You are on page 1of 173

The Diodicity Mechanism of

Tesla-Type No-Moving-Parts Valves

Ronald Louis Bardell

A dissertation submitted in partial fulfillment


of the requirements for the degree of

Doctor of Philosophy

University of Washington
2000

Program Authorized to Offer Degree: Mechanical Engineering

University of Washington
Graduate School

This is to certify that I have examined this copy of a doctoral dissertation by


Ronald Louis Bardell
and have found that it is complete and satisfactory in all respects,
and that any and all revisions required by the final
examining committee have been made.

Chairperson of the Supervisory Committee::

Fred K Forster

Reading Committee:

Fred K Forster
James J Riley
Karl F Bhringer

Date:

In presenting this dissertation in partial fulfillment of the requirements for the Doctorial
degree at the University of Washington, I agree that the Library shall make its copies
freely available for inspection. I further agree that extensive copying of the dissertation
is allowable only for scholary purposes, consistent with fair use as prescribed in the U.S.
Copyright Law. Requests for copying or reproduction of this dissertation may be referred to
Bell and Howell Information and Learning, 300 North Zeeb Road, Ann Arbor, MI 481061346, or to the author.

Signature

Date

University of Washington
Abstract

The Diodicity Mechanism of


Tesla-Type No-Moving-Parts Valves
by Ronald Louis Bardell
Chairperson of the Supervisory Committee:
Professor Fred K Forster
Mechanical Engineering
Microvalves are needed for micropumps that can move particulate-laden fluids in MEMS
(Micro-Electro-Mechanical Systems) devices. No-moving-parts (NMP) valves are especially qualified for this task, yet no knowledge of the mechanism that creates valve diodicity
in the low-Reynolds number regime so characteristic of microfluidics has been available.
As a result, the design of NMP valves has relied on the "build & test" method.
We have developed a numerical method that accurately predicts the diodicity and reveals the diodicity mechanism of NMP microvalves by combining analysis of field variables from numerical valve simulations with analysis of momentum and kinetic-energy
conservation in regional control-volumes. The numerical method is carefully validated
by comparison with known analytical solutions and with experimental data from physical
realizations of two distinct designs of Tesla-type NMP valves. It predicts their diodicity
within 4% of measured values. It reveals their low-Reynolds-number diodicity mechanism
as the viscous dissipation surrounding laminar jets that have flow-direction-dependent locations and orientations. This diodicity mechanism is dominated by viscous forces, unlike
the high-Reynolds-number mechanism of macro-scale valves that is solely due to inertial
forces. Understanding of the diodicity mechanism is encapsulated in design guidelines
for laying out valve geometry and is demonstrated by developing an enhanced-diodicity
valve design solely by following these guidelines. The numerical method predicts with
95% confidence that the diodicity of this new design is a 27-47% improvement over the
original design. Clearly, knowledge of the low-Reynolds-number diodicity mechanism in
Tesla-type NMP valves leads directly to an improved design.

TABLE OF CONTENTS

List of Figures

List of Tables

xii

Chapter 1:
1.1
1.2

1.3

Introduction

Nature and Scope of the Problem . . . . . . . . . . . . . . . . . . . . . . .


Background for NMP Microvalves . . . . . . . . . . . . . . . . . . . . . .

1
2

1.2.1

Diodicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.2.2
1.2.3

Laminar vs. Turbulent Flow . . . . . . . . . . . . . . . . . . . . .


Modeling the Dynamics of Microfluidic Systems . . . . . . . . . .

3
5

1.2.4

Steady-State Response . . . . . . . . . . . . . . . . . . . . . . . .

1.2.5 Transient Response . . . . . . . . . . . . . . . . . . . . . . . . . 7


Background for Tesla-Type NMP Valves . . . . . . . . . . . . . . . . . . . 11
1.3.1

Prior Research on the Diodicity Mechanism in Tesla-Type NMP


Valves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1.4

Research Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

1.5

Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

Chapter 2:

The Governing Equations

19

2.1

Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

2.2

Momentum Conservation . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.2.1 Dimensional Analysis . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2.2

2.3

Kinetic Energy Conservation . . . . . . . . . . . . . . . . . . . . . . . . . 22


2.3.1 Dimensional Analysis . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3.2

2.4

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.4.1 Momentum Perspective . . . . . . . . . . . . . . . . . . . . . . . . 27
2.4.2

Kinetic-Energy Perspective . . . . . . . . . . . . . . . . . . . . . . 28
i

Chapter 3:

The Numerical Method

29

3.1

Methods to Quantify the Diodicity Mechanism . . . . . . . . . . . . . . . 30

3.2

Numerical Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

3.3

Physical Grid Layout and Grid Independence . . . . . . . . . . . . . . . . 31

3.4

Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.5

Characteristic Parameters for Nondimensionalization . . . . . . . . . . . . 33

3.6

Calculation of the Terms in the Conservation Equations . . . . . . . . . . . 34

3.7

Verification via Analytical Solution for a 2-D Slot Flow . . . . . . . . . . . 35

Chapter 4:
4.1

4.2

4.3

Validation of the Numerical Method in Steady Flow

36

Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.1.1

Experimental Methods . . . . . . . . . . . . . . . . . . . . . . . . 36

4.1.2

Numerical Methods . . . . . . . . . . . . . . . . . . . . . . . . . . 39

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.2.1

Volume Flow-Rate . . . . . . . . . . . . . . . . . . . . . . . . . . 40

4.2.2

Diodicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

4.2.3

Prediction Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . 44

4.2.4

Evidence of Laminar Flow . . . . . . . . . . . . . . . . . . . . . . 47

Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

Chapter 5:

Validation of the Numerical Method in Transient Flow

51

5.1

Harmonic Response of a 2-D Slot . . . . . . . . . . . . . . . . . . . . . . 51

5.2

Harmonic Response of an NMP Valve . . . . . . . . . . . . . . . . . . . . 53

5.3

Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

Chapter 6:

Diodicity Mechanism of T45A Valve

55

6.1

Simulation Methods and Conditions . . . . . . . . . . . . . . . . . . . . . 55

6.2

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

6.3

6.2.1

Velocity Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

6.2.2

Pressure Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

6.2.3

Energy-Dissipation Field . . . . . . . . . . . . . . . . . . . . . . . 60

6.2.4

Momentum Conservation . . . . . . . . . . . . . . . . . . . . . . . 64

6.2.5

Kinetic-Energy Conservation . . . . . . . . . . . . . . . . . . . . . 69

Discussion

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
ii

Chapter 7:

Diodicity Mechanism of T45C Valve

77

7.1

Simulation Methods and Conditions . . . . . . . . . . . . . . . . . . . . . 77

7.2

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

7.3

7.2.1

Velocity Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

7.2.2

Pressure Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

7.2.3

Energy Dissipation Field . . . . . . . . . . . . . . . . . . . . . . . 82

7.2.4

Momentum Conservation . . . . . . . . . . . . . . . . . . . . . . . 86

7.2.5

Kinetic-Energy Conservation . . . . . . . . . . . . . . . . . . . . . 92

Discussion

Chapter 8:

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

Valve Design Guidelines

99

8.1

Preliminary Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

8.2

How To Lay Out a Tesla-Type Valve . . . . . . . . . . . . . . . . . . . . . 101

Chapter 9:

Enhanced Diodicity Mechanism of T45A-2 Valve

106

9.1

Simulation Methods and Conditions . . . . . . . . . . . . . . . . . . . . . 106

9.2

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

9.3

9.2.1

Velocity Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

9.2.2

Pressure Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

9.2.3

Energy Dissipation Field . . . . . . . . . . . . . . . . . . . . . . . 111

9.2.4

Momentum Conservation . . . . . . . . . . . . . . . . . . . . . . . 115

9.2.5

Kinetic-Energy Conservation . . . . . . . . . . . . . . . . . . . . . 118

Discussion

Chapter 10:

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

Conclusions

124

10.1 Develop a Numerical Method to Reveal the Diodicity Mechanism . . . . . 124


10.1.1 Verify Mathematically Correctness . . . . . . . . . . . . . . . . . . 125
10.1.2 Verify Steady-Flow Response Predictions . . . . . . . . . . . . . . 125
10.1.3 Verify Diodicity Prediction Accuracy . . . . . . . . . . . . . . . . 126
10.1.4 Verify Transient-Flow Response Predictions . . . . . . . . . . . . . 127
10.1.5 Reveal the Diodicity Mechanism in Low Reynolds Number Flow
in Tesla-Type NMP Valves . . . . . . . . . . . . . . . . . . . . . 127
10.2 The Low-Reynolds-Number Diodicity Mechanism is Dominated by Viscous Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
iii

10.3 Demonstrate Knowledge of the Diodicity Mechanism . . . . . . . . . . . . 129


10.3.1 Develop Valve Design Guidelines . . . . . . . . . . . . . . . . . . 129
10.3.2 Demonstrate the Effectiveness of the Guidelines . . . . . . . . . . 130
10.4 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
Bibliography
Appendix A:

131
Valve Resistance Modeling

134

Appendix B:
Valve Inertance Modeling
137
B.1 Step Response of a 2-D Slot . . . . . . . . . . . . . . . . . . . . . . . . . 137
B.2 Step Response of a 2-D Slot . . . . . . . . . . . . . . . . . . . . . . . . . 139
B.3 Step Response of an NMP Valve . . . . . . . . . . . . . . . . . . . . . . . 139
Appendix C:

Series Solution for Starting Flow in a Slot

141

Appendix D:

Diodicity From a Ratio of Flow Rates

145

Appendix E:

Valve Diodicity Measurements

147

Appendix F:

Valve Layout Points

149

F.1
F.2
F.3

T45A Valve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149


T45C Valve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
T45A-2 Valve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

iv

LIST OF FIGURES

1.1

Flow separation and recirculation at Re

0 01 based on cavity depth (Taneda




1979). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.2

Circuit diagram for linear model of complete micropump.

. . . . . . . . .

1.3

Velocity vector field shows center flow out-of-phase with flow nearer the
wall. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.4

Closeup photo of a single-element, Tesla-type (T45A) outlet valve connecting the pump chamber on the right and the outlet port on the upper left. The
white object is a human hair. . . . . . . . . . . . . . . . . . . . . . . . . . 11

1.5

Control volume for momentum analysis of F. Paul of reverse flow through


the T-junction of a Tesla-type valve. Reverse flow is from right to left. . . . 13

3.1

Typical residuals plot showing termination of interations and procession to


next time step, controlled by USRCVG . F. Note that all residuals have ceased
changing before a new time step begins: first the Mass residual, then the W
velocity residual, and finally the U and V velocity residuals. . . . . . . . . 31

4.1

T45A valve pressure drop vs. Reynolds number based on the hydraulic
diameters. Experimental and numerical data are shown as symbols. The
curves are the the fitted power-law relation (Eq. 4.1). The legends refer to
each valve name and its etch depth; the numerical simulations are marked
sim. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

4.2

T45C valve pressure drop vs. Reynolds number based on the hydraulic
diameters. Experimental and numerical data are shown as symbols. The
curves are the the fitted power-law relation (Eq. 4.1). The legends refer to
each valve name and its etch depth; the numerical simulations are marked
sim. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
v

4.3

Deep T45C valve pressure drop vs. Reynolds number based on the hydraulic diameters. Experimental and numerical data are shown as symbols.
The curves are the the fitted power-law relation (Eq. 4.1). The legends
refer to each valve name and its etch depth; the numerical simulations are
marked sim. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

4.4

Diodicity following Eq. 1.1 versus Reynolds number of the valves in the
T45A Test Group. The symbols are numerical and experimental data.
The curves are the ratio of the fitted power-law relations (Eq. 4.1) for the
reverse and forward flow directions. Legends refer to test name and etch
depth; numerical simulations are marked sim. . . . . . . . . . . . . . . 44

4.5

Diodicity following Eq. 1.1 versus Reynolds number of the valves in the
T45C Test Group. The symbols are numerical and experimental data.
The curves are the ratio of the fitted power-law relations (Eq. 4.1) for the
reverse and forward flow directions. Legends refer to test name and etch
depth; numerical simulations are marked sim. . . . . . . . . . . . . . . 45

4.6

Diodicity following Eq. 1.1 versus Reynolds number of the valves in the
Deep T45C Test Group. The symbols are numerical and experimental
data. The curves are the ratio of the fitted power-law relations (Eq. 4.1)
for the reverse and forward flow directions. Legends refer to test name and
etch depth; numerical simulations are marked sim. . . . . . . . . . . . . 45

4.7

Diodicity prediction error of the numerical method for all 11 tests of the
three test groups: the T45A Test Group, the T45C Test Group, and the
Deep T45C Test Group. . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

4.8

Pressure drop versus Reynolds number based on hydraulic diameter. Symbols are experimental and numerical data; curves are the fitted power-law
relation (Eq. 4.1). Legends refer to test name and etch depth; numerical
simulations are marked sim. . . . . . . . . . . . . . . . . . . . . . . . . 48

5.1

Comparison of the nondimensional velocity profiles in a slot with oscillating flow, 8 6 , from the numerical method (symbols) and the exact
solution (lines). The centerline of the slot is at zero slot height and the slot


wall is at slot height = 1. The legends note the phase of each profile with
respect to the applied pressure difference, a cosine function. . . . . . . . . 53
vi

5.2

Terms in the conservation of kinetic-energy equation (Eq. 2.12) for the


entire T45A valve over two cycles in a harmonic response simulation at
2.818 kHz, including: the energy dissipation rate (PHI), viscous work rate
(VWR), energy flux rate (EFR), pressure work rate (PWR), and the transient kinetic-energy (TKE). The dissipation and pressure work are dominant. 54

6.1

Division of the T45A valve into regional control volumes. . . . . . . . . . 56

6.2

Forward-flow velocity field on the centerplane of a single-element, Teslatype T45A valve with a volume flow rate of 3710 l/min corresponding to
Re=528 based on the hydraulic diameter of the main channel. One dimensionless unit equals 10 m/s. . . . . . . . . . . . . . . . . . . . . . . . . . . 57

6.3

Reverse-flow velocity field on the centerplane of a single-element, Teslatype T45A valve with a volume flow rate of 3710 l/min corresponding to
Re=528 based on the hydraulic diameter of the main channel. One dimensionless unit equals 10 m/s. . . . . . . . . . . . . . . . . . . . . . . . . . . 59

6.4

Pressure field [atm] on the centerplane of a single-element, Tesla-type T45A


valve with a volume flow rate of 3710 l/min corresponding to Re=528
based on the hydraulic diameter of the main channel. . . . . . . . . . . . . 61

6.5

Base 10 logarithm of the energy dissipation rate in forward flow on the


centerplane of a single-element, Tesla-type T45A valve with a volume flow
rate of 3710 l/min corresponding to Re=528 based on the hydraulic diameter of the main channel. One dimensionless unit equals 14 mW. . . . . . . 62

6.6

Base 10 logarithm of the energy dissipation rate in reverse flow on the centerplane of a single-element, Tesla-type T45A valve with a volume flow
rate of 3710 l/min corresponding to Re=528 based on the hydraulic diameter of the main channel. One dimensionless unit equals 14 mW. . . . . . . 63

6.7

Force vector terms in the integral form of the momentum conservation


equation for the T45A valve. Net pressure force and net momentum flux
into a control volume are positive. Viscous force is applied on the fluid
by the wall. X-vectors to the right and Y-vectors upward are positive and
consistent with the valve layout in Fig. 6.1 including the numbering of the
control volumes (blocks). . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
vii

6.8

Vector magnitudes of the terms in the momentum conservation equation


for the T45A valve. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

6.9

Magnitude of the pressure work rate and the energy flux rate terms in
the kinetic-energy conservation equation from the T45A valve simulations
with a volume flow rate of 3710 l/min corresponding to Re=528 based on
the hydraulic diameter of the main channel. . . . . . . . . . . . . . . . . . 70

6.10 Magnitude of the energy dissipation rate and the viscous work rate terms in
the kinetic-energy conservation equation from the T45A valve simulations
with a volume flow rate of 3710 l/min corresponding to Re=528 based on
the hydraulic diameter of the main channel. . . . . . . . . . . . . . . . . . 72
6.11 Magnitude of the terms in the kinetic-energy conservation equation from
the T45A valve simulations with a volume flow rate of 3710 l/min corresponding to Re=528 based on the hydraulic diameter of the main channel. . 73
7.1

Division of the T45C valve into regional control volumes.

. . . . . . . . . 78

7.2

Forward-flow velocity field on the centerplane of a single-element, Teslatype T45C valve with a volume flow rate of 3640 l/min corresponding to
Re=519 based on the hydraulic diameter of the main channel. One dimensionless unit equals 10 m/s. . . . . . . . . . . . . . . . . . . . . . . . . . . 79

7.3

Reverse-flow velocity field on the centerplane of a single-element, Teslatype T45C valve with a volume flow rate of 3640 l/min corresponding to
Re=519 based on the hydraulic diameter of the main channel. One dimensionless unit equals 10 m/s. . . . . . . . . . . . . . . . . . . . . . . . . . . 81

7.4

Pressure field [atm] on the centerplane of a single-element, Tesla-type T45C


valve with a volume flow rate of 3640 l/min corresponding to Re=519
based on the hydraulic diameter of the main channel. . . . . . . . . . . . . 83

7.5

Base 10 logarithm of the energy dissipation rate in forward flow on the


centerplane of a single-element, Tesla-type T45C valve with a volume flow
rate of 3640 l/min corresponding to Re=519 based on the hydraulic diameter of the main channel. One dimensionless unit equals 14 mW. . . . . . . 84
viii

7.6

Base 10 logarithm of the energy dissipation rate in reverse flow on the centerplane of a single-element, Tesla-type T45C valve with a volume flow rate
of 3640 l/min corresponding to Re=519 based on the hydraulic diameter
of the main channel. One dimensionless unit equals 14 mW. . . . . . . . . 85

7.7

Force vector terms in the integral form of the momentum conservation


equation. Net pressure force and net momentum flux into a control volume
are positive. Viscous force is applied on the fluid by the wall. X-vectors to
the right and Y-vectors upward are positive and consistent with the valve
layout in Fig. 7.1 including the numbering of the control volumes (blocks).

87

7.8

Vector magnitudes of the terms in the momentum conservation equation


for the T45C valve. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

7.9

Magnitude of the pressure work rate and the energy flux rate terms in the
kinetic-energy conservation equation from the T45C valve simulations with
a volume flow rate of 3640 l/min corresponding to Re=519 based on the
hydraulic diameter of the main channel. . . . . . . . . . . . . . . . . . . . 93

7.10 Magnitude of the energy dissipation rate and the viscous work rate terms in
the kinetic-energy conservation equation from the T45C valve simulations
with a volume flow rate of 3640 l/min corresponding to Re=519 based on
the hydraulic diameter of the main channel. . . . . . . . . . . . . . . . . . 94
7.11 Magnitude of the terms in the kinetic-energy conservation equation from
the T45C valve simulations with a volume flow rate of 3640 l/min corresponding to Re=519 based on the hydraulic diameter of the main channel. . 95
8.1

Sketch of generic Tesla-type NMP valve with dimensioning per design


rules for high diodicity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

8.2

Overlay of the T45A (solid red lines) and the T45C (dashed blue lines)
showing the variation in path lengths: inlet channel, outlet channel, and
side channel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

9.1

Division of the T45A-2 valve into regional control volumes. . . . . . . . . 107

9.2

Forward-flow velocity field on the centerplane of a single-element, Teslatype T45A-2 valve with a volume flow rate of 2987 l/min corresponding
to Re=500 based on the hydraulic diameter of the main channel. One dimensionless unit equals 10 m/s. . . . . . . . . . . . . . . . . . . . . . . . . 108
ix

9.3

Reverse-flow velocity field on the centerplane of a single-element, Teslatype T45A-2 valve with a volume flow rate of 2987 l/min corresponding
to Re=500 based on the hydraulic diameter of the main channel. One dimensionless unit equals 10 m/s. . . . . . . . . . . . . . . . . . . . . . . . . 110

9.4

Pressure field [atm] on the centerplane of a single-element, Tesla-type T45A2 valve with a volume flow rate of 2987l/min corresponding to Re=500
based on the hydraulic diameter of the main channel. . . . . . . . . . . . . 112

9.5

Base 10 logarithm of the energy dissipation rate in forward flow on the


centerplane of a single-element, Tesla-type T45A-2 valve with a volume
flow rate of 2987 l/min corresponding to Re=500 based on the hydraulic
diameter of the main channel. One dimensionless unit equals 14 mW. . . . 113

9.6

Base 10 logarithm of the energy dissipation rate in reverse flow on the


centerplane of a single-element, Tesla-type T45A-2 valve with a volume
flow rate of 2987 l/min corresponding to Re=500 based on the hydraulic
diameter of the main channel. One dimensionless unit equals 14 mW. . . . 114

9.7

Force vector terms in the integral form of the momentum conservation


equation. Net pressure force and net momentum flux into a control volume
are positive. Viscous force is applied on the fluid by the wall. X-vectors to
the right and Y-vectors upward are positive and consistent with the T45A-2
valve layout in Fig. 9.1 including the numbering of the control volumes
(blocks). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

9.8

Vector magnitudes of the terms in the momentum conservation equation


for the T45A-2 valve. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

9.9

Magnitude of the pressure work rate and the energy flux rate terms in the
kinetic-energy conservation equation from the T45A-2 valve simulations
with a volume flow rate of 2987 l/min corresponding to Re=500 based on
the hydraulic diameter of the main channel. . . . . . . . . . . . . . . . . . 119

9.10 Magnitude of the energy dissipation rate and the viscous work rate terms
in the kinetic-energy conservation equation from the T45A-2 valve simulations with a volume flow rate of 2987 l/min corresponding to Re=500
based on the hydraulic diameter of the main channel. . . . . . . . . . . . . 120
x

9.11 Magnitude of the terms in the kinetic-energy conservation equation from


the T45A-2 valve simulations with a volume flow rate of 2987 l/min corresponding to Re=500 based on the hydraulic diameter of the main channel. 122
A.1 Local-slope resistance to fluid flow vs. Reynolds number in T45A and
T45C valves from both experiment and numerical simulation following
Eq.A.1, which is based on the fitted power-law relation, Eq.4.1. . . . . . . 135
A.2 Fluid resistance vs. volume flow rate in a typical NMP valve. The timeaverage and average R are approximations of the local-slope R for use in
linear models where a single value is required. Note the similarity to the
characteristic curve of nonlinear friction for an object moving at low Re in
a fluid medium. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
B.1 The predicted volume flow-rate response to an applied pressure difference
in a 2-D slot from the numerical method (symbols) shows good agreement
with the exponential response from the series solution, Eq. C.8. The time
constant Re 2 is also shown. . . . . . . . . . . . . . . . . . . . . . . 140
B.2 Inertance versus Reynolds number from the step-response simulations via
Eq. B.5. Inertance shows some dependence on flow rate, but not on flow
direction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
D.1 Characteristic pressure drop versus volume flow rate curves for reverse and
forward flow in an NMP valve. Diodicity can be derived from either the
pressure-drop ratio or the flow-rate ratio. . . . . . . . . . . . . . . . . . . . 146

xi

LIST OF TABLES

2.1

2.2

Order of magnitude of the integrand in each term in the steady form of the
momentum conservation equation Eq. 2.5 for water in a straight duct and
an NMP valve at various Rep . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Order of magnitude of each term in the kinetic-energy equation, Eq. 2.11
for various Re p and . For water in an NMP valve with DH 100 m,
1


f


10 kHz corresponds to 8


25


. . . . . . . . . . . . . . . . . 27

4.1

Etch depths and deviations from the mean depth of the valves tested in the

4.2

T45A group. The tested devices are representative samples of the set since
the deviations are much less than the measurement accuracy of 6.5%. . . . 38
Etch depths and deviations from the mean depth of the valves tested in the
T45C group. The tested devices are representative samples of the set since
the deviations are much less than the measurement accuracy of 6.5%. . . . 38

4.3

Etch depths and deviations from the mean depth of the valves tested in the
Deep T45C group. The tested devices are representative samples of the set
since the deviations are much less than the measurement accuracy of 6.5%.

4.4

Standard deviation and correlation coefficient of the experimentallymeasured P with respect to the power-law fit of P versus volume flowrate following Eq. 4.1. The measurements in the forward and reverse flow

38

r2

directions were fitted separately. The overall mean standard deviation is


3.6%. A correlation coefficient of r2 1 0 is a perfect fit. . . . . . . . . . 43


4.5

Parameters and n of the power-law fit of P versus volume flow-rate


following Eq. 4.1. The measurements in the forward and reverse flow
directions were fitted separately. The units of are Pa sec m3 . . . . . . . . 43

4.6

Mean prediction errors and standard deviations of the volume flow-rates


and the diodicity for each Test Group, shown in percent. . . . . . . . . . . 46

xii

ACKNOWLEDGMENTS
There are many who have helped make my experience as a graduate student a rewarding
one. In particular I wish to thank Prof. Fred K. Forster for introducing me to microfluidics
and steadfastly supporting me throughout my PhD studies. Without his encouragement I
would never have taken up the challenge. I also wish to thank Prof. James J. Riley for his
scientific advice and eagerness to help, Prof. Martin A. Afromowitz for his helpful insights,
Prof. George Kosly for guiding me through the Masters program, Prof. Karl F. Bhringer
for being on my Reading Committee, and Prof. Michael J. Pilat for his perspective and
advice. Id like to thank all my fellow graduate students who made learning science a joy,
especially Nigel R. Sharma, Chris Morris, Brian Williams, Ling-Sheng Jang, James Pate,
Robert Penney, Bill Constantine, Bent Wiencke, Tina Toburen, and Paul Galambos. I am
tempted to quote the last remarks of Huck Finn.

xiii

Chapter 1
INTRODUCTION
1.1 Nature and Scope of the Problem
MEMS (Micro-Electro-Mechanical Systems) devices often contain microfluidic systems
that are designed to move very small quantities of fluid, such as microliters or even nanoliters, within the device. These devices are often testing units using chemical, electrical, or
optical sensors to analyze chemical compounds outside the traditional medical laboratory
setting so that minitiaturization for portability is of prime concern. Another application
currently generating great research interest is the temperature control of three-dimensional
semiconductor devices that would greatly benefit from internal cooling by a high heat capacity liquid. A variety of micropumps for moving liquids using active, passive, or nomoving-parts (NMP) valves have been developed in recent years, as reported by Shoji [26]
and Gravesen [11].
Many of these micro-scale devices have been developed to pump gases, but pumping
liquids entails additional difficulties. Most applications do not involve a closed-loop system in which the working fluid can be highly filtered, but require pumping of real-world
fluids that contain particles of several microns, or more, in diameter. These particles can
render the valve seats of active and passive microvalves dysfunctional; if the particles are
hard, they damage the valve seats; if they are soft, they can adhere to the seat and prevent
complete valve closure; if the particles are delicate, they can be damaged during passage
through the valve.
A more robust way to handle particle-laden flows is to use NMP valves that allow
free passage of particles and rely on fluidic instead of mechanical mechanisms to inhibit
reverse flow. There are many additional reasons to utilize NMP valves, such as ease of
manufacture, simplicity of operation, robustness due to lack of moving parts, low cost,
etc. However, the technology of NMP valves was developed for the macro-scale, where
the mechanism to inhibit reverse flow is based on fully-turbulent high-Reynolds-number
flow. This technology is of questionable value at the micro-scale, since microvalves are

2
low-Reynolds number devices by nature of the small dimensions of their channels. Thus,
efforts to design optimal NMP microvalves suffer from the lack of understanding of fluidic
mechanisms that inhibit reverse flow in laminar low-Reynolds-number flow, resulting in a
dependence on the build & test method.
The goal of this research was to develop an understanding of the fluidic mechanism
of the Tesla-type NMP valve in low-Reynolds number flow, an understanding that can be
utilized to design improved valves and break the dependency on the build & test method.

1.2 Background for NMP Microvalves


For the reader to understand the operation of all NMP microvalves, several subjects are
introduced in this section. The first is a parameter to characterize valve performance, the
valve diodicity, which is defined in Sec.1.2.1. The second subject is assessment of the flow
in the valve as laminar or turbulent, since each requires its own method of analysis. Microvalves are typically small enough in scale to contain laminar flow throughout their range
of operation. This is particularly true of the valves in this study. Thus, Sec. 1.2.2 describes
how to differentiate laminar from turbulent flow. The final subject is system dynamics.
Unlike classic macro-scale check valves, which attempt to prevent reverse flow under all
conditions, NMP valves provide a net forward flow in oscillatory flow conditions, and only
mildly inhibit reverse flow in steady flow conditions. Thus, they are always operated in
oscillatory conditions, and understanding the dynamics of flow in NMP valves is essential. Sections 1.2.3, 1.2.4, and 1.2.5 introduce characterization of the dynamic response of
fluid flow in simple channels. These discussions form the basis for the analysis of NMP
microvalve dynamics in later chapters.

1.2.1 Diodicity
The figure of merit that characterizes the ability to pass flow in the forward direction while
inhibiting flow in the reverse direction is the diodicity of the valve. Since NMP valves
have more resistance to flow in the reverse direction than in the forward direction, they
produce a unidirectional net flow in the downstream direction even in the presence of a
backpressure. The remaining portion of the instantaneous flow is the oscillatory slosh flow.
In an electrical analogy, the instantaneous current is a sum of an alternating current (slosh
flow) and a direct current (net flow). The diodicity, Di, is defined as

3


Di

Preverse
Pforward

(1.1)


in which the ratio of pressure loss in the reverse-flow direction to loss in the forward direction is taken at identical volume flow rates. The typical diodicity for micro-scale NMP
valves is relatively low, 1 Di 2. The pressure loss could be further broken down by the
dependence, or non-dependence, on flow direction as in


Di

Pindependent
Pindependent


Pdependent reverse
Pdependent forward


(1.2)

Q


which shows that direction-independent pressure drop dilutes the diodicity and should be
avoided. By definition, Di 1, or the specification of forward and reverse directions would
be interchanged. Note that a unity diodicity device produces zero net flow.


The differential pressure loss that creates the diodicity of a valve is due to inertial and
viscous forces. The inertial losses are proportional to the square of the velocity and are due
to acceleration of the flow, for example, altering a plug-flow profile at the valve inlet to a
fully-developed, or even strongly-distorted, velocity profile at the outlet. In addition, local
accelerations distort the velocity profile in regions of separated flow that typically occur
where there are rapid changes in the channel crossectional area. Even flows with Reynolds
numbers as low as 0.01 can exhibit separation, as shown in Fig. 1.1 from Taneda, et al,
[27]. Viscous losses are proportional to velocity, and relatively unimportant in turbulent
flow. They become significant in laminar, separated flow where the velocity gradients are
large, for example, where jet flow occurs near the valve wall. This dissertation will show
that viscous forces are a major source of valve diodicity in low-Reynolds-number flows.

1.2.2 Laminar vs. Turbulent Flow


It is neccessary to differentiate flow in a microvalve as laminar or turbulent flow. Each flow
regime requires its own method of analysis, as the manner in which pressure loss varies
with volume flow rate depends on the flow regime. In laminar flow the friction factor of
a channel is proportional to the volume flow rate; in fully-turbulent flow it is relatively
constant despite changes in flow rate. The transition from laminar to turbulent flow in
a pipe is generally considered to occur at approximately 2000 Re 2300. For non

circular channels of micro-scale dimensions there has been some concern whether this still
holds true. In an early review of microfluidic devices, Gravesen [11] noted that not only

Figure 1.1: Flow separation and recirculation at Re


1979).

0 01 based on cavity depth (Taneda




are the channel width and height of small scale, but typically the channel length is less than
the entrance length for fully-developed laminar or turbulent flow. If the ratio of length and
width was large enough L DH 70, he considered flows with Re 2300 to be turbulent.
He modeled shorter devices as orifices and labeled flows with more pressure drop due to
inertial losses than viscous losses to be turbulent. Very few of the devices he reviewed were
marked by these rules as turbulent. Olsson [20] developed a formula to model pressure drop
as a function of the flow rate in his NMP valves, three of which are micro-scale and three
others are an order of magnitude larger in size. His relation for pressure drop contains two
terms, one representing laminar flow and a second representing turbulent, and he appears
to apply both terms simultaneously to obtain least-squares fits of his experimental data.
There is no need for this uncertainty; a flow can be identified as laminar or turbulent by
either experimental or computational methods. Using experimental data, a laminar flow is
identified by a linear proportionality between the log of the pressure drop across the valve
and the volume flow rate, ie. a straight line on a log-log plot of pressure loss versus flow
rate. If the flow transitions to turbulence at higher flow rates, the same linear proportionality would no longer hold and the slope of the line would change at that flow rate. Transition
to turbulence can also be identified using numerical methods to simulate a flow, because
as Hinze [13] states, turbulence is defined as irregular flow with random variation of flow
properties (eg. velocity, pressure, etc.) in both time and space coordinates simultaneously.
A numerical simulation based on solving the Navier-Stokes equations will not converge
to a steady solution if the flow is randomly varying. Time-averaging of the flow properties, spectral methods, or some other technique must be used to model a turbulent flow.
If a Navier-Stokes-based simulation without Reynolds averaging converges to a steady so-

5
lution, the modeled flow is laminar. This also applies to the case of harmonic boundary
conditions; if the simulation produces a steady, harmonic solution, the flow is laminar.
By these means, the flow in a microvalve can be accurately differentiated as laminar or
turbulent, and analyzed accordingly.

1.2.3 Modeling the Dynamics of Microfluidic Systems


System dynamics is discussed here, because unlike classic macro-scale check valves, which
attempt to prevent reverse flow under all conditions, NMP valves operate only in oscillatory
flow conditions. Thus, modeling of the dynamics of flow in NMP valves is essential to
predict their performance; diodicity alone is insufficient to characterize an NMP valve.
There are additional reasons to use system dynamics modeling when microvalves are
implemented as components of a micropump with a piezoelectrically-activated membrane.
The pump diaphragm must be stiff enough to resist up to pressurize the pump chamber
to one atmosphere. Yet, it must be deformed sufficiently by the potential-driven strain
of the lead-zirconium-titanate (PZT) piezoelectric actuator to sweep out the neccessary
volume to supply the slosh flow through the valves. On the other hand, the electrical current
requirements should be kept low to minimize the size of the power supply and enhance
portability. As a result, the diaphragm displacement per volt of excitation of the PZT must
be maximized. This is achieved by designing the pump to operate at a minimally-damped
resonance.
According to Gravesen [11], the most common technique used to model the dynamics
of micropumps with microvalves of all types (active, passive, and NMP) is the lumpedparameter method using the electrical-hydraulic analogy. For example, Voigt [29] used
this method to model a flap-valve micropump. Zengerle [32] also used this method to
address the interaction between micropumps and their connected fluid system, and developed a methodology for modeling each of the components as well as the entire system.
The interaction was shown to be especially prominent for pumps with pulsatile flow, and
capacitive elements were suggested for decoupling the micropump from the measurement
system. In another example, a complete linear model of a micropump with NMP valves was
introduced by Bardell, et al. [3] and used to predict the resonance frequency, membranedisplacement amplitude and slosh-flow-rate amplitude. The circuit diagram is shown in
Fig. 1.2. Lumped-parameter resistance and inertance elements were used to represent the
NMP valves: Rvi and Ivi for the inlet valve and Rvo and Ivo for the outlet valve. Another example is Olssons [20] numerical design study based on a lumped-mass micropump model

Figure 1.2: Circuit diagram for linear model of complete micropump.

with six coefficients used to match performance data from experiments. In contrast, Morris
[17] shows that fitting the model to data is not necessary if the frequency-dependence of
the valve resistance and inertance elements is considered. Clearly the lumped-parameter
method is a useful technique, thus in addition to revealing the diodicity mechanism and
calculating diodicity, there are discussions of resistance and inertance in Tesla-type NMP
valves in Apps. A and B, respectively.
1.2.4 Steady-State Response
This section introduces the characterization of the steady-state response of flow in simple
geometries, such as a rectangular channel, a pipe, and a two-dimensional slot. The fluid
resistance appropriate for a lumped-parameter element is developed forming the basis for
later discussions of NMP valve dynamics.
The resistance to fluid flow characterizes the steady-state response of the volume flow
rate to an applied pressure gradient between channel inlet and outlet. Since NMP valves
are typically formed in silicon by DRIE, they are planar structures with channel walls that
are vertical and flow crossections that are rectangular. As a result, the aspect ratio of the
channel, AR height width 1, becomes important in determining the resistance. Clearly
a channel with very large aspect ratio will have higher resistance to flow than a channel of
the same cross-sectional flow area with a unity aspect ratio. This variation in fluid resistance

7
is often modeled by using the hydraulic diameter, DH , which is
DH

4 Area
wetted perimeter

2
1
height


(1.3)

1
width

and a friction factor that is a function of aspect ratio as well as Reynolds number.
The Darcy friction factor,
4w
1
2
2 U

P DH
L

(1.4)

1
2
2 U

is the ratio of the energy dissipated in shear and the kinetic energy, and is related to pressure
drop P analogous to the Darcy-Weisbach equation for head loss [31]. Using the method of
O.C. Jones, Jr. [14], the laminar flow friction factor is easily approximated for rectangular
channels by
f

64
Re

UDH

where Re

and

2
3


11
24 AR

2


AR

(1.5)


in which the factor accounts for the variation in aspect ratio and is within 2% of the
infinite series solution. Combining the definition of resistance as the ratio of pressure drop
and volume flow rate R

P Q

P AreaU with Eqs. 1.4 and 1.5 leads to




128 L
4 Area D2H

(1.6)

for resistance to fluid flow in a channel of rectangular crossection and length L. This relation holds for a circular pipe if DH is taken as the pipe diameter and is set to unity, and
it also holds for a two-dimensional slot if the usual values for a slot are taken: D H as twice
the slot height, Area as the slot height times unit width, and AR
resulting in
12 L
R
height3


0 since height


width,
(1.7)

1.2.5 Transient Response


This section introduces the characterization of the transient response of flow in simple
geometries, such as a rectangular channel, a pipe, or a two-dimensional slot. The inviscidflow inertance appropriate for a lumped-parameter element is developed, and the analytical
solutions for the velocity profiles and impedance of a harmonically-oscillating slot flow are
presented. These concepts form a basis for later discussions of NMP valve dynamics.

8
The fluid inertance is a measure of the transient response of the fluid in the channel
to a time-varying applied pressure gradient. The fluid inertance in the channel can be
approximated by analogy with either of two standard electrical circuit analysis methods:
transient response to a step input, or steady-state forced response to a sinusoidal excitation.

Step response of a channel


A one-dimensional model (based on Newtons second law F ma) for the inertance of an
inviscid liquid in a simple straight channel is given by Ogata [18] in which the pressure
force F PA applied to the fluid cross-sectional area A is balanced by the acceleration
a

dQ Adt of the mass of fluid in the channel m


P

LA as in

dQ
Adt

The inertance I is the ratio of the change in pressure and the resulting change in flow rate
given by
P
I
L A
(1.8)
dQ dt


If a channel filled with inviscid liquid is modelled as a first-order system, (ie. a resistor and inductor in series), the response of the volume flow rate to a step input pressure
difference across the channel is an exponential function of time given by
Qt
At Q Qmax
P Qmax

Qmax 1

0 632, the time constant is




exp

t


I R where the resistance R is the ratio of

These relations are valid for inviscid flow, but only approximate when viscosity is considered and the crossectional shape of the channel becomes important. The analytical solution for impedance in a two dimensional slot is developed in App. C.

Harmonic response of a two-dimensional slot


An exact analytical solution for the velocity profiles is given by Panton [21] for unsteady
flow in a slot driven by an oscillating pressure gradient. This discussion forms the basis for
the development of the exact solutions for the volume flow-rate and fluid impedance of an
oscillating slot flow in Chap. 5.

Figure 1.3: Velocity vector field shows center flow out-of-phase with flow nearer the wall.

If the viscous diffusion length is much less than the channel height 2h, the dimensionless parameter , sometimes referred to as the kinetic-Reynolds number, is described by
h

(1.9)

and the center-channel flow becomes out-of-phase with the flow near the wall, as shown in
Fig. 1.3.
The analytical solution assumes the streamwise velocity, u, is a function of time and
the cross stream direction, y, but not the streamwise direction, x. The cross stream velocity,
v, is assumed zero everywhere, because it is zero at the wall. As a result the x-direction
momentum equation becomes linear in u and is
u
t

K cos t


2 u
y2

(1.10)

where the amplitude of the oscillating pressure gradient is due to the peak differential pressure P across a slot of length L resulting in


dp
dx

max


P
L

(1.11)


Assuming no slip along the wall and symmetry about the centerline at y
y


h, the initial conditions become u y

h t


0 at the wall and

u
y

0, since
0 t


h


0 at the

10
centerline.
If the variables are normalized by T t, Y y h, and U u K , (which scales
the velocity by the amplitude and frequency of the pressure oscillation), the dimensionless
form of the governing equation becomes


U
t

cos T


2U
h2 Y 2

(1.12)

which contains in the third term the reciprocal of the square of the kinetic-Reynolds number
2 . The general solution for the velocity is
U

real

i expiT  1


cosh  iY 


(1.13)


cosh 

i  

Harmonic response of a general channel


If both the applied pressure difference and the resulting flow rate are sinusoidal, the fluid
impedance can be obtained from the ratio of their amplitudes. This method is appropriate
for determining impedance from the results of a harmonic numerical simulation of oscillating flow.
The pressure difference and the flow rate can be represented as the real parts in the
complex plane by
P

Pm Re
cos t


P


j sin t

Pm Re e jP e jt 

P


(1.14)

and
Q

Qm Re
cos t


Q


j sin t

Qm Re e jQ e jt 

Q


(1.15)

The fluid impedance is the ratio of the pressure difference and the flow rate in the complex
plane and is
Z j


R


jI

Pm e jP
Qm e jQ

Pm
cos P
Qm 


Q


j sin P


Q 


(1.16)

in which the fluid resistance is the real part of Z j and the fluid inertance is the imaginary
part and written as
Pm
cos P Q
R
(1.17)
Qm


11

Figure 1.4: Closeup photo of a single-element, Tesla-type (T45A) outlet valve connecting
the pump chamber on the right and the outlet port on the upper left. The white object is a
human hair.

and

Pm
sin P
Qm

Q


(1.18)

1.3 Background for Tesla-Type NMP Valves


There are a variety of NMP valve designs. Forster [9] presents techniques for design and
testing of NMP valves including the Tesla-type (discussed below) and the diffuser valve,
which is a simple flat diffuser oriented such that the forward flow sees diverging walls and
the reverse flow sees converging walls. Gerlach [10] and Olsson [19] focus their research
solely on diffuser valves. Other designs are also feasible, such as a micro-scale version of
the classic vortex diode [5].
The focus of this research is on Tesla-type NMP valves, which as explained by Forster
[9] are expected to provide higher diodicity than diffuser-type valves. The simplest configuration is shown in Fig. 1.4 which is roughly similar to that designed in the macro-scale by
Nicola Tesla [28] and patented in 1920. It has a bifurcated channel that reenters the main
flow channel perpendicularly when the flow is in the reverse direction. In the forward direction, the majority of the flow is carried by the main channel with reduced pressure losses.
The valve channels are typically 60 400 m deep, 114 m wide, and at least 15-18 widths
long. These smoothly curving shapes are etched in silicon by deep reactive ion etching


12
(DRIE) to attain independence from the crystal planes and achieve vertical sidewalls.
NMP valves are used in micropumps, and form the inlet and the outlet to the central
pump chamber, which is generally 3-10 mm in diameter and sealed by a sheet of anodicallybonded Pyrex. The Pyrex serves as the pump diaphragm and is typically actuated by a
piezoelectric lead-zirconium-titanate (PZT) wafer, generally 50-95% of the diameter of
the diaphragm, that is bonded to its outer surface. Applying an alternating voltage to the
PZT results in a moment loading of the diaphragm such that it bows in and out of the
chamber creating a membrane pump. The inlet and outlet valves are often connected by
13-18 gauge stainless steel needles with B-D fittings to tri-fluoro-ethylene (TFE) or silicone
rubber tubing.

1.3.1 Prior Research on the Diodicity Mechanism in Tesla-Type NMP Valves


Previous researchers have experimented with macro-scale NMP valves, in which velocitysquared losses are more significant than viscous losses. Paul [23] reported diodicity values up to 4.07 for 0.3 m diameter single-element, momentum-interaction valves and Reed
[24] reported values of 12.5 for six-element valves. However, at the micro-scale the low
Reynolds number flows result in lower diodicities, typically 1 Di 2.


In his 1920 patent, Tesla [28] claimed that the recesses in the walls of the valvular
conduit subjected the reverse flow to rapid reversals of direction resulting in friction
and mass resistance and causing violent surges and eddies which interfere very materially with the flow through the conduit. His test fluid was hot, compressible gas from a
high-pressure combustion engine resulting in very high-speed turbulent flow. He found the
efficacy of device was: first, the reverse flow resistance being larger than the forward; second, the number of valve elements; third, the character of the gas impulses. He also stated
that the ratio of reverse flow resistance to forward was as high as 200.
Paul [23] states that diodicity is achieved by maximizing the reverse flow total pressure
loss coefficient and minimizing the forward, and is a function of the valve geometry. He
concluded that the major pressure loss in the reverse flow direction is due to confined
jet interacting flows. He performed a control volume analysis of the junction where the
channels rejoin in reverse flow shown in Fig. 1.5. The pressure loss in the main channel
was proportional to the sum of the momentum flux out of the control volume and parallel
to the main channel as in
A P1


P3


m 3V3


m 1V1


m 2V2 cos

(1.19)

13

V2

V3

V1

Figure 1.5: Control volume for momentum analysis of F. Paul of reverse flow through the
T-junction of a Tesla-type valve. Reverse flow is from right to left.

where A is the crossectional flow area of the main channel, is the intersection angle of
the side channel with the main channel measured clockwise from horizontal, subscript 2
refers to a location in the side channel, and 1 and 3 are, respectively, upstream and


downstream locations in the main channel. He then assumed that the upstream locations
shared a stagnation zone, and thus had the same static and total pressures, which for incompressible flow resulted in V1 V2 . His pressure loss predictions agreed with experimental
data for prototype valves with
30% at 90 .

45 and overpredicted pressure loss by approximately

He did not model what he considered secondary losses: flow separation and turbulence
downstream of the intersection of the jets. His test fluid was air, and his valve was 3
orders of magnitude larger than NMP microvalves and within the turbulent flow regime, ie.
1 700 Re 17 000 and 0 01 Ma 0 13.


Reed [24] claimed that the reverse flow losses were due to a vena contracta formed in
the trunk channel at the junction of the trunk flow channel and the reversing flow channel
where the reversing flow impinged on the trunk channel flow. The vena contracta was
disabled in forward flow. His test fluid was water. He stated that the best overall head
loss performance occurred when the forward flow negotiated 45 bends, instead of the
10-20 bends suggested by Tesla. Eddies and other energy sinks were avoided to utilize
the available energy to contract the flow and obtain a head loss. He further teaches that
it was extremely critical to the performance of the valve to follow the exact shapes and
positioning of both the guide vane that separates the channels, and the cusp downstream of
the junction of the trunk flow channel and the reversing flow channel.
Reed also stated that, unexpectedly, the head loss performance of the valve improved

14
nonlinearly as additional valve elements were added, resulting in pressure ratios of 4, 10,
and 12 in 1, 3, and 6 element valves, respectively. He claimed this effect was caused by each
special jet contraction or vena contracta acting to maintain high velocity and momentum,
which then effected a sizeable contraction at the next downstream jet impingement. In
addition, each vena contracta directed part of the impinged flow into the next downstream
reversing channel.
Reed notes that a considerable vacuum is formed in the separation region downstream
of the jet impingement, and that relieving that vacuum with a vent caused an additional
shrinkage of the vena contracta there and improved the pressure ratio by more than 20% in
one instance.
Overall, the discussion in the literature of the diodicity mechanism of Tesla-type valves
tends to be qualitative and always focused on the role inertial losses play in creating diodicity, and correctly so, as they are dominant in turbulent flow.

1.4 Research Objectives


Since NMP valves originated as high-Reynolds-number devices, previous research has focused solely on inertial forces as the source of diodicity. The technology that was developed
is of questionable value for micro-scale devices, since microvalves are low-Reynolds number devices by virtue of the small dimensions of their channels. Thus, efforts to design
optimal NMP microvalves suffer from the lack of understanding of diodicity mechanisms
in laminar low-Reynolds-number flow, resulting in a dependence on build & test methods.
It is the hypothesis of this research that a numerical method employing momentum
and kinetic-energy conservation in regional control-volumes accurately predicts the diodicity and reveals the low-Reynolds-number diodicity mechanism of Tesla-type NMP microvalves, that this diodicity mechanism is unlike the mechanism of high-Reynolds-number
macro-valves in that viscous forces, not inertial forces, are dominant, and that this understanding of the diodicity mechanism is necessary and sufficient knowledge to establish
effective guidelines for the development of enhanced-diodicity valve designs. Attainment
of the following objectives is sufficient to substantiate this hypothesis:
1. develop a numerical method employing momentum and kinetic-energy conservation
in regional control-volumes, and verify that it:
(a) is mathematically correct,

15
(b) accurately predicts flow-rate response to applied pressure in steady-flow conditions,
(c) accurately predicts valve diodicity,
(d) accurately predicts transient flow-rate response to step input and harmonicallyvarying pressures,
(e) reveals the low-Reynolds-number diodicity mechanism of Tesla-type NMP microvalves,
2. show that this diodicity mechanism is dominated by viscous forces, unlike the highReynolds-number mechanism of macro-scale valves, which is solely due to inertial
forces,
3. demonstrate knowledge of the diodicity mechanism of Tesla-type NMP microvalves
by:
(a) developing guidelines for the creation of enhanced-diodicity valve designs,
(b) using these guidelines as the sole method to modify a valve design to significantly increase its diodicity.

1.5 Summary
To predict the diodicity and reveal the low-Reynolds-number diodicity mechanism of NMP
valves, we have developed a numerical method that combines analysis of field variables
from valve flow simulations with analysis of momentum and kinetic-energy conservation
in regional control-volumes. The numerical method is developed from the governing equations and the relative importance of each of their integral components is studied by dimensional analysis.
The numerical method is extensively validated. To verify the mathematical accuracy
of the numerical method, it is applied to the case of two-dimensional flow in a slot and
compared to the known analytical solution. To verify the predictive ability of the numerical
method, it is applied to two distinct designs of Tesla-type NMP valves and the predicted
volume-flow-rates resulting from applied pressure-gradients are compared to 242 experimental measurements from physically-realized valves. To verify the accuracy of the diodicity predictions of the numerical method, they are compared to 121 diodicity measurements
from the same physical devices. To verify that the numerical method is time accurate, its

16
predicted volume-flow-rate is compared to the exact analytical solution for oscillating flow
in a slot.
The numerical method is consistently able to discern the fluidic mechanism responsible
for the variation in pressure drop between forward and reverse flow in each valve design.
It reveals their low-Reynolds-number diodicity mechanism as the viscous dissipation surrounding laminar jets that have flow-direction-dependent locations and orientations. This
diodicity mechanism is dominated by viscous forces, unlike the high-Reynolds-number
mechanism of macro-scale valves, which is solely due to inertial forces. Understanding
of the low-Reynolds-number diodicity mechanism is employed in the development of effective design guidelines for the geometrical layout of improved Tesla-type NMP valve
designs. The value of these guidelines is demonstrated by using them as the sole method
to modify a valve design to enhance its diodicity mechanism. The numerical method predicts the diodicity of the improved design is 27-47% higher than the original with 95%
confidence.
Since NMP valves operate in oscillatory flow and lumped-parameter elements are often
used to model valves in system-dynamics analyses, valve resistance and inertance values
are derived. Valve resistance is direction-dependent and inertance is direction-independent.
For clarity, this dissertation is mainly composed of self-contained chapters, each with
its own methods, results, and discussion sections. The contents of the chapters are as
follows:
Chap. 1

presents an introduction to the problem, provides the background on NMP


valves and Tesla-type valves in particular, and lists the research objectives.

Chap. 2

presents the methods used to develop appropriate forms of the momentum


and kinetic-energy equations for regional-control-volume-based analyses, then
performs dimensional analysis. Results of approximation theory are presented
for each equation. Discussion of the evidence supporting objective #2 ends
the chapter.

Chap. 3

is a methods chapter that explains how the numerical simulations were performed, covering such topics as: algorithms, grid layout, boundary conditions,
and the characteristic parameters for nondimensionalization. We present the
numerical methods used to calculate each term of the equations developed in
Chap. 2. We verify the mathematical correctness of these methods to satisfy
research objective #1a.

17
Chap. 4

presents the methods used to verify the correct implementation of the numerical method by comparison of predictions and experimental measurements of
steady flow in the T45A and T45C, and displays the results. The discussion satisfies research objective #1b by demonstrating the accuracy of flowrate response predictions in steady flow and satisfies research objective #1c
by demonstrating the accuracy of the diodicity predictions of the numerical
method.

Chap. 5

presents the methods used to verify the numerical method as time-accurate by


comparison of predictions and experimental measurements of transient flow in
the T45A and T45C, and displays the results. The discussion satisfies research
objective #1d by showing the numerical method correctly models transient
flow.

Chap. 6

presents the results of the numerical simulations of the Tesla-type T45A valve
using the field variable approach and the regional control volume approach.
The following discussion assembles the results to prove reseach objectives
#1e and #2.

Chap. 7

presents the results of the T45C valve analogous to Chap. 6. The discussion
includes comparisons with the T45A and also satisfies reseach objectives #1e
and #2.

Chap. 8

contains a preliminary discussion that extends key points on the diodicity


mechanism discussed in previous chapters, then satisfies reseach objective #3a
by presenting methods to obtain improved Tesla-type NMP valve designs in
the form of design guidelines. The guidelines are used to lay out a new valve
design, the T45A-2, by modifying the T45A valve design to enhance its diodicity mechanism.

Chap. 9

presents the results of application of the numerical method on the T45A-2


valve design and shows the enhancement of the diodicity mechanism. The
discussion includes comparisons with the T45A and completes the accomplishment of reseach objective #3b.

Chap. 10

presents the overall conclusions of the research by focusing on the accomplishment of each of the research objectives.

18
Appendix A discusses lumped-parameter modeling of fluid resistance in NMP valves in
first-order system models and shows valve resistance is direction-dependent.
Appendix B discusses lumped-parameter modeling of fluid inertance in NMP valves in
first-order system models and shows valve inertance is direction-independent.
Appendix C derives a series solution for starting flow in a slot.
Appendix D shows how to calculate diodicity from a ratio of flow rates instead of pressure
drops.
Appendix E contains the measured values of Reynolds number and valve diodicity from
tests of the physical devices.
Appendix F contains the geometrical information used to establish the physical dimensions of the T45A, T45C, and T45A-2 in the numerical models.

19

Chapter 2
THE GOVERNING EQUATIONS
In this chapter, useful integral forms of the momentum conservation and kinetic-energy
conservation equations are developed. Then dimensional analysis and approximation theory are applied to both equations to estimate the significance of each of their terms and
provide evidence supporting objective #2, ie. that viscous forces become dominant at low
Reynolds numbers.

2.1 Assumptions
Though these NMP valves are micro-scale with channels widths on the order of 100 m,
this scale is still much greater than that required to hold a statistically significant number of
molecules, assuming the fluid is a liquid. (This would also hold true for a gas near standard
temperature and pressure.) Thus, the continuum hypothesis holds and molecular-averaged
properties of the fluid and of the flow can be defined at any point in the valve. Additionally,
since the scale of the valves is large enough, Knudsen number effects (slip velocity along
the wall) are negligible.
Since NMP valves operate only in oscillatory flow conditions as explained in Sec. 1.2.3,
this suggests additional assumptions about the fluid. These valves are typically used in micropumps that operate at resonance as open systems in standard atmospheric conditions
with water as the working fluid. Due to their low compression ratios, these pumps must
avoid creating gas bubbles by cavitation, or most of the pump stroke is lost compressing bubbles instead of moving fluid through the valves. Also, since these pumps operate at system resonance, it is reasonable to assume harmonic oscillation and a maximum
pressure amplitude of 0.9 atm, which limits the variation of density to less than 5.6% assuming isothermal conditions and water as the working fluid. Since viscosity is primarily
temperature-dependent, these assumptions define the fluid as having constant density and
viscosity, and the flow as incompressible.

20

2.2 Momentum Conservation


The integral form of the momentum equation was obtained following Panton [22] and Leal
[16] by application of Newtons second law on a material region, a control volume moving
with the flow. This law states that the time rate of change of linear momentum in a material
region moving with the fluid (no mass flux across its boundaries) is equal to the forces
applied on its surfaces, which is
d
dt

cv t


u dV

n dA


cs t

(2.1)


Then the Reynolds transport theorem,


d
dt

cv t


dV
cv t

dV


ur n dA


cs

which allows a material control volume moving at ur

u to be fixed in space by accounting

for the flux of across its boundaries, was applied to the left hand side of Eq. 2.1 with
u, resulting in

u dV
cv t

u u n dA


cs

n dA


cs

a vector force balance relation for a fixed control volume. The right hand side makes use
of the fact that the surface force concept is instantaneous and thus the surfaces forces are
identical on a moving material volume and a fixed control volume at the moment they are
coincident in time and space. The stress vector was separated into the thermodynamic
pressure and the viscous stress tensor ,


u dV
t

u u n dA


which for an incompressible fluid is




Pn dA


n dA


(2.2)

In steady flow the momentum in the control volume is constant, so that the pressure
force normal to the control volume surfaces is equal to the corresponding momentum flux
and shear force in the same direction.
Pn dA


u u n dA


u n dA


(2.3)

Note that if we had used the divergence theorem to transform the surface integrals into a

21
volume integral, the integrand would be the Navier-Stokes equation for steady, incompressible flow.
2.2.1 Dimensional Analysis
To determine their relative magnitudes, approximation theory following Kline [15] was
applied to each term in the momentum conservation equation, Eq. 2.3. The momentum
equation was normalized in a manner that assured the dependent variables were of order
unity, (O)=1, at the maximum value of their range. The magnitude of the resulting
groups and the magnitude of each term in the momentum equation were determined from
parameter values for typical operating conditions.
The dependent variables were first nondimensionalized by their characteristic values
in

x y z x
u v w u

x y z
u v w

A
t

A 2x
t t

V 3x
P p

V
P

(2.4)

x u

resulting in dimensional coefficients for each term, respectively, in Eq. 2.2 of




LHS


RHS#1

u 3x
t



2u 2x

RHS#2

RHS#3

  

  

p 2x

u x


Subsequently normalizing by the coefficient of term #2 produced the generic groups of


LHS




u x
t p

RHS#1 RHS#2 RHS#3




2u
p



  

u
p x


For the case of a velocity response to a step input in pressure, the characteristic values
(in S.I. units) for the coefficients were chosen to be
p
x
u

P
DH
P

x u

in which p was related to the maximum pressure differential applied across the valve, and

22
x was based on the hydraulic diameter of the main valve channel. The resulting groups
were
0
1
2
3

u x
t p

2u
1
p
1
u

p x DH P

1
Re p

corresponding to the terms in the momentum equation, Eq. 2.2. All are unity except 3
of the viscous force term, which is the reciprocal of a Reynolds number based on the characteristic velocity,
became

P . Inserting these groups, the normalized momentum equation

u dV
t


u u n dA


Pn dA

1
Re p


u n dA


(2.5)

in which the asterisk superscripts have been dropped. Since the velocity exhibits an expoas steady conditions
nential response to the step input in pressure, u u f 1 exp t


are approached the transient term becomes small, and the remaining terms can be gathered
into a single surface integral and their individual magnitudes assessed.

2.2.2 Results
Approximation theory was applied to determine the relative importance of each term in
the momentum equation, Eq. 2.5. Two cases were studied: fully-developed straight duct
flow and NMP valve flow in which the velocity gradients are an order-of-magnitude larger
due to the presence of separated flow with laminar jets. Applying the characteristic values
chosen above, the relative importance of each term shown in Table 2.1 depends strongly
on the Reynolds number. In both cases the momentum flux and pressure force terms are
dominant if the values of u and P 1, but at lower Reynolds numbers the viscous force
term becomes larger than the momentum flux term.

2.3 Kinetic Energy Conservation


The conservation of kinetic energy equation was derived from the scalar product of the
velocity and the transient momentum conservation equation, Eq. 2.2, after applying the

23
Table 2.1: Order of magnitude of the integrand in each term in the steady form of the
momentum conservation equation Eq. 2.5 for water in a straight duct and an NMP valve at
various Rep .
Rep u P

Momentum Pressure
flux
force
100
100
10 2
10 1
10 4
10 2

1000, 1, 1
100, 0.1, 0.1
10, 0.01, 0.01

Viscous force
duct valve
10 3 10 2
10 3 10 2
10 3 10 2


divergence theorem to the right hand side terms,


u


u dV
t

u u n dA


P dV

dV

Further development of the desired integral form of the kinetic energy equation followed
Panton [22] starting with
u2
dV
t 2

u2

u n dA
2


u P dV

 dV


(2.6)

where is the viscous stress tensor, ie. the stress tensor minus the thermodynamic pressure


terms on the diagonal. A more insightful form of the equation was obtained by applying
two identities
Pu

u P

u

P u
ui
 i j
x j

(2.7)

that separate the respective work rates into kinetic energy and heat energy components.
The first identity represents the pressure work rate, and the second, the viscous work rate.
The first term on the R.H.S. is the kinetic energy component, and the second term, the heat
energy component. Utilizing these identities to replace the integrands on the R.H.S. of Eq.
2.6 resulted in
u2
dV
t 2

u2
u n dA
 u  n dA
2
ui
i j
P u dV
dV
x j

P u n dA


(2.8)

24
after applying the divergence theorem to the pressure work rate and the viscous work rate
terms to obtain surface integrals. In steady flow the L.H.S. of Eq. 2.8 would be zero. In
incompressible flow the last term on the R.H.S., the compression work rate, is always zero
as it contains the divergence of the velocities. With these assumptions the kinetic energy
equation simplified to
P u n dA


u2
u n dA
2


i j


ui
dV
x j




u  n dA


(2.9)

Going term by term from left to right, the L.H.S. is the pressure work rate and it is balanced
by the energy flux rate, the energy dissipation rate, and the viscous work rate on the R.H.S.
The dominant terms are the pressure work rate and the dissipation rate. The energy flux
rate would be zero if, for example, the cross-sectional flow area and the velocity profile of
the inlet were identical to those of the outlet. The viscous work rate concerns the viscous
forces normal to the wall, which are rarely significant. The energy dissipation rate term is
the volume integral of , the dissipation function. For incompressible flow, it is described
by White [30] as

ui

x j

u
2
y

v
x


v
2
y

u
y

2


w
z


2


w
y


v
z




u
z


w
x

2





(2.10)

2.3.1 Dimensional Analysis


Approximation theory was also applied to the kinetic-energy equation to determine the relative magnitude of each term. The equation was normalized for the case of a harmonic velocity response to harmonic pressure boundary conditions. The magnitude of the resulting
groups and the magnitude of each term in the kinetic energy equation were determined
from typical values during operation of the valve.
The dependent variables of the kinetic-energy equation, Eq. 2.8, were normalized in
the same manner as the momentum equation, using Eq. 2.4. The approximation analysis
is more straightforward when performed on the volume integral form of the kinetic energy
equation, which is
u2
dV
t 2


u2
u dV
2


u  dV


Pu dV


i j


ui
dV (2.11)
x j

25
in which the divergence theorem has been applied to the surface integral terms, and the
compression work rate has been neglected assuming incompressible flow. The resulting
dimensional coefficients for each term, respectively, in Eq. 2.11 are


LHS


RHS#1

2u 3x
t



RHS#3


3u 2x



RHS#2 4
  

p u 2x

2u x


Normalizing these by the coefficient of term #3 produced generic groups for each term
as
LHS




RHS#1 RHS#3 RHS#2 4




u x
t p

  



2u
p

u
p x

which are identical to those obtained previously for the momentum equation.

For the case of harmonic velocity response to harmonic pressure boundary conditions,
the characteristic values (in S.I. units) for the coefficients were chosen to be
p

x D H
t 1 1 2 f
u
P
in which p was related to the maximum pressure differential applied across the valve, x
was based on the hydraulic diameter of the main valve channel, and t was based on the
time period of the pressure boundary oscillation.

To approximate the magnitude of the first term of Eq. 2.11, the basic assumption of harmonic velocity response, u uo sint, was utilized to evaluate the time-derivative, resulting
in a velocity oscillation at twice the frequency of the pressure boundary conditions with a
magnitude of one-half the velocity amplitude, as can be seen from
d u2
dt 2

1d
u0 sint
2 dt

2


u0 d
1
4 dt


cos 2t


uo
sin 2t
2


26
The resulting groups were
u x
t p
2u
1
p

3
2 4


2 Re p
1
1

u
p x

1 Re p

corresponding to the terms in the kinetic-energy equation, Eq. 2.11. The transient term
group 0 is a ratio of the square of an unsteadiness parameter, (sometimes referred to as the
kinetic-Reynolds number or the Wommersley parameter), DH
and a Reynolds
number Re p , which is based on the characteristic velocity u
P and the hydraulic
diameter of the main valve channel. The groups for the viscous work rate and the energy
dissipation rate, 1 and 2 are both the reciprocal of the Reynolds number Re p . Dropping
the superscript asterisks, the resulting form of the kinetic-energy equation is
2
Re p

u2
dV
t 2

u2
u dV
2

Pu dV


1
 u  dV
Re p
1
ui
i j
dV
Re p
x j


(2.12)

2.3.2 Results

Approximation theory was applied to determine the relative importance of each term in the
kinetic-energy equation, Eq. 2.12. Applying the characteristic values chosen above, the
dimensionless u and P are unity. In NMP valve flow the velocity gradients are an order-ofmagnitude larger than in fully-developed straight-duct flow due to the presence of separated
flow with laminar jets. The relative importance of each term in the kinetic-energy equation
depends on both Reynolds numbers (ie. Re p and ) as shown by Table 2.2. The dominant
term is the pressure work rate; the transient term and the viscous work rate are negligible;
and the remaining terms exchange importance depending on Re p : the energy flux rate is
more important in the higher range of Re p , while the dissipation rate is more important in
the lower range.

27
Table 2.2: Order of magnitude of each term in the kinetic-energy equation, Eq. 2.11 for
various Re p and . For water in an NMP valve with DH 100 m, 1 f 10 kHz corresponds to 8 25


Rep
103
102
101

Transient term Energy


1 kHz 10 kHz flux rate
10 1
101
10 2
10 3
10 2
10 2
4
3
10
10 5
10


Viscous
work rate
10 1
10 2
10 3


Pressure
work rate
101
100
10 1

Dissipation
rate
100
10 1
10 2

2.4 Discussion
2.4.1 Momentum Perspective
The momentum perspective offers a wealth of information about the interplay of forces
between the surfaces, flow boundaries and fluid. In each control volume it is possible to
determine the pressure force and shear force that each surface applies to the fluid, the pressure force applied by each flow boundary, and the resulting momentum flux in each of the
three coordinate directions. The pressure forces on the valve calculated from steady-flow
simulations of reverse and forward flow provide an estimate of diodicity if the geometry of
the valve is appropriate. The inlet and outlet boundaries must have equal areas and surface
normals that are parallel, since the pressure forces are vectors and diodicity is a scalar. In
this case the diodicity can be estimated by


Di

Preverse
Pf orward





Pn dA reverse
Pn dA f orward

(2.13)

Steady-flow momentum conservation, Eq. 2.3, shows clearly two ways to enhance
diodicity by increasing the pressure force in the reverse-flow direction over that in forward
flow: increase the reverse-direction shear force (proportional to velocity) or ensure that
more momentum flux is leaving than entering the control volume (proportional to velocity
squared). The opposite tactics could be employed to increase diodicity by decreasing the
pressure force in the forward-flow direction.
In Table 2.1 the momentum flux and pressure force terms are dominant if the values
of u and P are near unity, but at lower Reynolds numbers the viscous force term becomes
larger than the momentum flux term. This suggests that prior researchers (see p. 12) were
correct in their neglect of the viscous force term when analyzing their macro-scale valve

28
flows with Re 1 700. But NMP valves are micro-scale devices, and at typical operating
conditions the maximum slosh flow is in the range of 50 Re 500. Thus as the scale of
the valve decreases the viscous forces become important.


2.4.2 Kinetic-Energy Perspective


Since the Tesla-type NMP valves by design have multiple flow directions, it becomes somewhat complicated to utilize the vector-based momentum perspective to obtain a measure of
valve diodicity, a scalar quantity. The kinetic energy, also a scalar quantity, presents no such
difficulty. There is no involvement of surfaces or coordinate directions, only the transfer
of energy across flow boundaries, work done on the fluid, and dissipation within the fluid.
And there is a correspondence between the diodicity and the pressure work rate calculated
from steady-flow simulations of reverse and forward flow. If the inlet and outlet surfaces
are located such that the pressure on each surface has an approximately constant value,
then the ratio of the reverse and the forward-flow pressure work rates at the same flow rate
provides a good estimate of the diodicity, the ratio of the corresponding pressure drops, as
shown in
P u n dA reverse
Preverse
(2.14)
Di

Pf orward Q
P u n dA f orward


 


The approximation analysis of the kinetic-energy equation had two important outcomes. First, it suggested that although the viscous work rate was negligible, the energy
dissipation rate was not, and indeed dominates over the energy flux rate in lower Re p flows.
This supports hypothesis #2 that the diodicity mechanism is dominated by viscous forces
in valves of this scale. Secondly, this approximation analysis suggested that the transient
term was negligible even for oscillating water flows with frequencies as high as 10 kHz.
This is important in that it suggests that steady-state simulations are sufficient to understand the role of the terms in the kinetic-energy equation when modeling flows oscillating
at frequencies up to 10 kHz, including the relative importance of the dissipation and the
energy flux rate.

29

Chapter 3
THE NUMERICAL METHOD

The scale of microvalves presents challenges to understanding their physical behaviour


through experimentation. Flow visualization using particles, several microns in diameter
or less, will significantly affect the flow. Particles, one or two orders smaller, are difficult
to see. Fluorescent dyes are useful if they do not affect important fluid properties, ie.
viscosity. However, any flow visualization method is hampered by the time scale in which
whole-field data for a fluid flow oscillating at 1 5 kHz must be taken, typically 10 50 s.
Direct measurement of physical properties, such as instantaneous pressure and flow rate,


are difficult because miniaturized flow meters that do not disrupt the flow with sensing
elements, (ie. vanes or cantilevered beams), are not yet available. Pressure transducers are
on the same scale as a microfluidic system itself, and pressure measurement by the height
of a fluid column profoundly alters the load seen by the system. Fortunately, the same
micro-scale that makes direct physical measurement difficult, makes numerical simulation
by computational fluid dynamics (CFD) more accurate, because unlike the macro-scale,
realistic flows are not turbulent, but laminar, and the exact governing equations can be
solved.
Numerical simulations also have the advantage of offering complete velocity and pressure field information, so that energy flux and dissipation rate, as well as momentum flux
and viscous force can be determined in any arbitrary control volume in the flowfield. This
has the potential of leading to complete understanding of the fluid mechanic behavior, even
in the case of transient or harmonic boundary conditions.
This chapter discusses the numerical method employed to predict the valve diodicity
and reveal the diodicity mechanism, including: algorithms, grid independence, boundary
conditions, and characteristic parameters for nondimensionalization. To satisfy research
objective #1a, it verifies the mathematical correctness of the numerical method via a 2-D
slot flow that has an analytical solution.

30

3.1 Methods to Quantify the Diodicity Mechanism


To identify the diodicity mechanism of an NMP valve, two methods were employed: a
field-variable approach and a regional control-volume approach, both based on numerical
simulations of forward and reverse flow in the valve. The field variable approach focuses on
the pressure, velocity, and energy dissipation fields and how they vary between the forward
and reverse flow directions. Study of the pressure fields reveals locations of significant
pressure gradients, which are coincident with the highest pressure losses. The velocity
fields show flow separation and laminar jets occurring where channels separate or recombine and at the valve exit; the energy dissipation rate is most significant where the velocity
gradients are largest.
The regional control-volume approach determined the net value in each control volume
of each term in the momentum and energy conservation equations. In the momentum
perspective, the integral form of the momentum conservation equation was utilized in each
control volume to study the proportion of pressure force expended on creating momentum
flux in comparison to that applied to overcoming viscous force. A second perspective was
gained through the kinetic energy equation by comparing the rate at which pressure work
is expended in energy dissipation to the energy flux rate out of the control volume.
The regional control-volume approach enabled the assessment of the significance of
the flow features (eg. laminar jets, pressure gradients, energy-dissipation regions) seen in
the field-variable approach as momentum or energy loss mechanisms. From the correspondence between the field-variable and the regional control-volume analyses it was possible
to uncover the nature of the diodicity mechanism, perceive it in terms of flow features, and
determine if it was due to changes in momentum or to viscous force, due to redistribution
of energy or to dissipation of energy.

3.2 Numerical Algorithms


The simulations of fluid flow in NMP valves were produced with the finite-volume computational fluid dynamics package CFX 4.2 from AEA Technology Engineering Software,
Inc., Pittsburgh, PA, [7]. These simulations were performed using the laminar, isothermal, incompressible, transient flow model. The time-stepping algorithm was backwarddifference with fixed time steps of length t. During each time step, the number of iterations 50 n 1000 performed depended on the change in the residuals of mass m and
the three velocity components: u v and w A user routine USRCVG . F was written that com

31
U
V
W
Mass

residual

10

10

10

100

200
300
iteration

400

500

Figure 3.1: Typical residuals plot showing termination of interations and procession to next
time step, controlled by USRCVG . F. Note that all residuals have ceased changing before a
new time step begins: first the Mass residual, then the W velocity residual, and finally the
U and V velocity residuals.

putes the average residual of the last 10 iterations. Once the change in each of the average
residuals m u v w is less than 1% of their magnitude, the main program is signalled to
proceed to the next time step, as shown in Fig. 3.1 from a typical simulation. The govern

ing equations were the incompressible form of momentum conservation, which inherently
includes mass conservation, and a pressure correction following the SIMPLEC solution
algorithm. The discretization scheme CCCT was chosen, a quadratic upwind differencing
that uses two upwind points and one downwind. It is third-order accurate for the advection
terms and second order for the other terms including the diffusion terms. It is a modification
of the QUICK scheme in that it is bounded to prevent non-physical overshoots.

3.3 Physical Grid Layout and Grid Independence


The three-dimensional grids for the valve geometry were body-fitted grids in cartesian
coordinates with 22 by 15-20 cross-stream and 286-292 streamwise grid points, resulting
in totals of 94380-128480 finite volumes. The grid density was refined near walls and
at channel junctions, where large velocity gradients were expected. The dimensions of
the valve channels were extracted directly from the CIF format files containing the layout
drawings for the photolithography masks of micropumps with single valves at the inlet and
outlet. The sets of points defining the geometry of the T45A, T45C and T45A-2 valves are
included in Appendix F. The as-etched geometry varies from the mask design where cusps

32
and protruding corners exist, which etch at a slightly faster rate than the rest of the pattern.
The entrance to the inlet or outlet channel from the pump chamber or inlet/outlet port is not
sharp-edged, and was measured from a magnified image to have rounded corners of 15m
radius. The nominal etch depth of the valve channels was 120m, so the grid depth was
specified as 60m assuming the flow was symmetrical about the valve centerplane. This
assumption allowed a 50% reduction in memory usage and reduced the simulation run time
by a factor of 3-4 to less than 24 hours of wall clock time.
The assessment of grid independence of the simulations was based on the accuracy of
the calculation of kinetic-energy conservation in Eq. 2.11. Since ideally the terms should
sum to zero, the error KE was defined as the ratio of the sum of the terms and the root
of the sum of the squares of the terms. When the reduction in KE was less than 5%
for a refined grid with at least 20% additional finite volumes, the solution was considered
grid independent. Note that the effect of this 20% grid refinement on the predicted value
of diodicity was less than 0.5% which is an order-of-magnitude less than the difference
between the predicted diodicity and that measured experimentally (see Chap 4).
The grid for the slot flow simulation used to verify the mathematical correctness of the
numerical method was laid out as a three-dimensional grid with a very high aspect ratio, 1
unit high by 100 units wide by 10 units long, to ensure that identical FORTRAN routines
could be used for the slot flow and the valve flow simulations. The grid was in cartesian
coordinates with 22 by 18 cross-stream and 100 streamwise grid points, resulting in a total
of 39600 finite volumes. A grid independence study was not done since the numerical
solution was compared directly to an analytical solution.

3.4 Boundary Conditions


Pressure boundaries were used at both inlet and outlet of the valve. They were located
beyond the inlet and the outlet in large, goblet-shaped plenums designed so that the velocity
at the pressure boundary was small, since specifying pressure boundary conditions also
applies zero spatial velocity gradients at that surface. The goblet-shaped plenums are an
approximation of a half-section the cylindrically-shaped inlet/outlet ports of the physical
devices.
The pressure boundaries for the slot flow were applied directly to the ends of the channels since zero velocity gradients were assumed in the streamwise direction.

33
Step-Response Simulations
In these simulations a differential pressure was applied, via the pressure boundaries at the
inlet and outlet, to an initially-zeroed velocity field. The downstream boundary was set to
zero pressure, thus the pressure field in the simulations was in terms of gauge pressure.
The pressures applied at the upstream boundary in the reverse-flow simulations were 0.1,
0.5, and 0.9 atm. Sufficiently lower pressures were applied at the upstream boundary in the
forward-flow simulations so that the volume-flow-rate responses were within less than 1%
of the reverse-flow value and the diodicity could be calculated via Eq. 1.1.
Harmonic-Response Simulations
These simulations also start with an initially-zeroed velocity field and a downstream boundary set at zero pressure. The upstream pressure boundary was set each time step to the
average value over that time step following
Pupstream

Pmax
t

t t

sin t dt


Pmax
cos t
t


t


cos t


in the user-supplied subroutine USRBCS . F, where t is the fixed width of the time steps,
t is the current time during the time step, and is the radian frequency of oscillation.
To minimize the initial transient response and reach the long-term harmonic behavior
more quickly, Pmax is defined as a linear function of time during the first period, so that
Pmax 0

t


t


Pmax t 2


3.5 Characteristic Parameters for Nondimensionalization


The governing equations were nondimensionalized as described in Secs. 2.2 and 2.3 both
to generalize the solution to different geometries and to minimize roundoff error that would
have resulted from the using SI units for the spatial dimensions, which are on the order of
(O) = 10 4 . The characteristic values chosen were:


p = 101325 Pa = 1 atmosphere
x

2
1 h 1 w


116 9


10 6 m the hydraulic diameter of the main valve channel




10 066 m/s, a characteristic velocity




34
t

x u

11 616


10

6


s, a characteristic time.

Since the CFD software package (CFX 4.2) solves a dimensional equation set, values are
required for the following fluid properties in the CFX command file to implement the nondimensionalization scheme:
density

1
1
Re

viscosity

u x

8 497


10 4 , where


10


10 6 m2 s.


3.6 Calculation of the Terms in the Conservation Equations


To calculate the terms of the momentum and kinetic-energy conservation equations, the
CFD simulations were modified by adding FORTRAN subroutines to accomplish two main
tasks: to export information, and to perform auxiliary computations after each time step.
To calculate the terms of the momentum equation, subroutines were written to export
grid geometry, fluid velocity, mass flux, and fluid pressure at the surfaces of the control
volumes. Matlab routines [12] were written to calculate the pressure force and momentum
flux through each surface from the relevant data on that surface: surface location, surface
area, pressures, and velocities. From the x and y components of the surface area and the
velocities, the surface normal and the flow rate normal to the surface were calculated. The
pressure force on the surface in each spatial direction and the momentum flux through
the surface in each direction were calculated by integrating the respective quantities over
the surface following Eq: 2.5. These were combined to calculate the pressure forces and
momentum fluxes on each of the six faces of each of the control volumes that make up the
valve.
A different programming strategy was employed to calculate the magnitudes of the
terms in the kinetic-energy conservation equation. FORTRAN subroutines were written to
perform auxiliary computations within the simulation after each time step and the magnitudes of the terms were then exported to disk files. Matlab routines were used to gather
the data from the forward-flow and reverse-flow simulations and make comparison plots.
The form of the kinetic-energy conservation equation used to calculate the energy flux rate,
work rates and dissipation rate was the volume integral form of Eq: 2.12 with two exceptions. Using the fact that in incompressible flow u 0 the energy flux rate was

2 dV and the pressure work rate as u P dV . These formulations


calculated as u
reduced the use of velocity gradients, which are difficult to resolve accurately.


u2

35

3.7 Verification via Analytical Solution for a 2-D Slot Flow


We have added numerical methods to the incompressible Navier-Stokes solver in CFX
4.2 to calculate kinetic-energy conservation by employing user subroutines written in
FORTRAN. These added routines must be verified as mathematically correct. A stepresponse simulation of slot flow was performed to validate these routines. The differential
pressure applied was 0.01 atm and the resulting volume flow rate was equilibrated within
72 times steps at Re 115, based on the slot height.
The pressure work rate, energy flux rate, and the dissipation rate are the significant
terms in the kinetic-energy equation as discussed in Sec. 2.3. In incompressible flow in a
slot the energy flux rate is zero since the velocity profile is unchanged throughout a control
volume which contains only fully-developed flow. Thus the pressure work rate and the
dissipation rate must be equal and opposite.
The pressure work rate was calculated from
P u n dA


PLz

PLz

udy

where the dimensionless lengths are Lx 10 Ly


P 0 01 and the dimensionless viscosity is

P
 y
2Lx

Ly
0

y2  dy

1 Lz 100 the differential pressure is


8 497 10 4 using the nondimension-

alization scheme of Sec. 3.5. The analytical solution for the pressure work rate is 0.0981,
the numerical solution was 0.09808, which is within 0.02%.
Since the velocity u u y is only a function of one coordinate direction, the dissipation


rate (see Eq. 2.10) was calculated using


ui
dV
i j
x j

Ly

L x Lz

u y
y

2


dy


As expected, the dissipation rate was equal to the pressure work rate. The numerical solution was 0.09807, which is within 0.03%.
The energy flux, as explained above, should be zero. The energy flux calculated in the
numerical simulation was 4 067 10 6 , which is four orders-of-magnitude smaller than


the pressure work rate and the dissipation rate.


Thus the slot-flow simulation verifies that the added numerical methods used to calculate the terms of the kinetic-energy equation are mathematically correct and research
objective #1a is satisfied.

36

Chapter 4
VALIDATION OF THE NUMERICAL METHOD IN STEADY
FLOW
Chapter 3 has described and discussed the implementation of the numerical method
used in this research to predict the flow-rate response to an applied pressure difference,
predict the valve diodicity, and reveal the diodicity mechanism. The mathematical correctness of this method (research objective #1a) was addressed in that chapter; it is the
task of this chapter to verify the numerical method has been correctly implemented in this
research and thus accurately predicts the steady-flow response and the diodicity of Teslatype NMP valves as required by research objectives #1b and #1c. Attainment of objectives
#1d and #1e, proof that the numerical method is time-accurate and can reveal the diodicity
mechanism, is left to following chapters. The current task is achieved by comparing volume flow-rate and diodicity predictions to 242 independent experimental observations from
physical-realized valves of three distinct groups of etch depths and two distinct Tesla-type
designs, the T45A and the T45C. By basing the comparison on data from multiple valve
geometries, we determine the accuracy of the numerical methods diodicity predictions for
similar Tesla-type valves in a manner that is independent of geometry. In other words, we
are free to modify valve geometry (to the same extent as the variation between T45A and
T45C) and we still know the accuracy of the numerical methods diodicity predictions.

4.1 Methods
4.1.1 Experimental Methods
To obtain physical devices for experimental data, T45A and T45C Tesla-type valves were
produced in silicon by deep reactive-ion etching DRIE at the Stanford Nanofabrication Facility and covered by a layer of Pyrex attached by anodic bonding. The etch depths of the
valves vary from wafer to wafer (approx. 25%) as well as by location on the wafer (approx.
15%) resulting in a range of etch depths of 90 h 150 m and a mean of h 120 m.


A technique for measuring the etch depth was developed using an MTI Photonic displacement sensor (MTI, Latham, NY) and a Newport Corporation 3-axis stage with 855C pro-

37
grammable controller connected by GPIB interface to a 486PC running an existing FORTRAN motion-control program. For each valve a series of n measurements (7 n 10) at
each of three locations (two up on the chip surface and one down in the valve inlet/outlet
port) were processed with a geometric-analysis Matlab routine to determine the valve-port
etch depth to within 5 micrometers. The valve-port depth was measured since the valve
channel is narrower than the spot size of the photonic sensor and cannot be measured directly with this technique. The valve port diameter is large enough to contain the sensor
spot, yet only 6.5 times larger than the channel width. To determine if the etch depths of
the valve channels are equal to the valve-port etch depths, four devices were milled and
polished and investigated with a microscope. The valve-channel etch depths were within
5% of the measured valve-port etch depth. The total etch-depth measurement error etotal
for the average-depth channel is a product of two independent errors: the valve-port etch
measurement error e port and the variation between valve-port etch and valve-channel etch
channel ,

e port

and was determined by

etotal

e2port


e2port channel


5
120

052


065 or 6 5%


following Barlow [4].


An effort was made to select devices for testing that were close to the mean etch depth
of 120 m. The resulting range of measured etch depths of the chosen devices was 109-116
m, resulting in valve channels with slightly differing aspect ratios. To better understand
the impact of etch depth, three additional devices with depths of 144-152 m were also
tested. To account for the variation in etch depth and aspect ratio when comparing the
characteristics of different valves, the pressure drop and resistance data for the Tesla-type
T45A and T45C valves are plotted versus Reynolds number instead of volume flow-rate.
The valve tests were divided into three groups considering etch depth and valve design:
the T45A group, the T45C group, and the Deep T45C group. The etch depths and deviations from the mean depth of the valves in each test group are listed in Tables 4.1, 4.2, and
4.3. In each set the tested devices are representative samples since the deviations from the
mean etch are much less than the measurement accuracy of 6.5%.
The pressure drop P across the valve was measured by driving de-ionized water at
known volume-flow rates through the valve using a syringe pump (Model 200, KD Scientific, Boston, MA) and measuring the gauge pressure upstream of the valve using a
miniature pressure transducer (Model EPI-411-3.5B-/RTV, Entran, Fairfield, NJ) that was

38

Table 4.1: Etch depths and deviations from the mean depth of the valves tested in the T45A
group. The tested devices are representative samples of the set since the deviations are
much less than the measurement accuracy of 6.5%.
etch m

112
109
110
111
110.5

T45A Test Group


T45A Bi2
T45A Bo2
T45A Ti2
T45A To2
Mean

deviation [%]
1.36
-1.36
-0.45
0.45
0

Table 4.2: Etch depths and deviations from the mean depth of the valves tested in the T45C
group. The tested devices are representative samples of the set since the deviations are
much less than the measurement accuracy of 6.5%.
T45C Test Group
T45C Li2
T45C Lo2
T45C Li2t2
T45C Lo2t2
Mean

etch m

116
114
116
114
115

deviation [%]
0.87
-0.87
0.87
-0.87
0

Table 4.3: Etch depths and deviations from the mean depth of the valves tested in the Deep
T45C group. The tested devices are representative samples of the set since the deviations
are much less than the measurement accuracy of 6.5%.
Deep T45C Test Group
T45C Si
T45C So
T45C Ri2
Mean

etch m

144
149
152
148.3

deviation [%]
-2.92
0.45
2.47
0

39
calibrated to a mercury manometer. Downstream of the valve the outlet tubing was open
to the atmosphere. Effects such as the elevation change between inlet and outlet, as well as
the pressure drop from flow resistance in the tubing, were calculated as less than 1% of the
total measured pressure drop. Data were recorded and fitted to a power-law function for
flow in each direction, forward (subscript F) and reverse (subscript R), by
PF

F QnF and PR

R QnR

(4.1)

using a linear least-squares method applied to the natural logarithm of the equation, ie.
ln P

ln


n ln Q


4.1.2 Numerical Methods


Steady-state solutions of forward and reverse flow in the T45A and T45C valves were
obtained by the numerical method described in Chap. 3. These steady-state solutions are
the final solution of step-response simulations that were allowed to proceed until steady
flow was achieved.
For study of the diodicity mechanism, the numerical simulations of the T45A and T45C
valves were performed at the nominal etch depth of 120m. However, the flow rate in rectangular channels is dependent on channel width and height, and although the channel width
is 114m for all devices, the etch depths of the tested valves vary from the 120m etch used
for the diodicity mechanism study. In the T45A case the difference between the etch depth
of the numerical simulations and the mean etch depth of the T45A Test Group of experimental devices is 120 110 5 9 5m, or 8.5%. In the T45C Test Group the difference
is 120 115 5m or 4.3%, which is within the 6.5% etch measurement accuracy. If the


flow in NMP valves behaved like fully-developed flow in straight rectangular channels, we
would expect the flow rate, in terms of Reynolds number, to be proportional to the hydraulic
diameter cubed for a given pressure gradient as in
Re D3H

(4.2)

which was derived from Eqs. 1.5 and 1.6 in which is a function of the channel height to
width aspect ratio and DH follows Eq. 1.3. But in Tesla-type valves the pressure drop at
the entrance and the bends in the flow path are dominant and the exponent in Eq. 4.2 is less
than 3. Instead of using a modification of Eq. 4.2 to adapt the results of the numerical simulations at the nominal etch depth to the etch depth of the physical devices, three additional

40
sets of numerical simulations were performed using the mean etch depth of each of the
test groups: h 110 5m for the T45A Test Group, h 115m for the T45C Test Group,
and h 148 3m for the Deep T45C Test Group, so that experimental measurements and


numerical simulation results could be compared directly.


For each combination of valve design and etch depth, the numerical method employed
nine step-response simulations to characterize, via the power-law function of Eq. 4.1, the
flow-rate response to a range of applied pressure-differences limited by cavitation to P 1
atm. The diodicity relation, Eq. 1.1, requires forward and reverse-flow pressure differences


that produce the same volume flow rate. To achieve pairs of forward and reverse-flow simulations with matching volume flow-rates, the following technique was employed. First,
reverse-flow simulations were performed, each using one of three applied pressure differences, Preverse 0 1 0 5 and 0 9 atm. The mass flux m through the valve was recorded
for each of these three cases. Then, three corresponding forward-flow simulations were


performed by setting the normal velocity u at the inlet boundary to provide the correct
mass flux (ie. u m ) with Poutlet 0 at the outlet boundary. The average pressure developed at the inlet boundary Pinlet was recorded for each of these three cases. Finally,
three corresponding forward-flow simulations were performed using the applied pressure
differences Pforward Pinlet Poutlet . The mass flux through the valve of each of these


forward-flow simulations was checked to ensure that it matched within less than 1% the
mass flux of the corresponding reverse-flow simulation. The applied pressure differences
from the first three Preverse and the last three simulations Pforward were used directly


in the diodicity relation, Eq. 1.1, as each matched forward and reverse-flow pair produced
equal volume flow-rates.

4.2 Results
The four valve experiments in the T45A Test Group and the seven valve experiments in
the T45C Test Group and Deep T45C Test Group were conducted to validate the numerical
method. Each valve experiment contains 22 independent measurements of pressure drop
versus volume flow-rate for an overall total of 242 independent observations.
4.2.1 Volume Flow-Rate
The predicted valve-flow response to applied pressure from the numerical simulations at the
nominal etch depth of 120m and at the mean etch depths of the T45A, T45C, and Deep
T45C Test Groups are compared with the corresponding experimental measurements in

41
1
0.8

Pressure drop, atm

0.6
0.4

Bi2 112um
Bo2 109um
Ti2 110um
To2 111um
simA 120um
simB 110um

0.2
0
0.2
0.4
0.6
0.8
1
1000

500

0
Reynolds number

500

1000

Figure 4.1: T45A valve pressure drop vs. Reynolds number based on the hydraulic diameters. Experimental and numerical data are shown as symbols. The curves are the the fitted
power-law relation (Eq. 4.1). The legends refer to each valve name and its etch depth; the
numerical simulations are marked sim.

Figs. 4.1, 4.2, and 4.3 in terms of pressure drop versus Reynolds number. The experimental
and numerical data are plotted as symbols; the curves are the fitted power-law relation (Eq.
4.1).
A statistical analysis was done for the power-law function Eq. 4.1 as a model equation for the true relationship between pressure-drop P and volume flow-rate. A linear
least-squares method was were used to fit the logarithm of the function to the data from
each experiment. The measurements in the forward and reverse flow directions were fitted
separately. The standard deviation and correlation coefficient r 2 between the measured
P and the calculated P from the power-law function are shown in Table 4.4. The overall
mean standard deviation is 3.6%. The correlation coefficients are r 2 0 99 in all cases,
where r2 1 00 is a perfect fit. Figures 4.1, 4.2, and 4.3 allow visual comparison of the
data and the power-law fits. The power-law function is clearly a good representation of


the physical relationship between pressure-drop and flow-rate, and there is no justification
to seek a more complicated model equation. Thus the power-law function was also used
to represent the numerical methods predictions of flow-rate response to applied-pressure.
The parameters and n of the power-law fit are shown in Table 4.5.

42
1
0.8

Pressure drop, atm

0.6
0.4

Li2 116um
Lo2 114um
Li2t2 116um
Lo2t2 114um
simA 120um
simB 115um

0.2
0
0.2
0.4
0.6
0.8
1
1000

500

0
Reynolds number

500

1000

Figure 4.2: T45C valve pressure drop vs. Reynolds number based on the hydraulic diameters. Experimental and numerical data are shown as symbols. The curves are the the fitted
power-law relation (Eq. 4.1). The legends refer to each valve name and its etch depth; the
numerical simulations are marked sim.

1
0.8

Pressure drop, atm

0.6

Si 144um
So 149um
Ri2 152um
sim 148um

0.4
0.2
0
0.2
0.4
0.6
0.8
1
1000

500

0
Reynolds number

500

1000

Figure 4.3: Deep T45C valve pressure drop vs. Reynolds number based on the hydraulic
diameters. Experimental and numerical data are shown as symbols. The curves are the the
fitted power-law relation (Eq. 4.1). The legends refer to each valve name and its etch depth;
the numerical simulations are marked sim.

43

Table 4.4: Standard deviation and correlation coefficient r2 of the experimentallymeasured P with respect to the power-law fit of P versus volume flow-rate following
Eq. 4.1. The measurements in the forward and reverse flow directions were fitted separately. The overall mean standard deviation is 3.6%. A correlation coefficient of r 2 1 0
is a perfect fit.


Test
T45A Bi2
T45A Bo2
T45A Ti2
T45A To2
T45C Li2
T45C Lo2
T45C Li2t2
T45C Lo2t2
T45C Si
T45C So
T45C Ri2

forward [%]
4.098
3.478
2.994
5.087
2.702
3.224
4.439
3.980
4.728
6.621
1.212

r2forward
0.9986
0.9989
0.9997
0.9978
0.9995
0.9994
0.9988
0.9996
0.9993
0.9996
0.9998

reverse [%]
1.986
3.494
2.560
4.161
2.227
2.854
2.156
2.651
6.439
1.931
1.768

r2reverse
0.9999
0.9998
0.9995
0.9997
0.9997
0.9991
0.9996
0.9997
0.9968
0.9998
0.9996

Table 4.5: Parameters and n of the power-law fit of P versus volume flow-rate following Eq. 4.1. The measurements in the forward and reverse flow directions were fitted
separately. The units of are Pa sec m3 .
Test
T45A Bi2
T45A Bo2
T45A Ti2
T45A To2
T45C Li2
T45C Lo2
T45C Li2t2
T45C Lo2t2
T45C Si
T45C So
T45C Ri2

forward
4.753e+15
5.568e+15
8.29e+15
2.145e+15
6.442e+15
3.806e+15
2.957e+15
6.028e+15
4.345e+15
1.259e+16
6.988e+15

nforward
1.533
1.540
1.563
1.482
1.556
1.520
1.509
1.551
1.568
1.633
1.590

reverse
3.746e+16
2.471e+16
3.952e+16
4.852e+16
3.524e+16
4.263e+16
3.674e+16
4.154e+16
6.195e+16
6.642e+16
3.923e+16

nreverse
1.646
1.618
1.647
1.658
1.643
1.649
1.645
1.650
1.716
1.717
1.681

44
2
1.8
1.6

Diodicity

1.4
1.2
1

Bi2 112um
Bo2 109um
Ti2 110um
To2 111um
simA 120um
simB 110um

0.8
0.6
0.4
0.2
0
0

200

400
600
Reynolds number

800

1000

Figure 4.4: Diodicity following Eq. 1.1 versus Reynolds number of the valves in the T45A
Test Group. The symbols are numerical and experimental data. The curves are the ratio of
the fitted power-law relations (Eq. 4.1) for the reverse and forward flow directions. Legends
refer to test name and etch depth; numerical simulations are marked sim.

4.2.2 Diodicity
The predictions and experimentally-derived values of diodicity following Eq. 1.1 for the
120 m nominal etch depth and the T45A, T45C, and Deep T45C Test Groups are shown
in Figs. 4.4, 4.5, and 4.6. The increase in scatter over that seen in the pressure versus
flow-rate test data is due to the difficulty of taking the ratio of two large numbers that are
nearly equal in value, ie. the ratio is very sensitive to measurement accuracy.

4.2.3 Prediction Accuracy


The numerical predictions and experimental data in Figs. 4.1 through 4.6 show good agreement. To obtain a quantitative measure of the agreement, we utilized the power-law fits, our
representation of the true relationship between pressure drop and flow rate, as the basis for
comparison of experimental measurements and numerical predictions. One benefit of this
method is that it segregates the experimental data points from the numerical data points and
avoids conflating measurement errors with prediction errors, otherwise an experiment with
less scatter in the data would make the numerical prediction accuracy appear to improve.

45
2
1.8
1.6

Diodicity

1.4
1.2
1

Li2 116um
Lo2 114um
Li2t2 116um
Lo2t2 114um
simA 120um
simB 115um

0.8
0.6
0.4
0.2
0
0

200

400
600
Reynolds number

800

1000

Figure 4.5: Diodicity following Eq. 1.1 versus Reynolds number of the valves in the T45C
Test Group. The symbols are numerical and experimental data. The curves are the ratio of
the fitted power-law relations (Eq. 4.1) for the reverse and forward flow directions. Legends
refer to test name and etch depth; numerical simulations are marked sim.

2
1.8
1.6

Diodicity

1.4
1.2
1
0.8

Si 144um
So 149um
Ri2 152um
sim 148um

0.6
0.4
0.2
0
0

200

400
600
Reynolds number

800

1000

Figure 4.6: Diodicity following Eq. 1.1 versus Reynolds number of the valves in the Deep
T45C Test Group. The symbols are numerical and experimental data. The curves are the
ratio of the fitted power-law relations (Eq. 4.1) for the reverse and forward flow directions.
Legends refer to test name and etch depth; numerical simulations are marked sim.

46
Table 4.6: Mean prediction errors and standard deviations of the volume flow-rates and the
diodicity for each Test Group, shown in percent.
Test group
T45A
T45C
Deep T45C

Forward flow-rate Reverse flow-rate Diodicity


eRe
eRe
eRe
eRe
eDi eDi
-5.22
1.48
-7.94
1.06
5.43 1.95
-5.26
1.55
-6.17
1.55
2.21 1.19
-9.70
3.98
-11.45
3.68
4.78 2.93

The mean prediction errors (in percent) of the volume flow-rates and the diodicity for each
Test Group are shown in Table 4.6. Each flow-rate prediction error e Re is defined in terms
of Reynolds numbers by


Re predicted Remeasured
Remeasured


eRe

100

(4.3)


in which the Re values both correspond to the same pressure-drop P. Similarly, each
diodicity prediction error eDi is defined by


Di predicted Dimeasured
Dimeasured


eDi

100

(4.4)


Re

in which the diodicity values both correspond to the same Reynolds number. Table 4.6
also includes the mean of the standard deviation of the error e for each test in the test
group. The flow-rate response is underpredicted by the numerical method by 7.62% on
average over all three test groups: T45A, T45C, and Deep T45C, and the diodicities are
overpredicted by 4.14%.
Since determination of statistical significance of predicted diodicity improvement is
needed in later chapters, the confidence level estimates of diodicity prediction error were
calculated. There are N 11 valve tests contained in the three test groups: T45A, T45C,
and Deep T45C. Since N 25, the Students t distribution [4] was used to determine the
prediction error range that has a 95% probability of including the actual error of a diodicity prediction. Instead of finding the mean error as in Table 4.6, the diodicity prediction


error eDi was plotted as a function of Reynolds number in Fig. 4.7. The diodicity mechanism study in the following chapters utilizes numerical simulations in which the flow-rate
responses were within 5.6% of Re 500, so the 95% confidence band on diodicity prediction error at that flow rate has special significance. At Re 500, the diodicity prediction

47
20

95% Confidence Band, percent

15
10
5
0
5

Upper limit
Mean
Lower limit

10
15
20
200

300

400
500
600
Reynolds number

700

800

Figure 4.7: Diodicity prediction error of the numerical method for all 11 tests of the three
test groups: the T45A Test Group, the T45C Test Group, and the Deep T45C Test Group.

error with 95% confidence is eDi

4 42


5 13% or


0 708


eDi


9 55%.


4.2.4 Evidence of Laminar Flow


It is important to show that the flow in the valve is laminar despite the presence of separated
flow and recirculation regions, since the governing equations employed in the numerical
method are the exact equations for modeling laminar flow. Identification of laminar flow
was discussed in Sec. 1.2.2 in which two methods were suggested for identifying the flow
in the valve as laminar over the Reynolds number range of the data. First, the data from a
laminar flow exhibit a linear proportionality between the log of the pressure drop and the
log of the volume flow rate. Figure 4.8 shows that the linear proportionality does hold for
data from both experiments and numerical simulations. Second, a numerical simulation
based on solving the Navier-Stokes equations will not converge to a steady solution if the
flow is randomly varying, ie. turbulent. All the valve simulations in this dissertation converged to steady solutions without applying time-averaging of the flow properties, spectral
methods, or any other technique used to model a turbulent flow.

48
1

10

Bi2 112um
Bo2 109um
Ti2 110um
To2 111um
simA 120um
simB 110um

10

Reverseflow pressure drop, atm

Forwardflow pressure drop, atm

10

10

10

Bi2 112um
Bo2 109um
Ti2 110um
To2 111um
simA 120um
simB 110um

10

10

10

10

10

(b) Reverse flow in T45A valves

10

Li2 116um
Lo2 114um
Li2t2 116um
Lo2t2 114um
simA 120um
simB 115um

10

Reverseflow pressure drop, atm

Forwardflow pressure drop, atm

10

10

Li2 116um
Lo2 114um
Li2t2 116um
Lo2t2 114um
simA 120um
simB 115um

10

10

10

10

10

10
Reynolds number

(c) Forward flow in T45C valves

(d) Reverse flow in T45C valves

10

10

Si 144um
So 149um
Ri2 152um
sim 148um

Reverseflow pressure drop, atm

Forwardflow pressure drop, atm

10

Reynolds number

10

10

10

10
Reynolds number

(a) Forward flow in T45A valves

10

10

Reynolds number

Si 144um
So 149um
Ri2 152um
sim 148um
0

10

10

10

10
Reynolds number

(e) Forward flow in Deep T45C valves

10

10

10
Reynolds number

(f) Reverse flow in Deep T45C valves

Figure 4.8: Pressure drop versus Reynolds number based on hydraulic diameter. Symbols
are experimental and numerical data; curves are the fitted power-law relation (Eq. 4.1).
Legends refer to test name and etch depth; numerical simulations are marked sim.

49

4.3 Discussion
The primary task of this chapter was to demonstrate that the numerical method has been
correctly implemented in this research and thus accurately predicts the flow-rate response to
applied pressure and the diodicity as required by research objectives #1b and #1c. This validation was achieved by comparing volume flow-rate and diodicity predictions to 242 independent experimental measurements from physical-realized valves of three distinct groups
of etch depths and two distinct Tesla-type designs, the T45A and the T45C. The tests of the
physical-realized valves were divided into three groups considering etch depth and valve
design: the T45A group, the T45C group, and the Deep T45C group. In each group the
tested devices are representative samples of the set since the deviations from the mean etch
are much less than the measurement accuracy of 6.5%.
For study of the diodicity mechanism in the following chapters, the numerical simulations of the T45A and T45C valves were performed at the nominal etch depth of 120m.
To ensure that experimental measurements and numerical simulation results could be directly compared, three additional sets of numerical simulations were performed using the
mean etch depth of each of the valve test groups: h 110 5m for the T45A Test Group,


h 115m for the T45C Test Group, and h 148 3m for the Deep T45C Test Group.
It is also important to note that exactly the same numerical method was employed in all


cases. The only variation between simulations was the substitution of the physical grid
of the T45A or T45C scaled in the thickness dimension to match the appropriate mean
etch-depth.
Analysis of the experimental data showed that the power-law function of Eq. 4.1 provides a good estimate of the true functional relationship between measured pressure-drop
and volume flow-rate and there is no justification to use a more complicated function. Thus
the power-law function was used to represent the numerical methods predictions of flowrate response to applied-pressure. Knowledge of a good representative function allowed
flow-rate predictions to be produced from fewer numerical simulations.
Analysis of the numerical simulation results showed that the flow-rate response is underpredicted by the numerical method by 7.62% on average over all three test groups:
T45A, T45C, and Deep T45C. This is good agreement considering that the etch depth
measurement accuracy is 6.5% and that, in order to be a predictive tool, the numerical
model is gridded-up from the ideal mask layout drawing (see Ch. 3) instead of the asetched physical devices. Research objective #1b has been accomplished by verifying the
numerical method accurately predicts the steady-flow response to an applied pressure.

50
The diodicity is insensitive to any Reynolds-number underprediction that is a similar
proportion of both forward-direction and reverse-direction flow. The diodicity is overpredicted by 4.14% on average over all three test groups: T45A, T45C, and Deep T45C. The
diodicity prediction error with 95% confidence is eDi 4 42 5 13% or 0 708 eDi
9 55% in the numerical simultations utilized in the diodicity mechanism study in the fol

lowing chapters. By making comparisons between predicted and experimental values from
multiple valve geometries we demonstrated that we are free to modify valve geometry (to
the same extent as the variation between the T45A and T45C) and still obtain accurate flow
rate and diodicity predictions from the numerical method. Thus the numerical method is
validated as a firm basis on which to accept or reject valve designs for the enhancement
of diodicity. Research objective #1c has been accomplished by verifying the numerical
method accurately predicts valve diodicity. Proof that it can reveal the diodicity mechanism remains the task of following chapters.
Finally, both the experimental measurements and numerical simulation results showed
that the flow in the valve is laminar despite the presence of separated flow and recirculation
regions, thus the governing equations employed in the numerical method are exact.

51

Chapter 5
VALIDATION OF THE NUMERICAL METHOD IN TRANSIENT
FLOW
This chapter verifies that the numerical method is time-accurate in harmonic-response
simulations. The approximation analysis shown in Table 2.2 and discussed in Sec. 2.4
suggests that the transient term of the kinetic-energy conservation equation is negligible for
oscillating water flows with frequencies as high as 10 kHz. This allows investigation of the
diodicity mechanism using inexpensive steady-state solutions to calculate the terms in the
conservation equation, (the steady form of Eq. 2.12), instead of computationally-intensive
harmonic-response solutions. To use the numerical method to verify that the transient
term is negligible, we first need to demonstrate that the numerical method produces timeaccurate harmonic-response simulations.
Since there are no analytical solutions for transient flow in an NMP valve nor experimental methods available to directly measure it, the accuracy of the transient response
predictions of the numerical method was verified by modeling the harmonic response of
2-D slot flow for which an analytical solution does exist. Then the numerical method was
applied to model harmonic flow in a T45A valve and show that the kinetic-energy transient
term is negligible.

5.1 Harmonic Response of a 2-D Slot


The exact solution for the velocity response to an oscillating pressure gradient in a 2-D slot
was described in Sec. 1.2.5. The dimensionless velocity profiles are given by Eq. 1.13.
Proceeding with the nondimensional variables of Sec. 1.2.5, the volume flow-rate Q T in


the slot was obtained by integrating the velocity profile over the slot height as in
1

QT

U dY


real


cosh 

i expiT 2

1
0

iY 

cosh 

i 

dY

52
After integration, the volume flow rate is
QT


2 real

i expiT


tanh 

i 

in which the only component of Q T that has a dependence on time is the exponential.
The maximum value of the applied oscillating pressure gradient K cos T occurs at T
0 2 4
. The corresponding amplitude of the volume flow rate is


2 real

i


tanh 

i 

(5.1)


The fluid impedance per unit length is a complex number that is the ratio of the applied
pressure gradient and the flow rate amplitude given by




i


tanh




(5.2)


which represents the particular case of oscillating flow in a slot, compared to Eq. 1.16
which represents the general case.
A harmonic response simulation was obtained for a 2-D slot following the numerical
methods described in Chap. 3, taking the slot height as 2h 90m, the slot length as
L 1mm , and the viscosity of water 1 719 10 3 N s m2 for 273 K. The amplitude of
the oscillating pressure was Pa 1 atm applied at a frequency of 10 kHz. Convergence
was reached in 25 periods of 12 time steps each with 100 inner iterations per time step. The


predicted velocity profiles were compared to the analytical solution given by Eq. 1.13.
Following the nondimensionalization scheme of Sec. 1.2.5 as above, Fig. 5.1 shows
the excellent agreement between the nondimensional velocity profiles from the numerical
methods harmonic-response simulation (symbols) and the exact analytical solution of Eq.
1.13 (lines) at 60 phase intervals of one cycle. The kinetic-Reynolds number is 8 6 and
the motion of the fluid near the wall is clearly out-of-phase with the flow near the centerline.
The predicted dimensionless volume-flow-rate amplitude was 1.854, which is 100.47% of


the exact solution obtained from Eq. 5.1. The phase lag of the flow rate response with
respect to the applied pressure difference across the slot was calculated by normalizing the
flow rate and pressure difference by their amplitudes (they are both sinusoidal), and finding
a that in a least squares sense minimized the error e Pnorm t
Qnorm t over


53
1.5

060 sim
060 exact
120 sim
120 exact
180 sim
180 exact
240 sim
240 exact
300 sim
300 exact
360 sim
360 exact

Velocity, U

0.5

0.5

1.5
0

0.5
1
Distance from centerline, Y

1.5

Figure 5.1: Comparison of the nondimensional velocity profiles in a slot with oscillating
flow, 8 6 , from the numerical method (symbols) and the exact solution (lines). The
centerline of the slot is at zero slot height and the slot wall is at slot height = 1. The legends
note the phase of each profile with respect to the applied pressure difference, a cosine
function.


0


t


2 The phase lag computed in this way was




84 96 , which is within 0.09%




of the exact solution determined from arctan imag Z real Z using Eq. 5.2. Clearly the
numerical solution agrees very well with the exact solution.


5.2 Harmonic Response of an NMP Valve


A harmonic response simulation of the T45A valve with a pressure amplitude of P

05


atm applied at 2.818 kHz for 6 cycles was performed using the numerical method described
in Chap. 3, including the nondimensionalization parameters of Sec. 3.5. The terms in the
conservation of kinetic-energy equation (Eq. 2.12) were calculated, including: the energy
dissipation rate (PHI), viscous work rate (VWR), energy flux rate (EFR), pressure work
rate (PWR), and the transient kinetic-energy (TKE). Figure 5.2 displays them and shows
that the transient kinetic-energy term is insignificant compared to the pressure work rate
and the dissipation rate.

54
PHI
VWR
EFR
PWR
TKE

0.08

Work rate

0.06

0.04

0.02

0.02

0.04
4

4.5

5
Cycle

5.5

Figure 5.2: Terms in the conservation of kinetic-energy equation (Eq. 2.12) for the entire
T45A valve over two cycles in a harmonic response simulation at 2.818 kHz, including: the
energy dissipation rate (PHI), viscous work rate (VWR), energy flux rate (EFR), pressure
work rate (PWR), and the transient kinetic-energy (TKE). The dissipation and pressure
work are dominant.

5.3 Discussion
The modeling of a 2-D slot has shown that the predictions of the harmonic simulation
agree with the analytical solution. This accomplishes research objective #1d to show that
the numerical method accurately predicts transient flow-rate response.
The approximation analysis for NMP valve flow shown in Table 2.2 and discussed
in Sec. 2.4 suggested that the transient term was negligible even for oscillating water
flows with frequencies as high as 10 kHz. This was corroborated by a harmonic response
simulation run at 2818 Hz that shows that the transient kinetic-energy term is insignificant
compared to the pressure work rate and the dissipation rate. The insignificance of transient
effects allows the use of relatively inexpensive steady-state simulations to calculate the
terms in the momentum and kinetic-energy conservation equations, (the steady forms of
Eqs. 2.5 and 2.12, respectively), instead of computationally-intensive harmonic-response
simulations. Thus steady-state simulations were used to study the diodicity mechanism in
the following chapters.

55

Chapter 6
DIODICITY MECHANISM OF T45A VALVE

6.1 Simulation Methods and Conditions

The diodicity mechanism of the T45A valve was analyzed by studying the results of the numerical methods explained in Chap. 3 from two perspectives: first, analysis of the velocity,
pressure, and dissipation-rate fields, and second, analysis of the terms of the momentum
and kinetic-energy conservation equations applied in 9 regional control volumes as shown
in Fig. 6.1.
Steady-state solutions of forward and reverse flow in the T45A valve were obtained and
analyzed using the methods described in Chap. 3. These steady-state solutions were obtained from the final solution of step-response simulations that were continued until steady
flow was achieved. In all cases the simulations converged to steady solutions without applying Reynolds averaging to the Navier-Stokes equations. As discussed in Sec. 1.2.2,
this identifies the flow in the valve as laminar over the Reynolds number range simulated,
Re 953, despite the presence of separated flow and recirculation regions. To assess grid
independence, the grid was refined from 96360 to 115632 finite volumes, an increase of
20%, resulting in a reduction in the kinetic-energy conservation error KE 0 06 of less


than 5%. The accuracy of the flow-rate response predictions of the numerical method was
determined in Ch. 4.
To represent typical operating conditions a differential pressure of 0.5 atm was applied
to the pressure boundaries, producing an equilibrium volume flow rate in the reverse direction of 3710 l min corresponding to Re D 528 based on the hydraulic diameter of
the main channel. To achieve the same flow rate in the forward-flow direction required a
differential pressure of only 0.394 atm. This corresponds to a diodicity according to Eq.
1.1 of 1.27. According to the analysis of diodicity prediction accuracy in Ch.4, there is a
95% probability that the true diodicity is within 1 15 Di 1 28.


56

Figure 6.1: Division of the T45A valve into regional control volumes.

6.2 Results
6.2.1 Velocity Field
From the step-response simulations, Fig. 6.2 shows the forward and Fig. 6.3 the reverseflow velocity fields on the symmetric centerplane of the valve. The velocity fields in the
forward and reverse flow cases are radically different. Flow separation and laminar jets
occur in three locations: where the channels separate, the channels recombine, and at the
valve exit.
In the forward-flow case the flow accelerates rapidly as it enters the main channel from
the goblet-shaped inlet plenum. As the velocity profile is developing it reaches the Tjunction. The flow begins to veer slightly into the side channel, and a small portion impinging on the guide vane becomes a minor jet flowing into the side channel. However,
85% of the main channel flow is unperturbed and continues downstream becoming fullydeveloped. The side channel jet spreads as it proceeds around the bend in the side channel,
and approaches the Y-junction as a low-velocity stream filling the entire width of the side
channel, moving at less than 20% of the bulk velocity of the main channel. However, it
has sufficient momentum to cause the main channel flow to begin to turn before it reaches
the far wall of the Y-junction. This may have a small effect in improving the diodicity as
it tends to increase the radius of curvature of the main channel flow as it travels around the
45 bend into the outlet channel. Downstream of the Y-junction, the flow separates from
the inner wall and forms a narrow, hi-speed jet next to the outer wall. The high velocity
gradient between the jet and the wall produces additional dissipation that lowers the diod-

57

MENUWINDOW
Z-SC-ROT CLK
X-SC-ROT CLK
Y-SC-ROT CLK
LEFT MOVE
UP MOVE
NEARMOVE
EXPAND
RESET VIEW
BACKCOLOUR
MARKER SIZE
COARSENESS

8.5000E-01
7.0833E-01
5.6667E-01
4.2500E-01
2.8333E-01
1.4167E-01
0.0000E+00

Figure 6.2: Forward-flow velocity field on the centerplane of a single-element, Tesla-type


T45A valve with a volume flow rate of 3710 l/min corresponding to Re=528 based on the
hydraulic diameter of the main channel. One dimensionless unit equals 10 m/s.

58
icity of the valve. A significant portion of the channel is filled with a quiescent zone and
is essentially underutilized by the forward flow. It may improve diodicity to narrow this
section of the outlet channel so it cannot be utilized by the reverse flow. The flow leaves the
outlet channel and enters the goblet-shaped outlet plenum as a high-speed jet. Any of its
momentum flux that is dissipated increases the pressure drop in the forward-flow direction
and lowers diodicity.
In the reverse-flow case the flow accelerates rapidly as it enters the channel from the
goblet-shaped plenum, just as it does in the forward flow case. On reaching the Y-junction
the flow stream begins to veer slightly toward the main channel. As it impinges on the
cusp of the guide vane, 36% of the flow is deflected down the main channel where it travels
next to the guide vane wall, but the vast majority of the main channel is filled with a
large, slowly-moving recirculation zone. The side channel flow separates from the outer
wall and forms a laminar jet along the guide vane wall. The large velocity gradient there
increases dissipation, which improves diodicity. At the bend in the side channel, the jet
separates from the guide vane and attaches to the outer side-channel wall as it undergoes
radial acceleration. The flow emerges from the side channel as a laminar jet heading for
the opposite wall of the main channel. But the momentum of the smaller flow traveling
down the main channel turns the jet before it reaches the opposite wall. There is a large,
vigorous recirculation zone downstream of the T-junction, however the jet spreads to fill
the entire channel as it leaves the channel and enters the goblet-shaped plenum. Any of
the momentum flux at the channel exit that is eventually dissipated serves to increase the
diodicity.

6.2.2 Pressure Field


In contrast to the velocity fields there are similarities between the forward and reverse-flow
pressure fields of the T45A valve shown in Fig. 6.4, partly due to the symmetrical inlet
and outlet geometry. Significant pressure loss is apparent where the flow enters the channel
regardless of the flow direction. In fact this is a major source of pressure loss in either flow
direction. There is only one other signficant pressure loss in each case: in forward flow it
is at the 45 bend in the channel, in reverse flow it is just downstream of the T-junction.
Between these locations and their corresponding flow exits there is little additional pressure
drop; the momentum flux maintains the fluid velocity until it reaches the goblet-shaped
plenum and begins to diminish. Closer study of the pressure fields shows that pressure
drop at the 45 bend in forward flow is approximately 0.1 atm, and downstream of the

59

MENUWINDOW
Z-SC-ROT CLK
X-SC-ROT CLK
Y-SC-ROT CLK
LEFT MOVE
UP MOVE
NEARMOVE
EXPAND
RESET VIEW
BACKCOLOUR
MARKER SIZE
COARSENESS
8.5000E-01
7.0833E-01
5.6667E-01
4.2500E-01
2.8333E-01
1.4167E-01
0.0000E+00

Figure 6.3: Reverse-flow velocity field on the centerplane of a single-element, Tesla-type


T45A valve with a volume flow rate of 3710 l/min corresponding to Re=528 based on the
hydraulic diameter of the main channel. One dimensionless unit equals 10 m/s.

60
T-junction in reverse flow is approximately 0.2 atm. This is the main source of diodicity.
Thus, as suggested by Eq. 1.2, if the direction-independent losses at the channel entrances
could be eliminated, the diodicity would be increased from 1 27 to 2 0. Though they


cannot be completely eliminated, it would be helpful to adjust the mouth of the channel
entrances so that the forward-flow entrance loss is less than that of the reverse flow.

6.2.3 Energy-Dissipation Field


The base 10 logarithm of the energy dissipation rate is shown in Figs. 6.5 and 6.6. The
red and yellow contours are the most significant and occur where the velocity gradients are
largest. In the forward-flow case, dissipation occurs along the walls of the main channel
and on both sides of the laminar jet downstream of the 45 bend. In the outlet goblet
plenum, the dissipation of the momentum flux leaving the exit is clearly visible. But the
most significant dissipation occurs at the convex surfaces of the channel mouth where the
flow accelerates as it enters the channel, at the upstream cusp of the guide vane, at the
inner corner of the 45 bend, and along the wall downstream of the 45 bend all the way to
channel exit. These locations correspond well to the locations of large pressure drop in Fig.
6.4, but it is not conclusive that dissipation is the most important source of that pressure
loss.
The locations of dissipation in the reverse flow case are quite different than in forward
flow. The only similar region is the dissipation along the walls of the channel from where
the flow enters the channel to the 45 bend. As the flow bifurcates at the Y-junction, the
dissipation is spread more evenly in the side channel than elsewhere, but very significant
dissipation occurs along the surfaces of the guide vane, both in the side channel and the
main channel, and especially at its leading cusp. There is additional major dissipation in
the side channel as the separated jet follows the curvature of the outer wall. This dissipation continues alongside the jet proceeding from the side channel into the main channel,
collocated with the large velocity gradient between the jet and the recirculation zone in the
main channel downstream of the T-junction. There is also significant dissipation where the
jet exiting from the side channel impinges on the opposite wall of the main channel. As in
the forward-flow case, part of the momentum flux leaving the channel is dissipated in the
goblet-shaped plenum.

61
MENUWINDOW
Z-SC-ROT CLK
X-SC-ROT CLK
Y-SC-ROT CLK
LEFT MOVE
UP MOVE
NEARMOVE
EXPAND
RESET VIEW
BACKCOLOUR
MARKER SIZE
COARSENESS

AXES OFF
Z-SC-ROT ANT
X-SC-ROT ANT
Y-SC-ROT ANT
RIGHT MOVE
DOWNMOVE
FAR MOVE
SHRINK

MENUWINDOW
Z-SC-ROT CLK
X-SC-ROT CLK
Y-SC-ROT CLK
LEFT MOVE
UP MOVE
NEARMOVE
EXPAND
RESET VIEW
BACKCOLOUR
MARKER SIZE
COARSENESS

AXES OFF
Z-SC-ROT ANT
X-SC-ROT ANT
Y-SC-ROT ANT
RIGHT MOVE
DOWNMOVE
FAR MOVE
SHRINK

LINE WIDTH
TEXT SIZE

5.2000E-01
4.1811E-01
3.1623E-01
2.1434E-01
1.1245E-01
1.0566E-02
<-8.0000E-02

(a) Forward flow

LINE WIDTH
TEXT SIZE

> 5.2000E-01
4.2769E-01
3.2385E-01
2.2000E-01
1.1615E-01
1.2308E-02
<-8.0000E-02

(b) Reverse flow

Figure 6.4: Pressure field [atm] on the centerplane of a single-element, Tesla-type T45A
valve with a volume flow rate of 3710 l/min corresponding to Re=528 based on the hydraulic diameter of the main channel.

62

MENUWINDOW
Z-SC-ROT CLK
X-SC-ROT CLK
Y-SC-ROT CLK
LEFT MOVE
UP MOVE
NEARMOVE
EXPAND
RESET VIEW
BACKCOLOUR
MARKER SIZE
COARSENESS

-4.0000E-02
-1.3667E+00
-2.6933E+00
-4.0200E+00
-5.3467E+00
-6.6733E+00
-8.0000E+00

Figure 6.5: Base 10 logarithm of the energy dissipation rate in forward flow on the centerplane of a single-element, Tesla-type T45A valve with a volume flow rate of 3710 l/min
corresponding to Re=528 based on the hydraulic diameter of the main channel. One dimensionless unit equals 14 mW.

63

MENUWINDOW
Z-SC-ROT CLK
X-SC-ROT CLK
Y-SC-ROT CLK
LEFT MOVE
UP MOVE
NEARMOVE
EXPAND
RESET VIEW
BACKCOLOUR
MARKER SIZE
COARSENESS
-4.0000E-02
-1.3667E+00
-2.6933E+00
-4.0200E+00
-5.3467E+00
-6.6733E+00
-8.0000E+00

Figure 6.6: Base 10 logarithm of the energy dissipation rate in reverse flow on the centerplane of a single-element, Tesla-type T45A valve with a volume flow rate of 3710 l/min
corresponding to Re=528 based on the hydraulic diameter of the main channel. One dimensionless unit equals 14 mW.

64
6.2.4 Momentum Conservation
Diodic Effects of the Force Vector Interactions
The momentum equation is a vector relation, a summation of forces, so each of its terms
has both magnitude and direction sense. The X and Y components of these forces are
shown in Fig. 6.7 for both the forward and the reverse flow case. The figure caption
explains positive-negative direction sense and the X-Y coordinate system is shown in Fig.
6.1. Essentially, the figures show the viscous and pressure forces applied on the fluid by
the walls and inlet/outlet boundaries, and the fluids momentum flux response. Both the
momentum flux and viscous force are important in counterbalancing the pressure force.
The goblet-shaped inlet plenum (block 1) is oriented for flow in the X direction, so
as Fig. 6.7 shows, the Y-direction force components are negligible. In forward flow the
pressure force is to the right (positive) and it is expended by accelerating the fluid as it
approaches the inlet channel and by a small amount of viscous force. The momentum flux
is to the right (positive), but more is leaving the control volume than entering (defined as
negative), so its overall sign is negative. In reverse flow the pressure is already near ambient
when it enters the block 1 control volume, so the pressure force is negligible. The viscous
force applied by the walls is also very small, so the momentum flux is unchanged as the
flow passes through block 1.
The inlet channel (block 2) is also oriented in the X direction but due to its proximity
to the side channel (block 8) it has Y-direction force components during reverse flow. The
positive-Y pressure force is applied by the wall on the jet of fluid leaving the side channel, altering its direction so that it flows out of block 2 and into block 1. The momentum
of this jet is directed into the inlet channel (positive) but is flowing in the negative-Y direction (negative) resulting in a net negative Y-component of the momentum flux. In the
X-direction in both forward and reverse flow, the upstream flow boundary applies the pressure force needed to drive the fluid through the inlet channel offset by the viscous force
and momentum flux. In both flow directions the viscous force is larger than the momentum flux. Unfortunately, the X-direction pressure force is larger in forward flow than in
reverse, which has an adverse impact on valve diodicity. It is not clear from the momentum
perspective what impact the Y-direction pressure force has on diodicity.
The T-junction (block 3) does not have significant Y-direction forces, a surprising result
since this control volume connects the side channel (block 8) to the inlet channel and the
main channel (block 4). However, inspection of Fig. 6.3 shows that the jet from the side
channel during reverse flow does not significantly interact with the wall in block 3, but

65

Momentum Flux
Viscous Force
Pressure Force

0.2
0.15
0.1
0.05
0
0.05
0.1
0.15
0.2

0.25
ReverseFlow Xdirection Forces

ForwardFlow Xdirection Forces

0.25

0.25

Momentum Flux
Viscous Force
Pressure Force

0.2
0.15
0.1
0.05
0
0.05
0.1
0.15
0.2
0.25

4
5
6
Block number

(a) X-direction, forward flow

Momentum Flux
Viscous Force
Pressure Force

0.2

4
5
6
Block number

(b) X-direction, reverse flow

0.15
0.1
0.05
0
0.05
0.1
0.15
0.2
0.25

0.25
ReverseFlow Ydirection Forces

ForwardFlow Ydirection Forces

0.25

Momentum Flux
Viscous Force
Pressure Force

0.2
0.15
0.1
0.05
0
0.05
0.1
0.15
0.2
0.25

4
5
6
Block number

(c) Y-direction, forward flow

4
5
6
Block number

(d) Y-direction, reverse flow

Figure 6.7: Force vector terms in the integral form of the momentum conservation equation
for the T45A valve. Net pressure force and net momentum flux into a control volume are
positive. Viscous force is applied on the fluid by the wall. X-vectors to the right and Yvectors upward are positive and consistent with the valve layout in Fig. 6.1 including the
numbering of the control volumes (blocks).

66
passes through to block 2. However the X-direction forces are important in creating diodicity. In forward flow there is a small amount of pressure recovery as the momentum flux
drives the flow through the area expansion of block 3; this negative pressure force increases
diodicity as it lowers the net valve pressure force for forward flow. The reverse flow has a
much greater positive impact on diodicity; the momentum flux leaving block 3 (negative)
and flowing in the negative X direction is the jet from the side channel, and results in a
positive net momentum flux for block 3 as seen in Fig. 6.7b. A large pressure force is
needed to offset this momentum flux and this is a major source of diodicity.
The main channel (block4) has almost identical Y-direction force components in forward and reverse flow. In both cases the upper wall, which forms part of the guide vane
or island of the valve, applies a pressure force in the negative-Y direction to the impinging
flow, which has just passed an opening to the side channel. This impingement can be seen
in both Figs. 6.2 and 6.3. It is not clear from the momentum perspective if this affects
diodicity. In the X direction the force components are very different between forward and
reverse flow. In forward flow most of the volume flow is carried by the main channel,
and as a result significant pressure force is needed to counteract viscous force due to wall
friction. This has a negative impact on diodicity as most of the reverse flow occurs in the
side channel and almost no pressure force is required to push the remaining reverse flow
through the main channel. The viscous force due to wall friction is supplied by reduction
in momentum flux.
The Y-connection (block 5) has negligible X-direction force components. In forward
flow the net Y-direction flux of momentum is out of the control volume (negative) and in
the negative Y direction resulting in a positive value, which is counteracted by the pressure
force applied by the outer wall. In reverse flow some of the flow is directed down the main
channel (block 4) by the guide vane, thus there is a net flux of momentum into the control
volume in the positive Y direction, so the momentum flux again has a positive Y sense.
The outlet channel (block 6) is similar in form and function to the inlet channel (block2),
except that it is oriented at 45 cw with respect to the positive X direction. Because of different orientations it is difficult to compare the forces in blocks 2 and 6 using Fig. 6.7, but
a comparison is made below using the vector magnitudes shown in Fig. 6.8 to assess the
impact on diodicity. Essentially, the pressure force applied at the upstream flow boundary is roughly equally opposed by the viscous force applied by the walls and by the net
momentum flux leaving the control volume.
The goblet-shaped outlet plenum (block 7) is analogous to the inlet plenum (block 1)
except for its orientation of 45 cw with respect to the positive X direction. The forward flow

67
forces are small; the reverse-flow net pressure force accelerates the fluid as it approaches
the next downstream control volume (block 6). All the viscous forces are negligible.
The side channel (block 8) has negligible forces in forward flow because almost the
entire volume flow passes through the main channel instead. In reverse flow, Fig. 6.7d
shows a very large pressure force applied by the outer wall to alter the direction of the
impinging flow in the side channel. Accelerated by the outer wall (see Figs. 6.3 and 6.4), a
correspondingly large net momentum flux is propelled from the side channel into block 3.
Because these are Y-direction forces it is difficult to assess their impact on diodicity. The
last control volume (block 9) has significant forces only in the reverse-flow direction.

Forward-Flow vs. Reverse-Flow Force Vector Magnitudes


Another way to study the diodicity mechanism, especially for control volumes that are not
oriented along the X direction, was to study the vector magnitudes of the force components
in the control volumes as shown in Fig. 6.8, which directly compares the forward-flow and
reverse-flow components.
Especially pertinent is the comparison of the inlet and outlet channels (blocks 2 and 6)
and the inlet and outlet goblet-shaped plenums (blocks 1 and 7), which could not be directly
compared in the previous section because of their different orientations. Comparing the
goblet-shaped plenums, the viscous forces applied by the walls are small. The significant
forces are the pressure forces and momentum fluxes; those of the forward flow jet entering
block 1 and the reverse flow jet entering block 7 are an order of magnitude larger than all
other forces. Though it appears in Fig. 6.8a that slightly more pressure force is applied in
the plenums in forward flow than in reverse, the combined diodic effect of blocks 1 and 7
is actually slightly above unity, Di1 7


1 01


The inlet and outlet channels (blocks 2 and 6) function similarly as entrance and exit
channels due to their geometric similarity. One difference is the slightly greater net momentum flux into block 2 from the side channel jet during reverse flow compared to the
net momentum flux into block 6 from the main channel jet during forward flow; this can
be seen in Fig.6.8c. Another difference is the smaller viscous force applied by the walls of
block 2 during reverse flow, probably due to its isolation by the recirculation zone that can
be seen in Fig. 6.3. The net diodic effect is slightly above unity, though this is not obvious
on the pressure force magnitude plot of Fig.6.8a.
The main channel (block 4) and the Y-junction (block 5) have larger pressure force
magnitudes in forward flow than in reverse flow, and thus have a diodic effect less than

68

reverse flow
forward flow

0.2

0.15

0.1

0.05

reverse flow
forward flow

0.25
Viscous Force Magnitude

Pressure Force Magnitude

0.25

0.2

0.15

0.1

0.05

4
5
6
Block number

(a) Pressure force magnitude

4
5
6
Block number

(b) Viscous force magnitude

reverse flow
forward flow

0.25
Momentum Flux Magnitude

0.2

0.15

0.1

0.05

4
5
6
Block number

(c) Momentum flux magnitude

Figure 6.8: Vector magnitudes of the terms in the momentum conservation equation for the
T45A valve.

69
unity. In block 4 this is because almost all the forward flow proceeds through the main
channel instead of bypassing it as it does in reverse flow. As a result the viscous force
applied by the wall is much almost twice as large in forward flow as it is in reverse. In
block 5 the larger pressure force in forward flow is required offset the larger momentum
flux and viscous force, which are due to the forward flow changing direction in block 5,
whereas in reverse flow it proceeds straight through toward the side channel.
The major contributors to diodicity are the T-junction (block 3) and the side channel
(blocks 8 and 9). The T-junction is where the side channel jet emerges and must be redirected and forced out through block 2, the result is a large net increase in momentum flux
and additional viscous force applied by the walls that both must be overcome by the pressure force. This pressure force needed to redirect the jet from the side channel is not needed
in forward flow, since the forward-flow jet in block 6 is already pointed downstream. The
large increase in pressure across block 3 can be seen in Fig. 6.4b. The side channel requires
much larger pressure force in reverse flow than in forward. In part this is due to there being
very little forward flow through the side channel, so of course there is much more viscous
force applied by the walls in reverse flow and the possibility of greater change in the momentum flux in the side channel. But the great increase in momentum flux is due to the
fluid impinging on the outer wall as it curves back toward the main channel. The jet that
emerges from the side channel has much higher velocity than when it entered as can be
seen in Fig. 6.3.

6.2.5 Kinetic-Energy Conservation


The calculation of the kinetic energy terms for the entire valve showed that two-thirds of
the pressure work is dissipated in the valve. The ratio of the dissipation rates (reverse flow
/ forward flow) is 1.13 showing that viscous effects within the valve are a significant source
of diodicity. The ratio of energy flux rates is 1.91 showing that the reverse-flow jet contains
more energy, and momentum, than the forward-flow jet. However, the energy flux rates are
only one-third the magnitude of the dissipation rates, so the overall valve diodicity is 1.27.
The diodicity prediction from the ratio of the pressure work rates as in Eq. 2.14 is 1.25,
which is within 2% of 1.27.
There are two alternate ways to look at the data. Figs. 6.9 and 6.10 show the three most
important terms of the kinetic energy equation in steady incompressible flow: the pressure
work rate, the energy flux rate, and the dissipation rate, which are shown for each of the
control volumes for both forward and reverse flow. Blocks 3, 6, 7, 8, and 9 were identified

70

reverse flow
forward flow

Pressure Work Rate

0.05

0.05
1

4
5
6
Block number

(a) Pressure work rate

reverse flow
forward flow

Energy Flux Rate

0.05

0.05
1

4
5
6
Block number

(b) Energy flux rate

Figure 6.9: Magnitude of the pressure work rate and the energy flux rate terms in the
kinetic-energy conservation equation from the T45A valve simulations with a volume flow
rate of 3710 l/min corresponding to Re=528 based on the hydraulic diameter of the main
channel.

71
as sources of diodicity, since the pressure work applied in the reverse flow direction is
larger than that applied in the forward direction. On the other hand, blocks 1, 2, 4, and
5 cause a significant reduction in diodicity. The energy flux rate has opposite sign (+ is a
net increase, - is a net decrease in the control volume) depending on flow direction in all
blocks except block 2. The dissipation rate also exhibits direction dependence, especially
in blocks 2, 6, and 8. The viscous work rate is inconsequential in all control volumes.
An alternate view of the kinetic energy in the valve is the contribution of each term of
the kinetic energy equation to the forward-flow direction and the reverse-flow direction as
shown in Fig. 6.11. Note that the pressure work rate typically must balance all the other
energy rate terms.
Blocks 1-2 and 6-7 fullfil similar roles as valve entrance and valve exit regions in forward and reverse flow. Figure 6.11 shows that the pressure work rate is smaller in blocks 6
and 7 in reverse flow than in blocks 1 and 2 in forward flow. The ratio is 0.9, a reduction
in diodicity. They are geometric similar flow paths, both serving to accelerate the flow as
it enters the channel and develops its flow profile, but the forward-flow entrance, block 2 is
14% longer than block 6. This suggests block 2 should be shortened to improve diodicity.
However, blocks 2 and 6 also have a role as valve exits and in this role each contains a separated jet as can be seen in Figs. 6.2 and 6.3 with accompanying high rates of dissipation.
In fact the dissipation rate in the combined blocks 2 and 6 is 10% higher in forward than in
reverse flow; a negative impact on diodicity. Yet when the flow in the combined blocks 1,
2, 6, and 7 is considered, the ratio of pressure work rates in reverse flow relative to forward
flow is 1.08, an increase of diodicity. Study of Fig. 6.11 reveals that in forward flow the
energy flux rate into the outlet channel (block 6) is due to its laminar jet, whose momentum
flux is oriented approximately parallel to the downstream direction, but in reverse flow,
pressure work must be expended in block 2 to create energy flux. Figure 6.3 shows the
reason why; the laminar jet from the side channel is nearly perpendicular to the main channel, so the jets momentum flux does little to force the flow downstream. Thus diodicity
is created by the asymmetry; the forward-flow outlet-channel jet is aimed downstream but
the reverse-flow side-channel jet is not. If both blocks 2 and 6 were shortened, their large
dissipation rates would be diminished, and the positive diodic effect of the asymmetry of
energy flux rates and momentum flux directions, would not be so diluted.
The flow in the main channel (blocks 4 and 5) is radically different in the forward and
reverse directions. In the forward direction the main channel contains 85% of the valve
flow, and the pressure work is converted to energy dissipation and energy flux, 58% to
dissipation and 42% to flux. In the reverse direction only 36% of the valve flow is carried

72

reverse flow
forward flow

Dissipation Rate

0.05

0.05
1

4
5
6
Block number

(a) Energy dissipation rate

reverse flow
forward flow

Viscous Work Rate

0.05

0.05
1

4
5
6
Block number

(b) Viscous work rate

Figure 6.10: Magnitude of the energy dissipation rate and the viscous work rate terms in
the kinetic-energy conservation equation from the T45A valve simulations with a volume
flow rate of 3710 l/min corresponding to Re=528 based on the hydraulic diameter of the
main channel.

73

Energy Flux Rate


Dissipation Rate
Viscous Work Rate
Pressure Work Rate

ForwardFlow Power

0.05

0.05
1

4
5
6
Block number

(a) Forward flow

Energy Flux Rate


Dissipation Rate
Viscous Work Rate
Pressure Work Rate

ReverseFlow Power

0.05

0.05
1

4
5
6
Block number

(b) Reverse flow

Figure 6.11: Magnitude of the terms in the kinetic-energy conservation equation from the
T45A valve simulations with a volume flow rate of 3710 l/min corresponding to Re=528
based on the hydraulic diameter of the main channel.

74
by the main channel, and it is the energy flux that is converted to energy dissipation and
the pressure work needed to sustain the flow rate. The main-channel flow has a negative
impact on diodicity, but, as consequence of adopting a fixed geometry for the valve, it must
exist during reverse flow.
The most significant diodicity mechanism is located in the side channel and the Tjunction (blocks 3, 8, and 9) due to the radical asymmetry between the forward and reverse
flow there. In forward flow the only significant energy interchanges occur in the T-junction
(block 3) where 1/3 of the energy flux in the T-junction is dissipated and the remainder
is converted to pressure work to sustain the volume flow rate as the forward flow passes
through the area expansion of the T-junction and on into the main channel. In reverse
flow 1/3 of the pressure work rate of the entire valve is utilized to drive 64% of the valve
flow through the side channel and the T-junction. Virtually all the pressure work in the
side channel is dissipated in the large velocity gradient between the side-channel jet and
the outer wall of the side channel. The side-channel jet can be seen in Fig. 6.3, and the
corresponding dissipation in Fig. 6.6. Most of the pressure work (85%) in the T-junction is
converted to additional energy flux of the side-channel jet as it is turned to flow out through
block 2. The remainder of the pressure work is dissipated. The overall result is a strong
diodic effect as energy flux is transformed to pressure work in forward flow, but in reverse
flow 1/3 of the entire valves pressure work is dissipated or transformed to energy flux in
the side channel and T-junction alone. Considering the entire valve, 2/3 of the pressure
work is dissipated and 1/3 transformed to energy flux.

6.3 Discussion
From the preceeding qualitative analysis it is clear that the diodicity mechanism is located
in the side channel and T-junction (blocks 3, 8, 9), and that the rest of the valve adds to the
overall fluid resistance but has diodic effects near unity. As was suggested in Eq. 1.2, these
valve channels dilute the overall valve diodicity. For example, the goblet-shaped plenums
(blocks 1 and 7) and the inlet and outlet channels (blocks 2 and 6) exhibit a small amount
of diodicity, Di 1 1. Both the viscous forces and the dissipation rate are 10% higher in
forward flow than in reverse, but this adverse impact on diodicity is offset by the energy flux


rate and by the momentum flux because the laminar jet created in forward flow is pointed
downstream and the reverse flow jet is not. If these channels were not needed to direct the
fluid into and out of the valve, the overall valve diodicity would be closer to 2.0, instead of
1.27.

75
The underlying physical basis for the diodicity of the T45A valve is its ability to direct
the majority of the volume flow through alternate paths depending on flow direction. In
forward flow 85% of the fluid proceeds through the main channel (blocks 3 4 5); in
reverse flow 64% proceeds through the side channel (blocks 5 9 8 3). These two
alternate flow paths were compared quantitatively to determine the nature of the diodicity
mechanism. The mechanism is partly the viscous forces applied by the wall on the fluid;
the vector magnitude of the viscous force in reverse flow is 56% larger than in forward.
However, the energy dissipation rate is 97% larger in reverse flow than in forward. So the
diodicity mechanism is more due to dissipation in the fluid itself than to wall friction. The
energy flux rate is also 97% larger in reverse flow than in forward in these four blocks, but
it is only 12% of the rate of energy dissipation and its impact is small.
We satisfy research objective #1e, the description of the diodicity mechanism, by assembling the results of the field variable analysis, the momentum perspective and the
kinetic-energy perspective. The velocity field showed that there is a forward-flow jet in
the outlet channel and a reverse-flow jet in the inlet channel. The pressure field showed
that a large pressure drop occurs at the T-junction in reverse-flow, but not in forward flow.
The energy-dissipation-rate field showed substantial dissipation in the side channel and increased dissipation in the inlet channel during reverse flow. The momentum perspective
showed that substantial pressure force was needed to alter the direction of the reverse flow
jet, but not the forward flow jet, which is already pointed downstream. The kinetic-energy
perspective showed that most of the transformation of pressure work to energy dissipation
occurs in the shear gradient surrounding the laminar jets. Thus, the diodicity mechanism is
due to the asymmetry between a weaker forward-flow laminar jet pointed downstream and
a stronger reverse-flow laminar jet that is not pointed downstream and has a higher rate of
energy dissipation in the shear layer surrounding it.
Paul [23] performed a control-volume analysis of the region surrounding the T-junction
of a macro-scale Tesla-type valve with turbulent flow, see Sec. 1.3.1, in which he assumed
viscous forces were negligible and removed the viscous force term from the momentum
equation. He obtained an analytical solution for the pressure drop due solely to the change
in momentum flux, and agreement with his experimental data was good, showing the diodicity mechanism was due only to inertial forces. Lets compare his macro-scale case with
flow in this micro-scale Tesla-type valve. Study of the pressure field from the T45A numerical simulation shown in Fig. 6.4 reveals that the maximum pressure gradient occurs near
the T-junction where the channels recombine in reverse flow, the same location that Paul
assumed. In control volumes 2, 3, and 4, which are comparable to Pauls control volume,

76
43% of the pressure force is required to overcome the viscous force in the x-direction, leaving 57% to create momentum flux. Thus the viscous and inertial forces are of comparable
magnitude at Re 528 as anticipated by approximation theory and shown in Table 2.1.
In Pauls macro-valve at Re 10 000 there exists a turbulent jet instead of a laminar jet,
and the momentum flux would be four orders-of-magnitude larger than the viscous force,
but as the Reynolds number decreases below 1000, the viscous force becomes ever more
dominant over the momentum flux. Ignoring viscous forces in the low-Reynolds-number
flow of microvalves at this scale, ie. channel widths of 100 m, would not be valid. In
fact, if the entire valve is considered, fully 2/3 of the pressure work is dissipated and only
1/3 is transformed to energy flux. Clearly, viscous effects dominate inertial effects even at
Re

500 in Tesla-type valves of this scale, satisfying research objective #2.

77

Chapter 7
DIODICITY MECHANISM OF T45C VALVE
7.1 Simulation Methods and Conditions
The diodicity mechanism of the T45C valve was revealed by studying the steady-state
solutions of the numerical methods explained in Chap. 3 from two perspectives: first,
analysis of the velocity, pressure, and dissipation-rate fields, and second, analysis of the
terms of the momentum and kinetic-energy conservation equations applied in 9 regional
control volumes as shown in Fig. 7.1. It is important to note that the numerical methods
employed to analyze the T45C were exactly identical to those used in the T45A analysis.
The only variation is the substitution of the physical grid of the T45C for that of the T45A.
The flow behavior and characteristics are compared throughout this chapter with those of
the T45A valve.
Steady-state solutions of both forward flow and reverse flow in the T45C valve were
calculated using the numerical methods described in Chap. 3. These steady-state solutions
were obtained from the final solution of step-response simulations that were continued until
steady flow was achieved. In all cases the simulations converged to steady solutions without
applying Reynolds averaging to the Navier-Stokes equations. As discussed in Sec. 1.2.2,
this identifies the flow in the valve as laminar over the Reynolds number range simulated,
Re 748, despite the presence of separated flow and recirculation regions. To assess grid
independence, the grid was refined from 94380 to 127008 finite volumes, an increase of
34.6%, resulting in a reduction in the kinetic-energy conservation error KE 0 05 of less
than 5%. The accuracy of the flow-rate response predictions of the numerical method was


determined in Ch. 4.
To represent typical operating conditions a differential pressure of 0.5 atm was applied
to the pressure boundaries, producing an equilibrium volume flow rate in the reverse direction of 3640 l min corresponding to ReD 519 based on the hydraulic diameter of the
main channel. To achieve the same flow rate in the forward-flow direction required a differential pressure of only 0.377 atm. This corresponds to a diodicity Di 1 33 according to
Eq. 1.1. According to the analysis of diodicity prediction accuracy in Ch.4, there is a 95%


78

Figure 7.1: Division of the T45C valve into regional control volumes.

probability that the true diodicity is within 1 20 Di 1 34. The increase in predicted
diodicity over that of the T45A valve Di 1 27 is not statistically significant.


7.2 Results
7.2.1 Velocity Field
From the step-response valve simulations, Fig. 7.2 shows the forward and Fig. 7.3 the
reverse-flow velocity fields on the symmetric centerplane of the valve. The velocity fields
in the forward and reverse flow cases are very similar to those in the T45A valve. Flow separation and laminar jets occur in three locations: where the channels separate, the channels
recombine, and at the valve exit.
In the forward-flow case the flow accelerates rapidly as it enters the main channel from
the goblet-shaped inlet plenum. As the velocity profile is developing it reaches the Tjunction. The flow begins to veer slightly into the side channel, and a small portion impinging on the guide vane becomes a minor jet flowing into the side channel. However,
88% of the main channel flow (compared to 85% in the T45A) is unperturbed and continues downstream becoming fully-developed. The side channel jet spreads as it proceeds
around the bend in the side channel, and approaches the Y-junction as a low-velocity stream
filling the entire width of the side channel, moving at less than 11% of the bulk velocity
of the main channel (compared to 20% for the T45A). However, it has sufficient momentum to cause the main channel flow to begin to turn before it reaches the far wall of the
Y-junction. This may have a small effect in improving the diodicity as it tends to increase

79

MENUWINDOW
Z-SC-ROT CLK
X-SC-ROT CLK
Y-SC-ROT CLK
LEFT MOVE
UP MOVE
NEARMOVE
EXPAND
RESET VIEW
BACKCOLOUR
MARKER SIZE
COARSENESS

8.5000E-01
7.0833E-01
5.6667E-01
4.2500E-01
2.8333E-01
1.4167E-01
0.0000E+00

Figure 7.2: Forward-flow velocity field on the centerplane of a single-element, Tesla-type


T45C valve with a volume flow rate of 3640 l/min corresponding to Re=519 based on the
hydraulic diameter of the main channel. One dimensionless unit equals 10 m/s.

80
the radius of curvature of the main channel flow as it travels around the 45 bend. Downstream of the Y-junction, the main channel flow separates from the inner wall and forms a
narrow, hi-speed jet next to the outer wall. The high velocity gradient between the jet and
the wall produces additional dissipation that lowers the diodicity of the valve. However,
this channel downstream of the Y-junction is only 69% of the length of the corresponding
channel in the T45A and should contain relatively less dissipation. A significant portion of
the channel is filled with a quiescent zone and is essentially underutilized by the forward
flow. The flow leaves the channel and enters the goblet-shaped outlet plenum as a laminar
jet as in the T45A. Whatever momentum flux is dissipated increases the pressure drop in
the forward-flow direction and lowers diodicity.
In the reverse-flow case the flow accelerates rapidly as it enters the channel from the
goblet-shaped plenum, just as it does in the forward flow case. On reaching the Y-junction
the flow stream veers toward the main channel. As it impinges on the cusp of the guide
vane, 43% of the flow is deflected down the main channel (compared to 36% in the T45A)
where it travels next to the guide vane wall, but most of the main channel is filled with a
large, slowly-moving recirculation zone. The side channel flow separates from the outer
wall and forms a laminar jet attached to the guide vane wall. The large velocity gradient
there increases dissipation, which improves diodicity. At the bend in the side channel, the
jet separates from the guide vane, but unlike the jet in the T45A it does not attach to the
outer side-channel wall, but rather proceeds down the center of the side channel until it
reaches the T-junction. There it brushes past the protruding cusp and emerges from the side
channel as a laminar jet heading for the opposite wall of the inlet channel. The momentum
of the smaller flow traveling down the main channel turns the jet, but unlike the jet in the
T45A which never reaches the opposite wall, the jet in the T45C attaches to the opposite
wall of the inlet channel and remains attached through the rest of the valve. There is a large,
vigorous recirculation zone downstream of the T-junction, and the jet never spreads to fill
the entire channel as the jet in the T45A does.

7.2.2 Pressure Field


The pressure fields of the T45C valve shown in Fig. 7.4 are very similar to those of the
T45A. A major pressure loss is occurs where the flow enters the valve channel regardless of
the flow direction. There is only one other signficant pressure loss in each flow direction:
in forward flow it is at the 45 bend in the channel, in reverse flow it is just downstream
of the T-junction. Between these locations and their corresponding flow exits there is little

81

MENUWINDOW
Z-SC-ROT CLK
X-SC-ROT CLK
Y-SC-ROT CLK
LEFT MOVE
UP MOVE
NEARMOVE
EXPAND
RESET VIEW
BACKCOLOUR
MARKER SIZE
COARSENESS

> 8.5000E-01
7.2170E-01
5.7736E-01
4.3302E-01
2.8868E-01
1.4434E-01
0.0000E+00

Figure 7.3: Reverse-flow velocity field on the centerplane of a single-element, Tesla-type


T45C valve with a volume flow rate of 3640 l/min corresponding to Re=519 based on the
hydraulic diameter of the main channel. One dimensionless unit equals 10 m/s.

82
additional pressure drop; the momentum flux maintains the fluid velocity until it begins
to dissipate in the goblet-shaped plenum and beyond. Closer study of the pressure fields
shows that pressure drop at the 45 bend in forward flow is approximately 0.1 atm, and
downstream of the T-junction in reverse flow is approximately 0.24 atm, (compared to 0.20
in the T45A). This is the main source of diodicity. Thus, as suggested by Eq. 1.2, if the
direction-independent losses at the channel entrances could be eliminated, the diodicity
would be increased from 1 33 to 2 0. The outlet channel of the T45C is 30% shorter
than the T45A, and the inlet channel is 7% longer. The geometry of the mouths of the


channel entrances are identical.

7.2.3 Energy Dissipation Field


The base 10 logarithm of the energy dissipation rate is shown in Figs. 7.5 and 7.6. The
red and yellow contours are the most significant and occur where the velocity gradients are
largest. Energy dissipation in the forward-flow case is very similar to the T45A; dissipation
occurs along the walls of the main channel and on both sides of the laminar jet downstream
of the 45 bend. In the outlet goblet plenum, the dissipation of the momentum flux leaving
the exit is clearly visible. But the most significant dissipation occurs at the convex surfaces
of the channel mouth where the flow accelerates as it enters the channel, at the upstream
cusp of the guide vane, at the inner corner of the 45 bend, and along the wall downstream
of the 45 bend all the way to channel exit. These locations correspond well to the locations
of large pressure drop in Fig. 7.4, but it is not conclusive from this field variable analysis
that dissipation is the most important source of that pressure loss.
The locations of dissipation in the reverse flow case are somewhat different than in
the T45A valve. As the flow bifurcates at the Y-junction, the dissipation is spread more
evenly in the side channel than elsewhere, but very significant dissipation occurs along the
surfaces of the guide vane, both in the side channel and the main channel, and especially
at its leading cusp. Unlike in the T45A, there is no large dissipation along the outer wall
of the side channel since the separated jet does not attach to it. There is very significant
dissipation at the protruding cusp at the T-junction, and this dissipation continues alongside
the jet as it proceeds from the side channel into the main channel, collocated with the large
velocity gradient between the jet and the recirculation zone in the main channel downstream
of the T-junction. There is also significant dissipation where the jet exiting from the side
channel impinges on the opposite wall of the inlet channel. But unlike in the T45A, the jet
remains attached to the opposite wall and this high rate of dissipation persists all along the

83
MENUWINDOW
Z-SC-ROT CLK
X-SC-ROT CLK
Y-SC-ROT CLK
LEFT MOVE
UP MOVE
NEARMOVE
EXPAND
RESET VIEW
BACKCOLOUR
MARKER SIZE
COARSENESS

AXES OFF
Z-SC-ROT ANT
X-SC-ROT ANT
Y-SC-ROT ANT
RIGHT MOVE
DOWNMOVE
FAR MOVE
SHRINK

MENUWINDOW
Z-SC-ROT CLK
X-SC-ROT CLK
Y-SC-ROT CLK
LEFT MOVE
UP MOVE
NEARMOVE
EXPAND
RESET VIEW
BACKCOLOUR
MARKER SIZE
COARSENESS

AXES OFF
Z-SC-ROT ANT
X-SC-ROT ANT
Y-SC-ROT ANT
RIGHT MOVE
DOWNMOVE
FAR MOVE
SHRINK

LINE WIDTH
TEXT SIZE

5.2000E-01
4.2000E-01
3.2000E-01
2.2000E-01
1.2000E-01
2.0000E-02
-8.0000E-02

(a) Forward flow

LINE WIDTH
TEXT SIZE

5.2000E-01
4.1811E-01
3.1623E-01
2.1434E-01
1.1245E-01
1.0566E-02
<-8.0000E-02

(b) Reverse flow

Figure 7.4: Pressure field [atm] on the centerplane of a single-element, Tesla-type T45C
valve with a volume flow rate of 3640 l/min corresponding to Re=519 based on the hydraulic diameter of the main channel.

84

MENUWINDOW
Z-SC-ROT CLK
X-SC-ROT CLK
Y-SC-ROT CLK
LEFT MOVE
UP MOVE
NEARMOVE
EXPAND
RESET VIEW
BACKCOLOUR
MARKER SIZE
COARSENESS

-4.0000E-02
-1.3667E+00
-2.6933E+00
-4.0200E+00
-5.3467E+00
-6.6733E+00
-8.0000E+00

Figure 7.5: Base 10 logarithm of the energy dissipation rate in forward flow on the centerplane of a single-element, Tesla-type T45C valve with a volume flow rate of 3640 l/min
corresponding to Re=519 based on the hydraulic diameter of the main channel. One dimensionless unit equals 14 mW.

85

MENUWINDOW
Z-SC-ROT CLK
X-SC-ROT CLK
Y-SC-ROT CLK
LEFT MOVE
UP MOVE
NEARMOVE
EXPAND
RESET VIEW
BACKCOLOUR
MARKER SIZE
COARSENESS

-4.0000E-02
-1.3917E+00
-2.7434E+00
-4.0951E+00
-5.4468E+00
-6.7985E+00
<-8.0000E+00

Figure 7.6: Base 10 logarithm of the energy dissipation rate in reverse flow on the centerplane of a single-element, Tesla-type T45C valve with a volume flow rate of 3640 l/min
corresponding to Re=519 based on the hydraulic diameter of the main channel. One dimensionless unit equals 14 mW.

86
wall until the plenum is reached. As in the forward-flow case, part of the momentum flux
leaving the channel is dissipated in the goblet-shaped plenum.
7.2.4 Momentum Conservation
Diodic Effects of the Force Vector Interactions
The momentum equation is a vector relation, a summation of forces, so each of its terms has
both magnitude and direction sense. The X and Y components of these forces are shown in
Fig. 7.7 for both the forward and the reverse flow case. The figure caption explains positivenegative direction sense and the X-Y coordinate system is shown in Fig. 7.1. Essentially,
the figures show that the viscous and pressure forces applied on the fluid and the fluids
momentum flux response are similar to those of the T45A. The only differences are in the
outlet channel (block 6) in forward flow and the region of the T-junction (blocks 2, 3, 4,
and 8) in reverse flow.
The goblet-shaped inlet plenum (block 1) is oriented for flow in the X direction, so
as Fig. 7.7 shows, the Y-direction force components are negligible. In forward flow the
pressure force is to the right (positive) and it is expended by accelerating the fluid as it
approaches the inlet channel and by a small amount of viscous force. The momentum flux
is to the right (positive), but more is leaving the control volume than entering (defined as
negative), so its overall sign is negative. In reverse flow the pressure is already near ambient
when it enters the block 1 control volume, so the pressure force is negligible. The viscous
force applied by the walls is also very small, so the momentum flux is unchanged as the
flow passes through block 1.
The inlet channel (block 2) is also oriented in the X direction but due to its proximity
to the side channel (block 8) it has Y-direction force components during reverse flow. The
positive-Y pressure force is applied by the lower wall on the jet of fluid from the side channel, altering its direction so that it flows out of block 2 and into block 1. The momentum
of this jet is directed into the inlet channel (positive) but is flowing in the negative-Y direction (negative) resulting in a net negative Y-component of the momentum flux. In the
X-direction in both forward and reverse flow, the upstream flow boundary applies the pressure force needed to drive the fluid through the inlet channel offset by the viscous force and
momentum flux. In both flow directions the viscous force is larger than the momentum flux,
as in the T45A. However, the reverse-flow viscous force is 24% larger in the T45C than in
the T45A, because the laminar jet attaches to the lower wall instead of flowing nearer the
center of the channel as it does in the T45A. Unfortunately, the X-direction pressure force

87

ForwardFlow Xdirection Forces

0.15
0.1
0.05
0
0.05
0.1
0.15

Momentum Flux
Viscous Force
Pressure Force

0.2
ReverseFlow Xdirection Forces

Momentum Flux
Viscous Force
Pressure Force

0.2

0.2

0.15
0.1
0.05
0
0.05
0.1
0.15
0.2

4
5
6
Block number

(a) X-direction, forward flow

0.1
0.05
0
0.05
0.1
0.15
0.2

0.2
ReverseFlow Ydirection Forces

ForwardFlow Ydirection Forces

0.15

4
5
6
Block number

(b) X-direction, reverse flow

Momentum Flux
Viscous Force
Pressure Force

0.2

Momentum Flux
Viscous Force
Pressure Force

0.15
0.1
0.05
0
0.05
0.1
0.15
0.2

4
5
6
Block number

(c) Y-direction, forward flow

4
5
6
Block number

(d) Y-direction, reverse flow

Figure 7.7: Force vector terms in the integral form of the momentum conservation equation.
Net pressure force and net momentum flux into a control volume are positive. Viscous force
is applied on the fluid by the wall. X-vectors to the right and Y-vectors upward are positive
and consistent with the valve layout in Fig. 7.1 including the numbering of the control
volumes (blocks).

88
is larger in forward flow than in reverse, which has an adverse impact on valve diodicity.
It is not clear from the momentum perspective what impact the Y-direction pressure force
has on diodicity.
The T-junction (block 3) does not have significant Y-direction forces, a surprising result
since this control volume connects the side channel (block 8) to the inlet channel and the
main channel (block 4). However, inspection of Fig. 7.3 shows that the jet from the side
channel during reverse flow does not significantly interact with the wall in block 3, but
passes through to block 2. However the X-direction forces are important in creating diodicity. In forward flow there is a small amount of pressure recovery as the momentum flux
drives the flow through the area expansion of block 3; this negative pressure force increases
diodicity as it lowers the net valve pressure force for forward flow. The reverse flow has a
much greater positive impact on diodicity; the momentum flux leaving block 3 (negative)
and flowing in the negative X direction is the jet from the side channel, and results in a
positive net momentum flux for block 3 as seen in Fig. 7.7b. The pressure force needed to
offset this momentum flux is 22% larger in the T45C than the T45A. This is a major source
of the increased diodicity of the T45C.
The main channel (block4) has almost identical Y-direction force components in forward and reverse flow. In both cases the upper wall, which forms part of the guide vane
or island of the valve, applies a pressure force in the negative-Y direction to the impinging
flow, which has just passed an opening to the side channel. This impingement can be seen
in both Figs. 7.2 and 7.3. It is not clear from the momentum perspective if this affects
diodicity. In the X direction the force components are very different between forward and
reverse flow. In forward flow 88% of the volume flow is carried by the main channel, and
as a result significant pressure force is needed to counteract viscous force due to wall friction. This has a negative impact on diodicity as 57% of the reverse flow occurs in the side
channel and little pressure force is required to push the remaining reverse flow through the
main channel. The reverse-flow is 19% larger in the T45C than in the T45A, as a result
the viscous force applied by the wall in the T45C is correspondingly larger. Otherwise the
forces are essentially identical to those of the T45A.
The Y-connection (block 5) forces are essentially identical in the T45C to the T45A
forces. There are negligible X-direction force components. In forward flow the net Ydirection flux of momentum is out of the control volume (negative) and in the negative Y
direction resulting in a positive value, which is counteracted by the pressure force applied
by the outer wall. In reverse flow some of the flow is directed down the main channel (block
4) by the guide vane, thus there is a net flux of momentum into the control volume in the

89
positive Y direction, so the momentum flux again has a positive Y sense.
The outlet channel (block 6) is similar in form and function to the inlet channel (block2),
except that it is oriented at 45 cw with respect to the positive X direction. Because of
different orientations it is difficult to compare the forces in blocks 2 and 6 using Fig. 7.7,
but a comparison is made below using the vector magnitudes shown in Fig. 7.8 to assess
the impact on diodicity. In the T45A, the pressure force applied at the upstream flow
boundary was roughly equally opposed by the viscous force applied by the walls and by
the net momentum flux leaving the control volume, but in the T45C the viscous force is
reduced because the outlet channel is 31% shorter, and thus the pressure force required in
both flow directions is 22% less in the T45C than in the T45A. This has a positive influence
on diodicity according to Eq. 1.2 by reducing direction-independent losses.
The goblet-shaped outlet plenum (block 7) is analogous to the inlet plenum (block 1)
except for its orientation of 45 cw with respect to the positive X direction. The forward flow
forces are small; the reverse-flow net pressure force accelerates the fluid as it approaches the
next downstream control volume (block 6). All the viscous forces are small. The forwardflow momentum flux, though still small, is larger in the T45C than in the T45A, because the
outlet channel (block 6) is shorter and the velocity of the forward-flow jet entering block 7
is higher.
The side channel (block 8) has negligible forces in forward flow since it carries only
12% of the volume flow rate. In reverse flow, Fig. 7.7d shows a very large pressure force
applied by the outer wall to alter the direction of the flow in the side channel. Unlike in
the T45A, the flow is not attached to the outer wall (see Figs. 7.3 and 7.4), and in addition
there is 12% less flow in the side channel in the T45C than in the T45A. As a result, the
pressure force and momentum flux are 18% smaller in the T45C than in the T45A. Because
these are Y-direction forces it is difficult to assess their impact on diodicity. The last control
volume (block 9) has significant forces only in the reverse-flow direction.

Forward-Flow vs. Reverse-Flow Force Vector Magnitudes


Another way to study the diodicity mechanism, especially for control volumes that are not
oriented along the X direction, was to study the vector magnitudes of the force components
in the control volumes as shown in Fig. 7.8, which directly compares the forward-flow and
reverse-flow components.
The inlet and outlet goblet-shaped plenums (blocks 1 and 7) perform similarly in the
T45C as in the T45A valve. The viscous forces applied by the walls are small, and the

90

0.2
Viscous Force Magnitude

0.2
Pressure Force Magnitude

reverse flow
forward flow

reverse flow
forward flow

0.15

0.1

0.05

0.15

0.1

0.05

4
5
6
Block number

(a) Pressure force magnitude

4
5
6
Block number

(b) Viscous force magnitude

reverse flow
forward flow
Momentum Flux Magnitude

0.2

0.15

0.1

0.05

4
5
6
Block number

(c) Momentum flux magnitude

Figure 7.8: Vector magnitudes of the terms in the momentum conservation equation for the
T45C valve.

91
pressure forces and momentum fluxes are the significant forces with those of the forward
flow jet entering block 1 and the reverse flow jet entering block 7 an order of magnitude
larger than all other forces. The combined diodic effect of blocks 1 and 7 is again slightly
above unity, Di1 7


1 02


The inlet and outlet channels (blocks 2 and 6) function similarly as in the T45A. One
difference is a 20% reduction in the viscous forces in the shortened block 6 (31%) in the
T45C. Another difference is an increased viscous force applied by the walls of block 2
in the T45C during reverse flow, due to the reverse-flow jet that has attached to the wall
instead of remaining near the center of the channel as in the T45A; compare Figs. 7.3 and
6.3. The net diodic effect is slightly above unity, though this is not clear from the pressure
force magnitude plot of Fig. 7.8a.

The main channel (block 4) and the Y-junction (block 5) of the T45C perform identically as in the T45A. They have larger pressure force magnitudes in forward flow than in
reverse flow, and thus have a diodic effect less than unity. In block 4 this is because almost
all the forward flow proceeds through the main channel instead of bypassing it as it does in
reverse flow. As a result the viscous force applied by the wall is much almost twice as large
in forward flow as it is in reverse. In block 5 the larger pressure force in forward flow is
required offset the larger momentum flux and viscous force, which are due to the forward
flow changing direction in block 5, whereas in reverse flow it proceeds straight through
toward the side channel.

The major contributors to diodicity are the T-junction (block 3) and the side channel
(blocks 8 and 9). The T-junction is where the side channel emerges and must be redirected
and forced out through block 2, the result is a large net increase in momentum flux and
additional viscous force applied by the walls that both must be overcome by the pressure
force. The large increase in pressure across block 3 can be seen in Fig. 7.4b. The forwardflow forces in the side channel and T-junction of the T45C are nearly identical to those in
the T45A. In reverse flow the T45C requires less pressure force in the side channel than
in the T45A since the flow rate is 11% less and the momentum flux is correspondingly
reduced. This reduces overall diodicity. However the net momentum flux in the T-junction
of the T45C is 22% higher than in the T45A, and the pressure force is 27% higher as seen
in Fig. 7.8. The net result is an increase in the diodicity of the T45C over the T45A.

92
7.2.5 Kinetic-Energy Conservation
The kinetic energy terms for the entire valve were compared to those of the T45A. Over
70% of the pressure work is dissipated in the valve compared to 66% in the T45A. The ratio
of the dissipation rates (reverse flow / forward flow) is 1.35, instead of 1.13 in the T45A,
due to a 12% reduction in forward-flow dissipation because of the shortening of the outlet
channel (block 6), and a 6% increase in reverse-flow dissipation because of the attachment
of the reverse-flow jet to the wall in the inlet channel (block 2). The ratio of energy flux
rates is 1.33, instead of 1.91 in the T45A, because the shortened outlet channel allows a
higher-velocity forward-flow jet to pass into the outlet plenum (block 7). The energy flux
rates are only one-third the magnitude of the dissipation rates, as in the T45A. The overall
valve diodicity of the T45C is 1.33. The diodicity prediction from the ratio of the pressure
work rates as in Eq. 2.14 is 1.34, which is within 1% of 1.33.
There are two alternate ways to look at the data. Figs. 7.9 and 7.10 show the two most
important terms of the kinetic energy equation in steady incompressible flow: the pressure
work rate and the dissipation rate, which are shown for each of the control volumes for
both forward and reverse flow. As in the T45A, blocks 3, 6, 7, 8, and 9 were identified as
sources of diodicity, since the pressure work applied in the reverse flow direction is larger
than that applied in the forward direction. On the other hand, blocks 1, 2, 4, and 5 cause
a significant reduction in diodicity. The only significant difference in the pressure work
rates for the T45C and the T45A is the reduced direction dependence of block 2, due to
an increase in the reverse-flow pressure work rate. The energy flux rate has opposite sign
(+ is a net increase, - is a net decrease in the control volume) depending on flow direction
in all blocks. The dissipation rate also exhibits significant direction dependence, except in
blocks 3, 4, and 5. There are three differences between the dissipation rates of the T45C and
T45A: the reverse-flow dissipation rate in block 2 is increased (due to the wall attachment
of the reverse-flow jet), the forward-flow dissipation rate in block 6 is decreased (due to
the shortening of block 6), and the reverse-flow dissipation rate in block 8 is decreased
(due to the reduced flow rate in the side channel). Once again, the viscous work rate is
inconsequential in all control volumes.
An alternate view of the kinetic energy in the valve is the contribution of each term of
the kinetic energy equation to the forward-flow direction and the reverse-flow direction as
shown in Fig. 7.11. Note that the pressure work rate typically must balance all the other
energy rate terms.
Blocks 1-2 and 6-7 fullfil similar roles as valve entrance and valve exit regions in for-

93

reverse flow
forward flow

0.06

Pressure Work Rate

0.04
0.02
0
0.02
0.04
0.06
1

4
5
6
Block number

(a) Pressure work rate

reverse flow
forward flow

0.06

Energy Flux Rate

0.04
0.02
0
0.02
0.04
0.06
1

4
5
6
Block number

(b) Energy flux rate

Figure 7.9: Magnitude of the pressure work rate and the energy flux rate terms in the
kinetic-energy conservation equation from the T45C valve simulations with a volume flow
rate of 3640 l/min corresponding to Re=519 based on the hydraulic diameter of the main
channel.

94

reverse flow
forward flow

0.06

Dissipation Rate

0.04
0.02
0
0.02
0.04
0.06
1

4
5
6
Block number

(a) Energy dissipation rate

reverse flow
forward flow

0.06

Viscous Work Rate

0.04
0.02
0
0.02
0.04
0.06
1

4
5
6
Block number

(b) Viscous work rate

Figure 7.10: Magnitude of the energy dissipation rate and the viscous work rate terms in
the kinetic-energy conservation equation from the T45C valve simulations with a volume
flow rate of 3640 l/min corresponding to Re=519 based on the hydraulic diameter of the
main channel.

95

Energy Flux Rate


Dissipation Rate
Viscous Work Rate
Pressure Work Rate

0.06

ForwardFlow Power

0.04
0.02
0
0.02
0.04
0.06
1

4
5
6
Block number

(a) Forward flow

Energy Flux Rate


Dissipation Rate
Viscous Work Rate
Pressure Work Rate

0.06

ReverseFlow Power

0.04
0.02
0
0.02
0.04
0.06
1

4
5
6
Block number

(b) Reverse flow

Figure 7.11: Magnitude of the terms in the kinetic-energy conservation equation from the
T45C valve simulations with a volume flow rate of 3640 l/min corresponding to Re=519
based on the hydraulic diameter of the main channel.

96
ward and reverse flow. Figure 7.11 shows that the pressure work rate is smaller in blocks
6 and 7 in reverse flow than in blocks 1 and 2 in forward flow. The ratio is 0.85 compared
to 0.9 in the T45A, an even larger reduction in diodicity due to the shortening of block
6. They are geometric similar flow paths, both serving to accelerate the flow as it enters
the channel and develops its flow profile, but the forward-flow entrance, block 2 is 76%
longer than block 6 (compared to 14% longer in the T45A). This suggests block 2 should
be shortened as well to improve diodicity. However, blocks 2 and 6 also have a role as
valve exits and in this role each contains a separated jet as can be seen in Figs. 7.2 and 7.3
with accompanying high rates of dissipation. The dissipation rate in the combined blocks
2 and 6 is higher in reverse than in forward flow; a diodic effect of 1.28 compared to 0.89
in the T45A. Even when the flow in the combined blocks 1, 2, 6, and 7 is considered, the
ratio of pressure work rates in reverse flow relative to forward flow is 1.13 (compared to
1.09 in the T45A), an increase of diodicity.
In the T45A, there was an asymmetry between the forward-flow energy flux rate in
block 6 and the reverse-flow energy flux rate of block 2, see of Fig. 6.11, that had a
positive diodic effect. Since block 6 was shortened in the T45C, this asymmetry does
not exist (see Fig. 7.11), instead, additional momentum flux is expended in the high rate
of dissipation due to the wall attachment of the reverse-flow jet. In fact, block 2 has a
40% higher dissipation rate in the T45C than in the T45A, even though block 2 is only
7% longer in the T45C. This suggests that if block 2 was shortened as block 6 has been
(compared to the T45A), the forward-flow dissipation rate in block 2 would decrease more
than the reverse-flow dissipation rate, and some positive diodic effect of the asymmetry
of the dissipation rates may be retained. In any case, reduction of direction-independent
pressure losses should increase diodicity, following Eq. 1.2.
The flow in the main channel (blocks 4 and 5) of the T45C is very similar to that in the
T45A. In the forward direction the main channel contains 88% of the valve flow (3% more
than in the T45A), and the pressure work is converted to energy dissipation and energy
flux. In the reverse direction 43% of the valve flow is carried by the main channel (7%
more than in the T45A), and it is the energy flux that is converted to energy dissipation and
the pressure work needed to sustain the flow rate. The main-channel flow has a negative
impact on diodicity, but, as consequence of adopting a fixed geometry for the valve, it must
exist during reverse flow.
As in the T45A, the most significant diodicity mechanism is located in the side channel
and the T-junction (blocks 3, 8, and 9) due to the radical asymmetry between the forward
and reverse flow there. In forward flow the only significant energy interchanges occur

97
in the T-junction (block 3) where 1/3 of the energy flux is dissipated and the remainder
is converted to pressure work to sustain the volume flow rate as the forward flow passes
through the area expansion of the T-junction and on into the main channel. In reverse flow
35% of the pressure work rate of the entire valve (3% more than in the T45A) is utilized to
drive 57% of the valve flow through the side channel and the T-junction (7% less than in the
T45A). The increased pressure work rate in spite of a lower flow rate is due to the longer
path length of the side channel in the T45C. Unlike the T45A, in which virtually all the
pressure work in the side channel is dissipated by the large velocity gradient between the
side-channel jet and the outer wall, the T45C dissipates 76%, the remainder is transformed
into the energy flux of the side-channel jet. The side-channel jet is shown in Fig. 7.3,
and the corresponding dissipation in Fig. 7.6. Most of the pressure work (88%) in the Tjunction is converted to additional energy flux of the side-channel jet as it is turned to flow
out through block 2. The remainder of the pressure work is dissipated. The overall result is
a strong diodic effect as energy flux is transformed to pressure work in forward flow, but in
reverse flow 35% of the entire valves pressure work is dissipated or transformed to energy
flux.

7.3 Discussion
The identical numerical methods applied to the T45A were applied to the T45C. The only
change in the numerical computations was the substitution of the physical grid of the T45C
in place of the physical grid of the T45A. The T45C was shown to be functionally similar
to the T45A. The diodicity mechanism is located in the side channel and T-junction (blocks
3, 8, 9), and that the rest of the valve adds to the overall fluid resistance but has diodic
effects near unity. As was suggested in Eq. 1.2, these valve channels dilute the overall
valve diodicity. For example, the goblet-shaped plenums (blocks 1 and 7) and the inlet and
outlet channels (blocks 2 and 6) produce a diodic effect of 1.13 (compared to 1.09 in the
T45A). But if these channels were not needed to direct the fluid into and out of the valve,
the overall valve diodicity would be closer to 2.2, instead of 1.33.
As in the T45A, the underlying physical basis for the diodicity of the T45C valve is
its ability to direct the majority of the volume flow through alternate paths depending on
flow direction. In forward flow 88% of the fluid proceeds through the main channel (blocks
3
4
5); in reverse flow 57% proceeds through the side channel (blocks 5
9
8
3). The diodicity mechanism is partly due to the viscous forces applied by the wall
on the fluid; the vector magnitude of the viscous force in reverse flow is 60% larger than

98
in forward (compared to 56% in the T45A). However, the energy dissipation rate is 87%
larger in reverse flow than in forward, significantly less than the 97% of the T45A. So the
diodicity mechanism is again more due to dissipation in the fluid itself than to wall friction.
The reduced reverse-flow energy dissipation in the T45C is balanced by increased energy
flux; this additional energy in the side-channel jet enables it to cross the main channel and
attach to the far wall. Thus the T45A contains a reverse-flow jet that is attached to the
wall in the side channel with accompanying dissipation in the side channel; in the T45C
the reverse-flow jet doesnt attach to any wall until it reaches the main channel, so the
accompanying dissipation occurs in the main channel.
As for the T45A, research objective #1e has been achieved. The diodicity mechanism
of the T45C valve is a reverse-flow laminar jet that is not pointed in the downstream direction and has a high rate of energy dissipation in the shear layer surrounding it. This jet
requires significant pressure force to turn it downstream, and more pressure work to offset
the energy it dissipates than to offset the energy dissipated by the forward-flow jet. Due to
the combined effects of the shortening of the outlet channel (block 6) and the attachment
of the reverse-flow jet in the inlet channel (block 2), the T45C has an 11% lower dissipation rate in forward flow than the T45A, and a 6% increase in dissipation rate in reverse
flow, resulting in an overall increase in diodicity to 1.33. But, according to the analysis of
diodicity prediction accuracy in Ch.4, there is a 95% probability that the true diodicity is
within 1 20 Di 1 34. Thus the increase in predicted diodicity over that of the T45A
valve Di 1 27 is not statistically significant.


It is easy to answer research objective #2 for the T45C valve, that the diodicity mechanism is mainly due to viscous effects, since conveniently all four measures: the diodicity
and the ratios of the reverse-flow and forward flow pressure work, energy flux, and dissipation are all within 2% of 1.33. Over 2/3 of the pressure work is expended in energy
dissipation, thus 2/3 of the diodicity is due to dissipation. Clearly viscous effects are dominant.

99

Chapter 8
VALVE DESIGN GUIDELINES
This chapter is directed toward research objectives #3a: to demonstrate knowledge of
the diodicity mechanism of Tesla-type NMP microvalves by developing guidelines for the
creation of enhanced-diodicity valve designs, and #3b: to use these guidelines as the sole
method to modify a valve design to significantly increase its diodicity. We accomplish
this by modifying the lower-performing T45A valve according to the design guidelines to
produce a T45A-2 valve that has higher performance than the T45C.
Valve design is a difficult task, since most of the geometric design variables of a Teslatype NMP valve, variables a through h in Fig. 8.1, are interdependent; they impact the
distribution of flow between the main channel and the side loop, the foundation of the diodicity mechanism. Variation of any of these design variables will impact the optimal settings
of other variables. A parametric study of the 14-dimensional design space following Fig.
8.1 would require simulation of 14 14 196 valve designs even if only two values were
considered for each design variable. Even a fractional-factorial analysis would require the
simulation of at least 28 designs. To further add to the complexity of the design task, the


geometric layout determines not only the diodicity of the valve, but also its resistance and
inertance, which must be matched to the dynamics of the microfluidic system in which it
operates. A well-matched valve may provide higher performance than a poorly-matched
higher-diodicity valve. Thus, there is no single optimal valve design for all microfluidic
systems. Clearly, a more efficient method of valve design is needed than mapping diodicity
as each valve design variable is varied.
The solution is to use the knowledge gained from the kinetic-energy, momentum, and
field variable analyses of the diodicity mechanisms of the T45A and T45C valves to develop
valve design guidelines. These guide the design of a high-performing Tesla-type NMP
valve for the particular microfluidic system. Once the valve geometry has been laid out,
the impedance of the prototype valve design can be predicted by numerical simulations
and applied in lumped-parameter dynamics models of the complete microfluidic system to
predict performance [3][2][17].

100

8.1 Preliminary Discussion


To guide the layout of a valve, we reiterate and extend some of the key points from the
discussion sections of the previous chapters, in particular, Secs. 2.4, 6.3, and 7.3 in which
the diodicity mechanism of Tesla-type NMP valves was revealed.
In Tesla-type valves with their multi-directional flow, the vector nature of the momentum perspective inhibits its use in estimating the diodicity, a scalar. However, the kineticenergy, itself a scalar, has no such difficulty. In Eq. 2.14 it was shown that the diodicity
could be estimated from the ratio of the pressure work rates in reverse and forward flow.
This ratio has predicted the diodicity of the both the T45A and T45C within 2%. And since
other terms of the kinetic-energy equation are insignificant, the pressure work rate is equal
to the sum of the energy flux rate out of the control volume and the energy dissipation rate
in the control volume. Valuable insight can be gained by estimating the diodicity from


Di


u2
2
u2
2

u n dA


u n dA


1
Re p
1
Re p



ui
i j x
dV
j
reverse

ui
i j x j dV
f orward


(8.1)

in dimensionless variables. For high diodicity, the basic goal is to maximize the dissipation
rate and energy flux rate in reverse flow and minimize them in forward flow. This equation
2
also shows that the diodicity is a function of u3i and  ui x j  , so that for a given flow
cross-sectional area, energy losses are exhibited in order of increasing losses by plug flow,
fully-developed parabolic profiles, and highly-asymmetric velocity profiles. The relative
2
importance of u3i and  ui x j  is controlled by the Reynolds number; when Re p is high,
the energy flux rate dominates; when Re p is small, the dissipation rate dominates, as shown
in Table 2.2. Thus, diodicity is created by generating separated laminar jets with their
asymmetric velocity profiles and high velocity gradients (especially when the jet is attached
to the wall). To have high diodicity the reverse flow jet must dissipate more energy than
any, preferably nonexistent, forward-flow jet.
Though it is not as useful for estimating diodicity, additional insight into enhancing
diodicity is provided by the momentum perspective. Equation 2.13 relates diodicity to the
ratio of the pressure force in reverse and forward flow. Implicit is that this is a vector
equation. In other words, not only the velocity magnitude of the laminar jet is important,
but also its direction. Thus, a reverse-flow jet should be oriented counter to the downstream
direction, and if it is not possible to avoid creating a jet in forward flow, it should at least
be oriented downstream.

101
The diodicity relation Eq. 1.2 rewritten here
Di

Pdirection independent Preverse direction dependent


Pdirection independent Pf orward direction dependent


shows that direction-independent losses diminish diodicity. The most obvious implementation of this concept is the minimization of the lengths of the inlet and outlet channels.
Comparison of the T45A and T45C showed that shortening the outlet channel of the T45C
was partly responsible for its higher diodicity. Minimization of all fluid path lengths is
recommended when energy flux from the valve is certain to be dissipated.
It is significant that Eq. 8.1 assumes all energy flux will eventually be dissipated. In
other words, as the Reynolds number increases and less of the dissipation occurs within the
geometric boundaries of the valve itself, it becomes increasingly important to ensure that
channels upstream and downstream of the valve are dissipative not diffusive, (ie. change
cross-sectional flow area suddenly), transferring energy flux back into available pressure
work in the fluid.
An alternative strategy would choose the diodicity to be the ratio of the dissipation
within the valve, and attempt to recover the kinetic energy of the fluid by providing diffusers at both ends of the valves, (ie. transform the energy flux back to available pressure
work). This would be a particularly appropriate strategy when it is desirable to reduce the
valve resistance and increase the valve inertance. It is an advantageous characteristic of
the T45C that its diodicity would increase 6% from 1.33 to 1.35 in this scenario, while the
ratio of reverse and forward dissipation rates in the T45A is such that its diodicity would
decrease 48% from 1.27 to 1.13. In the T45C it is the interaction of the reverse-flow jet
with the long inlet channel and the forward-flow jet with the short outlet channel that provide this direction-dependent internal energy dissipation. Up to 20% of the pressure work
could be recovered with a corresponding decrease in valve resistance if effective diffusing
channels were added to both ends of the T45C. The addition of diffusing channels would
also increase overall channel length increasing valve inertance and lowering the corner frequency of the valve. This would be an appropriate strategy when energy flux from the valve
would not be dissipated in the surrounding structure.

8.2 How To Lay Out a Tesla-Type Valve


In this section we present design guidelines and demonstrate how to layout a Tesla-type
valve using a generic sketch of a Tesla-type valve shown in Fig. 8.1. As an example

102
f

sid

k
e

inlet
j

di

c
a
Y

main

tle

ou

do

Figure 8.1: Sketch of generic Tesla-type NMP valve with dimensioning per design rules
for high diodicity.

we will follow these guidelines to modify the geometry of the T45A valve to enhance its
diodicity mechanism resulting in a higher-performing valve we will call the T45A-2. This
new valve is analyzed in the following chapter by the numerical method to show that the
diodicity mechanism has indeed been enhanced. The sketch itself is an engineering tool;
there are other dimensioning schemes that could also be used, but the dimensioning used in
Fig. 8.1 is sufficient to allow us to manipulate the valve geometry to utilize our knowledge
of the diodicity mechanism.
Referring to Fig. 8.1, design of a Tesla-type valve begins with selection of angle a
between the main and side channels. The T45A and T45C both use angle a = 45 by
which they are named. Other angles have been used, from the 22 of the original design of
Tesla [28] up to 90 Angle a contains a design conflict; decreasing a weakens the forwardflow jet in the outlet channel and diminishes forward-flow losses, increasing a increases


the proportion of the reverse flow directed into the side channel and increases reverse-flow
losses. Comparison of the T45A and T45C showed that a decrease in the proportion of
reverse flow passing through the side channel from 64% to 57% did not adversely affect
the diodicity. For the T45A-2 we will maintain angle a at 90


103

Figure 8.2: Overlay of the T45A (solid red lines) and the T45C (dashed blue lines) showing
the variation in path lengths: inlet channel, outlet channel, and side channel.

Widths b and c and the etch depth of the valve are chosen to achieve a desired Reynolds
number range and to allow particles of a desired size to pass (if no smaller constrictions
are included in the valve design). These widths were historically chosen as 114 m and
the nominal etch depth as 120 m. Calculation of the resistance of an equivalent-length
rectangular duct and knowledge of the maximum pressure difference to be applied across
the valve gives an estimate of the operational Reynolds number range, (Re < 1000 for the
T45A and T45C). The Reynolds number, which is based on the hydraulic diameter of the
main channel at the T-junction, determines whether the inertial or viscous portion of the
diodicity mechanism is dominant per Eq. 8.1. For the T45A-2 we will maintain width c as
114 m and the nominal etch depth as 120 m, but we will increase width b to 135 m to
widen the main channel at the Y-junction and weaken the forward-flow jet.
Dimensions di and do are the offset of the inner side-channel radius e and the outer
side-channel radius f from the main channel. Together, these dimensions control the length
of the side channel, the angle between the side channel and main channel at the T-junction,
and the narrowing of the side channel as it approaches the T-junction. They also serve to
orient the jet emanating from the side channel counter to the downstream direction. Figure
8.2 shows an overlay of the T45A and T45C. In the T45A, di and do are approximately
zero so the side channel is a nearly constant width and is perpendicular to the main channel
at the T-junction. In the T45C, di and do are equal but nonzero such that the side channel
rejoins the main channel at the T-junction at a 45 angle. The increased path length of the

104
T45C side channel is clear, yet the reverse-flow jet from the side channel still has sufficient
momentum to reach the opposite wall of the main channel as seen in Fig. 7.3. Careful study
of the figure reveals a low velocity region where the inner wall of the side channel joins the
main channel. This suggests that di could be reduced in the T45C so that these walls join
perpendicularly, reducing the dimension g and diminishing the small jet that flows into the
side channel during forward flow seen in Fig. 7.2. Alternately, do could be increased to
narrow dimension g and orient the reverse-flow jet more counter to the downstream flow.
Note that 4 parameters are sufficient to define the side channel, so one of these dimensions
is redundant. For the T45A-2 we will maintain di at zero and e at 167 m, but will increase
do from zero to 90 m and decrease f from 284 to 225 m to narrow dimension g and create
a side channel with converging walls in the reverse-flow direction. This will increase the
momentum of the reverse-flow jet so that it can reach the opposite wall of the main channel
as in the T45C, and also to reduce flow into the side channel during forward flow.
The angle h and the width b define the main channel. Angle h is zero in both the
T45A and the T45C. The rationale for including it here is an understanding of the diodicity
mechanism. In forward flow it has been shown that a laminar jet (and its accompanying
dissipation) are created downstream of the Y-junction. Diffusing the forward flow before it
reaches the Y-junction would diminish the energy flux rate into the outlet channel, slow the
forward-flow jet, and lessen forward-flow energy loss. Conversely, accelerating the reverse
flow in the main channel would increase the energy flux rate into the inlet channel. Both
effects enhance diodicity. We will implement this strategy in the T45A-2 by increasing b
to 135 m (as previously mentioned) and setting h to 6 .
Length i and radii j and k define the inlet channel, and both radii are set as 15 m in
the T45A and T45C. In the T45A, the inlet channel is a source of direction-independent
energy loss and should be minimized (i is 532 m). The T45C exhibits a high dissipation
rate where the reverse flow jet is attached to the inlet channel wall. Shortening its inlet
channel (i is 566 m) would result in less energy flux transferred to dissipation within the
valve itself, though all energy flux developed in the valve by pressure work will eventually
dissipate if the channels upstream and downstream of the valve are long enough and do
not provide a diffusive effect. We can increase the diodicity due to energy dissipation
in the valve by giving different values to radii j and k. Specifically, radius k should be
large enough to allow a gradual acceleration of the forward flow in the inlet channel to its
maximum value at the T-junction. This will minimize the forward-direction dissipation rate
by avoiding high velocity-gradients near the walls. We dont need to worry about diffusing
the reverse-flow jet because we have already arranged via the side-channel geometry to

105
give the reverse-flow jet enough momentum to reach and attach to the wall on the opposite
side of the inlet channel. To increase the reverse-flow energy dissipation, we let the reverseflow jet dissipate itself along the wall of length i and set the radius j small enough to avoid
a diffusing effect. A reasonable value for j would be in the range of 50% to 100% of
dimension b, and i would be the difference between j and k. For the T45A-2 we will set j as
100 m (75% of b) and k as 400 m, and thus i will be 300 m, which is a 44% reduction
in length compared to the T45A.
Radius l, angle m, and length n define the outlet channel. In the T45A and T45C the
radius l is 15 m and angle m is 45 . The length n is 463 m in the T45A and 321 m
in the T45C. These dimensions address the forward-flow jet seen in Figs. 6.2 and 7.2.
The rounding of the corner via radius l minimizes the dissipation from the high velocity
gradient between the jet and the wall seen in Figs. 6.5 and 7.5. The radius l should be
large enough that the minimum outlet channel width is from 60% to 75% of width c, since
any additional flow cross-sectional area is not utilized by the forward flow and should be
made unavailable to the reverse flow. Angle m and length n define a diffusing section of the
outlet channel. The dissipation of the forward jet in the outlet channel would be more than
offset by the transfer of energy flux to pressure work, lessening the overall pressure loss in
the forward flow direction. In the reverse flow direction the increased energy flux into the
Y-junction due to the converging of the walls accomplished by angle m, combined with the
attachment of the flow to the outer wall of the side channel, would maximize the portion of
the flow proceeding into the side channel for any particular choice of angle a. The mouth
of the outlet channel may have sharp corners or protruding corners to further inhibit the
reverse flow. To follow these strategies in the T45A-2 we increase radius l to 85 m and
angle m to 52 We shorten length n to 125 m for the same reasons we shortened length i,


to reduce direction-independent dissipation. We choose sharp corners for the mouth of the
outlet channel.
We have accomplished research objective #3a by presenting design guidelines for the
geometrical layout of optimal Tesla-type NMP valves based on knowledge of the lowReynolds number diodicity mechanism. As an example we applied these guidelines to
modify the lower-performing T45A valve to produce a T45A-2 valve that has higher performance than the T45C. We have also pointed out an additional utility of the kinetic-energy
perspective, which through Eq. 8.1 not only accurately estimates diodicity, but also shows
the contributions due to the energy flux rate and the dissipation rate. In the following
chapter we complete accomplishment of research objective #3b by applying the numerical
method to reveal the enhancement of the diodicity mechanism in the T45A-2.

106

Chapter 9
ENHANCED DIODICITY MECHANISM OF T45A-2 VALVE
9.1 Simulation Methods and Conditions
The geometrical layout of the T45A-2 is a modification of the lower-performing T45A
valve. It was developed from the T45A by directly following the design guidelines presented in the previous chapter, which are based on knowledge of the low-Reynolds-number
diodicity mechanism revealed by the numerical method of Chap. 3. No other numerical
valve simulations of any kind were done varying the geometry of the T45A to determine
this new design. This design is also not an optimum version selected from a number of attempts. The geometry of the T45A-2 is from a single application of the design guidelines to
improve the T45A. The results show that the design guidelines are effective and sufficient
to create a valve with an enhanced diodicity mechanism.
It is also important to note that the numerical methods employed to analyze the T45A-2
were exactly identical to those used in the T45A and T45C analyses. The only variation is
the substitution of the physical grid of the T45A-2 for that of the T45A or T45C. As was
done for these other valves, momentum and kinetic-energy conservation were applied in 9
regional control volumes as shown in Fig. 9.1. The flow behavior and characteristics are
compared throughout this chapter with those of the T45A valve.
Steady-state solutions of both forward flow and reverse flow in the T45A-2 valve were
calculated using the numerical methods described in Chap. 3. These steady-state solutions
were obtained from the solution of the final time step of step-response simulations that
were continued until steady flow was achieved. In all cases the simulations converged to
steady solutions without applying Reynolds averaging to the Navier-Stokes equations. As
discussed in Sec. 1.2.2, this identifies the flow in the valve as laminar over the Reynolds
number range simulated, Re
regions.

683, despite the presence of separated flow and recirculation

To represent typical operating conditions a differential pressure of 0.5 atm was applied
to the pressure boundaries, producing an equilibrium volume flow rate in the reverse direction of 2987 l min corresponding to Re D 500 based on the hydraulic diameter of

MENUWINDOW
Z-SC-ROT CLK
X-SC-ROT CLK
Y-SC-ROT CLK
LEFT MOVE
UP MOVE
NEARMOVE
EXPAND
RESET VIEW
BACKCOLOUR
MARKER SIZE
COARSENESS

8
9
1

AXES OFF
Z-SC-ROT ANT
X-SC-ROT ANT
Y-SC-ROT ANT
107
RIGHT MOVE
DOWNMOVE
FAR MOVE
SHRINK
LINE WIDTH
TEXT SIZE

6
y

7
x

Figure 9.1: Division of the T45A-2 valve into regional control volumes.

the main channel at the T-junction. To achieve the same flow rate in the forward-flow direction required a differential pressure of only 0.287 atm. This corresponds to a diodicity

 

 

Di 1 74 according to Eq. 1.1. According to the analysis of diodicity prediction accuracy in Ch.4, there is a 95% probability that the true diodicity is within 1 57 Di 1 75

  

since the numerical method overpredicts the diodicity by 4 42 5 13%. To determine the
increase of diodicity over that of the T45A valve with 95% confidence, we take the ratio of
the diodicities and use the law of combination of errors [4] to obtain

DiT45A 2
DiT45A

1 74
1 27





e2T45A


2

100

e2T45A

  0  10

1 37

in which the 95% confidence band of the prediction errors are eT45A
Thus the diodicity increase is a statistically significant 27-47%.

eT45A

5 13%.

9.2 Results
9.2.1 Velocity Field
From the step-response valve simulations, Fig. 9.2 shows the forward and Fig. 9.3 the
reverse-flow velocity fields on the symmetric centerplane of the valve. The velocity fields
in the forward and reverse flow cases are similar to those in the T45A valve. Flow separation and laminar jets occur in three locations: where the channels separate, the channels
recombine, and where the flow leaves the valve.

108

MENUWINDOW
Z-SC-ROT CLK
X-SC-ROT CLK
Y-SC-ROT CLK
LEFT MOVE
UP MOVE
NEARMOVE
EXPAND
RESET VIEW
BACKCOLOUR
MARKER SIZE
COARSENESS
8.6448E-01
7.2040E-01
5.7632E-01
4.3224E-01
2.8816E-01
1.4408E-01
0.0000E+00

Figure 9.2: Forward-flow velocity field on the centerplane of a single-element, Tesla-type


T45A-2 valve with a volume flow rate of 2987 l/min corresponding to Re=500 based on
the hydraulic diameter of the main channel. One dimensionless unit equals 10 m/s.

109
In the forward-flow case the flow accelerates more slowly than in the T45A as it enters the main channel from the goblet-shaped inlet plenum due to the increased radii of
the channel mouth. Since we have shortened the inlet channel, and the walls are still converging, the velocity profile is still undeveloped as it reaches the T-junction. Because we
have narrowed the side channel at the T-junction, very little flow veers into the side channel; 94% of the main channel flow (compared to 85% in the T45A) is unperturbed and
continues downstream through the main channel. The diffusive strategy implemented in
the diverging walls of the main channel succeeds in decelerating the flow to minimize the
momentum of the forward-flow jet. Downstream of the Y-junction, the main channel flow
still separates from the inner wall and forms a forward-flow jet attached to the outer wall,
but the jet has 30% less momentum than the forward-flow jet in the T45A. Since we have
narrowed the outlet channel, a much smaller portion of the channel compared to the T45A
is filled with a quiescent zone that is essentially underutilized by the forward flow. The
flow leaves the channel and enters the goblet-shaped outlet plenum as a weaker jet than in
the T45A.
In the reverse-flow case the flow accelerates suddenly as it enters the channel from the
goblet-shaped plenum, just as it does in the T45A. On reaching the Y-junction the flow
stream veers toward the main channel. As it impinges on the cusp of the guide vane, 52%
of the flow is deflected down the main channel (compared to 36% in the T45A) where it
travels next to the guide vane wall, but most of the main channel is filled with a large,
slowly-moving recirculation zone. We have succeeded in creating a main channel that
is wholly available to the forward flow, but underutilized by the reverse flow. The side
channel flow separates from the outer wall and forms a laminar jet attached to the guide
vane wall. We have retained the large velocity gradient there that increases dissipation and
improves diodicity. As the side-channel flow approaches the T-junction it is accelerated by
the converging walls and emerges from the side channel as the reverse-flow jet heading for
the opposite wall of the inlet channel.This has increased the jets momentum sufficiently
to cross the main channel in spite of the momentum of the flow traveling down the main
channel and attach to the opposite wall of the inlet channel, unlike the reverse-flow jet in
the T45A that never reaches the opposite wall. There is a large separated zone downstream
of the T-junction, but the reverse-flow jet never spreads to fill the entire channel as the jet
in the T45A does, it remains attached to the lower wall until it reaches the plenum. This
high-velocity gradient between the wall and the jet is exactly the dissipative flow structure
the valve design guidelines anticipated.

110

MENUWINDOW
Z-SC-ROT CLK
X-SC-ROT CLK
Y-SC-ROT CLK
LEFT MOVE
UP MOVE
NEARMOVE
EXPAND
RESET VIEW
BACKCOLOUR
MARKER SIZE
COARSENESS
8.6448E-01
7.2040E-01
5.7632E-01
4.3224E-01
2.8816E-01
1.4408E-01
0.0000E+00

Figure 9.3: Reverse-flow velocity field on the centerplane of a single-element, Tesla-type


T45A-2 valve with a volume flow rate of 2987 l/min corresponding to Re=500 based on
the hydraulic diameter of the main channel. One dimensionless unit equals 10 m/s.

111
9.2.2 Pressure Field
The pressure fields of the T45A-2 valve shown in Fig. 9.4 are similar to those of the T45A
in their major features. Approximately two-thirds of the pressure loss in forward flow
occurs in the inlet channel before the T-junction is reached. The pressure loss at the 45
bend in the channel has been reduced from that in the T45A by deceleration of the main
channel flow. In reverse flow there is an abrupt pressure loss as the flow enters the sharpcornered valve mouth as in the T45A, and the pressure loss has also been increased at the
T-junction. Between these locations there is little additional pressure drop. Closer study of
the pressure fields of the T45A-2 shows that about two-thirds of the pressure drop in each
flow direction occurs at the T-junction.
9.2.3 Energy Dissipation Field
The base 10 logarithm of the energy dissipation rate is shown in Figs. 9.5 and 9.6. The
red and yellow contours are the most significant and occur where the velocity gradients are
largest. Energy dissipation in the forward-flow case is less than in the T45A, especially
along the diverging walls of the main channel and downstream of the 45 bend where we
have succeeded in reducing the momentum of the forward-flow jet. The dissipation rate is
reduced at the convex surfaces of the channel mouth where the flow accelerates as it enters
the inlet channel because we have provided such a large radius on the upper wall, knowing
that the reverse-flow jet would remain attached to the lower wall. The locations of high
dissipation rates correspond well to the locations of large pressure drop in Fig. 9.4.
The locations of dissipation in the T45A-2 in the reverse flow case are quite different
than in the T45A valve. High dissipation rates still occur along the wall as the flow enters
the valve where we chose sharp corners for the channel mouth and along the surfaces of the
guide vane both in the side channel and the main channel and especially at its leading cusp.
Unlike the T45A, there is no significant dissipation along the outer wall of the side channel
since the separated jet does not attach to it. As in the T45A there is a high dissipation rate
at the protruding cusp at the T-junction, and this dissipation continues alongside the jet as
it proceeds from the side channel into the main channel, collocated with the large velocity
gradient between the jet and the recirculation zone in the main channel downstream of the
T-junction. The most significant dissipation occurs where the reverse-flow jet exiting from
the side channel impinges on the opposite wall of the inlet channel. But unlike in the T45A,
the jet attaches to the opposite wall, as planned by the valve design guidelines, and this high
rate of dissipation persists all along that wall until the jet detaches as the plenum is reached.

112
MENUWINDOW
Z-SC-ROT CLK
X-SC-ROT CLK
Y-SC-ROT CLK
LEFT MOVE
UP MOVE
NEARMOVE
EXPAND
RESET VIEW
BACKCOLOUR
MARKER SIZE
COARSENESS

AXES OFF
Z-SC-ROT ANT
X-SC-ROT ANT
Y-SC-ROT ANT
RIGHT MOVE
DOWNMOVE
FAR MOVE
SHRINK

MENUWINDOW
Z-SC-ROT CLK
X-SC-ROT CLK
Y-SC-ROT CLK
LEFT MOVE
UP MOVE
NEARMOVE
EXPAND
RESET VIEW
BACKCOLOUR
MARKER SIZE
COARSENESS

AXES OFF
Z-SC-ROT ANT
X-SC-ROT ANT
Y-SC-ROT ANT
RIGHT MOVE
DOWNMOVE
FAR MOVE
SHRINK

LINE WIDTH
TEXT SIZE

5.0569E-01
4.0688E-01
3.0807E-01
2.0926E-01
1.1045E-01
1.1643E-02
-8.7166E-02

(a) Forward flow

LINE WIDTH
TEXT SIZE

5.0569E-01
4.0688E-01
3.0807E-01
2.0926E-01
1.1045E-01
1.1644E-02
-8.7166E-02

(b) Reverse flow

Figure 9.4: Pressure field [atm] on the centerplane of a single-element, Tesla-type T45A2 valve with a volume flow rate of 2987l/min corresponding to Re=500 based on the
hydraulic diameter of the main channel.

113

MENUWINDOW
Z-SC-ROT CLK
X-SC-ROT CLK
Y-SC-ROT CLK
LEFT MOVE
UP MOVE
NEARMOVE
EXPAND
RESET VIEW
BACKCOLOUR
MARKER SIZE
COARSENESS
1.0001E-01
-1.0851E+00
-2.2702E+00
-3.4553E+00
-4.6404E+00
-5.8255E+00
<-6.8789E+00

Figure 9.5: Base 10 logarithm of the energy dissipation rate in forward flow on the centerplane of a single-element, Tesla-type T45A-2 valve with a volume flow rate of 2987
l/min corresponding to Re=500 based on the hydraulic diameter of the main channel. One
dimensionless unit equals 14 mW.

114

MENUWINDOW
Z-SC-ROT CLK
X-SC-ROT CLK
Y-SC-ROT CLK
LEFT MOVE
UP MOVE
NEARMOVE
EXPAND
RESET VIEW
BACKCOLOUR
MARKER SIZE
COARSENESS
1.0001E-01
-1.0631E+00
-2.2263E+00
-3.3895E+00
-4.5526E+00
-5.7158E+00
-6.8789E+00

Figure 9.6: Base 10 logarithm of the energy dissipation rate in reverse flow on the centerplane of a single-element, Tesla-type T45A-2 valve with a volume flow rate of 2987
l/min corresponding to Re=500 based on the hydraulic diameter of the main channel. One
dimensionless unit equals 14 mW.

115
As in the forward-flow case, part of the momentum flux leaving the channel is dissipated
in the goblet-shaped plenum.
9.2.4 Momentum Conservation
Diodic Effects of the Force Vector Interactions
The momentum equation is a vector relation, a summation of forces, so each of its terms has
both magnitude and direction sense. The X and Y components of these forces are shown
in Fig. 9.7 for both the forward and the reverse flow case. The figure caption explains
positive-negative direction sense and the X-Y coordinate system is shown in Fig. 9.1.
Essentially, the figures show that the viscous and pressure forces applied on the fluid and
the fluids momentum flux response are significantly altered from those of the T45A. All
the forward-flow X-direction forces are reduced by as much as a factor of two, except for
those of the inlet channel (block2). The forward-flow Y-direction forces have been reduced
as well, including those in the outlet channel (block 6), the location of the forward-flow
jet, in which the pressure force has been reduced by 40%. The reverse-flow forces have
been reduced as well as a result of the reduction of direction-independent losses by the
shortening of the inlet and outlet channels. The only exceptions are the X-direction forces
in the T-junction (block 3), which are almost unchanged.
Forward-Flow vs. Reverse-Flow Force Vector Magnitudes
Another way to study the diodicity mechanism, especially for control volumes that are not
oriented along the X direction, is to study the vector magnitudes of the force components
in the control volumes as shown in Fig. 9.8, which directly compares the forward-flow and
reverse-flow components.
This perspective shows how much direction-independent flow restrictions have been
decreased by shortening the inlet and outlet channels, while retaining the flow restrictions
due to the diodicity mechanism. The pressure force magnitudes have been decreased by
25-50% except for the forward-flow in the inlet channel (block 2). Even though the reverseflow pressure-force magnitude in the side channel (block 8) has been halved, this is counterbalanced by the fact that the reverse-flow pressure-force magnitudes exceed those of
the forward-flow in all other control volumes except the inlet channel (block 2). The viscous force magnitudes are all diminished approximately by half as well. Since the viscous
forces are comparatively small to the other forces, the momentum fluxes are quite similar

116

Momentum Flux
Viscous Force
Pressure Force

0.15
0.1
0.05
0
0.05
0.1
0.15

0.2
ReverseFlow Xdirection Forces

ForwardFlow Xdirection Forces

0.2

0.2

Momentum Flux
Viscous Force
Pressure Force

0.15
0.1
0.05
0
0.05
0.1
0.15
0.2

4
5
6
Block number

(a) X-direction, forward flow

Momentum Flux
Viscous Force
Pressure Force

0.15

4
5
6
Block number

(b) X-direction, reverse flow

0.1
0.05
0
0.05
0.1
0.15
0.2

0.2
ReverseFlow Ydirection Forces

ForwardFlow Ydirection Forces

0.2

Momentum Flux
Viscous Force
Pressure Force

0.15
0.1
0.05
0
0.05
0.1
0.15
0.2

4
5
6
Block number

(c) Y-direction, forward flow

4
5
6
Block number

(d) Y-direction, reverse flow

Figure 9.7: Force vector terms in the integral form of the momentum conservation equation.
Net pressure force and net momentum flux into a control volume are positive. Viscous force
is applied on the fluid by the wall. X-vectors to the right and Y-vectors upward are positive
and consistent with the T45A-2 valve layout in Fig. 9.1 including the numbering of the
control volumes (blocks).

117

0.2

0.16
0.14
0.12
0.1
0.08
0.06

0.16
0.14
0.12
0.1
0.08
0.06

0.04

0.04

0.02

0.02

4
5
6
Block number

reverse flow
forward flow

0.18
Viscous Force Magnitude

Pressure Force Magnitude

0.2

reverse flow
forward flow

0.18

(a) Pressure force magnitude

4
5
6
Block number

(b) Viscous force magnitude

0.2

reverse flow
forward flow

0.18
Momentum Flux Magnitude

0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0

4
5
6
Block number

(c) Momentum flux magnitude

Figure 9.8: Vector magnitudes of the terms in the momentum conservation equation for the
T45A-2 valve.

118
in magnitude and location to the pressure forces.

9.2.5 Kinetic-Energy Conservation


To assess the diodicity enhancement of the T45A-2, the sum of each kinetic energy term
over the entire valve was compared with that of the T45A. The energy dissipation rate in the
T45A-2 is reduced to 49% of the T45A rate in forward flow while 64% is retained in reverse
flow. The energy flux rate in the T45A-2 is reduced to 41% of the T45A rate in forward
flow while 78% is retained in reverse flow. The pressure work rate applied to the fluid in the
T45A-2 is reduced to 52% of the T45A rate in forward flow while 72% is retained in reverse
flow, and predicts a valve diodicity of 1.74 following Eq. 2.14. The diodicity determined
by Eq. 1.1 is also 1.74 compared to the 1.27 of the T45A. The diodicity enhancement has
greatly reduced the pressure work rate required for forward flow.
There are two alternate ways to look at the data. Figs. 9.9 and 9.10 show the three most
important terms of the kinetic energy equation in steady incompressible flow: the pressure
work rate, the energy flux rate, and the dissipation rate, which are shown for each of the
control volumes for both forward and reverse flow.
The enhancement of the diodicity mechanism by reduction of unnecessary flow restrictions is seen in the pressure-work rate. The forward-flow pressure-work rate has been
nearly eliminated in all regional control volumes except the inlet channel (block 2) in which
the flow is accelerated to enter the valve, and the outlet channel (block 6), the location of
the forward-flow jet, which has been reduced by 35% compared to the T45A. The reverseflow pressure-work rate has also been reduced, but not in the inlet channel or the T-junction
(block 3) where the reverse-flow jet is located. Some of these reductions can be attributed
to reductions in the energy-flux rate where the flow in either direction enters the valve. The
T45A-2 has a 42% reduction in the rate of energy flux creation in the inlet plenum and
channel (blocks 1 & 2) in forward flow and the outlet channel and plenum (blocks 6 & 7)
in reverse flow. The dissipation rate also shows the reduction of flow restrictions except for
those that are the diodicity mechanism. The dissipation rate has been greatly reduced in
all regional control-volumes except the inlet channel (block 2) where the reverse-flow jet
is attached to the wall, which has been reduced by only 24%. In comparison, the reduction
in the dissipation rate associated with the forward-flow jet in the outlet channel (block 6) is
72%. Once again, the viscous work rate is inconsequential in all control volumes.
An alternate view of the kinetic energy in the valve is the contribution of each term of
the kinetic energy equation to the forward-flow direction and the reverse-flow direction as

119

reverse flow
forward flow

Pressure Work Rate

0.05

0.05
1

4
5
6
Block number

(a) Pressure work rate

reverse flow
forward flow

Energy Flux Rate

0.05

0.05
1

4
5
6
Block number

(b) Energy flux rate

Figure 9.9: Magnitude of the pressure work rate and the energy flux rate terms in the
kinetic-energy conservation equation from the T45A-2 valve simulations with a volume
flow rate of 2987 l/min corresponding to Re=500 based on the hydraulic diameter of the
main channel.

120

reverse flow
forward flow

Dissipation Rate

0.05

0.05
1

4
5
6
Block number

(a) Energy dissipation rate

reverse flow
forward flow

Viscous Work Rate

0.05

0.05
1

4
5
6
Block number

(b) Viscous work rate

Figure 9.10: Magnitude of the energy dissipation rate and the viscous work rate terms in
the kinetic-energy conservation equation from the T45A-2 valve simulations with a volume
flow rate of 2987 l/min corresponding to Re=500 based on the hydraulic diameter of the
main channel.

121
shown in Fig. 9.11. The forward-flow kinetic-energy terms are strikingly reduced from
those of the T45A. Only the terms associated with the acceleration of the fluid in the inlet
channel (block 2) and the forward-flow jet in the outlet channel and plenum (blocks 6
& 7) are of significance. Even the longest channel section, the main channel (block 4),
transforms very little pressure-work to the other forms. Its flow is driven by energy flux
created in the inlet channel (block 2). The reverse-flow kinetic-energy terms have been
somewhat reduced in comparison to the T45A, except for those associated with the reverseflow jet in the inlet region and T-junction (blocks 1, 2, & 3).

9.3 Discussion
The geometrical layout of the T45A-2 valve is a modification of the lower-performing
T45A obtained by directly following the valve design guidelines based on knowledge of the
low-Reynolds-number diodicity mechanism revealed by the numerical methods of Chap.
3. No other numerical valve simulations of any kind were done varying the geometry of
the T45A to determine this new design; this design is not an optimum version selected
from a number of attempts. The T45A-2 is the result of a single effort to apply the design
guidelines to improve the T45A. It is also important to note that the numerical methods
employed to analyze the T45A-2 were exactly identical to those used in the T45A and
T45C analyses. The only variation is the substitution of the physical grid of the T45A-2
for that of the T45A or T45C. The numerical results show that the design guidelines are
effective and sufficient to create a valve with an enhanced diodicity mechanism.
The T45A-2 was shown to be functionally similar to the T45A. The diodicity mechanism is located in the side channel and T-junction (blocks 3, 8, 9). The underlying physical
basis for the diodicity of the T45A-2 valve is its ability to direct the majority of the volume flow through alternate paths depending on flow direction. In forward flow 94% of the
fluid proceeds through the main channel (blocks 3 4 5); in reverse flow 48% proceeds
through the side channel (blocks 5 9 8 3).
Analysis of the field variables of the T45A-2 simulations shows that following the design guidelines has enhanced the diodicity mechanism. The redesign of the inlet channel
has resulted in more gradual acceleration of the forward flow as it enters the valve, reducing
the pressure loss and high dissipation rate along the wall. The diverging walls of the main
channel have reduced the forward-flow velocity as well as the energy dissipation along the
walls. This has slowed and widened the forward-flow jet in the outlet channel reducing its
momentum. In the reverse-flow direction the pressure losses as the flow enters the valve

122

Energy Flux Rate


Dissipation Rate
Viscous Work Rate
Pressure Work Rate

ForwardFlow Power

0.05

0.05
1

4
5
6
Block number

(a) Forward flow

Energy Flux Rate


Dissipation Rate
Viscous Work Rate
Pressure Work Rate

ReverseFlow Power

0.05

0.05
1

4
5
6
Block number

(b) Reverse flow

Figure 9.11: Magnitude of the terms in the kinetic-energy conservation equation from the
T45A-2 valve simulations with a volume flow rate of 2987 l/min corresponding to Re=500
based on the hydraulic diameter of the main channel.

123
have been increased by sharpening the corners at the mouth of the valve and narrowing the
outlet channel. The converging walls accelerate the reverse flow so that it impinges on the
guide vane with a high dissipation rate and creates a large recirculation in the main channel. The converging walls of the side channel accelerate and increase the momentum of the
reverse-flow jet sufficiently to cross the main channel and attach to the opposite wall of the
inlet channel. The diverging walls of the inlet channel provide a surface for dissipation of
the reverse-flow jet, yet offer a low pressure loss entry to the forward-flow.
The momentum perspective showed that the design guidelines correctly address the
minimization of direction-independent flow restrictions while retaining the flow restrictions
of the diodicity mechanism. The kinetic-energy perspective reinforces this conclusion by
showing that the pressure-work rate, energy-flux rate, and dissipation rate have all been
halved in forward flow in the T45A-2 compared to the T45A, while the reverse-flow rates
have only been reduced by one quarter. The direction-independent flow restrictions have
been minimized while those due to the diodicity mechanism have been maintained.
Research objective #3b has been completed by demonstrating the effectiveness of the
valve design guidelines of the previous chapter by using them as the sole method to obtain a
valve design in which the diodicity mechanism, and thus the diodicity, has been enhanced.
In the T45A-2 the direction-independent flow restrictions have been reduced, the forwardflow jet has been weakened by decelerating the flow before it reaches the Y-junction, and
the reverse-flow jet has been accelerated sufficiently to cross the main channel and induce
a high rate of energy dissipation where it is attached to the opposite wall. By applying the
valve design guidelines to modify the geometry of the T45A, the T45A-2 design has been
created with a diodicity of 1.74 compared to 1.27 for the T45A, a statistically significant
diodicity increase of 27-47% with 95% confidence.

124

Chapter 10
CONCLUSIONS
Previous researchers of Tesla-type NMP valves have focused on inertial forces as the
sole source of the diodicity mechanism because their devices were macro-scale valves containing high Reynolds-number flows. This level of understanding is of questionable value
at the micro-scale, since the physical dimensions of NMP microvalves lead to laminar
flows with Reynolds numbers well below 1000 at typical operating conditions. Thus, efforts to design optimal NMP microvalves suffer from a lack of understanding of the fluidic
mechanisms that inhibit reverse flow in laminar low-Reynolds-number flow, resulting in a
dependence on the build & test method.
A numerical method using field variable analysis and momentum and kinetic-energy
conservation in regional control-volumes was developed from the governing equations to
reveal the low-Reynolds-number diodicity mechanism. The numerical method was applied to two distinct designs of Tesla-type NMP valves and was consistently able to discern
the fluidic mechanism responsible for the variation in pressure drop between forward and
reverse flow in each valve design. It revealed their low-Reynolds-number diodicity mechanism as the viscous dissipation surrounding laminar jets that have flow-direction-dependent
locations and orientations. This diodicity mechanism is dominated by viscous forces, unlike the high-Reynolds-number mechanism of macro-scale valves, which is solely due to
inertial forces. Understanding of the low-Reynolds-number diodicity mechanism was employed in the development of effective design guidelines for the geometrical layout of optimal Tesla-type NMP valves. The value of these guidelines was demonstrated by using
them as the sole method to obtain a valve design in which the diodicity was increased by a
statistically significant amount.

10.1 Develop a Numerical Method to Reveal the Diodicity Mechanism


Analysis of the velocity, pressure, and energy dissipation fields from the numerical simulations provide qualitative information about the role that physical flow phenomena (laminar
jets, recirculation regions, regions of high pressure-gradient, and regions with high energy-

125
dissipation) play in creating valve diodicity. But a source of quantitative information is
neccessary to assess the significance of each of these phenomena to the diodicity mechanism. To this end, each valve was divided into nine regional control volumes and integral
forms of the momentum conservation and kinetic-energy conservation equations were developed and applied in these control volumes. The magnitudes of the components of the
conservation equations in each control volume provide the quantitation that reveals which
phenomena are important and how their effects combine to create the diodicity mechanism.
The momentum perspective provides information about the interplay of forces between
the surfaces, flow boundaries and fluid. In each regional control volume it is possible
to determine the pressure force and shear force that each surface applies to the fluid, the
pressure force applied by each flow boundary, and the resulting momentum flux in each of
the three coordinate directions. Since Tesla-type NMP valves inherently have multiple flow
directions and the force components of the momentum perspective are vectors, it is often
difficult to relate them directly to valve diodicity, a scalar quantity. The kinetic energy,
also a scalar quantity, presents no such difficulty. The kinetic energy perspective provides
information on the transfer of energy across flow boundaries, work done on the fluid, and
dissipation within the fluid. There is a direct correspondence between diodicity and the
ratio of forward and reverse pressure-work rates when the inlet and outlet surfaces are
located such that the pressure on each surface has an approximately constant value. The
energy transfers and interplay of forces in each regional control volume are instrumental in
determining how the physical flow phenomena create the diodicity mechanism.
10.1.1 Verify Mathematically Correctness
The numerical method includes not only the incompressible Navier-Stokes solver implemented in CFX 4.2, a mature commercial software package, but also user subroutines
written in FORTRAN to calculate kinetic-energy conservation. These were validated by
using the numerical method to model steady-flow in a slot and compute the magnitude of
the terms of the kinetic-energy conservation equation. These magnitudes agree very well
with those obtained from the analytical solution for steady-flow in a slot.
10.1.2 Verify Steady-Flow Response Predictions
The validation of the steady-flow modeling by the numerical method was achieved by comparing volume flow-rate and diodicity predictions to 242 independent experimental measurements from physical-realized valves of three distinct groups of etch depths and two

126
distinct Tesla-type designs, the T45A and the T45C. The tests of the physical-realized
valves were divided into three groups considering etch depth and valve design: the T45A
group, the T45C group, and the Deep T45C group. Numerical simulations were performed
using the mean etch depth of each of the valve test groups: h 110 5m for the T45A Test
Group, h 115m for the T45C Test Group, and h 148 3m for the Deep T45C Test


Group. It is important to note that exactly the same numerical method was employed in
all cases; the only variation between simulations was the substitution of the physical grid
of the T45A or T45C scaled in the thickness dimension to match the appropriate mean
etch-depth.
Analysis of the experimental data showed that the power-law function of Eq. 4.1 provides a good estimate of the true functional relationship between measured pressure-drop
and volume flow-rate and there is no justification to use a more complicated function. Thus
the power-law function was used to represent the numerical methods predictions of flowrate response to applied pressure.
Statistical analysis of the numerical simulation results showed that the flow-rate response is underpredicted by the numerical method by 7.62% on average over all three test
groups: T45A, T45C, and Deep T45C. This is good agreement considering that the etch
depth measurement accuracy is 6.5% and that, in order to be a predictive tool, the numerical model is gridded-up from the layout drawing of the photolithographic mask used in the
etching process instead of the as-etched shape of the physical devices.

10.1.3 Verify Diodicity Prediction Accuracy


The diodicity is insensitive to any Reynolds-number underprediction that is of similar proportions in both forward and reverse-direction flow. The diodicity is overpredicted by
4.14% on average over all three test groups: T45A, T45C, and Deep T45C. For the particular flow-rate of Re 500 that is characteristic of the numerical simultations for the diodicity
mechanism study, the diodicity prediction error with 95% confidence is e Di
or 0 708 eDi 9 55%.


4 42 5 13%


By making comparisons between predicted and experimental values from the multiple
valve geometries of the T45A, T45C and Deep T45C Test Groups we demonstrated that we
are free to modify valve geometry (to the same extent as the variation between the T45A
and T45C) and still obtain accurate flow rate and diodicity predictions from the numerical
method. Thus the numerical method was validated as a firm basis on which to accept or
reject valve designs for the enhancement of diodicity.

127
10.1.4 Verify Transient-Flow Response Predictions
Since there are no analytical solutions for transient flow in an NMP valve nor experimental
methods available to directly measure it, the transient response predictions of the numerical method were verified by modeling the harmonic response of 2-D slot flow for which
an analytical solution does exist. Since excellent agreement was shown, the numerical
method was applied to model harmonic flow in a T45A valve to show that the kineticenergy transient term is negligible. The insignificance of the transient term allowed the
investigation of the diodicity mechanism using steady-state numerical solutions instead of
the more computationally-intensive harmonic-response solutions.
10.1.5 Reveal the Diodicity Mechanism in Low Reynolds Number Flow in Tesla-Type
NMP Valves
The underlying physical basis for the diodicity of the T45A valve is its ability to direct
most of the forward flow (85%) through the main channel and most of the reverse flow
(64%) through the side channel. In each flow direction a laminar jet is created, surrounded
by a highly-dissipative shear layer. Most of the transformation of pressure work to energy
dissipation occurs in the shear gradient surrounding the laminar jets, and the shear layer
surrounding the reverse-flow jet in the side channel and the inlet channel is more dissipative than the shear layer surrounding the forward-flow jet in the outlet channel. Energy
dissipation in the fluid itself contributes more to diodicity than does wall friction. Energy
flux is also a contributor to diodicity, it is even more direction dependent than energy dissipation, but it is an order of magnitude smaller. The vector nature of jet momentum also
contributes to diodicity. Substantial pressure force is needed to alter the direction of the
reverse flow jet to turn downstream, but not the forward flow jet, which is already pointed
downstream.
The T45C is functionally similar to the T45A. However, the T45A contains a reverseflow jet that attaches to the wall in the side channel and then proceeds down the center of
the inlet channel with a dissipative shear layer surrounding it; in the T45C the reverse-flow
jet doesnt attach to the wall in the side channel and thus has enough additional energy to
cross the main channel and attach to the opposite inlet-channel wall. The dissipation rate
between the jet and the wall is greater than that experienced by the T45A jet at the center
of the channel. A second difference is the decreased length of the T45C outlet channel,
which decreases the energy dissipation between the forward flow jet and the outlet channel
wall where it is attached. As a result, the diodicity due solely to the ratio of reverse and

128
forward-flow dissipation rates is much higher in the T45C (1.35 instead of 1.13). But the
combination of effects is a statistically insignificant increase in diodicity from 1.27 in the
T45A to 1.33 in the T45C.
While macro-scale high-Reynolds-number valves also have direction-dependent jets,
the direction-dependent variation of energy dissipation rate, which is of such great importance in low Reynolds-number flow, would be negligible. Instead the directional variation
of the macrovalve jets inertial forces would be paramount.

10.2 The Low-Reynolds-Number Diodicity Mechanism is Dominated by Viscous Forces


In a classic analysis of a macro-scale Tesla-type valve with turbulent flow of Re 1 700,
Paul [23] performed a control-volume analysis assuming viscous forces were negligible.
His analytical solution for pressure drop due solely to the change in momentum flux agreed
with experimental data, showing the diodicity mechanism was due only to inertial forces.
But in micro-scale Tesla-type valves the typical operating conditions generate a maximum flow-rate of Re 500. To assess the effect of Reynolds number on the balance be

tween inertial and viscous forces, normalization and approximation theory were applied to
the momentum and kinetic-energy conservation equations. The results suggest that viscous
force in microvalve flows becomes larger than momentum flux for Re 100 and negligible
for Re 1000, and that energy dissipation is dominant over energy flux for Re 100 and
significant up to Re 1000. Thus as the scale of the valve decreases, the viscous forces


become ever more dominant.


Steady-flow numerical simulations of the T45A and T45C valves corroborate the results
of approximation theory. Even at Re 500, fully 2/3 of the pressure work is dissipated
and only 1/3 is transformed to energy flux. Ignoring viscous effects in the low-Reynoldsnumber flow in valves would lead to invalid results. This is particularly clear for the T45C
valve, since conveniently all four measures: the diodicity and the ratios of reverse and
forward-flow pressure work, energy flux, and energy dissipation are all within 2% of 1.33.
Over 2/3 of the pressure work is expended in energy dissipation, thus 2/3 of the diodicity
is due to dissipation. Clearly viscous effects are dominant.
A further effect of the dominance of viscous forces over inertial forces is the directiondependence of valve resistance (see App. A) and the direction-independence of valve inertance (see App. B).

129

10.3 Demonstrate Knowledge of the Diodicity Mechanism


Since the resistance and inertance of the valve must be matched to the dynamics of the
microfluidic system in which it operates, there is no single optimal valve design. However,
the knowledge gained from analysis of the diodicity mechanisms of the T45A and T45C
valves was applied to develop guidelines the design of high-performance Tesla-type NMP
valves. As an example we have applied these guidelines to modify the lower-performing
T45A valve to produce a T45A-2 valve that has higher performance than the T45C.
10.3.1 Develop Valve Design Guidelines
The diodicity can be cast in terms of the ratio of pressure work rate in the reverse and
forward direction, or cast in terms of the ratios of energy dissipation rate and energy flux
2
rate. In the latter case the diodicity is seen as a function of u3i and  ui x j  , so that for a
given flow cross-sectional area, energy losses are exhibited (in order of increasing losses)
by plug flow, fully-developed parabolic profiles, and highly-asymmetric velocity profiles.
2

The relative importance of u3i and  ui x j  is controlled by the Reynolds number; when
Re 1000 the energy flux rate dominates; when Re 100, the dissipation rate dominates.
Thus, the diodicity is strongly affected by the asymmetric velocity profiles of separated


laminar jets and their accompanying high velocity gradients (especially when the jet is
attached to the wall). And not only the velocity magnitude of the laminar jet is important,
but also its direction. Thus, a reverse-flow jet should be oriented counter to the downstream
direction, and if it is not possible to avoid creating a jet in forward flow, it should be oriented
downstream.
As the Reynolds number increases and less of the dissipation occurs within the geometric boundaries of the valve itself, it becomes increasingly important to ensure that channels
upstream and downstream of the valve are dissipative not diffusive, (ie. do not transfer
energy flux back into available pressure work in the fluid). If this is not possible, an alternative strategy is to choose to create diodicity from the direction-dependent dissipation
rates within the valve only and attempt to recover kinetic energy in both flow directions by
providing diffusers at the ends of the valve. This is a particularly appropriate strategy when
it is desirable to minimize valve resistance.
Valve design guidelines were presented based on knowledge of the low-Reynoldsnumber diodicity mechanism revealed by the numerical methods of Chap. 3. Their intent
is to enhance the diodicity mechanism by increasing the energy flux and dissipation rates
in the reverse-flow jet and decreasing them in the forward-flow jet, and also by decreasing

130
flow restrictions that are direction-independent.
10.3.2 Demonstrate the Effectiveness of the Guidelines
As an illustration of our understanding of the low-Reynolds-number diodicity mechanism
in Tesla-type NMP valves, a T45A-2 valve was created by modifying the geometry of the
T45A following the design guidelines. No additional technique was employed. The same
numerical methods employed to analyze the T45A and T45C were used to analyze the
T45A-2. The only variation was the substitution of the physical grid of the T45A-2 for
that of the T45A or T45C. Analysis of the T45A-2 simulations showed that the directionindependent flow restrictions were minimized while retaining the flow restrictions of the
diodicity mechanism. The forward-flow jet was weakened by decelerating the flow before
it reaches the Y-junction; the reverse-flow jet was accelerated sufficiently to cross the main
channel and induce a high rate of energy dissipation where it attaches to the opposite wall.
The new inlet channel geometry provides a wall surface for dissipation of the reverse-flow
jet, yet offers little flow restriction in the forward-flow direction. In the reverse-flow direction the flow restriction where the flow enters the valve was maintained by sharpening
the corners at the mouth of the valve and narrowing the channel. Overall, the pressurework rate, energy-flux rate, and dissipation rate have all been halved in forward flow in
the T45A-2 compared to the T45A, while the reverse-flow rates have only been reduced
by one quarter. By applying the valve design guidelines the diodicity mechanism has been
enhanced and the previous diodicity of 1.27 has been increased to 1.74, a statistically significant diodicity increase of 27-47% using a 95% confidence level. These results show that
the valve design guidelines are effective and sufficient to create a valve with an enhanced
diodicity mechanism. The low-Reynolds-number diodicity mechanism in Tesla-type NMP
valves has been revealed; there is no longer any need to rely on the build & test method.

10.4 Future Work


Since the numerical method developed in this research to reveal the low-Reynolds-number
diodicity mechanism of Tesla-type valves is a general method that applies to all NMP
microvalves, it is recommended that this method be applied to reveal the diodicity mechanisms and develop guidelines for the optimal design of other types of NMP valves, such
as: diffuser valves, asymmetric conduits, and vortex diodes.

131

BIBLIOGRAPHY
[1] Dale A. Anderson, John C. Tannehill, and Richard H. Pletcher. Computational Fluid
Mechanics and Heat Transfer, page 45. Hemisphere Publishing Corporation, 1984.
[2] R.L. Bardell and F.K. Forster. Impedances for design of microfluidics systems. In
Micro Total Analysis Systems 98 Workshop held in Banff, Canada, 13-16 October,
pages 299302, 1998.
[3] R.L. Bardell, N.R. Sharma, F.K. Forster, M.A. Afromowitz, and R.J. Penney. Designing high-performance micro-pumps based on no-moving-parts valves. In Microelectromechanical Systems (MEMS), DSC-Vol. 62/HTD-Vol. 354, pages 4753. ASME
IMECE, 1997.
[4] Roger J. Barlow. Statistics, a Guide to the Use of Statistical Methods in the Physical
Sciences, pages 5758,134138. John Wiley & Sons, 1989.
[5] Bichara and Orner. Analysis and modeling of the vortex amplifier. Journal of Basic
Engineering, pages 755763, December 1969.
[6] William E. Boyce and Richard C. DiPrima. Elementary Differential Equations and
Boundary Value Problems, page 483. John Wiley & Sons, third edition, 1977.
[7] CFX 4.1. User Guide, pages 117129. Computational Fluid Dynamics Services,
1995.
[8] B. W. Char, K. O. Geddes, G. H. Gonnet, B. L. Leong, M. B. Monagan, and S. M.
Watt. First Leaves: A Tutorial Introduction to Maple V, pages 4148. SpringerVerlag, 1992.
[9] F.K. Forster, R.L. Bardell, M.A. Afromowitz, N.R. Sharma, and A. Blanchard. Design, fabrication and testing of fixed-valve micro-pumps. In Proceedings of the ASME
Fluids Engineering Division, pages 479484, 1995.

132
[10] Torsten Gerlach. Microdiffusers as dynamic passive valves for micropump appplications. Sensors and Actuators A, 69:181191, 1998.
[11] P. Gravesen, J. Brandebjerg, and O. Sondergard Jensen. Microfluidics - a review. J.
Micromech. Microeng., 3:168182, 1993.
[12] Duane Hanselman and Bruce Littlefield. Mastering MATLAB, A Comprehensive Tutorial and Reference. Prentice-Hall, 1996.
[13] J. O. Hinze. Turbulence, 2nd ed., pages 14. McGraw-Hill, 1987.
[14] O. C. Jones Jr. An improvement in the calculation of turbulent friction in rectangular
ducts. J. Fluids Engineering, Trans. ASME, pages 173181, June 1976.
[15] Stephen J. Kline. Similitude and Approximation Theory, pages 68117. SpringerVerlag, 1986.
[16] L. Gary Leal. Laminar Flow and Convective Transport Processes, pages 1127.
Butterworth-Heinemann, 1992.
[17] C. J. Morris and F. K. Forster. The correct treatment of harmonic pressure-flow behavior in microchannels. In MEMS. ASME IMECE, 2000. Accepted for publication.
[18] Katsuhiko Ogata. System Dynamics, pages 258264. Prentice-Hall, 1978.
[19] A. Olsson, P. Enoksson, G. Stemme, and E. Stemme. A valve-less planar pump in
silicon. In Transducers 95, pages 291294, 1996.
[20] A. Olsson, G. Stemme, and E. Stemme. A numerical design study of the valveless
diffuser pump using a lumped-mass model. J. Micromech. Microeng., 9:3444, 1999.
[21] Ronald L. Panton. Incompressible Flow, pages 279283. John Wiley and Sons, 1984.
[22] Ronald L. Panton. Incompressible Flow, pages 128135. John Wiley and Sons, 1984.
[23] F.W. Paul. Fluid mechanics of the momentum flueric diode. In IFAC Symposium on
Fluidics, Section A, pages 115. Royal Aeronautical Society, 1968.

133
[24] Jay L. Reed. Fluidic rectifier. U.S. Patent No. 5,265,636, 1993.
[25] Frederick S. Sherman. Viscous Flow, pages 146149. McGraw-Hill, 1990.
[26] S. Shoji and M. Esashi. Microflow devices and systems. J. Micromech. Microeng.,
4:157171, 1994.
[27] S. Taneda. Visualization of separating Stokes flows. Journal of the Physical Society
of Japan, 46:1935, June 1979.
[28] E.H. Tesla. Valvular conduit. U.S. Patent No.1,329,559, February 1920.
[29] P. Voigt, G. Schrag, and G. Watchutka. Electrofluidic full-system modeling of a flap
valve micropump based on kirchoffian network theory. Sensors and Actuators A,
66:914, 1998.
[30] Frank M. White. Viscous Fluid Flow, 2nd ed., pages 72,120,133. John Wiley and
Sons, 1991.
[31] Frank M. White. Fluid Mechanics, 3rd ed., pages 307311. John Wiley and Sons,
1994.
[32] R. Zengerle and M. Richter. Simulation of microfluid systems. J. Micromech. Microeng, 4:192204, 1994.

134

Appendix A
VALVE RESISTANCE MODELING
The main purpose of this appendix is to show that the resistance to fluid flow through
a valve in response to an applied pressure difference is dependent on the flow direction in
the valve. This is anticipated by the dimensional analyses in Secs. 2.2 and 2.3 that show
that the diodicity mechanism of a microvalve is mainly due to viscous forces. In addition,
appropriate valve resistance values are suggested for linear and nonlinear system models.
Resistance as used in lumped-parameter system models was introduced in Sec. 1.2.3.
Fluid resistance is the change in pressure drop with respect to a change in volume flow rate,
so the local-slope valve resistance was calculated from the power-law relation Eq. 4.1 as
Rlocal slope

dP
dQ

nQn

1


(A.1)

in which the constants n and have different values for the forward and reverse flow directions, ie. nF nR and F R . It is refered to as local-slope resistance because the
fluid flow in the valve exhibits significant non-linear behavior, thus the slope dP dQ varies
dramatically over the typical operating range of flow rates.
Figure A.1 compares the local-slope resistances calculated by applying Eq. A.1 to
the numerical and experimental pressure versus flow-rate data of the T45A and the T45C
valves. The local-slope resistance of the T45A is overpredicted by 5% in the forward-flow
direction and by 7% in the reverse; the T45C by 5% in the forward and 9% in the reverse.
In each case the reverse-flow resistance is larger than the forward-flow resistance at any
measured Reynolds number, thus the dependence of resistance on flow direction, as well as
flow rate, is clear.
In linear system models a single constant resistance value is required and care should
be taken in its choice. If a fluid flow is oscillatory and the oscillation is a small perturbation
about a mean flow rate, then the local-slope R at the mean flow rate would be representative
of R across the flow-rate perturbation. However, in typical valve operation, the flow oscillation is much larger than the mean flow rate, (ie. the slosh flow >> net flow), and using
the value of R at the slosh flow amplitude would not be representative of valve resistance
over the entire range of flow rate. In this situation Ogata [18] recommends using an aver-

135
12

Bi2 112um
Bo2 109um
Ti2 110um
To2 111um
simA 120um
simB 110um

1
0.5

x 10

1.5
3

1.5
Localslope resistance Pa*s/m

12

x 10

Localslope resistance Pa*s/m

0
0.5
1
1.5

1
0.5

Li2 116um
Lo2 114um
Li2t2 116um
Lo2t2 114um
simA 120um
simB 115um

0
0.5
1
1.5

2
1000

500

0
Reynolds number

500

1000

2
1000

500

(a) T45A Valves

0
Reynolds number

500

1000

(b) T45C Valves

Figure A.1: Local-slope resistance to fluid flow vs. Reynolds number in T45A and T45C
valves from both experiment and numerical simulation following Eq.A.1, which is based
on the fitted power-law relation, Eq.4.1.

age resistance that is the slope of a line connecting the pressure P at the slosh flow rate
amplitude Qa to the origin, which using the power-law fit introduced above is
P
Qa

Raverage

Qna
Qa

Qan

1


(A.2)


A more representative value of R was obtained by considering the harmonic nature of valve
flow Q t
Qa sin t in typical operating conditions and computing the time-averaged
resistance R (simplified by a coordinate change from t to ) as


n Qa sin

n 1


Rlocal slope

1 n
sin

1


d


(A.3)

which considered 1/2 of a period (eg. the forward-flow portion of a cycle). The integral
is a function of the power-law exponent n; a look-up table for 1 n 2 was numerically
integrated via Maple V software [8]. Ogatas relation can be derived from a time-average


of Q t as a triangle wave. As can be seen in Fig. A.2, the time-averaged (sine) resistance
is higher than the average resistance, but significantly lower than the local-slope resistance.


For a linear system model containing an NMP valve with harmonic flow, the appropriate value for R is the time-averaged resistance at the expected flow rate amplitude. For

136
1
0.8
0.6

Local slope
Timeave (sine)
Average

Resistance

0.4
0.2
0
0.2
0.4
0.6
0.8
1
1

0.5

0
Flow rate

0.5

Figure A.2: Fluid resistance vs. volume flow rate in a typical NMP valve. The timeaverage and average R are approximations of the local-slope R for use in linear models
where a single value is required. Note the similarity to the characteristic curve of nonlinear
friction for an object moving at low Re in a fluid medium.

nonlinear system models in which resistance can be a function of the slosh flow rate, the
fitted parameters and n in Eq. A.1 are useful since the governing equation for a series
combination of resistor and inductor can be stated as
P t


Qn


dQ
dt


(A.4)

We have developed methods to obtain appropriate values to characterize valve resistance in both linear and nonlinear system dynamics models, and shown that the resistance
in a Tesla-type NMP valve is direction dependent.

137

Appendix B
VALVE INERTANCE MODELING
This appendix shows that the inertance that characterizes a Tesla-type NMP valve in a
lumped-parameter system dynamics model is independent of flow direction. Since there
are no analytical solutions for transient flow in an NMP valve and no experimental methods available to directly measure valve inertance, the accuracy of the transient response
predictions of the numerical method were verified by employing them to model the step response of flow in a slot since those numerical results could be compared with an analytical
solution. Then the numerical method was applied to model step response in the T45A and
T45C valves and show that the inertance is independent of flow direction.

B.1 Step Response of a 2-D Slot


Analytical Solution
Relations to characterize the step-response of a channel in lumped-parameter dynamics
modeling were introduced in Secs. 1.2.4 and 1.2.5. Following Eq. 1.7 the resistance of a
slot of height h and length L is
R 12L h3
(B.1)
and following Eq. 1.8 the inertance of inviscid flow in a slot with a cross-sectional area per
unit width of A h is
Iinviscid

L h

(B.2)


To rewrite these equations in nondimensional form we use the characteristic parameters


p u
p and t x u of Sec. 3.5 except for the characteristic length, which
was taken as x

116 9m. We find

L x R

p
and Iinviscid
x u

Iinviscid

p t
x u

138
remembering that R and Iinviscid are on a per-unit-width basis. Substituting these into Eqs.
B.1 and B.2 gives us the nondimensional forms as
12L
and Iinviscid
Re

R
in which Re

u x

(B.3)

But the inviscid-flow inertance is only approximate; when viscosity is considered the
crossectional shape of the channel becomes important. For starting flow in a pipe carrying viscous fluid White [30] presents an analytical solution utilizing Bessel functions.
Following his basic method a series solution for starting flow of a viscous fluid in a twodimensional slot was developed and is described in App. C. This series solution gives the
nondimensional inertance I in a slot as
I

2
L
8

1 23L

(B.4)

that is 23% larger than the inviscid flow inertance of Eq. B.3. Since the numerical method
considers viscosity, Eq. B.4 supplies the inertance value for comparison.

Numerical Solution
The step response of a 2-D slot was simulated using the numerical method described in
Chap. 3, the same method used to determine the step response of the NMP valves. Thus,
the slot simulation was performed as a 3-D rectangular channel with a 100 to 1 crossstream aspect ratio to ensure the 2-D character of the flow and a streamwise length of 10.
The same characteristic parameters used above for Eq. B.3 were used to nondimensionalize
the variables. A step pressure difference P was applied at time t
was allowed to procede until steady-state was reached.

0, and the simulation

The governing equation for a first-order model that represents a fluid channel as a resistor and an inductor in series is
P
At time t

RQ t

0, the volume flow-rate is Q 0


I

I Q t

0. Solving for the inertance gives


P
Q 0

(B.5)

139
The derivative of the volume flow-rate at t 0 was obtained from the numerically-derived
flow-rate vector Q t and time vector t via a O2 accurate three-point forward-differencing
method [1]
3Q1 4Q2 Q3
Q 0
t3 t1


in which the subscripts refer to index position in the vector.


The inertance prediction of Eq. B.5 was used for comparison with the analytical solution, Eq. B.4.

B.2 Step Response of a 2-D Slot


The predicted volume flow-rate response to an applied pressure difference in a 2-D slot is
compared in Fig. B.1 with the series solution of Eq. C.8. At equilibrium the nondimensional flow rate was Q tequil
0 098075 in response to an applied pressure difference of
P 0 01atm. This determined the proportionality constants and n of the power-law


relation Eq. 4.1 as P Q tequil


0 10196 and n 1, since pressure drop is linearly
proportional to volume flow rate for laminar flow in a slot. The resulting numerical value
for inertance following Eq. B.5 is I 1 217L, which is 99% of the analytical solution of


Eq. B.4.

B.3 Step Response of an NMP Valve


Step-response simulations of the T45A and T45C valves were performed using the numerical methods of Chap. 3. Because of the nonlinear response of the volume flow rate Q t ,
the inertance was determined from the numerically-calculated values by direct application


of Kirchoffs voltage law for a resistor and inductor in series following Eq. B.5. The constants and n that describe the nonlinear relation between pressure drop across the valve
and the volume flow rate were obtained for each valve and each flow direction from powerlaw fits of the numerical data from the step-response simulations following Eq. 4.1. Figure
B.2 shows that the inertance as a function of the Reynolds number of the T45A and T45C
valves has similar nonlinear behavior and magnitude. The inertance decreases slightly with
increasing Reynolds number but exhibits no significant dependence on flow direction.

140

Flow rate / Equilibrium flow rate

0.8

0.6

0.4

Predicted
Time constant
Exponential

0.2

0
0

0.2

0.4
0.6
0.8
Time / Equilibrium time)

Figure B.1: The predicted volume flow-rate response to an applied pressure difference in a
2-D slot from the numerical method (symbols) shows good agreement with the exponential
response from the series solution, Eq. C.8. The time constant Re 2 is also shown.

25

25

Reverse flow
Forward flow

20

20

15

15

Inertance

Inertance

Reverse flow
Forward flow

10

0
0

10

200

400
600
Reynolds number

(a) T45A valve

800

1000

0
0

200

400
600
Reynolds number

800

1000

(b) T45C valve

Figure B.2: Inertance versus Reynolds number from the step-response simulations via Eq.
B.5. Inertance shows some dependence on flow rate, but not on flow direction.

141

Appendix C
SERIES SOLUTION FOR STARTING FLOW IN A SLOT
For starting flow in a pipe carrying viscous fluid White [30] presents an analytical solution utilizing Bessel functions. Following his basic method a series solution for starting
flow of a viscous fluid in a two-dimensional slot was developed. This method plus a method
using image flows is also found in Sherman [25].
The starting point for the series solution for the starting flow of a viscous fluid in a
slot is the conservation of momentum equation assuming negligible cross-stream velocities
v w 0 and thus through continuity u x 0 resulting in
u
t

dP
dx


1 2 u
Re y2


(C.1)

a nondimensional equation in which the variables were nondimensionalized by the same


parameters as Eq. B.3 but we have dropped the asterisk notation. This partial differential equation models the development of the velocity profile over time at one streamwise
location in the slot.
The steady-flow solution for u was obtained by setting the transient term to zero and
integrating twice. Taking 0


1 as the dimensionless slot height, the steady velocity is




usteady

dP Re
 y
dx 2


y2 

(C.2)

leading to the maximum velocity of


umax

dP Re
dx 8


and a steady volume-flow rate of


Qsteady


dP Re
dx 2


0

y


y2  dy


dP Re
dx 12


(C.3)

To make Eq. C.1 homogeneous and utilize a transient heat conduction solution, the velocity

142
was separated into unsteady and steady parts using
u

4umax y


y2


(C.4)


Substituting this relation for u into Eq. C.1 results in


1 2 u
Re y2

u
t

(C.5)

which is analogous to unsteady heat conduction in a bar. Using separation of variables,


Boyce and DiPrima [6] give a series solution for initial value and boundary conditions of

u y 0
u 0 t

0


f y


u 1 t

0


for 0

for t

Thus the initial value is


f y


4umax  y


which negates the steady component of u at t

0 and the solution is

u y t


bn exp

y2 


n2 2t Re

sin ny

n 1

where the coefficients are


1

bn

8umax
3

f y sin ny dy


n sin n


2 cos n
n3


2


which was integrated with the aid of Maple V software [8]. Note that b n is not a function
of y.


We derive the unsteady portion of the volume-flow rate from the integral of u over the
slot height as in

1


u y t dy


bn exp

n2 2t Re


n cos n


1


n 1

When n is even, cos n 1


0 and the summation is zero. When n is odd, cos n
1
2 and the coefficients are bn
32umax n 3 Thus the unsteady portion of the


143
volume flow rate is also written as

64umax

n 1 3 5 


exp

n2 2 t Re


Because of the rapid growth of the denominator in the series expression and the rapid
decrease in the exponential with increasing n, only the first term of the series is significant.
Thus to good accuracy we have


dP
dx

8Re
exp
4


2t Re

(C.6)

Since the unsteady component of u is contained in u,


the time derivative of the volume-flow
rate can be obtained from the time rate of change of Q and is


Q
t

dP
dx


8
exp
2


2t Re

(C.7)




the steady and unsteady components of the volume flow rate, Eqs. C.3 and C.6, were
combined to produce the volume flow rate as a function of time as


Q
Acknowledging that 4 8

dP
1
Re
dx
12


12 18


Q


8
exp
4


2t Re





(C.8)

12, (within 1.5%), allows simplification to


dP Re
1
dx 12

exp


2t Re


(C.9)

The governing equation for a first-order model that represents a fluid channel as a resistor and an inductor in series is

At time t

RQ t

0, the volume flow-rate is Q 0


I

I Q t

0. Solving for inertance gives




P
Q 0


144
and substituting Eq. C.7 evaluated at t

0 gives


I


P
dP 8
dx 2

1 23L


since the constant pressure-gradient in fully-developed slot flow is dP dx

P L. This is

a 23% larger inertance than given by Eq. B.3 in which viscosity was neglected.

145

Appendix D
DIODICITY FROM A RATIO OF FLOW RATES
Diodicity can be calculated from a ratio of flow rates as well as from a ratio of pressure
drops. Figure D.1 shows the characteristic pressure drop versus volume flow rate curves
for reverse and forward flow in an NMP valve. Combining the power-law relation of Eq.
4.1 with Eq. 1.1 and referring to the points a, b, and c of the figure results in


Di

PR
PF


Qb

Pc
Pa

F Qnc F
F QnaF

nF

Qc
Qb


nF

QF
QR

(D.1)


Pb

where subscript F is forward and R is reverse, showing that diodicity is determined from
either a ratio of pressure drops at the same flow rate or a ratio of flow rates at the same
pressure drop.

ard
rw
Fo

Re

Pressure drop

ve

flo

rse

flo

146

c
a

Volume flow rate

Figure D.1: Characteristic pressure drop versus volume flow rate curves for reverse and
forward flow in an NMP valve. Diodicity can be derived from either the pressure-drop
ratio or the flow-rate ratio.

147

Appendix E
VALVE DIODICITY MEASUREMENTS
T45A Test Group

The measured values of Reynolds number and valve diodicity of the T45A Test Group
corresponding to the data shown in Fig. 4.4.
T45A Bi2

T45A Bo2

T45A Ti2

T45A To2

Re

Di

Re

Di

Re

Di

Re

Di

167.2
250.8

1.00
1.08

169.4
254.1

1.14
1.12

168.7
253.0

1.07
1.15

167.9
251.9

1.03
1.00

334.3
417.9
501.5

1.21
1.19
1.19

338.8
423.6
508.3

1.15
1.16
1.25

337.3
421.7
506.0

1.13
1.16
1.18

335.8
419.8
503.7

1.06
1.19
1.23

585.1
668.7

1.23
1.26

593.0
677.7

1.25
1.31

590.3
674.7

1.16
1.20

587.7
671.7

1.24
1.28

752.3
835.9
919.5

1.28
1.24
1.25

762.4
847.1
931.8

1.29
1.26
1.23

759.0
843.3
927.7

1.24
1.21
1.19

755.6
839.6
923.5

1.30
1.28
1.28

1003.0 1.22 1016.5 1.25 1012.0 1.24 1007.5 1.25

T45C Test Group

The measured values of Reynolds number and valve diodicity of the T45C Test Group
corresponding to the data shown in Fig. 4.5.

148
T45C Li2t2
Re

Di

T45C Li2
Re

Di

T45C Lo2t2
Re

T45C Lo2

Di

Re

Di

165.8 1.04 165.8 1.16

168.7

1.13 165.3 1.07

248.6 1.19 248.6 1.22


331.5 1.27 331.5 1.22
414.4 1.25 414.4 1.29

253.1
337.4
421.8

1.25 248.0 1.21


1.30 330.7 1.28
1.27 413.3 1.28

497.3 1.29 497.3 1.30


580.2 1.31 580.2 1.32

506.2
590.5

1.34 496.0 1.37


1.33 578.7 1.34

663.0 1.36 663.0 1.37


745.9 1.39 745.9 1.38
828.8 1.34 828.8 1.34

674.9
759.2
843.6

1.34 661.3 1.26


1.38 744.0 1.43
1.35 826.7 1.39

911.7 1.34 911.7 1.31 928.0 1.37 909.3 1.38


994.5 1.37 994.5 1.36 1012.3 1.41 992.0 1.39

Deep T45C Test Group


The measured values of Reynolds number and valve diodicity of the Deep T45C Test Group
corresponding to the data shown in Fig. 4.6.
T45C Ri2

T45C Si

Re

Re

Di

Di

T45C
Re

Di

213.1 1.19 113.1 1.06 147.9 1.07


284.1 1.16 150.8 1.00 221.9 1.30
355.1 1.25 226.2 1.06 295.8 1.31
426.1 1.23 301.5 1.19 369.8 1.43
497.1 1.23 376.9 1.15 443.7 1.27
568.1 1.27 452.3 1.13 517.7 1.28
639.2 1.31 527.7 1.15 591.6 1.34
710.2 1.31 603.1 1.29 665.6 1.32
781.2 1.31 678.5 1.28 739.5 1.32
852.2 1.32 753.8 1.32 813.5 1.36
829.2 1.33 887.4 1.37
-

904.6 1.45

149

Appendix F
VALVE LAYOUT POINTS
The x y points that define the geometry of the microvalves were directly extracted from
the CIF-format files from which the vendor created the photolithographic masks used in the
deep reactive-ion etching process that created the valves in silicon.

F.1 T45A Valve


All dimensions in millimeters.
inlet channel length = 0.531
outlet channel length = 0.462
71 points as (x,y,z)
-0.17 -0.001 0
-0.17 0.009 0
-0.168 0.026 0
-0.166 0.037 0
-0.161 0.054 0
-0.155 0.07 0
-0.146 0.087 0
-0.139 0.098 0
-0.126 0.114 0
-0.113 0.127 0
-0.102 0.136 0
-0.089 0.145 0
-0.072 0.154 0
-0.055 0.161 0
-0.041 0.165 0
-0.025 0.168 0
-0.019 0.169 0
-0.007 0.17 0

150
0.008 0.17 0
0.026 0.168 0
0.044 0.164 0
0.068 0.156 0
0.083 0.148 0
0.096 0.14 0
0.109 0.13 0
0.121 0.119 0
0.206 0.204 0
0.189 0.22 0
0.173 0.233 0
0.16 0.242 0
0.141 0.252 0
0.127 0.258 0
0.112 0.264 0
0.094 0.27 0
0.067 0.277 0
0.045 0.281 0
0.024 0.283 0
0.009 0.284 0
-0.009 0.284 0
-0.024 0.283 0
-0.042 0.281 0
-0.058 0.278 0
-0.075 0.274 0
-0.096 0.268 0
-0.13 0.255 0
-0.153 0.243 0
-0.179 0.227 0
-0.199 0.211 0
-0.211 0.199 0
-0.223 0.186 0
-0.236 0.169 0
-0.249 0.149 0
-0.26 0.128 0

151
-0.269 0.106 0
-0.274 0.088 0
-0.277 0.07 0
-0.28 0.049 0
-0.283 0.026 0
-0.284 -0.001 0
0.114 0.126 0
0.734 -0.494 0
0.904 -0.494 0
0.204 0.206 0
-0.94 -0.114 0
-0.284 -0.114 0
-0.284 0.0 0
-0.94 0.0 0
-0.631 -0.114 0
0.354 -0.114 0
0.354 0.0 0
-0.631 0.0 0

F.2 T45C Valve


All dimensions in millimeters.
inlet channel length = 0.566
outlet channel length = 0.321
112 points as (x,y,z)
-0.1 -0.141 0
-0.116 -0.128 0
-0.129 -0.115 0
-0.138 -0.104 0
-0.147 -0.091 0
-0.156 -0.074 0
-0.163 -0.057 0
-0.167 -0.043 0
-0.17 -0.027 0
-0.171 -0.021 0

152
-0.172 -0.009 0
-0.172 0.006 0
-0.17 0.024 0
-0.166 0.042 0
-0.158 0.066 0
-0.15 0.081 0
-0.142 0.094 0
-0.132 0.107 0
-0.121 0.119 0
-0.206 0.204 0
-0.222 0.187 0
-0.235 0.171 0
-0.244 0.158 0
-0.254 0.139 0
-0.26 0.125 0
-0.266 0.11 0
-0.272 0.092 0
-0.279 0.065 0
-0.283 0.043 0
-0.285 0.022 0
-0.286 0.007 0
-0.286 -0.011 0
-0.285 -0.026 0
-0.283 -0.044 0
-0.28 -0.06 0
-0.276 -0.077 0
-0.27 -0.098 0
-0.257 -0.132 0
-0.245 -0.155 0
-0.159 -0.149 0
-0.089 -0.144 0
-0.17 -0.001 0
-0.17 0.009 0
-0.168 0.026 0
-0.166 0.037 0

153
-0.161 0.054 0
-0.155 0.07 0
-0.146 0.087 0
-0.139 0.098 0
-0.126 0.114 0
-0.113 0.127 0
-0.102 0.136 0
-0.089 0.145 0
-0.072 0.154 0
-0.055 0.161 0
-0.041 0.165 0
-0.025 0.168 0
-0.019 0.169 0
-0.007 0.17 0
0.008 0.17 0
0.026 0.168 0
0.044 0.164 0
0.068 0.156 0
0.083 0.148 0
0.096 0.14 0
0.109 0.13 0
0.121 0.119 0
0.206 0.204 0
0.189 0.22 0
0.173 0.233 0
0.16 0.242 0
0.141 0.252 0
0.127 0.258 0
0.112 0.264 0
0.094 0.27 0
0.067 0.277 0
0.045 0.281 0
0.024 0.283 0
0.009 0.284 0
-0.009 0.284 0

154
-0.024 0.283 0
-0.042 0.281 0
-0.058 0.278 0
-0.075 0.274 0
-0.096 0.268 0
-0.13 0.255 0
-0.153 0.243 0
-0.179 0.227 0
-0.199 0.211 0
-0.211 0.199 0
-0.223 0.186 0
-0.236 0.169 0
-0.249 0.149 0
-0.26 0.128 0
-0.269 0.106 0
-0.274 0.088 0
-0.277 0.07 0
-0.28 0.049 0
-0.283 0.026 0
-0.284 -0.001 0
0.114 0.126 0
0.734 -0.494 0
0.904 -0.494 0
0.204 0.206 0
-0.994 -0.256 0
-0.338 -0.256 0
-0.338 -0.142 0
-0.994 -0.142 0
-0.484 -0.256 0
0.501 -0.256 0
0.501 -0.142 0
-0.484 -0.142 0

155

F.3 T45A-2 Valve


All dimensions in millimeters.
27 points as (x,y,z)
0.000000 0.000000 0.000000
0.118087 0.118087 0.000000
0.236174 0.000000 0.000000
0.316784 0.080610 0.000000
-0.167000 0.000000 -0.000000
0.148295 0.249099 0.000000
-0.010804 0.090000 -0.000000
-0.217020 0.000000 -0.000000
0.371174 -0.135000 0.000000
0.451784 -0.054390 0.000000
-0.167000 -0.078746 -0.000000
-0.217020 -0.078746 -0.000000
-0.467000 -0.078746 -0.000000
-0.217020 0.400000 -0.000000
-0.467000 -0.178746 -0.000000
-0.567000 -0.178746 -0.000000
-0.567000 0.206315 -0.000000
0.371174 -0.220000 0.000000
0.438155 -0.167669 0.000000
0.514539 -0.265436 0.000000
0.588684 -0.191290 0.000000
-0.888803 0.013785 -0.000000
-0.888803 -0.361215 -0.000000
-0.888803 0.388785 -0.000000
0.814172 -0.490924 0.000000
0.549007 -0.756089 0.000000
1.079337 -0.225759 0.000000

156

VITA
Born in Minneapolis in 1950, Ronald Louis Bardell began study of the classical piano
and string bass at the age of eight. In 1965 he switched to percussion and played rock&roll
and rhythm&blues at school dances, bars, and nightclubs. He left the music business at age
28 to attend the University of Minnesota and in 1983 received a Batchelor in Mechanical
Engineering degree. He worked in Minneapolis for the Onan Corporation, a division of
Cummins Engines, as a Design Engineer until 1986 when he moved to Sunnyvale, California, to work for Westinghouse Marine Division as a fluids mechanist in the Performance
Analysis group. At age 41 he moved to Seattle to attend the University of Washington, and
in 1994 received a Master of Science degree in Mechanical Engineering. He then became
involved in microfluidics research under Professor Fred K Forster, and was an author on
several proceedings papers and co-inventor on a patent. Since 1999 he has worked as a microfluidics consultant for biotechnology applications and has several patents pending. He
received the PhD in Mechanical Engineering from the University of Washington in 2000.

You might also like