You are on page 1of 8

Photoelectric Effect

Advanced Lab: Photoelectric Effect


Conner Herndon and Nick Lucas
(Dated: 3 February 2015)
Following the discovery that light is electromagnetic radiation, physicists in the late 19th and early 20th
century discovered light can interact with metals to expel photoelectrons. The emission of electrons was
then found dependent not on intensity of light, but wavelength. Einstein, using Planck’s concept of quanta,
explained the effect using a corpuscular theory of light with quantized energy. In this paper we review the
photoelectric experiment using both the traditional apparatus and modern LEDs and apply Einstein’s Nobel-
winning theory of photons. We receive values for Planck’s constant over electronic charge of 3.74029 · 10−15
eV·s for the traditional experiment and 6.9567 · 10−15 eV·s for the modern experiment as compared with the
actual value 4.135667516 · 10−15 eV·s.

I. BACKGROUND
which we now know as Planck’s constant.4,5 Although
James Clerk Maxwell published his electromagnetic the- Planck had introduced the constant h, he did not
ory of light in 1864. In his Treatise published in 1873, fully recognize its importance.6 In 1905, Einstein used
Maxwell states, Planck’s theoretical tool to explain the photoelectric
effect,7 and his explanation of the photoelectric effect in
“When light is emitted, a certain amount of terms of a corpuscular theory of light earned him the
energy is expended by the luminous body, 1921 Nobel prize in physics. Light was then found to be
and if the light is absorbed by another body, a particle, in the form of a photon, with energy E = hν,
this body becomes heated, shewing [sic] that where h is Planck’s constant and ν the frequency.
it has received energy from without. [...] Ac-
cording to the theory of emission, the trans- The explanation of the photoelectric effect in terms of
mission of energy is effected by the actual photons broke ground leading to the creation of the
transference of light-corpuscules from the lu- old quantum theory. Wave particle duality, although
minous to the illuminated body, carrying a highly unintuitive concept in classical mechanics,
with them their kinetic energy, together with proved necessary to explain experiment and thus became
any other kind of energy of which they may a staple in quantum theory. Later experiments such
be the receptacles.”1 (Ch. XX). as Compton scattering8 and electron diffraction9 also
proved the necessity of wave particle duality in modern
Physicists thus realized the strong connection between physics.
the newly established electromagnetic theory and light

with c = 1/ µ0 0 , where µ0 and 0 are the permeability
and permittivity of free space, respectively, and c the
speed of light. Following Maxwell’s work in electromag- II. THEORY
netism, Hertz and Hallwachs found that shining light on
a metallic surface released electrons with certain kinetic A. Traditional
energies.2,3
The interesting phenomenon observed in the photoelec- Classically, one would expect the kinetic energy delivered
tric effect was that electron emission was not dependent to metallic electrons to be proportional to the intensity
on intensity of light but rather the frequency. If light of light shined on the surface. The amount of kinetic
were a wave, as determined by electromagnetic theory, energy delivered to an electron is equal to the energy
then the electron emission should be dependent on delivered by the incoming minus the energy of the bound
intensity of light. In another branch of physics, the state, or
thermodynamics of black body radiation was in a state
of turmoil over the ultraviolet catastrophe. When
radiation was assumed to be continuous, an ideal black T = Einc − ϕ0 , (2)
body would emit radiation of infinite power. Max Planck
posited a combinatorial approach to finding the energy
density emitted by a blackbody. His combinatorics where T is the kinetic energy of the photoelectron, Einc
required a constant the energy of the incoming radiation, and ϕ0 the “work
function”, or amount of energy necessary to break the
electron free. Suppose the surface upon which we are
h ' 6.626 · 10−34 m2 kg/s, (1) shining our light is one side of a capacitor within a
Photoelectric Effect 2

circuit. Then for a photoelectron to escape, it must have


more energy than the product of its charge with the stop-
ping potential Vs . Then we can modify equation 2 to read

eVs = Einc − ϕ0 (3)

for electrons at the stopping threshold. We can then


find the linear relationship for Vs by dividing through
by the electron charge e

FIG. 1. Two-band energy struc-


Einc
Vs (Einc ) = − ϕ00 , (4) ture.
e
energy states are filled, and the number of filled states
where we have scaled ϕ00 ≡ ϕ0 /e. At this point, Einstein
corresponds with the number of electrons in the system.
assumed that the energy of light was proportional to
The ground state of this system with N particles has
its frequency and not its intensity. Thus Einc ∝ ν, and
the lowest N single-particle states occupied in a Fermi
therefore
“sea” state:

Vs ∝ ν, (5)
ĉ†n,k,σ |0i ,
Y
|FSi = (9)
n,k,σ
n,k <µ
where ν is the frequency of the incoming light. Let h,
Planck’s constant, be the constant of proportionality for
where µ, the chemical potential, is chosen such that
the energy of incoming light to its frequency, or
|FSi contains N electrons. This ground state has the
following property
Einc = hν. (6)
D E
FS ĉ†nkσ ĉnkσ FS = Θ (µ − nk ) , (10)

Then the stopping voltage will have the relationship
where Θ is the Heaviside function. Consider the two-
dVs h band model with an energy gap between the bands and
= . (7)
dν e µ inside the gap as shown in figure 1.

The stopping voltage is linearly dependent on frequency


with a slope of h/e. If we know the electronic charge Then the ground state is
e, we can determine Planck’s constant h by varying the
frequency of light.
ĉ†kσ |0i
Y
|GSi = (11)
k,σ
B. Modern
For the following it is useful to discuss the combination
The modern experiment uses LEDs (light emitting Ĥ − µN̂ , where N̂ is the total number of particles, which
diodes) which rely on semiconductor technology. To appears in the Boltzmann weight of the grand canonical
understand the general process in a semiconductor, ensemble. To express this quantity in second quantized
we will discuss here the transition energies in a band notation, Ĥ − µN̂ becomes
structure. In solids, we have several bands with index n
without two-body interactions. The Hamiltonian is then
(nk − µ) ĉ†nkσ ĉnkσ .
X
(12)
nkσ
n,k ĉ†n,k,σ ĉn,k,σ .
X
Ĥ = (8)
n,k,σ It is useful to consider the tipping point between filled
and unfilled levels as the vacuum state and write all
for some energy n,k with spin states σ and creation/an- states in terms of this state. To create a state above this
nihilation operators ĉ† and ĉ, respectively. All lowest level, we create a particle. To lower the state, we create
Photoelectric Effect 3

a “hole”. The new vacuum state is described the two-band energy structure in terms of
new variables for hole creation/annihilation. The band
gap ground state has an energy difference that may be
0̃ = |GSi (13) expressed in terms of a diffusion potential VD such that

and “hole operators” describing the removal/addition of eVD = Eg (21)


electrons from the upper band (band 1) are defined as
Using Einstein’s assumption of the energy of light
ĥ†k↑ ≡ ĉ1,−k,↓ , creation of spin-down hole E = hν, we have the relation
ĥ†k↓ ≡ −ĉ1,−k,↓ , creation of spin-up hole
(14) eVD = hν, (22)
ĥk,↑ = ĉ†1,−k,↓
ĥk,↓ = −ĉ†1,−k,↑ .
or

The commutation relations are then


dVD h
= , (23)
n o dν e
ĥ†kσ , ĥk0 σ0 = δkk0 δσσ0 . (15)
just as before in the traditional experiment version.

We now want to check how this new state functions in


terms of the newly defined variables. Since all states in III. EXPERIMENT
band 1 are occupied,
We performed the experiment using two different ap-
proaches: traditional and modern. The traditional ex-
ĉ†1kσ

0̃ = 0, (16) periment is of the variety used during the discovery of the
photoelectric effect, circa 1887. The modern experiment
uses light emitting diodes (LEDs). Although modern ap-
then
proaches yield cleaner results, the traditional apparatus
proves highly instructive in how physics was done before

ĥkσ 0̃ = 0 (17) current technology.

and also A. Traditional

The traditional apparatus featured a mercury lamp as a


ĉ1kσ 0̃ = 0. (18) light source. For different wavelengths, a colored filter
was placed in front of the light. In the path of the light
were two lenses, an adjustable slit, and an interference fil-
Therefore, 0̃ is the vacuum state in the new variables. ter as shown in figure 2. The resulting light is captured in
Use now a detector. This detector contains a vacuum tube, within
which a wire is housed. This wire is biased by a retard-
ing voltage that acts as the stopping voltage described in
ĉ†1kσ ĉ1kσ = ĥ−k,−σ ĥ†−k,−σ
(19) equation 3. This voltage will only allow photoelectrons
= 1 − ĥ†−k,−σ ĥ−k,−σ with sufficient kinetic energy to pass.

and assume that 1k = 1,−k for simplicity. Then our


combination becomes We collected current vs. voltage data for 5 different
wavelengths: 465nm, 405nm, 436nm, 546nm, and 578nm.
Figure 3 features one such so-called IV curve for 578nm
(µ − 1k ) ĥ†kσ ĥkσ +(1k − µ) ĉ†2kσ ĉ2kσ +const.
X
Ĥ−µN̂ = wavelength light. There are 10 total runs shown, and the
kσ data are quite noisy due to rudimentary equipment and
(20) light pollution in the room.

The constant on the right is the energy of the vacuum


state (the completely filled Fermi sea). We have thus
Photoelectric Effect 4

FIG. 2. Traditional apparatus. Mercury lamp


emits light through a lens, adjustable slit, inter-
ference filter, then another lens before reaching the
detector.
FIG. 4. IV curve for 6 different diodes. Black:
950nm, red: 660nm, yellow: 590nm, maroon:
630nm, green: 525nm, and blue: 480nm.

FIG. 3. IV curve for 578nm wavelength scan with FIG. 5. Stopping voltage versus frequency. The lin-
traditional apparatus. 10 total runs are shown. ear regression yields dVs /dν = 4.135667516 · 10−15 .

B. Modern and to retrieve useful information from the dataset, we


need to have a much smoother function. To smooth
The modern experiment features a light emitting diode the noise, we first convoluted each experimental data
(LED) instead of mercury lamp. For the LED to light set with a normalized, 20-point Gaussian window, and
up, there must exist a sufficient voltage for the electrons the resulting plot is shown in figure 7. Following the
to jump across the band gap. We ran different voltages convolution, averages of currents were taken within 0.3
across the different color diodes and measured the corre- V window bins, which is shown in figure 8.
sponding current. We used LED wavelengths of 950nm,
660nm, 630nm, 591nm, 525nm, and 472nm. The result- The important portion of the scan is the when the
ing IV curves are shown in figure 4. To determine the plot begins to curve in a section known as the “knee”.
diffusion potential (the potential that prevents electrons To determine the voltage of interest, we need to find
or holes from jumping), we extrapolate the linear regime when the IV curve begins to curve, which would be
to the x-intercept. Figure 12 shows the linear regression when the second derivative reaches zero. We then found
over the infrared 950nm diode IV curve. the second derivative of the resulting curve shown in
figure 9. Figure 10 features a zoomed in version of figure
9 to better see the zero intersect. This zero intersection
is recorded for all wavelengths. After performing this
noise reduction and finding the extrema of IV curve
first derivatives, we may plot these important points for
stopping voltages versus frequency as shown in figure 5.
IV. ANALYSIS

A. Traditional Fitting a linear regression to these data yield a slope of

The traditional experiment features a great noise dVs


= 3.74029 · 10−15 eV · s. (24)
contribution due to the bulb, lenses, and measuring dν
apparatus. The data from the 578nm scan are shown in
figure 3. As one may see, the data are not very smooth, The actual value we are searching for is
Photoelectric Effect 5

For the modern experiment, we used nearly pris-


tine circuitry, LEDs, and measurement equipment. We
h
= 4.135667516(19) · 10−15 eV · s (25) expect the error contribution from these devices to be
e much less than the traditional experiment. We received
a value for h/e with an error of ' 68%, quite large
according to the National Institute for Standards and compared to the traditional set-up. The error in this
Technology. Then our traditional experiment yields situation may be attributed to the linear regressions
9.56% error. used and the heating of the LEDs. The diffusion voltage
for each LED is temperature dependent, and some LEDs
were left on longer than others.
B. Modern

For the modern experiment, the diffusion voltages were VII. REFERENCES
found for the various wavelengths and plotted as shown
in figure 6. From our linear regression, we found a slope 1 J. C. Maxwell, A Treatise on Electricity and Magnetism, Vol. 2
of (MacMillan and Co., 1873).
2 Annual Report of the Board of Regents of the Smithsonian

(Smithsonian Institution, 1913).


dVD 3 H. Hertz, “Ueber den einfluss des ultravioletten lichtes auf die
= 6.9567 · 10−15 eV · s. (26) electrische entladung,” Annalen der Physik 8 (1887).
dν 4 M. Planck, “Zur theorie des gesetzes der energieverteilung im

normalspektrum,” Verhandlungen der Deutschen Physikalischen


Our error is then 68.2%. Gesellschaft 2, 237–245 (1900).
5 M. Jammer, The Conceptual Development of Quantum Mechanics

(McGraw-Hill, Inc., 1966).


6 M. Planck, “Kurz zusammengefasst kann ich die ganze tat als
V. CONCLUSION einen akt der verzweiflung bezeichnen,” (1931).
7 A. Einstein, “On a heuristic point of view concerning the produc-

tion and transformation of light,” Annalen der Physik , 132–148


Classical intuition provides an erroneous foundation for
(1905).
intrinsically quantum phenomena. At the turn of the 19th 8 A. Compton, “A quantum theory of the scattering of x-rays by
century, physicists were just discovering the importance light elements,” Phys. Rev. 483 (1923).
9 Davisson and Germer, “Diffraction of electrons by a crystal of
and necessity of Planck’s constant. Maxwell’s equations
explain light as an electromagnetic wave; however, the nickel,” Phys. Rev. 30 (1927).
photoelectric effect proves light has corpuscular nature
as well with energy proportional to the frequency and
not intensity. The two experiments outlined in this pa-
per derive Planck’s constant by traditional and modern
methods. The traditional method relies on shining a light
on a surface to emit photoelectrons, and the determina-
tion of the stopping voltage across various frequencies
provides the relation dV /dν = h/e. The modern exper-
iment uses LEDs to reverse this process and emit light
based on electrons passing a pn-junction. Both methods
provide Planck’s constant within reasonable error. From
the traditional experiment we received a h/e quotient of
3.74029 · 10−15 eV · s. For the modern experiment with
the LED we received h/e = 6.9567 · 10−15 eV · s.

VI. UNCERTAINTY

The traditional experiment suffers from many sources


of error. Since we are working in a lit room and we’re
trying to measure photons, we have considerable light
pollution. Errors also could come from the arrangement
of the lens system. Once our data were collected, error
also is introduced through our linear regression fitting.
Despite these potential error contributions, our resulting
value for h/e is within 10% of the actual value.
Photoelectric Effect 6

FIG. 6. Diffusion voltages versus frequency.

FIG. 7. Noise reduced IV curve for 578nm wavelength scan


with traditional apparatus. 10 total runs are shown.
Photoelectric Effect 7

FIG. 8. Averaged, noise reduced IV curve for 578nm wavelength scan with traditional apparatus. 10 total runs are shown.

FIG. 9. Second derivative of averaged, noise reduced IV curve for 578nm wavelength scan with traditional apparatus. 10 total
runs are shown.
Photoelectric Effect 8

FIG. 10. Zoomed in on second derivative of averaged, noise reduced IV curve for 578nm wavelength scan with traditional
apparatus. 10 total runs are shown.

FIG. 11. IV curve for modern experiment with 525nm LED. Two scans are shown.

FIG. 12. Linear regression over 950nm LED IV curve.

You might also like