R. O. Popovych, J. Phys. A: Math. Theor. 41,185202 (2008).
96. B. Kruglikov, Acta Appl. Math. 101, 145 (2008).
97. N. N. Yanenko, Compatibility theory and integration methods for systems of nonlinear
partial differential equations, in Proc. Fourth All-Union Math. Congr. in Leningrad
1961, Vol. II (in Russian) (Nauka, Leningrad, 1964) p. 247.
98. A. F. Sidorov, V. P. Shapeev and N. N. Yanenko, The method of differential constraints
and its applications in gas dynamics, (in Russian) (Nauka, Novosibirsk, 1984).
99. V. G. Dulov, S. P. Novikov, L. V. Ovsyannikov, B. L. Rozhdestvenskii, A. A.
Samarskii, Yu. I. Shokin, Russ. Math. Surv. 39, 99 (1984).
100. S. V. Meleshko, Methods for Constructing Exact Solutions of Partial Differential
Equations: Mathematical and Analytical Techniques with Applications to Engineering
(Springer-Verlag, New York, 2005).
101. V. A. Galaktionov, Diff. Int. Eqns. 3, 863 (1990).
102. J. R. King, Physica D 64, 35 (1993).
103. M. C. Nucci, Physica D 78, 124 (1994).
104. G. Guthrie, Constructing Miura transformations using symmetry groups, Research
Report No. 85 (1993).
April 1, 2015
1:18
Maria*Nucci
105.
106.
107.
108.
109.
110.
111.
112.
29
E. M. Vorobev, Differ. Eqs. 25, 322 (1989).

P. J. Olver, Proc. R. Soc. Lond. A 444, 509 (1994).
A. S. Fokas and Q. M. Liu, Theor. Math. Phys. 99, 263 (1994).
R. Z. Zhdanov, J. Phys. A: Math. Gen. 28, 3841 (1995).
M. C. Nucci, J. Phys. A: Math. Gen. 29, 8117 (1996).
J. Goard, Appl. Math. Lett. 16, 481 (2003).
M. C. Nucci, J. Math. Anal. Appl. 279, 168 (2003).
H. Berestycki, O. Diekmann, C. J. Nagelkerke, P. A. Zegeling, Bull. Math. Biol. 71,
399 (2009).
113. M. S. Hashemi and M. C. Nucci, J. Nonlinear Math. Phys. 20, 44 (2013).
114. M. C. Nucci, J. Math. An. Appl. 178, 294 (1993).
115. F. Allassia and M. C. Nucci, J. Math. An. Appl. 201, 911 (1996).
116. S. Martini, N. Ciccoli and M. C. Nucci, J. Nonlinear Math. Phys. 16, 77 (2009).
April 1, 2015
1:18
Asghar*Qadir


Higher Dimensional Systems of Differential Equations

Obtainable by Iterative Use of Complex Methods
Asghar Qadir
School of Natural Sciences,
National University of Sciences and Technology
H-12, 44000, Islamabad, Pakistan.
aqadirmath@yahoo.com
Fazal M Mahomed
DST-NRF Centre of Excellence in Mathematical and Statistical Sciences; School of
Computational and Applied Mathematics, University of the Witwatersrand, Johannesburg, Wits
2050, South Africa and School of Mathematics and Statistics, University of New South Wales,
Sydney NSW 2052 Australia
fazal.mahomed@wits.edu.sa

A procedure had been developed to solve systems of two ordinary and partial differential
equations (ODEs and PDEs) that could be obtained from scalar complex ODEs by splitting into their real and imaginary parts. The procedure was extended to four dimensional
systems obtainable by splitting complex systems of two ODEs into their real and imaginary parts. As it stood, this procedure could be extended to any even dimension but
not to odd dimensional systems. In this paper, the complex splitting is used iteratively
to obtain three and four dimensional systems of ODEs and four dimensional systems
of PDEs for four functions of two and four variables that correspond to a scalar base
equation. We also provide characterization criteria for such systems to correspond to the
base equation and a clear procedure to construct the base equation. The new systems of
four ODEs are distinct from the class obtained by the single split of a two dimensional
system. The previous complex methods split each infinitesimal symmetry generator into
a pair of operators such that the entire set of operators do not form a Lie algebra. The
iterative procedure sheds some light on the emergence of these Lie-like operators. In
this procedure the higher dimensional system may not have any or the required symmetry for being directly solvable by symmetry and other methods although the base
equation can have sufficient symmetry properties. Illustrative examples are provided.
Keywords: complex symmetries; linearization
1. Introduction
Lie 1 developed his study of the symmetry of differential equations for complex
functions. Though analyticity was assumed, it was not used in the sense of treating
Principal
corresponding author; FM is visiting professorial fellow at UNSW for 2014

1
April 1, 2015
1:18
Asghar*Qadir
Asghar Qadir & Fazal M Mahomed
the complex function as the real and imaginary parts connected to each other by the
Cauchy-Riemann (CR) conditions. Thus, while the complex nature of the function
was very important for the topological properties of the Lie groups that arose, it
was not used directly for the differential equations. More recently 2 the splitting of
a scalar ODE, into its real and imaginary parts, was exploited to obtain methods to
deal with systems of ODEs and PDEs. This was called complex symmetry analysis
(CSA). Of particular interest was its application to dealing with the variational
principle for systems of ODEs 3,4 and to linearization (conversion of the equation to
linear form by transformation of the dependent and independent variables) of ODEs
5,6,7
. The latter was of special interest because of the connection between geometry
and the symmetries of systems of ODEs 8,9,10,11 . This enabled one to use geometric
methods to linearize, and hence solve, systems (including the scalar case) of ODEs
12,13
. This geometric method allowed one to not only write down the linearizing
transformation but also to directly provide the solution. The procedure also led
to some new insights regarding standard systems of equations and to methods for
solving systems that were not amenable to solution by standard symmetry analysis
4,14
.
The essential idea of the CSA approach is to check if a system of equations
corresponds to a complex scalar ODE that can be solved by linearization or other
symmetry methods and do so if it can. One would like to extend the original CSA to
deal with higher dimensional systems. In fact, it has been extended to 4-dimensional
systems 7 by looking for correspondences of the system to a complex 2-dimensional
system that is solvable. That procedure can only be used for even dimensional
systems. Thus odd dimensional systems seemed to be beyond the reach of CSA. In
this paper we use iterative splitting of the scalar ODE to obtain 4 and 3 dimensional
systems of ODEs and systems of 4 PDEs of 4 functions of 4 and 2 variables. While
the odd dimensions could have been obtained by projecting down from the even
dimensions, the subsequent procedure of identification would become unwieldy (if
not impossible).
One might have expected that a system of four ODEs or PDEs would be
amenable to solving by quaternion symmetry analysis. This turns out to be impossible. The reason is that while algebra works for quaternions, the usual calculus,
which relies on the Leibnitz rule, does not. Since we could not find this fact in the
literature, we present a short proof. Consider q = w + ix + jy + kz, subject to the
usual quaternion rules that i2 = j 2 = k 2 = 1 , ij = k = ji , jk = i = kj , ki =
j = ik. For dq/dq = 1 the derivative operator is
d
=a
+ ib
+ jc
+ kd
,
dq
w
x
y
z
(1)
subject to the condition that a b c d = 1. Now ask that dq 2 /dq = 2q. This
requires the above condition along with the additional conditions, a b = a c =
a d = 1, which are inconsistent with the earlier requirement. One has to obtain
the four dimensional system by other means, such as using a complex system of two
April 1, 2015
1:18
Asghar*Qadir
Higher Dimensional Systems of Differential EquationsObtainable by Iterative Use of Complex Methods
ODEs and splitting it into a system of four ODEs.

To obtain the system of ODEs we start with a scalar ODE and regard the
dependent variable as a complex function of a real variable, as was done for (CSA),
and split it into a system of two ODEs. Now regard the two dependent variables
as themselves complex functions of a real variable. This provides a system of four
ODEs. Of course, if both steps had been merged into one, only a system of two
equations would again have been obtained. (This would be the reduced system
obtainable from the new system of four equations.) Instead, we first close our eyes
to the fact that we are going to treat the two dependent variables as complex, and
only after obtaining the system of two equations do we treat each of the dependent
variables as themselves complex. The result of repeating the CSA procedure is
different from treating the dependent variable of the original scalar equation as a
function of two complex functions of a real variable as the symmetry structure of
the systems is different.
In the CSA procedure it was found that one may be able to linearize the base
scalar equation even though the corresponding real system does not have enough
symmetries to allow linearization, or even to permit of solution by usual symmetry
methods. The system of four ODEs could then have too few symmetries to be able
to obtain a solution by symmetry analysis while the scalar equation is linearizable
14
. It is to be expected that with iterative splitting we will not only get the same
sort of cases, but that there would be some even stranger cases arising than for
simple CSA.
One has another option. For the system obtained by the first split, one can treat
one of the two dependent variables as real and the other as complex. This provides
the system of three ODEs. In fact, it provides two systems of three ODEs, as we can
choose either of the dependent variables to be real and the other complex. The two
systems obtained are dual to each other in some sense. That sense will be clarified
by some examples that we will provide.
Of course, one could take the CSA system of PDEs and treat each of the dependent variables as complex functions of the two real variables. This would yield
a system of four PDEs for four functions of two variables. Alternatively, we could
have started with the system of two ODEs and now treated both the dependent and
independent variables as complex. This split also gives a system of four PDEs for
four functions of two variables. The two systems obtained are dual to each other in a
fairly obvious way. Splitting the system of two PDEs by treating the dependent and
independent variables as complex gives a system of four PDEs for four functions of
four variables.
The role of the CR-equations was fairly obvious in the original CSA and had
not been spelled out explicitly. A geometric description of these equations was given
later 14 but the detailed requirements for a general scalar ODE were not. Since the
whole idea is to be able to identify the systems of ODEs that correspond to a scalar
(complex) ODE, clear criteria should have been provided. This had not been done
even for the original CSA. Here we state the explicit criteria as theorems for the
April 1, 2015
1:18
Asghar*Qadir
original CSA. On double use of the splitting procedure the conditions become more
thoroughly coupled and need to be spelled out explicitly. This has been done in the
form of further theorems here.
The original CSA, when applied to the infinitesimal symmetry generators yielded
pairs of operators that do not form a Lie algebra. These operators were called Lielike operators 7 . This odd behaviour could have been anticipated. The Lie algebra
of the free-particle (complex) scalar ODE consists of 8 (complex) symmetry generators, satisfying the Lie algebra sl(3, CI ), which has 8 complex infinitesimal symmetry generators, yielding 16 real operators on splitting. However, the corresponding
free-particle system of two ODEs satisfies an sl(4, IR ) Lie algebra, which has 15
generators. Nor can one expect the 16 to be the 15 for the system with one surplus,
as the complex generators come in conjugate pairs. What happened there was that
we lost two generators (as Lie-like) and gained one. In this paper we follow the
complex splitting of the symmetry generators to see what, and how many, Lie-like
operators arise. These are not symmetries of the system and as such the higher
dimensional system may not have any symmetry but be integrable.
We have limited the discussion to second order equations only. One can, of
course, go to higher order equations but that complicates the expressions without
providing any further understanding of the procedures being developed. Also, there
is no direct equivalent of the geometric connection and procedures for the higher
order equations. (There is an indirect connection provided by differentiating the
second order ODE and requiring that the original equation hold 15,16 , but that will
not be followed up here.) On the other hand, one could have limited oneself to first
order equations but that is the degenerate case and results for it will not hold more
generally.
The plan of the paper is as follows. In the next section we present the basics of the
CSA splitting procedure and briefly mention symmetries of differential equations.
We also present two characterization theorems that had not been provided earlier.
In the subsequent section we give the split into a system of three ODEs. In section
four we give the split into a system of four ODEs. The section after that deals
with the split into a system of four PDEs for four functions of two variables. In
section six we present the split into a system of four PDEs for four functions of four
variables. For each of the systems we provide some examples in the same sections.
In the concluding section we give a brief summary and discussion of the results.
One might wonder why one should bother with this new procedure to get the
four dimensional system by complex methods when we could have done it directly.
Of course, the same objection could have been raised to using the complex splitting
when one could have directly get the solution by considering a four-dimensional
system without complex splitting or even the complex splitting of a scalar equation
into a two-dimensional system. The point is that new features appear, as is evidenced by the appearance (and disappearance) of Lie-like operators. This signals
the possibility arising of finding solutions by Lie point transformations by the new
methods, which could not have been found by the more direct methods. We will not.
April 1, 2015
1:18
Asghar*Qadir
here, explore this matter further, leaving it for subsequent investigation. Further,
as mentioned earlier, by the double-split we access the odd-dimensional systems
as well. Another motivation for the double-split was to face squarely the Lie-like
operators so as to obtain a better understanding of their significance. At present we
have only been able to see how these new features arise and have not obtained the
desired insights. It is hoped that as the behaviour is seen more fully a pattern will
emerge. This, also, is left for later investigation.
2. Complex Splitting and Symmetries
Consider a general second order ODE
u00 (r) = f (r; u, u0 ) .
(2)
We can now take either u to be a complex function of the real variable r or take
both u and r to be complex. Let u = p + q in the former case and put
f (r; u, u0 ) = f r (r; p, q; p0 , q 0 ) + f i (r; p, q; p0 , q 0 ) .
(3)
The resulting system of ODEs is

p00 (r) = f r (r; p, q; p0 , q 0 ) , q 00 (r) = f i (r; p, q; p0 , q 0 ) .
(4)
The CR-equations for this system are

fpr = fqi , fqr = fpi ; fpr0 = fqi0 ; fqr0 = fpi0 ;
(5)
with no conditions on p and q other than second differentiability.

For the latter case let r = s + t as well. Then
f (r; u, u0 ) = f r (s, t; p, q; ps , qs , pt , qt ) + f i (s, t; p, q; ps , qs , pt , qt ) .
(6)
Splitting the equation into its real and imaginary parts then gives the system of
two PDEs
pss ptt + 2qst = 4f r , qss qtt 2pst = 4f i .
(7)
The CR-equations for this system include conditions for p and q, apart from the
previous ones (which are now more complicated),
ps = qt , pt = qs ;
fsr = fti , ftr = fsi ;
fpr = fqi , fqr = fpi .
(8)
The derivative of the functions with respect to the derivatives is more complicated
to state. The problem is that
2u0 (
)(p + q) = (ps + qt ) (pt qs ) .

s
t
(9)
April 1, 2015
1:18
Asghar*Qadir
Thus, for f (r; u, u0 ) to be analytic, f r and f i cannot depend arbitrarily on ps ,qs ,pt
and qt but must depend on ps + qt and pt qs . If we call these variables and ,
respectively, then the last conditions are
fr = fi , fr = fi .
(10)
Though the CR-equations were taken as obvious, they are needed to characterize
systems of (real) ODEs and PDEs that can arise by splitting a scalar ODE by CSA
methods. To complete the CSA procedure we state the following two theorems. If
the conditions for these theorems are satisfied for a system, we can construct the
corresponding scalar ODE and check if it can be solved by symmetry analysis. If
so, we have a solution of the system.
Theorem 1: A system of two second order ODEs (4) corresponds to a scalar
second order ODE (2) if and only if it satisfies the system of CR-conditions (5).
Theorem 2: A system of two second order PDEs (7) corresponds to a scalar
second order ODE (2) if and only if it satisfies the system of CR-conditions (8) and
(10) where = ps + qt and = pt qs .
Symmetry analysis uses the infinitesimal generators that leave the differential
equation invariant under point transformations (r, u) (R, U ) say,
+ (r, u)
.
(11)
r
u
To be able to apply the operators to differential equations these generators have to
be prolonged or extended to include the higher derivatives. For the first derivative
X = (r, u)
X[1] =
+
+ [1] (r; u, u0 ) 0 ,
r
u
u
(12)
where
[n] =
d [n1]
d
un (r)
,
dr
dr
(13)
( [0] = ), d/dr stands for the total derivative,

d
=
+ u0
+ u00 0
dr
r
u
u
(14)
and un (r) stands for the nth derivative. The prolonged generators for higher order
differential equations can be similarly obtained using (13).
We can extend the analysis to systems of ODEs by replacing the scalar u by
a vector u and the corresponding partial derivative by u . Consequently we must
replace the scalar and [1] by the vectors and [1] . The extension to the PDEs is
more complicated. The scalar r has to now also be replaced by a vector s and along
with it the scalar by the vector , but now the derivative of u becomes s u and
the derivative with respect to this vector of partial derivatives becomes too messy to
read. As such, we write s u = u1 . Of course, we need to also bear in mind that the
variables that the functions depend on will not be the derivatives given but linear
combinations as we saw for the CR-conditions. The second derivative can then be
April 1, 2015
1:18
Asghar*Qadir
written as u2 and so on. For the complex case the X was written as Z and similar
changes were made for the coefficients but it will be more convenient to use the
same notation throughout here.
The set of all (prolonged) symmetry generators for a (system of) differential
equation(s) form a Lie algebra. However, on splitting the operators there is no
guarantee that they will form a Lie algebra for the split systems. Consequently, we
have to drop some of these Lie-like generators and introduce others to close the
algebra.
3. System of Three ODEs by Double Splitting
Consider (2) as the base scalar equation with the split of the function given by (3)
leading to the system (4) subject to the CR-equations (5). Now regard p as a real
variable x and q as the complex variable y + z. We run into a problem here. There
are three second order ODEs but the number of functions to be obtained from f
must be even. To avoid this problem we put
f (r; u, u0 ) = g(r; p, p0 ) + G(r; p, q; p0 , q 0 ) .
(15)
The choice of what part to put into g and what part to put into G is clearly arbitrary.
For definiteness, we choose g to consist of all those terms that do not involve q or
q 0 . Then G consists of all those terms that do. Now we proceed with the second
split by putting
G(r; p, q; p0 , q 0 ) = k(r; x, y, z; x0 , y 0 , z 0 ) + l(r; x, y, z; x0 , y 0 , z 0 ) ,
(16)
yielding the system of three ODEs

x00 = h(r; x, x0 ) , y 00 = k(r; x, x0 ) , z 00 = l(r; x, x0 ) ,
(17)
where x = (x, y, z), subject to the CR-equations

ky = lz , kz = ly ; ky0 = lz0 , kz0 = ly0 .
(18)
The first prolonged symmetry generator is:
+ x (r, x)
+ y (r, x)
+ z (r, x)
r
x
y
z
+ x[1] (r; x, x0 ) 0 + y[1] (r; x, x0 ) 0 + z[1] (r; x, x0 ) 0 .

x
y
z
X[1] = (r, x)
(19)
Instead of this procedure, at the second split we could have taken p(r) = x(r) +
y(r) and set q(r) = z(r). This will give a dual system in some sense. We would
now have to set
f (r; u, u0 ) = g(r; x, y, z; x0 , y 0 , z 0 ) + h(r; x, y, z; x0 , y 0 , z 0 ) + k(r; z, z 0 ) ,
(20)
and the slightly modified CR-equations

gy = hz , gz = hy ; gy0 = hz0 , gz0 = hy0 .
(21)
April 1, 2015
1:18
Asghar*Qadir
The prolonged symmetry generator remains unaltered in form except that now the
coefficient of the last term is pure imaginary, to account for the last imaginary term
in (20). The sense of the duality will be clarified by the examples.
For completeness we state a theorem for the characterization of systems of three
ODEs that correspond to a scalar ODE by double splitting.
Theorem 3: The system of three ODEs (17) corresponds to the scalar ODE (2),
for any consistent identification of the function given by (15) and (16), provided the
CR-conditions (18) hold.
Remark: The dual procedure gives the same system but now we require that the
split given by (20) and (21) holds.
Example 1: Consider the free-particle scalar ODE, u00 = 0. It clearly yields
the system x00 = y 00 = z 00 = 0. The splitting of the functions is obviously trivial.
The system must clearly have 24 generators, as the algebra is sl(5, IR ). However,
the double splitting gives only 23 operators of which 15 are symmetries and 8 are
Lie-like 2 ; five of the nine dilations coming from the dependent variable and the
four local projective symmetries are missing. We have to use the actual symmetries
obtained here and require closure of the algebra to generate the full 24 Lie operators.
The dual system also gives the same symmetry structure, as expected. However, due
to the fact that the k in (20) has a coefficient with iota, even though k itself is zero,
the operator for the symmetry carries an imaginary. If we do not put in the iota for
the operator, we lose some of the symmetry generators.
Example 2: Consider the system of generalized Emden-Fowler system of ODEs:
x00 = 2s5 yz ;
y 00 = 2s5 zx ;
z 00 = 2s5 xy ;
(22)
admitting the symmetry algebra generated by
+ 3x
+ 3y
+ 3z
;
s
x
y
z
Y2 = s2
+ sx
+ sy
+ sz
.
s
x
y
z
Y1 = s
(23)
This corresponds to a double splitting of the generalized Emden-Fowler scalar equation 17

u00 = s5 u2 ,
(24)
x2 + y 2 = z 2 ,
(25)
subject to the algebraic constraint
which has the symmetry generators

X1 = s
+ 3u
; X2 = s2
+ su
s
u
s
u
(26)
April 1, 2015
1:18
Asghar*Qadir
and is completely integrable. Splitting these symmetry generators yields 8 Lie-like

operators and no symmetries of the system. These are
3
3
3
3
3
Y1 = ss + xx + yy + zz , Y2 = zy yz ,
2
4
4
4
4
3
3
3
3
Y3 = zx xz , Y4 = yx xy ,
2
4
2
4
1
1
1
1
1
2
Y5 = s s + sxx + syy + szz , Y6 = szy syz ,
2
4
4
4
4
1
1
1
1
Y7 = szx sxz , Y8 = syx sxy .
2
4
2
4
The Lie-lie operators come from X1 and X2 by two splits; the first by taking
u p + iq and then p ix and q y + iz just as one obtains the system of ODEs
(22) from (24) above.
Our system (22) is completely integrable although it has few symmetries. We
have also seen this to be the case in the standard CSA 14 . The first example of an
integrable system of two ODEs devoid of symmetry was given in 18 . Here we have an
integrable system of three ODEs with with only two symmetries. The integrability of
(24) follows from that of (22). Indeed we have that (22) is integrable by quadratures
using the standard Lie approach as is well-known for this system 17 . One has the
first integral of the equation (22) given by
I=
1 2 02 1 2 1 3 3
s u + u s u suu0
2
2
3
which has the symmetry X2 20 . An invariant function corresponding to X2 is v = u/s

and invoking this in the integral I = C1 reduces it to quadrature as
r
2
2 0
s v = 2C1 + v 3 .
3
The solution to the system (24) is traced backwards from v, u, p, q and then x, y, z.
Example 3: Consider the scalar Emden-Fowler equation 19
u00 + 5s1 u0 + u2 = 0 .
(27)
It splits into the following system of ODEs:

x00 + 5s1 x0 2yz = 0 ;
y 00 + 5s1 y 0 2zx = 0 ;
z 00 + 5s1 z 0 + 2xy = 0 ;
(28)
subject to the further algebraic constraint

x2 + y 2 = z 2 .
(29)
The Emden-Fowler equation has only the one scaling symmetry

X=s
2u
.
s
u
(30)
April 1, 2015
10
1:18
Asghar*Qadir
The split system has 4 Lie-like operators, none of which are symmetries of the
system. However, the system admits the scaling
Y=s
2x
2y
2z
.
s
x
y
z
(31)
In this case the scalar base Emden-Fowler equation is completely integrable by

the first integral approach given in 20 and therefore the iteratively split system
considered is also integrable although it possesses just one symmetry.
In this case the Lie-like operators are
1
1
1
1
Y1 = ss xx yy zz , Y2 = zy + yz ,
2
2
2
2
1
1
xz zx , Y4 = xy + yx ,
2
2
deduced in the same manner as in the previous example. Here the first integral of
the ODE is
1
1
I = s6 u02 + 2s5 uu0 + s6 u3 ,
2
3
in exactly the same way as the above gives the necessary reductions for our system
(28). One has an invariant v = us2 of the operator X which upon substitution in
I = C1 gives the quadrature
r
2
0
sv = 2C1 4v 2 v 3 .
3
Y3 =
The solution to our system is traced backwards as before.

The Lie-like operators arise if there exists one or more complex symmetry/ies of
the base complex system. We have criteria in Theorem 3 that enable us to identify
which system of three ODEs comes from a complex ODE. Moreover, this procedure
also enables one to construct the Lie-like operators from the complex symmetry
generator of the base ODE. However, the Lie-like operators are not symmetries
of the split or iteratively split system. We have seen this in the above examples.
There is a proposition in 21 which gives criteria when these are symmetries of a split
system. This occurs, in general, if the symmetries are real or pure imaginary. The
same result applies if we have an iteratively split system as the symmetries need to
be inherited each time from the complex to the iteratively split system which arises
from the base complex ODE.
4. System of Four ODEs by Double Splitting
For the four dimensional system, after the first split of (2) given by (3), (4) and (5)
we could set p(r) = w(r) + x(r) and q(r) = y(r) + z(r) and
f r (r; p, q; p0 , q 0 ) = g(r; w, w0 ) + h(r; w, w0 ) ,
f i (r; p, q; p0 , q 0 ) = k(r; w, w0 ) + l(r; w, w0 ) ,
(32)
April 1, 2015
1:18
Asghar*Qadir
yielding the system of four ODEs

w00 (r) = g(r; w, w0 ) , x00 (r) = h(r; w, w0 ) ,
y 00 (r) = k(r; w, w0 ) z 00 (r) = l(r; w, w0 ) ,
(33)
subject to the CR-conditions

gw + hx = ky + lz , gx hw = kz ly ,
gy + hz = kw lx , gz hy = kx + lw ,
gw0 + hx0 = ky0 + lz0 , gx0 hw0 = kz0 ly0 ,
gy0 + hz0 = kw0 lx0 , gz0 hy0 = kx0 + lw0 ,
(34)
where w = (w, x, y, z). The prolonged symmetry generator can now be written as
+ (r, w).w + [1] (r; w, w0 ).w0 .

(35)
r
Writing this equation out in detail makes it too unwieldy to convey much wisdom.
There is no dual system to this as the splitting is direct. We could have obtained
a system of four ODEs by a three stage splitting, setting one of the dependent
variables in the systems of three ODEs as a complex pair. For each of the three
dimensional systems obtained, one splitting would give the above system and one
would give a new system. We are, here, limiting our discussion to a two-step splitting
only.
The characterization theorem is:
Theorem 4: The system of four second order ODEs (33) corresponds to the scalar
ODE (2) by double complex splitting if and only if the CR-conditions (34) hold with
the splitting (32).
The examples will illustrate our procedure further.
Example 4: The corresponding free particle scalar equation obviously yields a 4d free-particle system of four equations, whose symmetry algebra must be sl(6, IR ),
which has 35 generators. However, there are a total of 18 symmetries and 8 Lie-like
operators. Again, the missing ones come from the dilations involving the dependent
variables and local projective symmetries. Again, the closure of the algebra starting
with the derived symmetries generates the full sl(6, IR ).
Example 5: The system of four ODEs
X = (r, w)
w00 = s5 (w2 x2 y 2 + z 2 ) ;
x00 = s5 (2wx 2yz) ;
y 00 = s5 (2wy 2xz) ;
z 00 = s5 (2wz + 2xy) ,
has the symmetry generators
+ 3w
+ 3x
+ 3y
+ 3z
;
Y1 = s
s
w
x
y
z
Y2 = s2
+ sw
+ sx
+ sy
+ sz
.
s
w
x
y
z
(36)
(37)
11
April 1, 2015
12
1:18
Asghar*Qadir
It corresponds to the scalar ODE (24), whose double splitting yields only 8 Lie-like
operators and no actual symmetries. Notice that in this case we cannot generate
the algebra by requiring closure, but have to resort to direct calculation.
This system has 8 Lie-like operators of which none are symmetries of the system.
However, the system of four ODEs does admit the two symmetries and is completely
integrable as the base generalized Emden-Fowler equation which has two symmetries
is.
Example 6: Again consider the scalar Emden-Fowler equation (27). It splits
into the following system of ODEs:
w00 + 5s1 w0 + w2 x2 y 2 + z 2 = 0 ;
x00 + 5s1 x0 + 2wx 2yz = 0 ;
y 00 + 5s1 y 0 + 2wy 2xz = 0 ;
z 00 + 5s1 z 0 + 2wz + 2xy = 0 .
(38)
The split system again has 4 Lie-like operators, none of which are symmetries of
the system. However, the system admits only the scaling
Y=s
2w
2x
2y
2z
s
w
x
y
z
(39)
and is integrable due to (27) being integrable.

5. System of Four PDEs for Four Functions of Two Variables by
Double Splitting
For the PDEs, we need to treat the dependent variable as complex. However, there
are two ways of doing so if we are to obtain functions of only two variables. We could
retain the independent variable as real in the first step and then make it complex in
the second step, or first treat it as complex and then retain the same independent
variables in the second step. We follow the former procedure first. Starting with
(2), with the first step splitting given by (3), (4) and (5), we put r = s + t and
proceed with treating the dependent variables as complex exactly as in the case for
the system of four ODEs.
The system of equations is now
wss wtt + 2xst = 4g(s; w; ws , wt ) , xss xtt 2wst = 4h(s; w; ws , wt ) ;
yss ytt + 2zst = 4k(s; w; ws , wt ) , zss ztt 2yst = 4l(s; w; ws , wt ) , (40)
subject to the CR-equations
ws = xt , wt = xs , ys = zt , yt = zs ;
gw = hx , gx = hw , gy = hz , gz = hy ;
kw = lx , kx = lw , ky = lz , kz = lw .
(41)
April 1, 2015
1:18
Asghar*Qadir
Now, as for (10), defining

= ws + xt , = wt sx , = ys + zt , = yt zs ,
(41)
the conditions for the derivatives with respect to the derivatives can be written as
g = h , g = h ; k = l , k = l .
(41)
Writing (s, t) = s, for the infinitesimal symmetry generator we will now have
two components for , namely ( s , t ) and will have four components as for the
system of four ODEs. The additional feature is that the prolonged derivatives will
be for ws and the coefficients will be [1] . Thus the prolonged generator can be
written as
X = (s, w).s + (s, w).w + [1] (s, w).s w ,
(41)
where [1] (s, w) is the generalization of [1] for the case of PDEs.
The characterization theorem here is:
Theorem 5: The system of four PDEs for four functions of two variables (40)
corresponds to the scalar ODE (2) by double complex splitting, if and only if the
CR-conditions (41) - (43) hold, provided that g, h, k, l depend on the derivatives only
in the combinations given by (43).
Instead, if we had put r = s + t in the first step and then proceeded to split
the dependent variables twice, we would have got the dual system
wss wtt + 2yst = 4g(s; w; ws , wt ) , xss xtt + 2zst = 4h(s; w; ws , wt ) ;
yss ytt 2wst = 4k(s; w; ws , wt ) , zss ztt 2xst = 4l(s; w; ws , wt ) . (41)
The CR-conditions are considerably more involved. The simple ones are
ws = yt , wt = ys , xs = zt , xt = zs ;
gw + hx = ky + lz , gx hw = kz ly ;
gy + hz = kw lx , gz hy = kx + lw .
(40)
For the derivatives with respect to derivatives we have to define the new variables
= ws + yt , = xs + zt , = wt ys , = xt zs
(40)
to get
g + h = k + l , g h = k l ,
g + h = k l , g h = k + l .
(40)
The form of the symmetry generator remains unchanged. Once again, we rely on
the examples to illustrate the procedure.
Here we need a separate theorem because the systems are apparently different,
though they are dual to each other in some sense.
Theorem 6: The (dual) system of four PDEs for four functions of two variables
(45) corresponds to the scalar ODE (2) by double complex splitting, if and only if the
13
April 1, 2015
14
1:18
Asghar*Qadir
CR-conditions (46) - (48) hold, provided that g, h, k, l only depend on the derivatives
in the combinations given by (48).
Example 7: The symmetry generators for the free particle equation split into
28 Lie-like operators of which 20 are symmetries. The system, of course, admits
infinitely many symmetries. The dual system is very similar. More precisely, it is
the system (45) with the right side set equal to zero. Since the CR-conditions are
trivial there is no significant difference between the original and the dual system.
Example 8: Again consider the scalar Emden-Fowler equation (27). It splits
into the following system of four PDEs of two independent variables:
s
t
(ws + xt ) + 10 2
(xs wt )
2
+t
s + t2
+4(w2 x2 y 2 + z 2 ) = 0 ;
s
t
xss xtt 2wst + 10 2
(xs wt ) 10 2
(ws + xt ) + 8(wx yz) = 0 (39)
;
2
s +t
s + t2
s
t
yss ytt + 2zst + 10 2
(ys + zt ) + 10 2
(zs yt ) + 8(wy xz) = 0 ;
2
s +t
s + t2
t
s
(zs yt ) 10 2
(ys + zt ) + 8(wz + xy) = 0(39)
.
zss ztt 2yst + 10 2
2
s +t
s + t2
wss wtt + 2xst + 10
s2
This split system again has 4 Lie-like operators, none of which are symmetries of
the system. However, it admits a scaling symmetry. The solution properties for the
system (49) and (50) can be deduced from the scalar base equation as before.
6. System of Four PDEs for Four Functions of Four Variables by

Double Splitting
This is the most straightforward (and the most complicated) of the various possibilities considered. At the first step we regard both the independent and the dependent
variables as given by (6) to (8). For the second step we run short of symbols for the
variables. As such, we now write the first variable (previously written as s), as the
complex variable s + t and the second variable (previously written as t), as the
complex variable u + v and write s for (s, t, u, v). Further, we put
p(s, t) w(s) + x(s) , q(s, t) y(s) + z(s) ,
(39)
f r (s, t; p, q; ps , qs , pt , qt ) = g(s; w, s w) + h(s; w, s w) ;

f i (s, t; p, q; ps , qs , pt , qt ) = k(s; w, s w) + l(s; w, s w) .
(39)
April 1, 2015
1:18
Asghar*Qadir
The system of equations is

wss wtt + 2xst wuu + wvv 2xuv + 2ysu 2ytv
+2zsv + 2ztv = 4g(s; w, s w);
xss xtt 2wst xuu + xvv + 2wuv + 2zsu 2ztv
2ysv 2ytv = 4h(s; w, s w);
yss ytt + 2zst yuu + yvv 2zuv + 2wsu 2wtv
+2xsv + 2xtv = 4k(s; w, s w);
zss ztt 2yst zuu + zvv + 2yuv + 2xsu 2xtv
2wsv 2wtv = 4l(s; w, s w);
(33)
subject to the CR-conditions

ws + xt = yu + zv , wt xs = yv zu ,
wu + xv = ys zt , wv xu = yv + zu ;
gs + ht = ku + lv , gt hs = kv lu ,
gu + hv = ks lt , gv hu = kt + ls ;
gw + hx = ky + lz , gx hw = kz ly ,
gy + hz = kw lx , gz hy = kx + lw .
(29)
The derivatives in the CR-conditions require the variables

= ws + xt + yu + zv , = wt xs + yv zu ;
= wu + xv ys zt , = wv xu yt + zs .
(29)
Then the rest of the CR-conditions are

g h = k l , g + h = k + l ;
g h = k + l , g + h = k l .
(29)
The characterization theorem in this case is:

Theorem 7: The system of four second order PDEs for four functions of four variables (53) corresponds to the scalar second order ODE (2) by double complex splitting
provided the functions g, h, k, l depend on the derivatives only in the combinations
given by (56) and the CR-conditions (54) and (56) hold.
The prolonged symmetry generator for the system is
X[1] = (s, g).s + (s, g).g + [1] (s, g, s g).s g .
(30)
The derivatives with respect to derivatives are to be taken bearing in mind the discussion for the CR-equations. However, even if we ignore it in taking the derivatives,
no error will ensue.
We again rely on examples to illustrate our systems.
Example 9: The free-particle system of equations is given by (53), with the
right side set equal to zero. The CR-conditions are trivial. There are now 32 Lie-like
15
April 1, 2015
16
1:18
Asghar*Qadir
operators, of which only 24 are symmetry generators. As before the local projective
symmetries are lost. Here there are 8 such. However, the dilations are not lost
here. The system, itself, has an infinite number of symmetry generators. The 24
symmetries do form a Lie algebra, say A24 , which can serve as a core for the
L
system of PDEs, in that one could write the full Lie algebra A = A A24 and
throw away the infinite dimensional algebra A to be left with a well-defined
finite dimensional Lie algebra that could be used to generate a solution with 24
arbitrary constants.
Example 10: Consider the double splitting of the Emden-Fowler equation (27)
into a system of four PDEs for four functions of four variables. It is given by (53)
with
g = C[(sA + tB) + (tA sB) + (uA + vB) + (vA uB)]
w2 + x2 + y 2 z 2 ;
h = C[(sA + tB) (tA sB) + (uA + vB) (vA uB)]
2wx + 2yz ;
k = C[(sA + tB) + (tA sB) (uA + vB) (vA uB)]
2wy + 2xz ;
l = C[(sA + tB) + (tA sB) + (uA + vB) (vA uB)]
2wz 2xy ;
(24)
where
C=
A2
5
, A = s2 t2 + u2 v 2 , B = 2st + 2uv .
+ B2
(24)
There are 4 Lie-like operators none of which are symmetries. The system, like the
scalar equation, has a scaling symmetry. It is integrable as (27) is.
7. Summary and Further Lines of Investigation
In this paper we have considered systems of three and four second order ODEs,
and systems of four second order PDEs for four functions of two or four variables,
that correspond to a scalar equation (which we call a base equation) by a specific
procedure (which we call double complex splitting). We have also provided characterization criteria for such systems to correspond to the base equation and a
clear procedure to be able to construct the base equation. Thus, in principle, we
could write a computer code that could take any such system given and check if it
corresponds to a base equation. It could then construct the base equation.
What is the advantage of having such base equations and constructing them?
The point is that it is much easier to deal with the base equation than with the
system. Thus, for example, if the base system has two infinitesimal symmetry generators it could be solved by symmetry methods. In fact it could have eight symmetry
generators and thus be linearizable. In that case we could write down the solution
April 1, 2015
1:18
Asghar*Qadir
directly. Following the double splitting procedure by which the system corresponds,
we could then write down the solution for the system of ODEs or PDEs. Note that
in this procedure the system need not have the required symmetry for being directly
solvable by symmetry methods.
The procedure adopted for CSA could only give an even dimensional system as
it simply split n (complex) equations into 2n (real) equations. Also, the number
of independent variables in the PDE equals the number of dependent variables. In
the complex double split procedure we are being considerably more adventurous.
Having obtained the system of two real equations we conveniently forget that
they arose from a scalar complex equation, treat it as complex and promptly split
the equations again. Now we have the earlier freedom of choosing the independent
variable to be either real or complex, while treating the dependent variables to be
complex, but we have the additional freedom to choose to treat one of the dependent
variables as real and the other as complex. This provides the possibility of obtaining
an odd dimensional system. Further, to obtain the PDEs, we could choose to treat
the independent variables as real first and then complex, complex first and then real
or complex both times. Thus we also get the system of four PDEs for four functions
of two variables.
In the cases of full double complex splitting, where either the dependent variables
were fully split or both the dependent and independent variables were double split,
giving the system of four ODEs or four PDEs for four functions of four variables,
there were no complications of additional dual systems arising. However, in the
case of the system of three ODEs or the system of four PDEs for four functions of
two variables, we got dual systems arising. The duality in the former case was very
obvious but in the latter it was considerably more involved on account of the CRconditions. Even in the case of the system of three ODEs the symmetry generators
had to carry an iota to make sense. The duality of these systems needs to be
better understood. Note that in (39) is simply s .w. It would be interesting to
find out what the operators for the other variables are. Presumably, they would
be dual divergence operators in some sense. This may shed some light on the
structure of the double split systems.
For the system of three ODEs obtained by double splitting, it would be of interest
to consider the ambiguity due to the choice of the function g in (15), or k in (20)
for the dual system. In some sense all choices must be equivalent. The question
is whether one gets an equivalence class. Further, would they be equivalent under
point transformations or possibly some more general transformations like contact
or higher order transformations 22 .
The algebraic constraint that arises in the system of three ODEs was not apparent in setting up the system but was found in the examples. It is interesting to
note that it geometrically amounts to the solution lying on a cone. This seems to be
generic for the three dimensional system. It also shows that it can be written as a
system of two ODEs. However, that system is much more complicated. It would be
of interest to understand the complex projective geometry that leads to the solution
17
April 1, 2015
18
1:18
Asghar*Qadir
lying on a cone.
In the original CSA the splitting procedure applied to the infinitesimal symmetry
generators yielded pairs of Lie-like operators. While it is understandable that such
operators arise, it is not so clear how they do so. There one never lost many Lie
generators and replaced by Lie-like ones. However, the second iteration can manage
to get rid of all the Lie operators and produce a number of Lie-like ones. It is worth
considering the number of generators and the number of operators that arise. For
the maximum symmetry case for a scalar equation we saw that the procedure causes
a loss of one generator. However, for a system of four ODEs the maximal algebra is
sl(4, IR ), which has 35 generators, while 4 8 = 32. Thus one has gained 3 in some
sense. As one goes to higher dimensions with repeated splitting, there would be
ever greater gains. While this paper has shed some light on the Lie-like operators,
a much clearer understanding is needed.
One would have hoped that for the system of PDEs corresponding to a base ODE
one could use the symmetries of the ODE to obtain a core set of symmetries for
the system of PDEs. This could provide a core general solution of the system of
PDEs as occurred in the case of the free-particle equation for the four functions of
four variables (Example 9) . Even if the system has infinitely many symmetries, the
base equation can only have a finite number. However, the examples show that we
can lose all the Lie symmetry generators and be left only with Lie-like ones. For
the Emden-Fowler equation, we are left with no Lie symmetries from the Lie-like
operators, though the equation has a scaling symmetry and so does the double-split
system. In general, we obtain Lie-like operators and not Lie-symmetry generators
that would form an algebra. The Lie-like operators somehow encode the symmetries
of the base equation. It would be most important to learn how they do so. It may
be that the CR-conditions will enable us to re-construct the Lie from the Lie-like
symmetry.
It is of interest to note that not only for the PDEs but also for the systems of
ODEs, we get Lie-like operators arising and lose some Lie-symmetry generators. It
would be worth while to see this encoding of symmetry as distinct from the PDE
case.
We hope that in future the use of these systems for the variational principle
and with linearization will be followed up. It should lead to interesting and useful
results.
References
1. Lie S 1967 Differential Equations, (Chelsea, New York).
2. Ali S, Mahomed FM and Qadir A (2009) Complex Lie symmetries for differential
equations J. Nonlinear Analysis B: Real World Applications, 10, 3335 - 3344.
3. Ali S, Mahomed FM and Qadir A (2008) Complex Lie symmetries for variational
problems J. Nonlinear Mathematical Physics, 15, 124-133.
4. Farooq U, Ali S and Qadir A (2010) Invariants of two-dimensional systems via complex Lagrangians with applications Commun. Nonlinear Science and Numerical Sim-
April 1, 2015
1:18
Asghar*Qadir
ulations, 16 1804-1810.
5. Ali S, Mahomed FM and Qadir A (2011) Linearizability criteria for systems of two
second-order differential equations by complex methods Nonlinear Dynamics, 66, 77
- 88.
6. Ali S, Qadir A and Safdar M (2011) Inequivalence of classes of linearizable systems of
second order ordinary differential equations obtained by real and complex symmetry
analysis Mathematical and Computational Applications, 16, 923-934.
7. Safdar M, Mahomed FM and Ali S (2011) Linearization of systems of four second-order
ordinary differential equations Pramana, 77, 581 - 594.
8. Aminova AV and Aminov NA-M (2000) Projective geometry of systems of differential
equations: general conceptions Tensor N S, 62, 65-86.
9. Aminova AV, Aminov NA-M (2006) Projective geometry of systems of second-order
differential equations Sbornik: Math, 197, Number 7, 951955.
10. Feroze T, Mahomed FM and Qadir A (2006) The connection between isometries and
geodesic equations Nonlinear Dynamics, 45, 65-74.
11. Qadir A (2007) Geometric linearization of ordinary differential equations Symmetry,
Integrability and Geometry: Methods and Applications (SIGMA), 3, 103-109.
12. Mahomed FM and Qadir A (2007) Linearization criteria for a system of second order
quadratically semi-linear ordinary differential equations Nonlinear Dynamics, 48, 417422.
13. Mahomed FM and Qadir A (2009) Invariant linearization criteria for systems of cubically semi-linear second order ordinary differential equations J. Nonlinear Mathematical Physics, 16, 283-298.
14. Safdar M, Ali S and Qadir A (2011) On the symmetry solutions of two-dimensional
systems not solvable by standard symmetry analysis, arXiv:1104.3837v2 [math.CA].
15. Mahomed FM and Qadir A (2008) Conditional linearizability criteria for third order
ordinary differential equations J. Nonlinear Mathematical Physics, 15, 25-35.
16. Mahomed FM and Qadir A (2009) Conditional linearizability of fourth-order semilinear ordinary differential equations J. Nonlinear Mathematical Physics, 16, 165-178.
17. Leach PGL, Maartens R and Maharaj SD (1992) Self-similar solutions of the generalized Emden-Fowler equation Int. J. Non-Linear Mech., 27, 583-592.
18. Gonzalez-Gascon F and Gonzalez-Lopez A (1988) Newtonian Systems of Differential
Equations, Integrable via Quadratures, with Trivial Group of Point Symmetry Phys.
Lett. A, 129(3), 153-156.
19. Rosenau P (1984) A note on the integration of the Emden-Fowler equation Int. J.
Non-Linear Mech., 19, 303-308.
20. Mahomed KS and Momoniat E (2014) Symmetry classification of first integrals for
scalar dynamical equations Int. J. Non-Linear Mech. 59, 5259.
21. Naz R and Mahomed FM (2014) Lie and Noether symmetries of systems of complex
ordinary differential equations and their split systems Pramana J. of Phys., to appear.
22. Ibragimov NH (1985) Transformation Groups Applied to Mathematical Physics, (D.
Reidel Publishing Company, Dordrecht).
19
April 1, 2015
1:19
M.*Tsamparlis


Geometrization of Lie and Noether symmetries and applications
Michael Tsamparlis
Faculty of Physics, Department of Astrophysics - Astronomy - Mechanics, University of Athens,
Panepistemiopolis, Athens 157 83, GREECE
mtsampa@phys.uoa.gr
We derive the Lie and the Noether conditions for the equations of motion of a dynamical
system in an ndimensional Riemannian space. We solve these conditions in the sense
that we express the symmetry generating vectors in terms of the collineation vectors
of the space. More specifically we give two theorems which contain all the necessary
conditions which allow one to determine the Lie and the Noether point symmetries of a
specific dynamical system evolving in a given Riemannian space in terms of the projective
and the homothetic collineations of space. We apply these theorems to various interesting
situations in Newtonian Physics.
1. Introduction
In theoretical Physics one has two main tools to study the properties of evolution of
dynamical systems: (a) Symmetries of the equations of motion; and (b) Collineations
(symmetries) of the background space where evolution takes place. It is well known
that both these tools have the following common characteristics:
a. They form a Lie algebra.
b. They do not fix uniquely either the dynamical system or the space.
The natural question to be to asked is if these two algebras are related and if so
in what way. Equivalently, one may state the question as follows:
To what degree and how does space modulate the evolution of dynamical systems
in it? Is a dynamical system free to evolve at will in a given space or is it constrained
to do so by the very symmetry structure of space?
This question was answered many years ago by the Theory of Relativity with
the Equivalence Principle, that is, the requirement that free motion in a given
gravitational field occurs along the geodesics of space. However as obvious as this
point of view may appear to be it is not easy to comprehend and accept! So let us
take a second more mathematical view of the situationa .
In a Riemannian space the affinely parameterized geodesics are determined
uniquely by the metric. The geodesics are a set of homogeneous ordinary differential
a This
point has been raised in an illuminating discussion with Prof P G Leach a few years back
in Athens while we were driving to the Poseidon Temple in cape Sounion.
1
April 1, 2015
1:19
M.*Tsamparlis
Michael Tsamparlis
equations (ODE) linear in the highest order term and quadratically non-linear in
the first order terms. A system of such ODEs is characterized (not fully) by its Lie
point symmetries. On the other hand a metric is characterized (again not fully)
by its collineations. Therefore it is reasonable to expect that the Lie point symmetries of the system of geodesic equations of a metric will be closely related to the
collineations of the metric. That such a relation exists is easy to see by the following
simple example. Consider a family of straight lines parallel to the xaxis on the
Euclidian plane. These curves can be considered either as the integral curves of the
d2 y
2
2
ODE dx
2 = 0 or as the geodesics of the Euclidian metric dx + dy . Subsequently
consider a symmetry operation defined by a reshuffling of these lines without preserving necessarily their parametrization. According to the first interpretation this
d2 y
symmetry operation is a Lie symmetry of the ODE dx
2 = 0 and according to the
second interpretation it is a (special) projective symmetry of the Euclidian two
dimensional space.
What has been said for a Riemannian space can be generalized to a space in
which there is only a linear connection. In this case the geodesics are called autoparallels (or paths) and they comprise again a system of ODEs linear in the highest
order term and quadratically non-linear in the first order terms. In this case one
is looking for relations between the Lie symmetries of the autoparallels and the
projective collineations of the connection.
A Lie point symmetry of an ordinary differential equation (ODE) is a point
transformation in the space of variables which preserves the set of solutions of the
ODE 1,2,3 . Regarding these solutions as curves in the space of variables, then we
may equivalently consider a Lie point symmetry as a point transformation which
preserves the set of the solution curves. Applying this observation to the geodesic
curves in a Riemannian (affine) space, we infer that the Lie point symmetries of the
geodesic equations in any Riemannian space are the automorphisms which preserve
the set of these curves. However we know from Differential Geometry that the
point transformations of a Riemannian (affine) space which preserve the set of
geodesics are the projective transformations. Therefore it is reasonable to expect a
correspondence between the Lie point symmetries of the geodesic equations and the
projective algebra of the metric defining the geodesics.
The equation of geodesics in an arbitrary coordinate frame is a second order
ODE of the form
x
i + ijk x j x k + F (xi , x j ) = 0,
(1)
where F (xi , x j ) is an arbitrary function of its argument and the functions ijk are the
connection coefficients of the space. The dot indicates differentiation with respect
to the geodesic affine parameter. Equivalently equation (1) is also the equation of
motion of a dynamical system moving in a Riemannian (affine) space under the
action of a velocity dependent force. According to the above argument we expect
that the Lie point symmetries of the ODE (1) for a given function F (xi , x j ) will be
related with the collineations of the metric. In the following we assume that F is a
April 1, 2015
1:19
M.*Tsamparlis
function of xi only. As it will be shown in this case the Lie symmetries of (1) are a
subalgebra of the special projective algebra of the space. The specific subalgebra is
selected by means of certain constraint conditions involving geometric quantities of
the space and the function F (xi , x j ) 4,5,6,7,8 .
The determination of the Lie point symmetries of a given system of ODEs consists of two steps: (a) the determination of the conditions which the components of
the Lie symmetry vector must satisfy; and (b) the solution of the system of these
conditions. Step (a) is formal and is outlined in e.g. 1,2,3 . The second step is the
key one and, for example, in higher dimensions where one has a large number of
simultaneous equations the solution can be quite involved and perhaps prohibitive.
However if one expresses the system of Lie symmetry conditions of (1) in terms of
collineation (i.e. symmetry) conditions of the metric, then the determination of Lie
point symmetries is transferred to the geometric problem of determining the special
projective group of the metric. In this field there is a significant amount of work
from Differential Geometry waiting to be used. Indeed the projective symmetries
are already known for many spaces or they can be determined by existing general
theorems. For example the projective algebra and all its subalgebras are known for
the important case of spaces of constant curvature 12 and in particular for the flat
spaces. This implies that, for example, the Lie symmetries of all Newtonian dynamical systems as well as those of Special Relativity can be determined by simple
differentiation from the known projective algebra of these spaces!
What has been said for the Lie point symmetries of (1) applies also to Noether
point symmetries (provided (1) follows from a Lagrangian). The Noether point
symmetries are Lie point symmetries which satisfy the additional constraint
X [1] L + L
df
d
= .
dt
dt
(2)
The Noether point symmetries form a closed subalgebra of the Lie point symmetries
algebra. In accordance to the above this implies that the Noether point symmetries
will be related with a subalgebra of the special projection algebra of the space where
motion occurs. As it will be shown this subalgebra is the homothetic algebra of
the space. It is well known that to each Noether point symmetry it is associated
a conserved current (i.e. a Noether first integral). This leads us to the important
conclusion that the (standard) conserved quantities of a dynamical system depend
on the space in which it moves and the type of force F (xi ) which modulates the
motion. In particular in free fall, that is when F (xi , ) = 0 the orbits are affinely
parametrized geodesics and the geometry of the space is the sole factor which determines the conserved quantities of motion. This conclusion is by no means trivial and
means that the space where motion occurs is not a pathetic carrier of motion but it
is the major modulator of the evolution of a dynamical system. In other words there
is a strong and deep relation between Geometry of the space and Physics (motion)
in that space!
In what follows we discuss the aforementioned ideas in a systematic way. More
April 1, 2015
1:19
M.*Tsamparlis
Michael Tsamparlis
specifically we do the following

a. We derive the general expression for the Lie symmetry conditions of (1) assuming a general (i.e. not necessarily symmetric) connection in terms of geometric
quantities;
b. For the case that the function F depends only on the coordinates, i.e. F (xi )
we give two theorems which establish the exact relation between the (special) projective/homothetic algebra of the space and the Lie/Noether symmetry algebra of
(1) respectively;
c. We consider applications in Newtonian Physics;
d. We use the above general results to generalize the Ermakov- Pinney dynamical
system to n dimensional Riemannian space.
2. Collineations of Riemannian spaces
For convenience we give a brief account of the basic definitions concerning
collineations in a Riemannian space. A collineation in a Riemannian space is a
vector field X which satisfies an equation of the form
LX A = B
(3)
where LX denotes Lie derivative, A is a geometric object (not necessarily a tensor)

defined in terms of the metric and its derivatives (e.g. connection, Ricci tensor,
curvature tensor etc.) and B is an arbitrary tensor with the same tensor indices as
A. The collineations in a Riemannian space have been classified by Katzin et al. 10 .
In the following we use only certain collineations.
A conformal Killing vector (CKV) is defined by the relation

LX gij = 2 xk gij .
(4)
If = 0, X is called a Killing vector (KV), if is a non-vanishing constant a
homothetic vector (HV) and if ;ij = 0, a special conformal Killing vector (SCKV).
A CKV is called proper if it is not a KV, HV or a SCKV.
A projective collineation (PC) is defined by the equation
i
LX ijk = 2(,j k)
.
(5)
If = 0 the PC is called an affine collineation (AC) and if ;ij = 0 a special

projective collineation (SPC). A proper PC is a PC which is not an AC, HV or KV
or SPC. The PCs form a Lie algebra whose ACs, HV and KVs form subalgebras. It
has been shown that if a metric admits a SCKV then also admits a SPC, a gradient
HV and a gradient KV 11 . We summarize the above in Table 1.
In the following we shall need the symmetry algebra of spaces of constant curvature. In 12 it has been shown that the PCs of a space of constant non-vanishing
curvature consist of proper PCs and KVs only and if the space is flat then the algebra of the PCs consists of KVs/HV/ACs and SPCs. For convenience we summarize
these results in Table 2.
April 1, 2015
1:19
M.*Tsamparlis
Collineation
Killing vector (KV)
Homothetic vector (HV)
Conformal Killing vector (CKV)
Affine Collineation (AC)
Projective collineation (PC)
A
gij
gij
gij
ijk
ijk
ijk
Special Projective collineation (SPC)
Collineation
Killing vectors (KV)
Homothetic vector (HV)
Affine Collineation (AC)
Special Projective collineation (SPC)
B
0
gij , ,i = 0
gij , ,i 6= 0
0
i
2(,j k)
, ,i 6= 0
i
2(,j k)
, ,i 6= 0 and,jk = 0
Gradient
SI = Ii i
H = xi i
AIJ = xJ Ii i
Non-gradient
j i
j] xj i
XIJ = [I
PI = SI H.
The Lie point symmetries of all Newtonian dynamical systems are amongst the
vectors in Table 2. Also the Noether point symmetries of all Newtonian dynamical systems (and more generally all systems moving in a flat space) follow from
the elements of the first two rows of Table 2 (apart form some differences in sign
depending on the signature of the metric).
3. The Lie point point symmetry conditions in an affine space

We consider the system of ODEs:
x
i + ijk x j x k +
n
X
Pji1 ...jm x j1 . . . x jm = 0
(6)
m=0
where ijk are the connection coefficients of the space and Pji1 ...jm (t, xi ) are smooth
polynomials completely symmetric in the lower indices and derive the Lie point
symmetry conditions in geometric form using the standard approach. Equation (6)
is quite general and covers most of the standard autonomous and non autonomous
dynamical systems and in particular equation (1). Furthermore because the ijk s
are not assumed to be symmetric, the results are valid in a space with torsion.
Obviously they hold in a Riemannian space where the connection coefficients are
given in terms of the Christofell symbols.
The detailed calculation can be found in 8 and shall not be repeated here. In
the following we summarize the results.
April 1, 2015
1:19
M.*Tsamparlis
Michael Tsamparlis
The terms x j1 . . . x jm for m 4 give the equations:

L P i + 2,t P i + P i ,t + i ,tt + j ,t P.ji = 0
(7)
L Pji
,t Pji
Pji ,t
,k ji
2,j ki
P + 2
,t|j ,tt ki
+ 2
i
,t P.jk
=0
(8)

i
i
i
d
i
i
i
+ ,(k |d|
P.j)
+ P.kj,t
+ 3 d ,t P.dkj
L Pjk
+ L ijk + ,d (k
2,t(j k)
=0
(9)
i
i
i
e
i
i
i
=0
+ P.jkd,t
+ 4 e ,t P.jkde
(,j|k d)
P.dj)
L P.jkd
,t P.jkd
+ ,e (k
(10)
and the conditions due to the terms x j1 . . . x jm for m > 4 are given by the following
general formula:
L Pji1 ...jm + Pji1 ...jm ,t + (2 m) ,t Pji1 ...jm +
j
+ ,r (2 (m 1)) Pji1 ...jm1 jrm + (m + 1) Pji1 ...jm+1 ,tm+1 + ,j Pjj1 ...jm1 jim = 0.
(11)
Note the appearance of the term L ijk in these expressions.
Eqn (1) is obtained for m = 0, P i = F i in which case the Lie symmetry
conditions read:
L P i + 2,t P i + P i ,t + i ,tt = 0

,k ji + 2,j ki P k + 2 i ,t|j ,tt ki = 0
L ijk
i
2,t(j k)
=0
i
(,j|k d)
= 0.
(12)
(13)
(14)
(15)
If F i = 0 we obtain the Lie symmetry conditions for the geodesic equations (see
and 9 ) .
4. The autonomous dynamical system moving in a Riemannian

space
We solve the Lie symmetry conditions (12) - (15) for an autonomous dynamical
system in the sense that we express them in terms of the collineations of the metric
of the space where motion occurs.
Equation (15) means that ,j is a gradient KV of gij . This implies that the metric
gij is decomposable. Equation (14) means that i is a projective collineation of the
metric with projective function ,t . The remaining two equations are constraint
conditions, which relate the components , ni of the Lie symmetry vector with the
vector F i (xj ). Equation (12) gives

i
L g ij Fj + g ij L Fj + 2,t g ij Fj + ,tt
= 0.
(16)
April 1, 2015
1:19
M.*Tsamparlis
This equation is an additional restriction for i because it relates it directly to the

symmetries of the metric. Finally equation (13) gives

i
k
ji ,tt + ,j ki + 2ji ,k F k + 2,tj
+ 2ijk ,t
= 0.
(17)
We conclude that the Lie symmetry equations is given by equations (16), (17)
where (t, x) is a gradient KV of the metric and i (t, x) is a special projective
collineation of the metric with projective function ,t . We state this result in theorem
4.1 below 8 .
Theorem 4.1. The Lie point symmetries of the system of equations of motion of
an autonomous system under the action of the force F j (xi ) in a general Riemannian
space with metric gij , namely
x
i + ijk x j x k = F i
(18)
are given in terms of the generators Y i of the special projective algebra of the metric
gij .
We consider next the case that the force F i is derivable from a potential V (xi ),
so that the equations of motion follow from the standard Lagrangian

1
L xj , x j = gij x i x j V xj
(19)
2
with Hamiltonian

1
E = gij x i x j + V xj
(20)
2
The Noether conditions are
V,k k + V ,t = f,t
(21)
i
,t
gij
(22)
,j V = f,j

1
L gij = 2
,t gij
2
,k = 0.
Thew last equation implies = (t) and reduces the system as follows:

1
L gij = 2
,t gij ,
2
V,k k + V ,t = f,t ,
i,t = f,i .
i
(23)
(24)
(25)
(26)
(27)
Equation (25) implies that is a conformal Killing vector of the metric provided
i
,t 6= 0. Because gij is independent of t and =
(t) thei must be is a HV of thei
i
i
j
metric. This means that (t, x) = T (t) Y x where Y is a HV. If ,t = 0 then
is a Killing vector of the metric. Equations (26), (27) are the constraint conditions,
which the Noether symmetry and the potential must satisfy for the former to be
admitted. The above lead to the following theorem 8 .
April 1, 2015
1:19
M.*Tsamparlis
Michael Tsamparlis
Theorem 4.2. The Noether point symmetries of the Lagrangian (19) are generated
from the homothetic algebra of the metric gij .
More specifically, concerning the Noether point symmetries, we have the following.
All autonomous systems admit the Noether symmetry t whose Noether integral
is the Hamiltonian E given in equation (20). For the rest of the Noether point
symmetries we have the following cases
Case I Noether point symmetries generated by the homothetic algebra.
The Noether symmetry vector and the Noether function G t, xk are

X = 2Y tt + Y i i , G t, xk = pt
(28)
where Y is the homothety factor of Y i (Y = 0 for a KV and 1 for the HV), and
p is a constant, provided the potential satisfies the condition
LY V + 2Y V + p = 0.
(29)
Case II Noether point symmetries generated by the gradient homothetic Lie

algebra i.e. both KVs and the HV must be gradient.
In this case the Noether symmetry vector and the Noether function are
Z
Z

X = 2Y
T (t) dtt + T (t) H i i , G t, xk = T,t H xk + p T dt,
(30)
where H i is the gradient HV or a gradient KV, the function T (t) is computed from
the relation T,tt = mT where m is a constant and the potential satisfies the
condition
LH V + 2Y V + mH + p = 0.
(31)
Concerning the Noether integrals we have the following result (not including the
Hamiltonian)
Corollary 4.1. The Noether integrals of Case I and Case II are respectively
ICI = 2Y tE gij Y i x j + pt,
Z
ICII = 2Y
T (t) dt E gij H ,i x j + T,t H + p
(32)
Z
T dt.
(33)
where E is the Hamiltonian given in (20).

We remark that theorems 4.1 and 4.2 do not apply to generalized or dynamical
symmetries.
Let us see a first application of the above general theorems.
April 1, 2015
1:19
M.*Tsamparlis
5. The Lie point symmetries of geodesic equations in an Einstein

space
Spaces of constant curvature are Einstein spaces whose curvature tensor is of the
R
(gac gbd gad gbc ) where R is a constant. A proper Einstein
form Rabcd = n(n1)
R
b
Rabcd 6= n(n1)
(gac gbd gad gbc ). The
space is a space in which Rab = R
n gab but
12
projective collineations of Einstein spaces can be found in .
Suppose X a is a projective collineation with projection function (xa ) so that
LX abc = ,b ca + ,c ba . For a proper Einstein space :
n(1 n)
;ab .
R
It follows that if X a generates either an affine or a special projective collineation
then ;ab = 0 hence X a reduces to a KV. This means that proper Einstein spaces do
not admit a homothetic vector, affine collineations or special projective collineations.
The above results and Theorem 4.1 lead to the following conclusion:
LX gab =
Theorem 5.1. The Lie point symmetries of the geodesic equations in a proper
Einstein space of curvature scalar R (6= 0) are given by the vectors
X = (Kt + L)t + Di (x)xi ,
(34)
where Di (x) is a KV of the metric and K, L are constants.

Theorem 5.1 extends and amends the conjecture of
of Einstein spaces.
13
to the more general case
6. Applications in Newtonian Physics

One important problem to consider is to find all 2-d and 3-d dimensional autonomous Newtonian dynamical systems, that is, all dynamical systems moving
in flat Euclidian space with equation of motion x
i = F i (xj ) where i, j = 1, 2 or
1, 2, 3 respectively which admit Lie and Noether point symmetries and of course to
determine the admitted symmetries. The answer to this question consists of finding
all the forces F i (xj ) which result in this property.
The importance of this problem is twofold. Indeed the knowledge of the Lie symmetries makes possible the computation of the invariants of the dynamical system;
also the knowledge of Noether symmetries gives the Noether integrals (conserved
currents) of the system. Both these data can be used to simplify the equations of
motion and even lead to their analytic solution.
The two dimensional case has been considered originally in 14 and the 3d case
15,16
in
. Both cases have been considered anew following the geometric approach of
the previous paragraphs. More specifically the 2-d case has been considered in 17
the contracted Bianchi identity (Rij 12 Rg ij );j = 0 it follows that in an Einstein space of
dimension n > 2 the curvature scalar R =constant.
b Using
April 1, 2015
10
1:19
M.*Tsamparlis
Michael Tsamparlis
Lie symmetry
d
2 tt +
d
2 tt + ()
d
2 tt
d
2 tt
d
2 tt
F (x , x , x )
edx f (x , x )

ed()f r(), x
x1d
f
+ RR
x1d
f (x
h , x )
+ x
+ x
d x
Lie symmetry
t
1
x3 f
1
t
m
e
t
m
x
x g (x , x )
+e
t m
i
+ f (x , x )
RR
Noether Symmetry
x x
2tt + RR
T (t)

R
2 T (t) dt t + T (t) RR
x1d
h (x , x )
1
x3 f
x1d
h
d x
1
x3 g
mx + f (x, x )
+
x1d
g (x , x )
x x
x , x
x x
x , x
F (x , x , x )
h (x, x )
x x
x , x
1
x3 h
g (x , x )

m
4 x
x x
x , x
m
4 x +
2
m
2 x px + f (x , x )
1
R2 f
x x
x , x
h (x , x )
1
x3 g
V(x, y, z) / T,tt = mT
px + f (x , x )

p() + f r() ,x

x x
1
1
,
f
(,
)
or
f
2
2
R
x
x x
2
m
8R +
x x
x , x
ed x h (x , x )
g (x , x )
F (x , x , x )
g (x, x )
x x
x , x
m
4 x
F (x , x , x )
edx h (x , x )

ed()h r(), x
x1d
g
F (x , x , x )
f (x, x )
t2 t + tRR
et
x x
x , x
F (x , x , x )
edx g (x , x )

ed()
g r(), x
2
(, ) or m
8R +
1
x2 f
x x
x , x
and the 3d case in 18 . In both cases it has been shown that the existing results were
incomplete.
It is to be noted that from the applications point of view the 2d case is important
because it applies to the mini super space which is particularly useful in cosmology.
Indeed in the standard cosmological model and many of its extensions the resulting
dynamical system reduces to a two dimensional dynamical system in a flat 2-d
Lorentz space. For convenience we collect the results of the 3-d case in Tables 3 and
4. Concerning the determination of all 3-d Newtonian dynamical systems, which
admit Noether point symmetries and subsequently the ones which are integrable
via Noether integrals we have the results of the Table 5 and Table 6.
In order to show the use of these tables we consider the dynamical system defined
by the equations of motion

1
m
x x
x
= x + 3 f,,
,
4
x
x x
This system is the well known Ermakov system which has a long history in the
literature 19,20,21,22,23 .
From Table 5 we find without
any calculations
that this system admits the Lie
point symmetry X = 1m et m t +et m R R .
1
x3 h
x x
x , x
April 1, 2015
1:19
M.*Tsamparlis
Noether Symmetry
a + b
a + b (x x )
a + b (x x )
a (x x ) +
+ b (x x )
2btt + a + bRR
2btt + a() () + bRR
V(x, y, z)

ap x + f
x ab x ,x
|b|x
p
|b|
arctan |(a+bx )| + f
1
2 r()
+ ab x , x
11

p
|b|
() + f r() , x ab ()

p
ax
+bx
+
2
2
a arctan
x a +b

2
x
1
a
2
2
+
x
+ a f x b x , x 1 ab + 2b
a x

px(2a+bx )
x
x
1
2 a+bx2 + a+bx2 f a+bx , a+bx
( )
(
)

x
1
a
f () b ln r() , r()
r2
()
In the following section we combine all the above results in order to generalize
the Ermakov dynamical system.
7. The Ermakov dynamical system
The Ermakov system has its roots in the study of the one dimensional time dependent harmonic oscillator
x
+ 2 (t)x = 0.
Ermakov
equation
24
(35)
obtained a first integral J of this equation by introducing the auxiliary

+ 2 (t) = 3
(36)
eliminating the 2 (t) term and multiplying with the integrating factor x x
:

1
J=
(x x)
2 + (x/)2 .
(37)
2
A short review of these studies and a detailed list of relevant references can be
found in 25 . Earlier reviews of the Ermakov system and its numerous applications
in divertive areas of Physics can be found in 26 , 27 .
The general Ermakov system does not admit Lie point symmetries. The form
of the most general Ermakov system which admits Lie point symmetries has been
determined in 28 and it is called the Kepler - Ermakov system 2930 . It is well known
that these Lie symmetries are a representation of the sl(2, R) algebra.
In an attempt to generalize the Kepler - Ermakov system to higher dimensions
Leach 30 used the sl(2, R) invariance of the two dimensional (Euclidean) Kepler Ermakov system and demanded that the generalized Kepler - Ermakov system will
posses two properties (a) A first integral, the Ermakov invariant; and (b) sl(2, R)
invariance. It has been shown that the invariance group of the Ermakov invariant
is reacher than sl(2, R) 31 . In this section we follow Leachs proposal and use
the theorems of the previous sections to generalize the Kepler - Ermakov system in
two directions: (a) to higher dimensions using the sl(2, R) invariance with respect to
April 1, 2015
12
1:19
M.*Tsamparlis
Michael Tsamparlis
Noether symmetries (provided the system is Hamiltonian); and (b) in a Riemannian

space which admits a gradient homothetic vector (HV).
8. The two-dimensional Kepler Ermakov system
The most general 2-d Newtonian Ermakov system is defined by the equations 28 :
1 y
x
+ 2 (t, x, y, x,
y)x
= 2f
,
(38)
yx
x

1
x
y + 2 (t, x, y, x,
y)y
= 2g
.
(39)
xy
y
This system admits the Ermakov first integral
Z y/x
Z y/x
1
f ( ) d +
g ( ) d.
2+
I = (xy y x)
2
By means of the transformation:

1
x
2 = 2 3 g
,
xy
y
y
y x2 x
F
,
=f
2g
x
x
y
y
(40)
equations (38) and (39) take the form:

x
+ 2 (x, y, x,
y)x
=
y
1
F
,
x2 y
x
y + 2 (x, y, x,
y)y
= 0.
(41)
(42)
Due to the second equation, except for special cases, the new function is
independent of t and depends only on the dynamical variables x, y and possibly on
their derivative. The Ermakov first integral in the new variables is:
Z y/x
1
2+
F ( ) d.
(43)
I = (xy y x)
2
The system of equations (41), (42) defines the most general 2-d Ermakov system
and produces all its known forms for special choices of the function . For example
the weak Kepler - Ermakov system 30 defined by the equations 29
x
1 y
x
+ 3 H (x, y) 3 f
=0
(44)
r
x
x

y
1
y
y + 3 H (x, y) 3 g
=0
(45)
r
y
x
where H, f, g are arbitrary functions of their argument, is obtained by the function
2 (x, y) = H (x, y) /r3
(46)
with Ermakov first integral

I=
1
2
(xy y x)
+
2
y
x

f () 3 g () d.
(47)
April 1, 2015
1:19
M.*Tsamparlis
13
The weak Kepler Ermakov system does not admit Lie point symmetries, however
the property of having a first integral prevails.
The system of equations (44) - (45) admits Lie point symmetries only for
h( y )
H (x, y) = 2 r3 + xx where C. This is the (standard) Kepler - Ermakov
system defined by the equations 19,32
1 y
1 y
x
2 x + 3 h
3f
= 0,
(48)
r
x
x
x
1 y
1 y y
3g
= 0.
(49)
y 2 y + 3 h
r x
x
y
x
The admitted Lie symmetry vectors are

1 2t
1 2t
X = c1 c2 e + c3 e
t + c2 e2t + c3 e2t RR
( 6= 0) (50)
and
X 1 = t , X 2 = 2tt + RR , X 3 = t2 t + tRR
( = 0)
(51)
where RR = xx + yy is the gradient HV of the 2-d Euclidian metric. Each set of

vectors in (50) and (51) is a representation of the sl(2, R) algebra. It is important
to note that these vectors are constructed from the vector t and the gradient HV
RR of the Euclidian two dimensional space E 2 .
Equations (48), (49) follow from the Lagrangian 32
L=
2
C ()
1 2
r + r2 2
r2
,
2
2
2r2
(52)
where C() = c + sec2 f (tan ) + csc2 g(tan ) provided the functions f, g satisfy
the constraint:
sin2 f 0 (tan ) + cos2 g 0 (tan ) = 0.
The Ermakov invariant in this case is
32
(53)
J = r4 2 + 2C () .
(54)
If a Kepler - Ermakov system is Hamiltonian then the Lie symmetries are also
Noether point symmetries with Noether integrals the Hamiltonian 33
1
1
1 2
r + r2 2 + 2 r2 + 2 F ()
(55)
E=
2
2
r
and two additional time dependent Noether integrals as follows:
=0
I1 = 2tE rr ,
(56)
1
I2 = t2 E trr + r2 ;
2
(57)
April 1, 2015
14
1:19
M.*Tsamparlis
Michael Tsamparlis
6= 0

1
E rr + r2 e2t , ,

1
I20 =
E + rr + r2 e2t .
I10 =
(58)
(59)
In a Hamiltonian Kepler - Ermakov system the Ermakov invariant (54) is constructed 19 from the Hamiltonian and Noether invariants (58),(59) as follows:
J = E 2 I10 I20 .
Finally in 19 it is shown that the Ermakov invariant is generated by a dynamical
Noether symmetry of the Lagrangian (52), a result which is also confirmed in 34 .
9. Generalizing the Kepler - Ermakov system

In the defining characteristics of the Kepler - Ermakov system we distinguish three
different properties of reduced generality: The property of having a first integral;
the property of admitting Lie/Noether symmetries with sl(2, R) invariance and
the property of being Hamiltonian and admitting sl(2, R) invariance via Noether
symmetries.
We first generalize the Kepler - Ermakov system to higher dimensions by the
requirement:
The generalized (Euclidian!) Kepler Ermakov system admits the sl(2, R) algebra
as a Lie symmetry algebra.
9.1. The 3-d Euclidian Kepler Ermakov system

Depending on the value 6= 0 or = 0 we consider the three dimensional Hamiltonian Kepler Ermakov systems of type I and type II.
9.1.1. The 3-d Hamiltonian Kepler Ermakov system of type I ( 6= 0)

For 6= 0 the admitted Noether symmetries are required to be (see (50)):
X 1 = t , X =
1 2t
e
t e2t RR .
(60)
From table 6, line 2 of 18 we find that for these vectors the potential is
2
V (R, , ) = 2 R2 + R12 f (, ) hence the Lagrangian is
L=

2 2
1
1 2
R + R2 2 + R2 sin2 2 +
R 2 f (, ) .
2
2
R
(61)
April 1, 2015
1:19
M.*Tsamparlis
15
The equations of motion, that is, the equations defining the generalized dynamical
system are
R 2 R sin2 2 2 R 2 f = 0,
R
R3
2
1
+ R sin cos 2 + 4 f, = 0,
R
R
2
1
+ R + cot + 4 2 f, = 0.
R
R sin
(62)
(63)
(64)
The Noether integrals corresponding to the Noether vectors are

2 2
1 2
1
R + R2 2 + R2 sin2 2
E=
R + 2 f (, ) ,
2
2
R
1
I+ = e2t E e2 t RR + e2t R2 ,
1
I = e2t E + e2t RR + e2t R2 ,
(65)
(66)
(67)
where E is the Hamiltonian. The Noether integrals I are time dependent. Following
19
we consider the time independent combined first integral
J = E 2 I+ I = R4 2 + R4 sin2 2 + 2f (, ) .
(68)
Using (68) the equation of motion (62) becomes

2 R = J
R
R3
(69)
which is the autonomous Ermakov- Pinney equation. Therefore J is the Ermakov

invariant 30 .
An alternative way to construct the Ermakov invariant (68) is to use dynamical
Noether symmetries 35 . Indeed one can show that the Lagrangian (61) admits the
dynamical Noether symmetry XD = Kji x j i where Kij is a Killing tensor of the
second rank whose non-vanishing components are K = R4 ,K = R4 sin2 . The
+
with gauge function
dynamical Noether symmetry vector is XD = R2
2f (, ).
9.1.2. The 3-d Hamiltonian Kepler Ermakov system of type II ( = 0)
For = 0, the Noether symmetries are required to be (see (51))
30
X 1 = t , X 2 = 2tt + RR , X 3 = t2 t + tRR .
(70)
From table 5, line 3 and table 6, line 2 of 18 we find that the potential is V (R, , ) =
1
R2 f (, ) hence the Lagrangian is

1 2
1
L0 =
R + R2 2 + R2 sin2 2 2 f (, ) .
(71)
2
R
April 1, 2015
16
1:19
M.*Tsamparlis
Michael Tsamparlis
The equations of motion are (62) - (64) with = 0. The Noether invariants of the
Lagrangian (71) are

1
1 2
E0 =
R + R2 2 + R2 sin2 2 + 2 f (, ) ,
(72)
2
R
I1 = 2tE 0 RR,
(73)
1
I2 = t2 E 0 tRR + R2 .
(74)
2
From the time dependent first integrals I1,2 we define the time independent first
integral J = 4I2 E 0 I12 which is calculated to be
J 0 = R4 2 + R4 sin2 2 + 2f (, ) .
(75)
J0
R3
= 0 i.e. the one

Using (75) the equation of motion for R (t) becomes R
0
dimensional Ermakov-Pinney equation, hence J is the Ermakov invariant 30 . As
was the case with the 3-d Hamiltonian Kepler - Ermakov system of type I, the

+

Lagrangian (71) admits the dynamical Noether symmetry XD = R2
whose integral is the J 0 .
The 3-d Hamiltonian Euclidian Kepler - Ermakov systems defined by the Lagrangians (61) and (71) need three independent first integral in involution in order
to be Liouville integrable. As we have shown, each system has the two Noether integrals (E, J) and (E 0 , J 0 ) respectively, hence we need one more Noether symmetry.
Such a symmetry exists only for special forms of the arbitrary function f (, ) .
These functions and other material relevant to the study of the integrability if the
3-d Hamiltonian Euclidian Kepler - Ermakov system are given in 36
10. The ndimensional Riemannian Kepler - Ermakov system
As it has been noted in section 8 that the Kepler - Ermakov system considered so
far is the Euclidean Kepler - Ermakov system. In this section we make a drastic
step forward and introduce the Riemannian Kepler Ermakov system of dimension
n. The generalization we consider is based on the following definition:
Definition 10.1. The ndimensional Riemannian Kepler - Ermakov system is an
autonomous dynamical system which:
a. It is defined on a Riemannian space which admits a gradient HV;
b. Admits a first integral, which we name the Riemannian Ermakov first integral and
it is characterized by the requirement that the corresponding equation of motion
takes the form of the Ermakov Pinney equation;
c. It is invariant at least under the sl(2, R) algebra, which is generated by the
gradient HV of the space.
There are two types of ndimensional Riemannian Kepler - Ermakov systems.
The ones which are not Hamiltonian and admit the sl(2, R) algebra as Lie symmetries and the ones which are Hamiltonian and admit the sl(2, R) algebra as Noether
symmetries.
April 1, 2015
1:19
M.*Tsamparlis
17
10.1. The non-Hamiltonian Riemannian Kepler - Ermakov system

Consider a ndimensional Riemannian space which admits a gradient HV. It is well
known that the metric of this space can always be written in the form 37,38
ds2 = du2 + u2 hAB dy A dy B ,
(76)
where the
Latin capital indices A, B, .. take the values 1, ..., n 1 and hAB =
hAB y C is the generic n 1 metric. The gradient HV of the metric is the vector
H i = uu ( = 1) generated from the function H = 21 u2 . We note the relation
hDA A
BC =
1
hDB,C .
2
(77)
The equations of motion Dx

Dt = F of a particle moving in that space under the
+ F A (u, y C ) yA are
action of the force F i = F u (u, y C ) u
u00 uhAB y 0A y 0B = F u ,
(78)
2
0B 0C
y 00A + u0 y 0A + A
= F A.
BC y y
u
(79)
Because the system is autonomous it admits the Lie point symmetry t . Using
the vector t and the gradient HV, H i = uu , we construct two representations of
the sl(2, R) by means of the sets of vectors (see (60),(70) )
t , 2tt + uu , t2 t + tuu
( = 0),
(80)
1 2t
e
t e2t uu
(2 6= 0),
(81)
t ,
and require that the vectors in each set be Lie symmetries of the system of equations
(78) and (79). It is found (see 36 ) that the requirement of the invariance of the force
under this group of symmetries demands that the force must be of the form

1
1
F i = 2 u + 3 Gu y C u + 4 GA y C A .
(82)
u
u
Replacing F i in the system of equations (78),(79) we find
u00 uhAB y 0A y 0B = 2 u +
y 00A +
1 u
G ,
u3
2 0 0A
1
0B 0C
u y + A
= 4 GA .
BC y y
u
u
(83)
(84)
Multiplying the second equation with 2u4 hDA y 0D and using (77) we have
u4

d
hDB y 0D y 0B + 4u3 hDA u0 y 0A y 0D = 2GD y 0D ,
ds
(85)
from which follows

d
u4 hDB y 0D y 0B = 2GD y 0D .
ds
(86)
April 1, 2015
18
1:19
M.*Tsamparlis
Michael Tsamparlis
The first integral is

J = u4 hDB y 0D y 0B + 2 y C .
(87)
A
We note that J involves the arbitrary metric hAB and the function (y ).
Finally equations (83) and (84) become

1
u00 uhAB y 0A y 0B = 2 u + 3 Gu y C ,
(88)
u

2
1
0B 0C
(89)
y 00A + u0 y 0A + A
= 4 hAB y C ,B .
BC y y
u
u
These are the equations defining the ndimensional Riemannian Ermakov - system.
Using the first integral (87) the equation of motion (88) is written as follows

yC
G
(90)
u00 = 2 u +
u3

u = J + Gu y C 2 y C . This is an equation of the Ermakov-Pinney
where G
form hence we identify (87) as the Riemannian Ermakov integral of the Riemannian
Kepler Ermakov system.
10.2. The Hamiltonian Riemannian Kepler - Ermakov system

In this section we assume that the force is derived from the potential V u, y C .
Then the equations of motion follow from the Lagrangian

1 02
L=
u + u2 hAB y 0A y 0B V u, y C ,
(91)
2
where u0 = du
ds and s is an affine parameter. The Hamiltonian is

1 02
u + u2 hAB y 0A y 0B + V u, y C .
(92)
E=
2
The equations of motion are
u00 uhAB y 0A y 0B + V,u = 0,
(93)
2
1
0B 0C
y 00A + u0 y 0A + A
+ 2 hAB V,B = 0.
(94)
BC y y
u
u
The demand that the Lagrangian (91) admits Noether point symmetries which
are generated from the gradient HV leads to the following cases (see theorem 4.2
and for details see 33 )
Case A: The Lagrangian (91) admits the Noether point symmetries (80) if the
potential is of the form

1
V u, y C = 2 V y C .
(95)
u
The Noether integrals of these Noether point symmetries are

1
1 02
EA =
u + u2 hAB y 0A y 0B + 2 V y C ,
(96)
2
u
I1 = 2tE uu0 ,
(97)
1
(98)
I2 = t2 E tuu0 + u2 ,
2
April 1, 2015
1:19
M.*Tsamparlis
19
where EA is the Hamiltonian.

Case B: The Lagrangian (91) admits the Noether point symmetries (81) if the
potential is of the form
V (u, y c ) =

2 2
1
u + 2 V 0 yC .
2
u
(99)
The Noether integrals of these Noether point symmetries are

2 2
1 02
1
u + u2 hAB y 0A y 0B
u + 2 V 0 yC ,
2
2
u
1 2t
2 t
0
2t 2
I+ = e E e
uu + e u ,
1
I = e2t E + e2t uu0 + e2t u2 ,
EB =
(100)
(101)
(102)
where EB is the Hamiltonian.

Using the Noether integrals we construct the Riemannian Ermakov invariant for
both Case A and Case B. The Riemannian Ermakov invariant is

JG = u4 hDB y 0D y 0C + 2V 0 y C .
(103)
This coincides with the invariant first integral defined in (87). We note that with
the use of the first integral (103) the Hamiltonian (100) takes the form
E=
J
1 02 2 2
u
u + 2,
2
2
2u
(104)
which is the Hamiltonian for the Ermakov Pinney equation.

As it was the case with the Euclidian case of section 8, it can be shown that the
Riemannian Ermakov invariant (103) is due to the existence of a dynamical Noether
symmetry35 .
11. Conclusion
The Lie and the Noether point symmetry vectors of the equations of motion of
an autonomous dynamical system involved in a general Riemannian space are determined completely by the geometry of the space. More specifically, Theorem 4.1
and Theorem 4.2 show that the Lie point symmetries are elements of the special
projective algebra of the space and the Noether point symmetries elements of the
homothetic algebra of the space. The selection of the particular vectors for a given
force modulating the motion of a particular system in this space is done by means
of certain compatibility conditions of a differential nature which can be easily managed. Using the above theorems we have determined all 2-d and 3-d Newtonian
dynamical systems which admit Lie and Noether point symmetries, we have derived in a straightforward manner the Lie point symmetries of the heat equation
and finally we have generalized the Kepler - Ermakov system in a Riemannian space
of dimension n.
April 1, 2015
20
1:19
M.*Tsamparlis
Michael Tsamparlis
References
1. Bluman W G and Kumei S 1989 Symmetries and Differential Equations (Springer
Verlag New York)
2. Olver P J 1986 Application of Lie groups to differential equations (Springer Graduate
texts in Mathematics, New York: Springer)
3. Stephani H 1989 Differential Equations: Their Solutions using Symmetry (Cambridge
University Press)
4. Prince G E and Crampin M (1984) Gen Rel Grav. 16, 921 942.
5. Aminova A V and Aminov N A 2006 Sbornic Mathematics 197 951
6. Aminova A V and Aminov N A 2010 Sbornic Mathematics 201 631
7. Tsamparlis M and Paliathanasis A (2010) Gen. Relativ. Gravit. 42 2957-2980;
(Preprint arXiv:1101.5769)
8. Tsamparlis M and Paliathanasis A (2011) Gen. Relativ. Gravit. 43 1861
9. Tsamparlis M, Paliathanasis A (2010), Nonlinear Dynamics 62, 203
10. Katzin G H Levine J and Davis R W 1969 J. Math. Phys. 10 617
11. Hall G S and Roy I M 1997 Gen. Relativ. Gravit. 29 827
12. Barnes A. 1993, Class. Quantum Grav. 10, 1139 - 1145
13. Feroze T Mahomed F M and Qadir A (2006) Nonlinear Dynamics 45 65
14. Sen T 1987 Phys. Lett. A 122 327
15. Damianou P A and Sophocleous C (1999) J. Math. Phys. 40 210
16. Damianou P A and Sophocleous C (2004) Nonlinear Dynamics 36 3
17. Tsamparlis M. and Paliathanasis A (2011) J. Phys. A: Math. and Theor. 44 175202;
(Preprint arxiv:1205.4114)
18. Tsamparlis M and Paliathanasis A (2012) J. Phys. A: Math and Theor. 275201;
(Preprint arXiv:1111.0810)
19. Moyo S and Leach P G L 2002 J. Phys. A: Math and Gen 35 5333
20. Haas F and Goedert J 2001 Phys. Lett. A 279 181
21. Athorne C 1991 Phys. Lett. A 159 375
22. Govinder K S Athorne C Leach P G L 1993 J. Phys. A: Math and Gen 26 4035
23. Govinder K S Leach P G L 1993 Quaestiones Mathematicae 16 405
24. Ermakov V 1880 Univ. Isz. Kiev Series III 9 (translated by Harin A O)
25. Leach P G L and Andriopoulos K 2008 Appl. Anal. Discrete Math. 2 146
26. Rogers C Hoenselaers C and Ray J R 1993 J. Phys. A: Math. and Gen. 26 2625
27. Schief W K Rogerts C and Bassom A P 1996 J. Phys. A: Math. and Gen. 29 903
28. Goedert J and Haas F. 1998 Phys. Lett. A 239 348
29. Athorne C 1991 J. Phys. A: Math and Gen 24 L1385
30. Leach P G L 1991 Phys. Lett. A 158 102
31. Govinder K S and Leach P G L 1974 Phys. Lett. A 186 391
32. Leach P G L and Karasu A 2004 J. Nonlinear Math. Phys. 11 269
33. Tsamparlis M and Paliathanasis A 2011 J. Phys. A: Math. and Theor. 44 175202
34. Haas F and Goedert J 1996 J. Phys. A: Math. Gen. 29 4083
35. Kalotas T M and Wybourne B G 1982 J. Phys A: Math and Gen 15 2077
36. Tsamparlis M and Paliathanasis A (2012) J. Phys. A: Math and Theor. 275202
37. Colley A A Tupper B O J 1989 J. Math. Phys 30 11
38. Colley A A Tupper B O J 1990 J. Math. Phys 31 3
April 1, 2015
1:20
G*Unal
STOCHASTIC SYMMETRIES OF WICK TYPE STOCHASTIC

ORDINARY DIFFERENTIAL EQUATIONS
GAZANFER UNAL
Department of International Finance, Yeditepe University,
Kayisdagi Caddesi, 26 August Campus,
Atasehir, 34755, Istanbul,Turkey
gunal@yeditepe.edu.tr
We consider Wick type stochastic ordinary differential equations with Gaussian white
noise. We define the stochastic symmetry transformations and Lie equations in Kondratiev space (S)N
1 . We derive the determining system of Wick type stochastic partial differential equations with Gaussian white noise. Stochastic symmetries for stochastic Bernoulli, Riccati and general stochastic linear equation in (S)N
1 are obtained. A
stochastic version of canonical variables is also introduced.
1. Introduction
Symmetry methods provide a systematic way to treat both deterministic1, 2 and
stochastic differential equations driven by Wiener process.3, 4 The latter is a continuous but nowhere differentiable stochastic process, therefore, its physical interpretation poses difficulties. White noise analysis, introduced by Hida5, 6 remedies this
problem. We aim to extend Lies symmetry analysis to white noise driven stochastic
equations. Gaussian white noise W (t, ) is a generalized process with the property
W (t, ) =
dB (t, )
in (S) ,
dt
where B (t, ) is the Brownian motion and (S) is the Hida distribution space (in
which the Wick product applies). Gaussian white noise probability space defined
on the set B of Borel sets of = S 0 in which S 0 = S 0 (R) is the space of tempered
distributions and it is the dual space of Scwartz space S = S 0 (R). The probability
measure, , is given by the Bochner-Minlos theorem7

Z
1
2
eih,i d () = exp kkL2 (R) ,
2
S 0 (R)
0
where S and S. Stochastic modeling by employing white noise is more

realistic than other methods.7 In this work we introduce symmetry analysis to treat
a system of stochastic ordinary differential equations which involves white noise
functionals.
1
April 1, 2015
1:20
G*Unal
Gazanfer Unal
We consider Wick type system of stochastic ordinary differential equations

(WSODE)
dXi (t)
= Fi (X1 , ..., Xn , t, W1 , ..., Wm )
i = 1, ..., n in (S)N
(1)
1 .
dt
Here is the Wick product on a suitable space of stochastic distributions, called
N
the Kondratiev space (S)N
1 . Defining the problem on (S)1 has two benefits: 1)
it comprises the solutions of many stochastic differential equations, both ordinary
and partial and in arbitrary dimension, 2) differentiation can be considered in the
usual strong sense in (S)N
1 when the components of such equations are regarded
N
7
as (S)N
-valued.
Kondratiev
space (S)1 is given by
1
(
N
(S)1
)
x=
b H , b R ; kxk1,k =
b2
(2N )
< for some q N .
Here H are Hermite functionals. Furthermore, for [0, 1] we have

N
(S)1 (S) (S)0 L2 (m ) (S)0 (S) (S)1 .

N
Note that (S)0 and (S)0 are different spaces and, they coincide with the Hida
spaces (S) and (S) , respectively. One of the many useful properties of (S) is that
it contains the singular or pointwise white noise. Although many other important
Brownian functionals belong to (S)6 , Hida distribution space (S) is too small for
the purpose of solutions of stochastic ordinary and partial differential equations.
7
There are equations with no solution in (S) but a unique solution in (S)N
1 . One
of the many useful properties of (S) is that it comprises the singular or pointwise
white noise W (x, ).
In this work we give definitions for stochastic transformations group and stochastic symmetry of Wick type equations. Furthermore, Wick type determining system
for the stochastic infinitesimals are obtained for the system of Wick type stochastic
ordinary differential equations. Several one-dimensional linear and nonlinear examples are also included.
2. Symmetries of Stochastic Ordinary Differential Equations of
Wick type
A salient feature of the Wick type stochastic ordinary differential equations
(WSODE ) is that the random coefficient functions can be anticipated5, 6 . This
property relaxes the adapted random coefficient function requirement of Ito type
SODEs (ISODE). In other words WSODEs are more general than ISODEs. We already know that symmetry transformations of ISODEs do not involve the Wiener
process, hence, they are deterministic in character (see for instance3, 4 ). By considering the symmetry transformations of WSODEs we anticipate that these transformations would be stochastic in nature. Therefore, our main objective is to derive
April 1, 2015
1:20
G*Unal
Stochastic Symmetries of Wick type Stochastic Ordinary Differential Equations
the determining equations for the WSODE. This requires several definitions and
intermediate results.
Definition 2.1. When the Wick version of infinitesimal transformations (or nearidentity change of variables) Eq. (3): leave Eq. (1) form invariant, i.e., the set of
transformations
i (t, ) = A (X, t, ; ) ; t(t, ) = B (X, t, ; ) in (S)N
G:X
1
i
i
(2)
is called one-parameter stochastic transformation group G provided

(1) G1 (inverse of eG) exists
(2) the composition law G (1 ) G (2 ) = G (1 + 2 ) hold in (S)N
1 .
Expanding A
i (X, t, ; ) and Bi (X, t, ; ) in the neighbourhood of = 0 leads
to
i (t, ) Xi + (X, t, ),
X
i
t(t, ) t + (X, t, )
in (S)N
1 ,
(3)
where
i (X, t, ) =
B (X, t, ; )
A
i (X, t, ; )
|=0 , (X, t, ) =
|=0 .in (S)N
1 .
We call i (X, t, ) and (X, t, ) the Wick version of the infinitesimal transformations. The latter leads us to define the Wick version of the Lie equations
i (t, )
dX
t, ), dt (t, ) = (X,
t, )
= i (X,
d
d
i (t, )|=0 = Xi , t(t, ) |=0 = t
X
Definition 2.2. When the Wick version of infinitesimal transformations (or nearidentity change of variables) Eq. (3) leave Eq. (1) form invariant, i.e.,
i (t)

dX
1 , ..., X
n , t, W
1 , ..., W
m
= Fi X
i = 1, ..., n in (S)N
1 .
dt
then G is called a stochastic symmetry of the WSODE given in Eq. (1).
Lemma 2.1.
(4)
(t, ) = W (t, ) + dW (t,) in (S)1 .

W
dt
Proof. The Hermite transformation H of W (t) is defined by7

(t, z) =
H (W (t, )) = W
k (t) zk ,
k=1
where z = (z1 , z2 ,, ) (CN )c . It transforms under t t + (X, t, z) as

(t + (X, t, z), z) =
W
k (t +
(X, t, z)) z k =
k=1
(t, z)
dW
(t, z) +
=W
,
dt
X
k=1
k (t) z k +
X
dk (t)
k=1
dt
zk
April 1, 2015
1:20
G*Unal
Gazanfer Unal
the inverse Hermite transformation

(t, z) = W (t, ), H1
H1 W
H
(t, z)
dW
dt
!
=
(t, z)
dW
dt
!
=
dW (t, )
,
dt
dW
.
dt
This prove the result.

Equipped with these tools we now state the main result of this paper.
Theorem 2.1. Wick version of infinitesimals i (X, t, ) and (X, t, ) of the
symmetry group of system of first order WSODE given in Eq. (1) must satisfy the
stochastic PDEs of Wick type (Wick version of determining system)

i

+ Fj i Fi
+ Fj
=
t
Xj
t
Xj
m
F
F X Fi dW

(5)
j i + i +
Xj
t
W
dt
=1
It can be proved by direct calculation.
3. Stochastic Symmetries of One Dimensional WSODEs
We start by considering the general one dimensional linear Wick type Skorohod
stochastic differential equation in
m
X
dX(t, ) = g(t, )dt + r(t, )X(t)dt +
i (t, )X(t, )Bi (t) X(0) = X0 (),
i=1
(6)
where g(t, ), r(t, ) and i (t, ); 1 i m are continuous (S)1 -processes and
that X0 (S)1 . That is to say that g(t, ), r(t, ), X0 () and i (t, ); 1 i m,
are random functions, possibly anticipating. B(t, ) = (B1 (t, ), ..., Bm (t, )) is mdimensional Brownian motion. In view of the relation between Skorohod integrals
and Wick products with white noise,
Z
Z
U (t) B (t, ) = U (t) Wi (t)dt ,
we rewrite Eq. (6) as
m
X
dX(t, )
= g(t, )+r(t, )X(t)+
i (t, )X(t, )Wi (t)
dt
i=1
X(0) = X0 ().
In this setting it is natural to assume that X0 and X(t, ) the coefficients involved
are (S)1 -valued. Wick type determining system has the form

X
m

f f

f dWi
+f

+f
+ f
=

.
t
x
x
t
t
x
Wi dt
i=1
(7)
April 1, 2015
1:20
G*Unal
Here
f = g(t, ) + K(t, )X(t, ), K(t, ) = r(t, ) +
m
X
i (t, )Wi (t).
i=1
Taking the H-transform of the Wick type determining system Eq. (7) we obtain

X
m
fi
f

f
f dW
+f

+f
+ f
=
.
(8)
f dt
t
x
x
t
t
x
i=1 Wi
Now
z)x, K(t,
z) = re (t) +
f = ge (t, z) + K(t,
m
X
fi (t, z),
ei (t, z)W
i=1

e z), = H , = H .
x = X(t,
Solving the determining system for each z (CN )c we obtain the infinitesimals as

Z
Z
K(t, z)dt .
K(t, z)dt , = K(t, z)x + ge (t, z) + C1 exp
= exp
The inverse Hermite transform

(t, X, ) = H1 (t, X, z) , (t, X, ) = H1 ( (t, X, z)) ,
yields stochastic infinitesimals of Wick type to be written as
! !
Z
m
X
(t, ) = exp
r(t, ) +
i (t, )Wi (t) dt
i=1
m
X
(t, ) = (t, )
r(t, ) +
(9)
!
i (t, )Wi (t)
!
X(t, ) + g(t, )
i=1
Z
+C1 exp
r(t, ) +
m
X
!
i (t, )Wi (t)
!
dt .
i=1
The Wick type stochastic canonical variables are

! !
Z
m
X
= exp
i (t, )Wi (t) dt X.
t = t, X
r(t, ) +
i=1
Notice that
d
dX
=
dt
dt
Z
exp
r(t, ) +
m
X
!
i (t, )Wi (t)
!
dt X
i=1
Z
= exp
r(t, ) +
m
X
!
i (t, )Wi (t)
!
dt g(t, ).
i=1
This verifies that

Z
exp
r(t, ) +
m
X
i=1
!
i (t, )Wi (t)
!
dt
in (S)1
(10)
April 1, 2015
1:20
G*Unal
Gazanfer Unal
is the Wick type stochastic integrating factor. This leads us to (S)1 -valued solution
! !
Z t
m
X
r(u, ) +
i (u, )Wi (u) du
X(t, ) = X0 () exp
0
Z
+
r(u, ) +
exp
0
i=1
m
X
i (u, )Wi (u) du g(s, )ds in (S)1 .
i=1
We next consider one dimensional nonlinear WSODE of the form

dX(t, )
= gq (t) X(t, )q + gs (t) X(t, )s
dt
!
m
X
+ r (t) +
i (t, )Wi (t, ) X(t, ) X(0) = X0
(11)
i=1
where q, s N and, X0 () and i (t, ); 1 i m, are random functions, possibly

anticipating. The Wick version of the determining system reads as

X
m
F dWi

F
F

+F

+F
+F
.
=

t
x
x
t
t
x
Wi dt
i=1
(12)
Here,
!
m
X
q
s
F = gq (t) X(t, ) +gs (t) X(t, ) + r (t) +
i (t, )Wi (t, ) X(t, ).
i=1
Taking the H-transform of the Wick type determining equation ( 12)we obtain

X
m
fi
F
F

F dW
+F

+F
+F
.
(13)
fi dt
t
x
x
t
t
x
W
i=1
Here
z)x, K(t,
z) = re (t) +
F = geq (t) xq + ges (t) xs + K(t,
m
X
fi (t, z),
ei (t, z)W
i=1

e z), = H , = H .
x = X(t,
We next identify several cases.
Case 1 : s = 0 q = 2 Stochastic Riccati Equation The solution of the determining equation (13) reads
Z

1
(t, x, z) = ge2 (t)
ke
g2 (t) dt + C1 ,

d
1
(t, x, z) = K + C2 x + K
ge0 k ge0 .
dt
The inverse Hermite transformation

(t, X, ) = H1 (t, X, z) , (t, X, ) = H1 ( (t, X, z)) ,
April 1, 2015
1:20
G*Unal
yields Wick type stochastic infinitesimals which involve the singular white noise
Wi (t, )
Z

1
(t, X, ) = g2 (t)
kg2 (t) dt + C1 ,
!
!
m
X
(t, X, ) = r (t) +
i (t, )Wi (t, ) + C2 X
i=1
r (t) +
m
X
!1
g0 k
i (t, )Wi (t, )
i=1

d
g0 .
dt
(14)
Case 2: geq = 0 Stochastic Bernouli Equation

The solution to the determining equation (13) has the form

Z
C0
x, z)x.
exp (s 1) K(t, x, z)dt , (t, x, z) = K(t,
(t, x, z) =
ge (t)
The inverse Hermite transformation

(t, X, ) = H1 (t, X, z) , (t, X, ) = H1 ( (t, X, z))
leads to Wick type stochastic infinitesimals
1
(t, X, ) = g2 (t)
Z
exp (s 1)
r (t) +
m
X
i (t, )Wi (t, ) dt ,
i=1
(t, X, ) = (t, X, ) r (t) +
m
X
!
i (t, )Wi (t, ) X(t, ).
(15)
i=1
Case 2 :i) s = 2
Similarly we obtain the following Wick type stochastic infinitesimals.
!
!
m
X
(t, X, ) =
r (t) +
i (t, )Wi (t, ) + C3 X
i=1
+ C1 exp
Z
r (t) +
m
X
i (t, )Wi (t, ) dt ,
i=1
(t, X, ) = g (t)
Z

k (t) g (t) dt + C2 .
(16)
4. Conclusions
We have defined the stochastic symmetry transformation group. We have shown that
Wick type stochastic ordinary differential equations admit stochastic symmetries in
Kondratiev space (S)N
1 .Stochastic symmetries for stochastic Bernoulli, Riccati and
general stochastic linear equation in (S)N
1 are obtained.
April 1, 2015
1:20
G*Unal
Gazanfer Unal
References
1. P. J. Olver, Applications of Lie groups to differential equations (Springer, 2000).
2. G. W. Bluman, A. F. Cheviakov and S. Anco, Applications of symmetry methods to
partial differential equations (Springer, 2009).
3. G. Unal,
Nonlinear Dynamics 32, 417 (2003).
4. E. Fredericks and F. M. Mahomed, Journal of Nonlinear Mathematical Physics 15, 44
(2008).
5. T. Hida, Brownian motion (Springer, 1980).
6. T. Hida, White noise: an infinite dimensional calculus (Springer, 1993).
7. H. Holden, B. ksendal, J. Ube and T. Zhang, Stochastic partial differential equations
(Springer, 1996).
April 1, 2015
1:21
Sergei*Zuev


A FINITE DIFFERENCE APPROACH TO FIND EXACT

SOLUTION OF DIFFERENTIAL EQUATIONS
SERGEI ZUEV
Department of the Computer Software and Automated Systems, Belgorod State Technological
University named after V.G. Shoukhov, 46 Kostyukova Str.
Belgorod, 308012, Russia
Email: sergey.zuev@bk.ru
This paper contains the background and samples of an approach to construct exact solutions of a wide range of differential equations (DEs). This approach is based on the
finite difference equation which corresponds to the given DE. There are three cases considered: linear partial differential equation (PDE) with constant coefficients and at least
one non-zero root of characteristic equation, linear PDE with constant coefficients and
completely zero roots of the characteristic equation, and a case of nonlinear autonomous
dynamical system of second order. Each of these cases is illustrated by an example.
Keywords: finite difference equation, exact solutions, linear PDE, nonlinear autonomous
dynamical systems
1. Introduction
There exist many physical systems that have a differential equation description that
looks quite simple. However it is difficult to find the exact expression for a motion
and even to define the phase trajectories of such simple systems. For example, the
general consideration of a one-dimensional heat equation and the corresponding
Cauchy problem was carried out by Canon (1984)1 . Instead of finding the exact
solution of differential equations which could be very challenging at times, it is
now common to investigate properties of the physical systems using other schemes.
For example one such approach is to examine the solution of DEs by numerical
techniques. But in some cases it is not enough to only know the solution behavior and
its numerical solution; it is critical to find the exact solution. For instance, in some
problems of hydrodynamics2 where the basic PDE is non-linear, the system requires
that the condition of continuity (linear PDE) must be obeyed. In a dynamical
system with nonlinear ODE, there are also many open problems and challenges. In
particular, the well-known van der Pol oscillator3 4 , which represents the simplest
model with dynamic chaos, has no exact solution yet.
As a proposition, in the nonlinear case, the linear nature in differentiation itself
could be a source of difficulties. As for the linear PDE case, the reason for difficulties could be a broken scale invariance of the system. Both of these issues can be
1
April 1, 2015
1:21
Sergei*Zuev
S. Zuev
overcome by means of finite difference equations (FDE) or, according to another

definition, lattice equations.
Remark 1.1. For any function u(x1 , . . . , xn ) on n real variables xl , l = 1, . . . , n,
in some neighborhood of the point p0 = (x01 , . . . , x0n ) which contains point
(x01 , . . . , x0l + l , . . . , x0n ), the finite difference derivation in p0 with respect
to xl is defined as
Dxl u(p0 ) =
u(x01 , . . . , x0l + l , . . . , x0n ) u(x01 , . . . , x0n )

.
l
(1)
Remark 1.2. Any functional equation which contains finite difference derivation
is called finite difference equation.
The first order FDE has the following general form
F (Dx1 u, . . . , Dxn u, u, x1 , . . . , xn ) = 0.
(2)
This equation determines the value u(x1 + 1 , . . . , xn + n ) of the unknown function

in the neighborhood of the point p0 when the initial and boundary conditions are
specified.
The l parameters are minimal values of the finite differences. Eq. (2) transforms to PDE when l l, tend to zero and liml 0 Dl u, exist l. If the boundary
conditions are given, then the function u(x1 , . . . , xn ) inside the boundary is determined by that PDE. Therefore, when the unknown function is differentiable (this
is necessary for PDE formulation), it is possible to find a solution of PDE
v
lim
1 ,...,n =0
(3)
from the solution u of FDE if all the relevant limits exist.

For linear PDE the following theorem holds.
Theorem 1.1. If Eq. (2) is linear, i.e.
n
X
Pi (x1 , . . . , xn )Dxi u + Q(x1 , . . . , xn )u + R(x1 , . . . , xn ) = 0,
(4)
i=1
then its solutions are in one-to-one correspondence with solutions v(x1 , . . . , xn ) of

the linear PDE
n
X
Pi (x1 , . . . , xn )
i=1
v
+ Q(x1 , . . . , xn )v + R(x1 , . . . , xn ) = 0,
xi
(5)
and lim1 ,...,n =0 u = v.

Proof. When 1 , . . . , n go to zero, Eq. (4) becomes Eq. (5) and v = lim1 ,...,n =0 u
obeys Eq. (5). Therefore it is evident that for any u we can find v as the Theorem
requires.
April 1, 2015
1:21
Sergei*Zuev
A new approach to solve DE exactly
Now we prove the converse statement. Let us do it for n = 2 (for arbitrary n it

can be made by analogy). Eqs. (4) and (5) will take the forms:
P1 D1 u + P2 D2 u + Qu + R = 0
(6)
P1 1 v + P2 2 v + Qv + R = 0.
(7)
and
Let v is given and obeys Eq. (7). Let us find u which has the limit lim1 ,2 0 u = v
and in the same time obeys (6).
We can represent u as
u=
1i 2j uij (x1 , x2 ),
u00 v.
(8)
i,j=0
From here
1k u =
1i 2j 1k uij (x1 , x2 ),
2k u =
i,j=0
1i 2j 2k uij (x1 , x2 )
i,j=0
for any k N. Since 1 and 2 are small, it is possible to expand function u to the
Taylor range near (x1 , x2 ):
u(x1 + 1 , x2 ) = u(x1 , x2 ) + 1 u
2
1
+ 12 u 1 + . . . ,
1!
2!
u(x1 , x2 + 2 ) = u(x1 , x2 ) + 2 u
2
2
+ 22 u 2 + . . . .
1!
2!
Therefore
D1 u = 1 u + 12 u
1
+ ...,
2!
D2 u = 2 u + 22 u
2
+ ....
2!
Substituting it into Eq. (6) and writing down terms with factors 1i 2j , we get
i = 0, j = 0 :
P1 1 v + P2 2 v + Qv + R = 0,

1
1
i = 1, j = 0 :
P1 1 u10 + 12 v + P2 2 u10 + 22 v + Qu10 = 0,
2
2
and so on. The first equation of this series is exactly Eq. (6). The others equations
determine all functions uij . So we have found all components in the expansion given
by Eq. (8). The function u is completely determined.
Similar propositions were presented by Gelfond5 . So, for the linear case, the
functions u and v, connected by Eq. (3), always exist. That is why we have the
following approach to find exact solution of linear PDE:
Write down FDE from the given PDE by just changing derivatives to finite difference derivatives;
Solve FDE and find the form of u(x1 + 1 , . . . , xn + n );
April 1, 2015
1:21
Sergei*Zuev
S. Zuev
Let all differences l tend to zero and find the solutions of the linear PDE by
calculating its limits.
The above results for FDEs of scalar functions can easily be generalized to
vector-functions. In which case, we need to examine a system of equations instead
of a single equation.
For FDE, it is normally assumed that the subject is numerical. In reality it is
not the case. For example in the case of linear FDE, once it is possible to find exact
FDE solution then we can get the exact solution of the linear PDE which is evident
from above. This is the core of the proposed approach in this paper.
2. Linear PDEI: Heat equation
Let us consider the case of a homogeneous rod without any heat sources, modeled
by the heat equation
2u
u
2 =0
(9)
t
x
which we solve by using the above mentioned approach. As a first step, write down
the FDE assuming that all the derivatives in Eq. (9) become finite. In general, the
finite difference values are different for x and t. Let us denote the finite differences
of x and t with and , respectively. Using these notations,
y = x,
we find the FDE which corresponds to the same differential equation (9)
Dt u Dy2 u = 0,
(10)
where = 2 / 2 . The differences for both independent variables is equal to . This

equation is linear and the linear combination of its solutions is the solution. Now
we consider the following situation, the ranges of the values of unknown function
for both independent variables coincide with each other while the other variable is
fixed. This allows us to find the solution of (10) in the following form
y
u(t, y) = C1 ( 1 + 1) ( 1 + 1) + C2 ( 1 + 1) ( 2 + 1) +
t
+C3 ( 2 + 1) ( 1 + 1) + C4 ( 2 + 1) ( 2 + 1) ,
where 1 and 2 are solutions of the characteristic equation
2 = 0,
which has the following roots
1 = 0, 2 =
1
.
The solution will take the form

y

t

t
y
u(t, y) = C1 + C2
+1
+ C3
+1
+ C4
+1
+1
.
(11)
April 1, 2015
1:21
Sergei*Zuev
Substituting this into Eq. (10), we get

C2 = C3 = 0,
while C1 , C4 are arbitrary constants. Finally
u(t, y) = C1 + C4 (
y
t
+ 1) ( + 1) .
Now, recall that = 2 and y = x, therefore

u(t, x) = C1 + C4
2
+1
t + x
.
(12)
This is the exact solution of finite difference heat equation for homogeneous rod
without heat sources. However, and are still undetermined. To get the solution
of corresponding PDE it is necessary to put , 0. Let us denote
lim
,0
= k.
After taking the limit we arrive at the solution

u(t, x) = C1 + C4 e
k2 t
kx
e.
(13)
The known general solution of the heat equation is determined for x (0; L),
u
(0, x) = (x) and u
(t, 0) = u
(t, L) = 0 and presented in Cannon1 as
u
(t, x) =
X
n=1
2
L
Z
0
n
() sin
d
l
!
sin
n
n 2
x e( L ) t .
L
If we put kn = i n
L in Eq. (13), then this solution will be the imaginary part of
the sum of the solutions of Eq. (13) form, i.e.
u
(t, x) =
An e
2t
kn
kn x
n=1
where
2
An =
L
() sin
0
n
d.
L
This above example illustrates the approach in the case when the characteristic
equation has at least one non-zero root. This result was published in Russian by
Dorokhova6 .
April 1, 2015
1:21
Sergei*Zuev
S. Zuev
3. Linear PDEII: Homogeneous continuity equation

3.1. The problem description
A well-known PDE, which represents the case of zero eigenvalue is the homogeneous
continuity equation
vx
vy
vz
+
+
= 0.
(14)
x
y
z
This equation determines the so called solenoidal vector fields
v(x, y, z) = (vx (x, y, z), vy (x, y, z), vz (x, y, z))
in the set D R3 , where the functions vx , vy vz are continuously differentiable at
least once.
In this section the general solution of Eq. (14) is constructed for the vector
fields which are determined over R3 and obey the physical assumptions expressed
in Carlson2 (a further application of the solution in Euler and Navier-Stokes equations). These assumptions are:
(1) vector field v is determined, continuous and differentiable
on the whole R3 ;
R
2
(2) vector field v obeys the condition of limited energy: R3 |v(x, y, z, t)| dxdydz <
C for any given t.
We follow the same procedure to obtain the solution of above equation with respect
to the linearity of Eq. (14).
3.2. FDE for homogeneous continuity equation
The FDE corresponding to (14) is
Dx ux + Dy uy + Dz uz = 0,
(15)
where Dx is finite difference derivation for x (the same for y and z). The name of
the vector field components is changed to u in order to separate it from the target
field v which is the solution of Eq. (14).
In the introduction we mentioned that in the case of linear PDE the reason
of difficulties may be connected with broken scale invariance of the system. The
exponential change in the dependent variable allows us to avoid this breaking of
scale invariance using new unknown functions. That is why the components of the
vector field u are re-defined in terms of the exponential counterparts
ul = el ,
l = x, y, z.
The functions l , are now C-valued because ul might be negative. Moreover, l are
also multi-valued, i.e., if their values change by 2ki, k Z, i2 = 1, then we get
the same values of ul . The finite difference derivations Dl ul in the new variables
take the form
eDl l l 1
Dl ul = el
,
l
April 1, 2015
1:21
Sergei*Zuev
and Eq. (15) becomes

y z (ex +Dx x x ex )+x z (ey +Dy y y ey )+x y (ez +Dz z z ez ) = 0. (16)
The main idea to find solutions to above equation is to explore exponent indices
x ; y ; z ; x + Dx x x ; y + Dy y y ; z + Dz z z .
(17)
From a simple observation we note that each of these indices can have one of the two
possible forms, constant or variable, one can derive that there are only 7 possible
sets {l }:
(1) x = cx , y = cy , z = z (x, y, t)
(2) x = cx , y = y (x, z, t), z = cz
(3) x = x (y, z, t), y = cy , z = cz
cy
)
(4) x = cx , y = cy yy + cz zz + yz (x, t), z = y + ln zy (1e
(ecz 1)
(5)
cx
)
x = cx xx + cz zz + xz (y, t), y = cy , z = x + ln zx(1e
(ecz 1)
cy
(1e )
y
x = y + ln xy (e
+ cx xx + yx (z, t), z = cz
cx 1) , y = cy
y
x = cx xx + cy yy + cz zz + x (t),
y = cx xx + cy yy + cz zz + y (t), z = cx xx + cy yy + cz zz + z (t),
and the following condition applies
(6)
(7)
z y ex (t) (ecx 1) + x z ey (t) (ecy 1) + x y ez (t) (ecz 1).

The sets (1)-(3) lead to the solution of Eq. (15) in the form
fx (y, z, t)
u(1) = fy (x, z, t) ,
fz (x, y, t)
which appear in any other solutions as a sum.
The sets (4)-(6) give the solutions
cy3 y
cz2 z
cx3 x
cx2 x
(1ecy3 )
y
e x
fxz (y, t)e x e z + xy (e
cx3 1) fyx (z, t)e
cy1 y
cy3 y
cz1 z
cx3 x
u(2) =
fyz (x, t)e y e z + fyx (z, t)e y e x
cy1 y
cz1 z
cx2 x
cz2 z
cx2
z (1ecy1 )
)
y
x e z
e z + zx(1e
y (ecz1 1) fyz (x, t)e
(ecz2 1) fxz (y, t)e
(18)
where cij are constants.

As for Eq. (7), the solution is
u(3) = e
cx x
x
cy y
y
cz z
z
gx (t)
gy (t) .
gz (t)
(19)
The general solution of Eq. (15) is the sum u(1) + u(2) + u(3) and with the
condition
ec x 1
ecy 1
ecz 1
gx (t)
+ gy (t)
+ gz (t)
=0
(20)
x
y
z
for every summand (range indices are omitted for clarity).
April 1, 2015
1:21
Sergei*Zuev
S. Zuev
3.3. The PDE solution

Let us denote
qx =
qxj =
ecx 1
ecy 1
ecz 1
, qy =
, qz =
,
x
y
z
ecxj 1
ecyj 1
eczj 1
, qyj =
, qzj =
, j = 1, 2, 3.
x
y
z
Now let x , y , z 0 and every summand of the solution of (14) will be the limit
of the sum
u=
lim
x ,y ,z 0
u(1) + u(2) + u(3) .
More precisely
ux = gx (t)eqx x+qy y+qz z + fxz (y, t)eqx2 x+qz2 z
qy3
fyx (z, t)eqy3 y+qx3 x + fx (y, z, t)
qx3
uy = gy (t)eqx x+qy y+qz z + fyz (x, t)eqy1 y+qz1 z +
uz
+fyx (z, t)eqy3 y+qx3 x + fy (x, z, t)

qy1
= gz (t)eqx x+qy y+qz z
fyz (x, t)eqy1 y+qz1 z
qz1
qx2
fxz (y, t)eqx2 x+qz2 z + fz (x, y, t).

qz2
(21)
The Eq. (20) takes the form

gx (t)qx + gy (t)qy + gz (t)qz = 0
for every summand.
After integrating (21) over R3 , the final derivation comes out to be
Gx + F1 H2 + fx (y, z, t)
v = Gy + F3 + F2 + fy (x, z, t) ,
Gz H3 H1 + fz (x, y, t)
(22)
April 1, 2015
1:21
Sergei*Zuev
where functions Gx , Gy , Gz , F1 , F2 , F3 , H1 , H2 , H3 are determined as follows:

Z
Gx =
gx (, , , t)e2i(x+y+z) ddd,
R3
Z
Gy =
gy (, , , t)e2i(x+y+z) ddd,
R3
Z
gx (, , , t)e2i(x+y+z) ddd
Gz =
3
ZR
gy (, , , t)e2i(x+y+z) ddd,
R3
Z
F1 =
fxz (, y, , t)e2i(x+z) dd,
R2
Z
fxz (, y, , t)
H1 =
2
ZR
F2 =
e2i(x+z)
dd,
fyx (, , z, t)e2i(x+y) dd,
R2
Z
H2 =
fyx (, , z, t)
2
ZR
F3 =
e2i(x+y)
dd,
fyz (x, , , t)e2i(y+z) dd,
R2
Z
H3 =
fyz (x, , , t)
R2
e2i(y+z)
dd.
(23)
Here, functions gx , gy , fxz , fyx and fyz are integrable and differentiable over R3
(0, a). It is clear that these functions are connected with the functions on the righthand side of Eq. (22) by Fourier transformation. In particular for gx , gy the transformation is 3-dimensional and 2-dimensional for others.
The physical condition (2) leads to the special form of functions gx , gy , fxz , fyx
fyz . It is possible to construct the ansatz for the solution of the Euler equation for
ideal fluid in the same way as presented here and by application of the physical
condition (2).
4. Nonlinear autonomous ODE
Let us consider the well-known case of second order autonomous dynamical system
(which is called as Lienard system):
x
+ (x)x + (x) = 0 ,
(24)
for t [0, a] R. The goal is to derive a general solution of Eq. (24) or find a
way to solve the Cauchy problem. The dynamical system determined by Eq. (24)
is autonomous, i.e. its left-hand side does not contain the time t explicitly. That is
why the order of the system could be decreased
y 0 + (x) +
(x)
= 0,
y
(25)
April 1, 2015
10
1:21
Sergei*Zuev
S. Zuev
where y = y(x) = x 6= 0.
One can consider first-order ODE of the general type
y 0 = F (x, y),
(26)
instead of (25) and make some relevant propositions. Determine the FDE corresponding to (26):
D
y = F (x, y) + O(),
(27)
where lim=0 y(x, ) = y(x).

Let us denote x0 x(0) and x = x(t). By using the definition of D operation in
ordinary way (D
y = 1 (
y (x + ) y(x))), we have
y(x0 + ) = y(x0 ) + F (x0 , y(x0 )) + O( 2 ).
(28)
Theorem 4.1. Let Eq. (27) have a solution y such that y(x, ) has the same sign
as y(x0 , ) for every x (x(0), x(a)). Let w(x,
) ln y(x, ). If the limit
lim w(x,
) = w(x)
(29)
=0
exists and the sequence

{Sn }
( n
X
Ckn
i=1
x x0
n
i
)
(i)
w (x0 ) ,
(30)
where Ckn are binomial coefficients, is convergent for every appointed variable x,
then the new dependent variable
w(x) = ln y(x)
(31)
integrates the Eq. (26), but may be as a convergent range. Note that
w0 =
F
,
y
w(2) =
F 2 F 0y
,
y2
, w(3) =
2F 3 + 3F 0 F y F 00 y 2
,....
y3
Proof. By splitting the segment [x(0), x(t)], t (0, a], into n equal parts, let each
of them have length . When go to zero, n go to . Using (28) n times, we have
y(x) = y(x0 + n) = y(x0 ) +
n1
X
+
F x0 + j, y(x0 ) +
j=0
j1
X

F x + k, y(x0 ) + F (x + k) + O( 2 ) + O( 2 ).
k=0
It is evident that
y(x0 + ) = Dy(x0 ) + y(x0 ) = y(x0 ) + F (x0 , y(x0 )) + O( 2 ).
(32)
April 1, 2015
1:21
Sergei*Zuev
11
At the same time

F (x0 , y(x0 )) = D
y (x0 ) O() =
F (x0 + , y(x0 ) + F (x0 ) + O()) =

0)
ew(x
0)
eDw(x
1 ,

0 +)
ew(x
0 +)
eDw(x
1 ,
and so on. Using identities

Dw(x
+ ) = D2 w(x)
+ Dw(x),
w(x
+ ) = Dw(x)
+ w(x),
we derive
y(x, ) = y(x0 + n) =
h
2 D2 w
= y(x0 ) 1 + eDw 1 + e2Dw
i
eDw + + O( 2 ) ,
(33)
where w
= w(x
0 ). The sum in Eq. (33) has a result
e
Pn+1
i=1
Cin+1 i D i w
That is why, (33) transforms to

ln
n+1
X
y(x, )
Cin+1 i Di w.
=
y(x0 )
i=1
(34)
Eq. (34) when 0 comes to

ln

i
n+1
X
y(x)
x x0
Cin+1
=
w(i) ,
y(x0 )
n
i=1
(35)
0
because = xx
n , and it is assumed that n . Equation (35) represents the
limit of the sequence Sn+1 in (30) and hence it is convergent. That is, we have
the relation (35) which is connecting y(x) and y(x0 ) through w(i) (x0 ), depending
on F (i) (x0 ) and y(x0 ) only. Consequently, we can say that Eq. (35) represents a
Cauchy problem solution in the form of convergent range.
Let us use the result of this theorem to find the general solution of the Eq.
(24). Note that it is evident that any solution of Eq. (26) could be presented as a
range in some neighborhood of the point x0 , but the above Theorem shows that
the variable change w = ln y gets to the solution of the Cauchy problem not
over the neighborhood but for any interval (x(0), x(a)) (up to sign of the function
y(x)). That is why it is possible to use the new variable w in order to solve the Eq.
(25) over any interval which is in correspondence with the Theorem 4.1 conditions.
Making the change of variables we obtain
w0 = ew + e2w .
(36)
April 1, 2015
12
1:21
Sergei*Zuev
S. Zuev
For Lienard Equation (25) we have

F (x, y) = (x)
(x)
y
and Eq.(27) takes the form

D
y = (x)
(x)
+ O().
y
Using the theorem 4.1, we can derive that solution y(x) of Eq. (25) has the following
form
"
#
i

n+1
X
x x0
n+1
(i)
y(x) = y(x0 ) exp lim
w
Ci
,
(37)
n=
n
i=1
where w0 = (x0 )/y(x0 ) + (x0 )/y 2 (x0 ) and
w(i) =
dw(i1)
(x0 ),
dx
Fx0 (x0 ) = 0 (x0 )
0 (x0 )
,
y(x0 )
y 0 (x0 ) = F (x0 , y(x0 )).
Note that if the limit exists for given functions , and the initial condition is
y0 = y(x0 ), then Eq. (37) yields an exact solution of the Cauchy problem for Lienard dynamical system. This is convenient for computer modeling of the dynamical
system. Of course, the case of Lienard system is a simple example. The result of
the theorem 4.1 could be applied to any autonomous dynamical system of second
order or to any first order dynamical system.
5. Acknowledgments
I appreciate Prof. Asya Aminova of the Kazan State University, Russia, for supervising my work several years ago. I also thank Prof. Asghar Qadir and Dr. Sajid
Ali for inviting me to SDEA-II, paying attention to my work and initiating useful
discussions.
References
1. J.R. Cannon, Encyclopedia of Mathematics and Its Applications, 23 1st ed., (1984).
2. J. Carlson, A. Jaffe, A. Wiles (eds.), The Millennium Prize Problems, Clay Mathematics
Institute (American Mathematical Soc., 2006).
3. B. Van der Pol, Phil. Mag., ser. 6, 43, (1922)
4. B. Van der Pol, Phil. Mag., ser. 7, 2, (1926)
5. A.O. Gelfond, Calculus of Finite Differences, (Hindustan Pub. Corp., Delhi, 1971)
6. K.Yu. Dorokhova, S.V. Zuev II Proceedings of young scientists and specialists of the
Belgorod region, 2, (Belgorod: Pub. Comp. Constanta, 2013) (in Russian)

Sdea Ii PDF

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Sdea Ii PDF

Uploaded by

Copyright:

Available Formats

Proceedings of the International Conference

Symmetries, Differential Equations and Applications

Courtesy: Google (Maxwells Fields)

Commemorating the sesquicentennial of the presentation by J C Maxwell

A Dynamical Theory of the Electromagnetic Field

Proceedings of the International Conference

Symmetries, Differential Equations and Applications (SDEA-II)

Journal: International Journal of Modern Physics Conference Series (IJMPCS)

Integrable systems with unitary invariance from non-streching geometric curve

A review on symmetries for certain Aedes Aegypti Models

Lie symmetry analysis of some time fractional partial differential equations

What symmetries can do for you

Higher dimensional systems of differential equations obtainable by iterative use of

viii) Geometrization of Lie and Noether symmetries and applications

Stochastic symmetries of Wick type stochastic ordinary differential equations

A finite differences approach to finding the exact solutions of differential equations

International Journal of Modern Physics: Conference Series

Sesquicentennial of the Presentation by James Clerk Maxwell of his

Asghar Qadir & David Mason

on thermodynamics. Tait was T0 and William Thomson, Lord Kelvin,who became

Asghar Qadir & David Mason

Tiles laid by Maxwell on the floor of the entrance hall to Glenlair.

Asghar Qadir & David Mason

Asghar Qadir & David Mason

Letter written by Maxwell on improvement to his Glenlair estate.

Asghar Qadir & David Mason

Asghar Qadir & David Mason

verse is from Cato: A Tragedy (1713) by Joseph Addison (1672 - 1719).

Asghar Qadir & David Mason

which is a constant (called the Boltzmann constant, normally denoted by k) times

fp (p) = (2mkT )3/2 ep /2mkT .

and thence the Maxwell energy distribution

Asghar Qadir & David Mason

Asghar Qadir & David Mason

Asghar Qadir & David Mason

namely Statistical Mechanics. It is worth mentioning one aspect of the introduction

I do not approve of the plan of a physical society considered as an instrument for

Asghar Qadir & David Mason

International Journal of Modern Physics: Conference Series

INTEGRABLE SYSTEMS WITH UNITARY INVARIANCE FROM

Stephen C. Anco & Esmaeel Asadi & Asieh Dogonchi

Integrable systems in the Hermitian symmetric space Sp(n)/U (n)

A Hermitian symmetric space is a Riemannian symmetric space that possesses a

while the remaining Lie brackets

Stephen C. Anco & Esmaeel Asadi & Asieh Dogonchi

obey the general relations (1).

The matrix representation of g = sp(n) is given as

in which A, B gl(n, C). The matrix representation of the vector spaces m =

Integrable systems in the Hermitian symmetric space Sp(n)/U (n)

(3) The restriction of Cartan-Killing form on sp(n) to m = sp(n)/u(n) is a

(4) The vector space m = sp(n)/u(n) is of dimension n2 + n and of rank n. The

where R is a normalization constant. We choose this constant so that e has unit

In the following lemma, we give an explicit matrix representation of the perp

Stephen C. Anco & Esmaeel Asadi & Asieh Dogonchi

(1) The matrix representation of mk and m in m = sp(n)/u(n) is given as

in which c iR, c Cn1 , and Ck u(n 1).

In particular the linear operator ad(e) gives an isomorphism of m and h :

To write out the explicit Lie bracket relations on m = mk m and h = hk h ,

is the Hermitian inner product. Also, we define

Integrable systems in the Hermitian symmetric space Sp(n)/U (n)

[mk , hk ] =[(ak , Ak ), (Ck )] = (0, Ak Ck Ck Ak ) mk

[hk , hk ] =[(Ck1 ), (Ck2 )] = (Ck1 Ck2 Ck2 Ck1 ) hk