See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/319974638

Mach-Effect thruster model

Article in Acta Astronautica · September 2017

DOI: 10.1016/j.actaastro.2017.09.021

CITATIONS READS

0 1,758

1 author:

Martin Tajmar
Technische Universität Dresden
230 PUBLICATIONS 1,267 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Advanced Electric and Breakthrough Space-Propulsion View project

NanoFEEP - electric propulsion system for CubeSats View project

All content following this page was uploaded by Martin Tajmar on 07 October 2017.

The user has requested enhancement of the downloaded file.

 Mach-Effect Thruster Model

M. Tajmar1

Technische Universität Dresden, 01307 Dresden, Germany

Abstract

The Mach-Effect thruster is a propellantless propulsion concept that has been in development

by J.F. Woodward for more than two decades. It consists of a piezo stack that produces mass

fluctuations, which in turn can lead to net time-averaged thrusts. So far, thrusts predictions had

to use an efficiency factor to explain some two orders of magnitude discrepancy between model

and observations. Here, a detailed 1D analytical model is presented that takes piezo material

parameters and geometry dimensions into account leading to correct thrust predictions in line

with experimental measurements. Scaling laws can now be derived to improve thrust range and

efficiency. An important difference in this study is that only the mechanical power developed

by the piezo stack is considered to be responsible for the mass fluctuations, whereas prior works

focused on the electrical energy into the system. This may explain why some previous designs

did not work as expected. The good match between this new mathematical formulation and

experiments should boost confidence in the Mach effect thruster concept to stimulate further

developments.

1
Professor and Chair for Space Systems, Director of Institute of Aerospace Engineering, Email: martin.tajmar@tu-
dresden.de
1
Nomenclature

a = Acceleration

C = Capacity

c = Speed of Light (=3×108 m/s)

d = Diameter

d33 = Piezoelectric Constant

ρ = Density

E = Energy

ε = Energy Density

ε0 = Electric Constant (=8.854×10-12 F/m)

εr33 = Relative Permittivity

F = Force

f = Frequency

f0 = Resonance Frequency

φg = Gravitational Potential

G = Newton’s Gravitational Constant (=6.67×10-11 m³/kg-s²)

g = Gravitational field

η = Efficiency

I = Current

k, kp = Electromechanical Coupling Coefficient

l = Length
2
M33 = Electrostrictive Constant

m = Mass

NPZT,Screw = Number of PZT Discs and Screws

ω = Angular Frequency

P = Power

ϕ = Phase

Qm = Mechanical Quality Factor

t = Time

tan δ = Dissipation Factor @ 1 kHz

V = Voltage

v = Velocity

Y = Young Modulus

3
Introduction

Propellantless propulsion is a concept which is traditionally associated with tethers, solar sails

or photon rockets. With an on-board power source, such as a nuclear reactor, the photon rocket,

which converts energy into radiation and uses radiation pressure to produce thrust, is the only

propellantless propulsion that is independent of external sources. This makes it in principle

interesting for interstellar travel. However, the thrust F=P/c² is very small and requires

Megawatts to produce milli-Newtons of thrust.

Since the 1990s, James F. Woodward has been developing an alternative approach called Mach-

Effect thruster [1]–[5]. It is based on the well-motivated idea by Sciama [6] that inertia is due

to the interaction of mass with the gravitational background from the whole universe. This is in

fact one of the interpretations of Mach’s principle [7] (“mass out there influences inertia here”)

which was a guideline for Einstein to develop his theory of general relativity. Although

Einstein’s theory is not fully Machian, there are well-known and experimentally verified Mach-

type-effects such as frame-dragging [8],[9], which can be described by the same weak-field

approximation of general relativity as used by Sciama [6]. Woodward used Sciama’s result to

show that time-changing energy content of a body is causing Machian mass fluctuations that

are much larger than one would expect from E=m⋅c². Woodward then devised a method to use

these mass fluctuations for a novel propulsion scheme: Push the mass when it is heavy and pull

it back when it is lighter. This cycle can create a time-averaged net linear impulse in one

direction that satisfies the definition of a propellantless thruster. Apart from Woodward’s own

thrust measurements (e.g. see [1] for a review), in 2016 Buldrini independently replicated this

effect [10]. Recently, it has been shown explicitly that such a scheme does not violate

conservation of momentum [11].

Of course, energy must still be spent to vary the mass and accelerate it. The power-to-thrust

ratio is an important figure of merit to compare it against photon (P/F=3⋅105 W/mN) and other

4
electric thrusters (P/F=20-60 W/mN). At present, typical experimental values for the Mach-

Effect thruster [1] are an order of magnitude better than the photon rocket (P/F=3⋅104 W/mN).

Woodward is using Piezo crystals both as capacitors and actuators to oscillate their energy and

to push/pull them. Both processes must appear at a proper phase between them to produce

thrust.

Unlike a rocket, the thrust for a Mach-Effect thruster is not due to the expulsion of a reaction

force. Instead, the anticipated magnitude of mass fluctuation and the thrust that can result from

those fluctuations is simply calculated using Newton’s 2nd law F=∆m⋅a. The important question

of course is: “How large is the mass fluctuation?”, to calculate the correct thrust and to

benchmark this propulsion scheme against photon rockets.

So far, the predictions and the observed thrust values differ by some orders of magnitudes. It

was suggested that this may be due to material efficiencies that were not properly considered

[1]. The thrust equation used up to now even predicts a dependence on the frequency to the 6th

power, which is not observed (power electronics limitations in tests so far). The only trend that

was experimentally verified by Woodward and coworkers is that the (on/off transient) effect

seems to scale with the fourth power of the applied voltage to the piezo stack (although only 4

data points have been taken up to now) [12]. We will use the same set of data to compare against

our model.

After significant improvements of the experimental techniques, the observed thrusts are in the

sub-µN - µN range, which requires micro thrust balances with high resolution. Proper analysis

and shielding is necessary to rule out possible artifacts such as thermal effects, outgassing or

magnetic interactions as demonstrated by Woodward and coworkers [1],[13]. Apart from the

need for further testing to consolidate the reality of the effect, the large discrepancy between

theory and experimental results persists after some 27 years of development and thus raises

doubts if the observed effects are due to mass fluctuations. Even more, the lack of a correct

model prohibits the development of scaling laws to amplify the effect beyond any doubt.
5
The most sophisticated model was recently developed by Rodal [14], who describes the

movement of the piezo stack by a set of differential equations with over 200 analytical terms

taking material properties into account. His model gives exact predictions; however, he must

assume an empirical efficiency factor of 0.4% to match experimental data. Moreover, no

analytical scaling laws are given in his paper.

Here, a fully analytical model of the Mach-Effect thruster is presented whose predictions match

experimental data and allows the design of optimized thrusters based on mass fluctuations by

taking both design and material properties into account. The model gives an important insight

into how mass fluctuations appear and why the present design works but other designs failed.

Mach-Effect Thruster Design

Fundamentals

The current embodiment of the Mach-Effect thruster consists of a stack of piezo discs that is

similar in design to typical actuators using ferroelectric (PZT=Lead Zirconate Titanate)

materials, which are sold by many suppliers e.g. for ultrasonic applications. In general, if an

electric field is applied across such PZT discs, they expand and contract depending on the field

strength and direction of the field. The piezo/PZT stack is made of several discs that are

mechanically connected in series but electrically connected in parallel (i.e. all discs have the

same electric potential applied between their electrodes). This is achieved by always switching

the polarity from disc to disc such that every electrode faces another electrode with the same

polarity to avoid electric short circuits. Woodward uses brass electrodes which are glued with

epoxy between each disc. The whole assembly is clamped with stainless steel screws between

two end caps, a larger one made from brass with threaded holes and a smaller one made from

aluminum. The screws are tightened to ensure that the piezo stack is well compressed between

the stiff end caps. A schematic sketch as well as an actual thruster is illustrated in Figs. 1 and

2.
6
Clamping is necessary to generate a force. If no clamping is applied, piezos generate maximum

movement but no force. On the other hand, if the stiffness of the clamping is equal to the

stiffness of the piezo stack, no movement will occur but maximum force will be generated. This

situation applies to both the acoustic applications of PZTs as well as the analytical model

developed here. Most actuators choose a clamp stiffness that is well below the piezo stiffness,

as it is the case for the present Mach Effect thruster. The whole assembly is connected with an

aluminum bracket on the opposite side of the larger brass cap to the test structure – for

measurement purposes, that’s a thrust balance. A rubber pad (e.g. Sorbothane) is placed in

between this connection to damp out vibration artifacts and to mechanically de-couple high

frequency vibrations in the piezo-stack assembly from the balance arm.

Basic Concept of Getting Thrust from a Variable Mass

Let’s assume that the mass of a body m0 can change with a certain angular frequency ω. If we

push and pull on this mass with the same frequency, it is easy to see that a net force is generated

if both mass oscillation and actuator oscillation are in phase or at a phase of 180° (which then

results in a change of the direction of force). We simply assume sinusoidal oscillations and use

Newton’s 2nd law like

m(t ) = m0 sin(ωt )
x(t ) = x0 sin(ωt + ϕ )
(1)
d 2 x(t )
a(t ) = 2
= − x0ω 2 sin(ωt + ϕ ) = −a0 sin(ωt + ϕ )
dt

where m0 is the stationary mass, x0 the amplitude of the actuator oscillation and ϕ is the phase

between mass and actuator oscillation. We get a non-zero force for a 0° phase and a zero force

for a 90° phase by making a time-average over one cycle as

7
 F0°− Phase = m(t )a(t ) = −m0 sin (ωt )a 0 sin (ωt ) = −m0 a 0 sin 2 (ωt )
2π
ω
ω m0 a 0
F0°− Phase =
2π ∫F
0
0° − Phase
dt = −
2
 π (2)
F90° − Phase = m(t )a (t ) = −m0 sin (ωt )a0 sin ωt +  = −m0 a 0 sin (ωt ) cos(ωt )
 2
2π
ω
ω
F90°− Phase =
2π ∫F
0
90° − Phase
dt = 0

The phase is therefore very important. This basic concept shows that a net time-averaged thrust

is possible without using propellants.

A Simple Mach-Effect Mass Fluctuation Model

The core of the Mach effect thruster are the mass fluctuations. Several different derivations

have been proposed by Woodward [1],[5] and most recently by Fearn et al [12],[15]. Here we

will summarize a simple approach by Tajmar [16] which gives in similar equations as obtained

by Woodward [1] using some main assumptions which are necessary for our thruster model.

One frequently derives analytical solutions from general relativity by applying the so-called

weak-field approximation. The main assumptions here are a flat background (a good

assumption for our neighborhood) and stationary solutions which leads to Newton’s

gravitational force law. To consider mass fluctuations, the last assumption must be dropped,

allowing the following time-varying solution,

v 1 ∂ φg
2
 1 ∂ 2φg 
∇g = −∆φg = −4πGρ0 − 2 2 = −4πG ρ0 + 
2 
(3)
c ∂t  4πGc2
∂ t 

where g is the gravitational field, φg the gravitational potential and ρ0 the stationary mass

density. Note the additional term next to the density that is time varying. One may express the

delta density as

8
 1 ∂ 2φ g φg ∂ 2 m0 φg ∂2ρ0
δρ 0 = = = (4)
4πGc 2 ∂t 2 4πGc 2 m 0 ∂t 2 4πGc 2 ρ 0 ∂t 2

where we used the simple gravitational scalar potential of the point mass m0, φg=-G.m0/r, to

arrive at the Woodward mass formula [1]. Woodward then applied Sciama’s inertia model [6],

where the effect of the surrounding mass of the universe follows φg/c²=-1. Contrary to

Woodward, a negative sign is used as in Sciama’s paper because a gravitational potential is

always negative. The energy density is taken to be ε=ρ0⋅c² to get

φg ∂2 ρ0 1 ∂ 2ε
δρ 0 = = − (5)
4πGc 2 ρ 0 ∂t 2 4πGc 2 ρ 0 ∂t 2

It’s important to realize that this term leads to much larger mass changes than expected from

the classical E=m⋅c². The amplification factor may be expressed as

δρ 0 1 ∂ε
=−
1 ∂ε 4πGρ 0 ∂t (6)
c 2 ∂t

For frequencies as used by Woodward in the kHz range and typical metal or piezo mass

densities, the mass fluctuation can be some 11 orders of magnitude higher compared to classical

expectations. However, in contrast to the classical energy-mass balance, no steady mass

changes but only fluctuations are possible with this amplification.

In contrast to prior Mach thruster analysis, the assertion in this analysis is that only the

mechanical (inertial) energy contributions to the Mach fluctuations, whereas the prior

interpretations focused on the electrical energy in the capacitors (or coils). This makes sense as

Sciama’s model describes inertial and hence only inertial (=mechanical) energy. In some

previous experiments, mechanical oscillation was replaced by ion/lattice movements that were

thought to be much more efficient because they can oscillate at much higher frequencies.

However, although early papers reported thrusts up to the mN range, no net thrusts were seen

when proper electrical shielding and setups were used in subsequent measurements [1],[17]. As

a result, it was thought the bulk acceleration is necessary for the effect to occur [1], however as
9
we will see, it is not only bulk acceleration but pure mechanical energy that is responsible for

the correct thrust values observed.

By integrating over the volume, the mass fluctuation term, as used by Woodward, is found but

with a negative sign.

1 ∂P
∆m0 = − (7)
4πGc ρ 0 ∂t
2

where dP/dt is now the time-derivate of the mechanical power produced by the piezo stack. The

negative sign is important to get the observed thrust direction. The new analysis offered here

uses this as a starting point.

Analytical Piezo Model using Clamping

We are going to develop a one-dimensional model along the thrust/electric field axis. First, we

will concentrate on one single PZT element. If we apply an AC voltage to the piezo with a

frequency f=2πω, we will get the following varying voltage, current and power given by

V(t)= V0 sin(ωt), I(t)= I0 cos(ωt), P(t)= V0 I0 sin(ωt)cos(ωt) (8)

where we assumed a pure capacitive load and therefore a 90° phase shift between voltage and

current which is very reasonable using a typical configuration. As shown before, mass

fluctuations depend on the time derivative of the power fed into the piezo stack. Using

trigonometry, we get

∂P(t)
=V0I0 cos( 2ωt) (9)
∂t

This shows that mass fluctuations will then have double the frequency and a phase shift of 90°

compared to the applied voltage signal. We therefore need a similar signal that accelerates the

stack to get a non-zero force such that mass fluctuation and actuator are perfectly in phase with

respect to each other as shown above.

10
Fortunately, the piezo material itself can act both as capacitor due to the high relative dielectric

permittivity, and as an actuator. Even more, as an actuator we see the superposition of two

effects that together produce an acceleration that partly has the correct waveform (double the

driving frequency with a 90° phase shift):

1. Piezo-Effect: This well-known effect produces a change of dimensions of the PZT material

that is proportional to the applied electric field strength. Different axes have different

material constants.

2. Electrostriction: This is a general property of all dielectrics caused by a slight displacement

of ions in the crystal lattice upon exposure of an electric field. It scales with the square of

the applied electric field strength and is much weaker. Also, different axes feature different

constants. Electrostriction is usually much weaker compared to the piezo effect in typical

PZTs.

We can now express the time-varying change in length of an unclamped PZT using the applied

electric field as

 V (t ) V (t ) 
2

ΔlPZT(t) = l PZT  d 33 +M33 2  (10)

 l PZT l PZT 

where lPZT is the stationary length, and d33 and M33 are the piezo and electrostrictive constants

respectively along the electric field direction for a disc shaped piezo crystal. Both effects always

appear simultaneously. It should be noted that all piezo constants are dependent on the electric

field strength, temperature etc. [18]. This model assumes that the piezo constants are static – a

reasonable first order assumption. In addition, the pre-stress on the PZT, due to clamping,

influences piezo characteristics. However, a typical torque on the screws of 4 in-lbf results in a

pre-stress of 9 MPa on the PZT disc which is still small enough to consider static values [19].

Higher pre-stress increases the d33 piezo constant up to a critical peak, which may be used to

increase piezo movement in the future.

11
The actual movement of the piezo however depends on the clamping stiffness as written above

and is expressed along with the velocity and acceleration as

Δl PZT (t)ηClamp
Δx(t) =
2
d
v(t ) = ∆x(t )
dt
V0ω cos(ωt )η Clamp [d 33l PZT + 2M 33V0 sin(ωt )]
= (11)
2l PZT
d2
a(t ) = ∆x(t )
dt 2

=
{
V0ω 2η Clamp − sin (ωt )[d 33 l PZT + 2M 33V0 sin (ωt )] + 2V0 cos 2 (ωt )M 33 }
2l PZT

where the factor of ½ is necessary such that ∆x characterizes the amplitude of the oscillation.

As shown in Fig. 1, the thruster is mounted on the side of the rubber pad/brass disc. Therefore,

when the piezos expand, the center of mass will shift to the right towards the small end cap

side. We therefore define the positive x direction in the direction from the fixed towards the

free side.

The clamping efficiency acting on a single PZT can be calculated [20] from the stiffness of the

PZT itself and the stiffness of the clamp as

k PZT
ηClamp = (12)
k PZT + kClamp

Looking at Fig. 1, we can calculate the clamping stiffness from all contributing elements

defining the clamp (other PZTs, screw and aluminum end cap) by properly taking into account

elements that add up in parallel (screws) or series (everything else) like springs as

12
 1
k Clamp =
1 1 1  1 1 1 
+ + + (N − 1) ⋅  + +
N Screw k Screw k Alu k Brass
PZT  k Electrode k Epoxy k PZT 
 
2
YScrew d Screw π Y Alu d Cap π
2
YBrass d Cap π
2
(13)
k Screw = , k Alu = , k Brass =
4(l PZT N PZT + l Alu ) 4l Alu 2l Brass
2
YElectrode, Epoxy , PZT d PZT π 2
Y33 d PZT π
k Electrode , Epoxy , PZT = , k PZT =
4l Electrode, Epoxy , PZT 4l PZT

where Y is the young modulus and other elements are determined by the geometry (Y33 is again

for the PZT in the longitudinal direction). The brass mass is assumed to contribute with half of

its length to the overall stiffness due to the clamping screw as shown in Fig. 1.

Calculation of Thruster Force

Now we are ready to calculate the force of the Mach Effect Thruster. Woodward and coworkers

[1],[12] started off with Equ. (7) and a simple mechanical approach. Using P=F⋅v and Newton’s

second law, they approximated the mass fluctuation (with a positive sign) as

1
∆m0 ≈ m0 a 2 , (14)
4πGc ρ0
2

but without taking any spring/stiffness into account. In general, piezo actuators are usually

simulated by using a quasi-static model (low frequency regime, typically well below any

resonance) and a dynamic model (at resonance frequencies). It’s possible to use quasi-static

models also at higher frequencies, however, they do not take the inertia of attached masses like

the large brass mass into account. Moreover, the Young modulus is increasing at higher

frequencies which causes the quasi-static model to overpredict the actual force [21]. Our

calculations will show that these two different models have a different frequency dependence

(ω4 and ω6), however for the geometry and electrical values used, both have similar amplitudes.

The models will deviate for higher frequencies or larger attached masses. Typical thrust

signatures from recent tests are shown in Fig. 3, where a switch on/off transients as well as a

13
steady thrust component is seen. The origin of the transient thrust is not fully clarified up to

now (a first idea is discussed in [22],[23]).

Quasi-Static Model

For the quasi-static case, we will pursue with two approaches that accurately tackle the problem

by either using the real mechanical force generated by our piezo actuator stack or the energy

content stored inside the piezo-capacitor. Both will eventually lead to a similar result.

Quasi-Static Thruster Model using Piezo Force

The mechanical force [20] and power generated by a piezo actuator is given by

FPZT (t ) = k PZT ∆l PZT (t )(1 − η Clamp )

(15)
PPZT (t ) = FPZT (t )v(t )

The thrust generated can now be calculated by inserting the mass fluctuations term and the

mechanical power generated by all PZTs in the stack given by NPZT as

N PZT ∂PPZT (t )
FThrust,PZT (t ) = ∆mPZT a(t ) = − a(t ) (16)
4πGc 2 ρ PZT ∂t

We can average this force over one complete cycle to get a time-average given by
2π
ω
ω
FThrust , PZT = ∫F (t ) dt
2π
Thrust , PZT
0

=
N PZT V04ω 4 k PZT M 33η Clamp
2
(
(η Clamp − 1) 11d 332 l PZT
2
+ 8V02 M 332 ) (17)
128πGc ρ2
l3
PZT PZT

≈
11N PZT V04ω 4 k PZT M 33 d 332 η Clamp
2
(η Clamp − 1)
128πGc 2 ρ PZT l PZT

Here, we dropped the second term as it is several orders of magnitude smaller and only gets

comparable to the first term at a field strength of 25⋅106 V/m, which is well above our values.

We can see the experimentally observed V4 dependence. All relevant material and thruster data

are summarized in Tables 1 and 2 which may be used to calculate thrust. The recent thrusters

use SM111 piezo discs from Steiner&Martins, which is a hard-PZT material with low dielectric

14
losses and a high Curie temperature of 320°C. In addition to the usual PZTs in the stack,

Woodward usually integrates thin piezo discs used as accelerometer, which we did not model

here (but that can be easily implemented following our approach, it modifies the results

presented here by less than one percent). Using typical values such as a voltage of V0=200 V

and a frequency of f=36.3 kHz, we get a net-time average thrust of FThrust,PZT=-3.4 µN. Note

that the thrust value is negative and therefore points from the PZTs towards the brass mass

along the negative x-direction (see Fig. 1).

Quasi-Static Thruster Model using Energy stored in Capacitor

Another very intuitive approach is to start from the energy that is stored in the piezo-capacitor

due to the high relative permittivity of the PZT material. The electric power here is simply

Pe (t) = V(t)I (t ) = V0 sin(ωt)I 0 cos(ωt ) (18)

The corresponding average force is calculated accordingly as

N PZT ∂PEnergy (t )
FThrust , Energy (t ) = ∆m Energy a (t ) = − a
4πGc ρ PZT
2
∂t
2π
ω
ω
FThrust , Energy = ∫F (t ) dt (19)
2π
Thrust , Energy
0

N PZT V03 I 0ω 3 M 33η Clamp

=−
8πGc 2 ρ PZT l PZT

Here, the force scales linearly with the current and therefore the electric power supplied to the

thruster. At first sight, this does not look like our approach using the PZT force in Equ. (17).

Recall, however, that key assumption in these analysis is that the mechanical, not electrical

power pertains to the Mach effect. For our capacitive load, the current depends on the capacity

of our electric circuit Ce. In addition, we need to multiply that with a correction factor that

separates the mechanical from the electric part. We can summarize that by the following

equations valid for a single PZT element as

15
 2
d PZT π
C e = ε 0 ε r 33
4l PZT
C = C eη Clamp (1 − η Clamp )k p2
(20)

I 0 = V0 ω C

Wolff et al [24] studied the mechanical-electric energy split for multilayer PZT actuators and

also identified the (1-ηClamp) correction factor for the capacity. From our PZT model before, we

know that this was to correct the force generated by the actuator. Another ηClamp is necessary to

correct for the position of the stack. The last factor kp is the so-called electromechanical

coupling factor in the radial direction of thin piezo discs. Such discs have different resonance

frequencies as we will see later. The first resonance, followed by several harmonics, is causing

oscillations in the radial direction only, therefore the kp coupling factor must be used [25]. The

next resonance causes longitudinal oscillations along the thickness of the disc, which are

characterized by the kt factor which is much smaller. In this case, even a reduced (clamped)

electric constant must be used that can further reduce the thrust effect. These factors cause a

degradation of the effect at higher frequencies that can also explain why a ω2- ω3 dependence

of the effect is observed so far instead of higher powers [12]. For different geometries, instead

of thin discs (larger rods, rectangular shapes, etc.), different electromechanical coupling factors

need to be applied. The radial expansion works well for the mass fluctuation part because as

the resulting mass change is in any case isotropic, however the push-pull must occur in the

longitudinal direction which would need the kt factor in that case. We therefore do not expect a

perfect match when comparing to the PZT model.

We can now re-write Equ. (19) as

N PZT V 04ω 4 M 33 ε 0 ε r 33 d PZT

2
k p2η Clamp
2
(η Clamp − 1)
FThrust , Energy = (21)
32 Gc 2 ρ PZT l PZT
2

which has again the same voltage and angular frequency dependence as our PZT force model.

In fact, both equations are very similar as the electromechanical coupling factor depends on the

piezo constants, Young modulus and electric permittivity. Using again the same conditions, the
16
energy thrust model predicts a net-time averaged force of -2.8 µN, which is close to the PZT

model force.

The small difference between both models is due to the different approach. For example, due

to the structure of the equations in the PZT models, a small part of the piezoelectric acceleration

contributes to a net-averaged thrust too whereas it does not for the energy model. Additionally,

the energy model here predicts large radial oscillations for our transient mass fluctuations due

to the disc geometry whereas the PZT model has no geometry correction factor. Both

approaches have their strengths and future test into higher frequencies or PZT geometry stacks

will tell what is better.

Fig. 4 shows the output of the models for this case and Fig. 5 compares both models to published

measurements which shows excellent agreement (using the correct geometry, material

parameters, driving frequencies and voltages). This demonstrates once again that it is a pure

inertial effect – only mechanical work causes transient mass fluctuations that can be used to

produce a net time-average force. The experimental data in Fig. 4 is based on measurements by

Woodward who used only switch-on/off transient thrusts which is an overprediction of the

actual expected steady-state thrust. Nevertheless, the comparison confirms the right order of

magnitude and correct scaling with voltage.

We see that the delta mass signal has double the frequency with a 90° phase shift as expected.

The amplitude is close to 0.4 mg which is a huge value that could be measured with standard

analytical balances – although not at this high frequency. This shows that transient mass

fluctuations are not small such as the usual effects in general relativity like frame-dragging that

needs dedicated space missions to detect them. The acceleration signal is dominated by the

piezoelectric effect which has the same frequency and a 180° phase shift with respect to the

applied signal. Therefore, Fig. 4 only shows the much smaller electrostrictive signal which is

responsible for most of the net time-average thrust value because it has the right double-

frequency and 90° phase shift (in the piezo force model, also the piezoelectric acceleration
17
contributes slightly to thrust, in the energy model only the electrostrictive component

contributes). We see that the thrust oscillates up to -5.5 µN which is significant and similar in

magnitude to existing micropropulsion thruster such as FEEP or µPPTs.

Of course, there remain a lot of improvement areas. For example, the need for mass fluctuations

and acceleration may be split and driven by two different frequency generators to create a much

better result (in the past, Woodward was using a mix of the standard frequency together with a

double frequency and a proper phase shift – some optimization could be expected in that case).

We can also calculate the total electric power going to the thruster and the power loss which is

important for the thruster as this can raise the temperature and cause significant degradation of

the material’s performance. Neglecting Ohmic losses or inductive/capacitive loads from the

power supply, we get the effective electric power for the whole PZT stack as

V02ωC e
Peff = N PZT
2
(22)
 1 
Ploss = N PZT V ωC e  tan δ +
2

Qm 
0


where tan δ is the dissipation factor (that is low for hard PZT material) and Qm is the mechanical

quality factor of the stack. Although Qm can be high for individual PZT discs, it is quite low for

a stack with epoxy and electrode material in between. The values are determined by spectrum

analysis and are typically around 60. Again, using our example, we get an effective power of

63 W and a power loss of 2.6 W and a total capacity of 14 nF.

18
Dynamic Model

For high frequency applications, the mass of the vibrating system due to its inertia can generate

a large oscillating force. This dynamic force [20] is given by

F dyn ( t ) = m eff a ( t ) (23)

where the effective mass consists of spring-type masses (PZTs) that expand and contract as well

as external fixed masses as expressed by

m spring
m eff = + m external (24)
3

In our geometry, the spring-type PZT masses will accelerate to the right (positive x-direction),

whereas the brass mass will be pushed into the opposite direction to the left (negative x-

direction) due to Newton’s 3rd law. At resonance, we assume that the acceleration of the brass

mass has the same amplitude as the one from the PZTs. In reality, this depends on the stiffness

of the mounting of the brass mass to the holding structure (screws and rubber) as well as on the

stiffness of the support structure. As we assume that the rubber pad mounting is effectively

isolating the brass mass vibration from the mounting structure, our assumption of similar

acceleration amplitudes of brass mass and PZTs is justified. However, future numerical

simulations and measurements will be necessary for better modelling. The brass mass therefore

enters our equation with a negative sign. We will neglect the aluminum cap, mounting structure

and screws as the have nearly equal positive (PZT side) and negative (brass side) contributions

that cancel out.

We can now express the dynamical force and its associated power by

N m 
FDyn (t ) =  PZT PZT − m Brass  a (t )
 3  (25)
PDyn (t ) = Fdyn (t ) v (t )

The corresponding average force is calculated as

19
 N PZT m PZT + m Brass
ρ avg =
N PZT m PZT m Brass
+
ρ PZT ρ Brass
1 ∂PDyn (t )
FThrust , Dyn (t ) = ∆m Dyn a (t ) = − a
4πGc ρ avg
2
∂t
2π
(26)
ω
ω
FThrust , Dyn = ∫F (t ) dt
2π
Thrust , Dyn
0

V ω 6 M 33 d 332 η Clamp
0
4 3
[N PZT m PZT + 3m Brass ]
=−
128πGc 2 ρ avg l PZT

Here we introduced the average density ρavg for the variable dynamic mass ∆mDyn, since it

contains contributions from all vibrating parts. This is however a small correction because the

density of the piezos and brass are very similar. The dynamic force has the same sign as the

quasi-static models (negative x-direction) which is also experimentally observed (personal

communication with J. Woodward and N. Buldrini) and contains the contribution from the

piezo stack as well as the attached brass mass. We also see that the mass of the piezo and

especially the end caps have a significant influence on thrust as well. For example, a larger

brass mass will increase the thrust, however, long/heavier brass caps will also reduce the

clamping efficiency ηClamp. Therefore, an optimum brass mass exists for each configuration

which was indeed found experimentally and in Rodal’s model as well [14]. In case brass masses

are used as end caps on both ends and the mounting structure is aligned with the middle of the

piezo stack instead of its position at the end, then there would be equal dynamic forces on both

sides resulting in zero thrust. This has been demonstrated in a recent experiment [26].

The predicted dynamic thrusts seem more realistic as shown in the comparison of Fig. 5. Fig. 6

shows a comparison with the best steady thrust measurements so far, which is in good

agreement. In general, the dynamic model should be preferred, especially if it is used to

extrapolate the thrust for larger and more massive configurations. However, the quasi-static

models show interesting insights how the energy is used to perform mechanical work which is

the basis of the generated effect.

20
Modeling of Resonances

So far, we have not taken resonances into account. This is for sure a very important aspect

which is however beyond the scope of our analytical model. We expect damping of our force

outside resonance frequencies that scale with the usual ω²/( ω0²- ω²) dependence. We can

analytically estimate the resonance frequencies of our piezo actuator using the following

approach. By calculating the masses of the individual PZT discs, end caps and the screws, we

can find the resonance frequencies using a multi mass-spring model [27] as

1 k PZT  iπ 
f 0i = ⋅ sin   , i = 1 ... N PZT (27)
π N PZT ⋅ m PZT + m Brass + m Alu + N Screw ⋅ m Screw  2 ( N PZT + 1) 

Using again our model example, we get seven resonance frequencies as f0= 14.7 kHz, 29 kHz,

42.4 kHz, 54.5 kHz, 65 kHz ,73.5 kHz, 79.7 kHz and 83.6 kHz which is again in good agreement

with experimental results as shown from a recent impedance spectrum recorded with an SR 780

dynamic signal analyzer for a similar device (supplied by H. Fearn). A comparison of a

spectrum analysis and our resonance frequencies are shown in Fig. 7. Both resonance and

thruster models should be used together as new geometries will also affect resonance

frequencies that must be considered when calculating thrusts.

Conclusions

We have developed a 1D analytical model that can accurately predict the thrust from Mach-

Effect thrusters taking design and materials parameters into account. It compares well to

experimental data and allows for further optimization to obtain higher thrusts and efficiencies.

Apart from the well-known voltage and frequency scaling, it predicts higher thrusts e.g. for

larger disc diameters and higher stiffness. For example, if the PZT discs are increased to a

diameter of 25 mm, the 2nd resonance frequency should rise to 51 kHz. Both should lead to an

increase in thrust to 12 µN at an amplitude of 200 V. Of course, there are several shortcomings

and simplifications that may be corrected in future iterations such as implementing resonances

21
into thrust model, use of electric field dependent piezo material parameters, include temperature

degradation effects and adding the influence of clamping torque from screws.

The model is flexible enough to be modified for different geometries (e.g. piezo rings instead

of discs with one single screw in the middle). One of the main conclusions of this analysis is

that the thrust is only accurately calculated if only the mechanical power is used in the transient

mass equation. This can explain why some previous designs (Mach-Lorentz thrusters) did not

work as expected. It is hoped that the model and its fit to experimental results adds further

confidence into Mach effect thrusters and stimulates further research in that area.

Acknowledgements

Part of this work was funded by DLR (50RS1704) which is greatly appreciated. I would like to

thank especially J.F. Woodward and H. Fearn for their great support and many helpful insights.

Lots of discussions most notably with J. Rodal, M. Kössling, M. Weikert, W. Stark and M.

Monette were also greatly appreciated. I would also like to acknowledge the comments from

one referee that greatly improved the clarity of the paper.

22
References

[1] Woodward, J. F., Making Starships and Stargates, New York, NY: Springer New York,

2013. doi:10.1007/978-1-4614-5623-0

[2] Woodward, J. F., “A New Experimental Approach to Mach’s Principle and Relativistic

Graviation,” Foundations of Physics Letters, vol. 3, Oct. 1990, pp. 497–506.

doi:10.1007/BF00665932

[3] Woodward, J. F., and Vandeventer, P., “Mach’s Principle, Flux Capacitors, and

Propulsion,” AIP Conference Proceedings, AIP, 2006, pp. 1341–1351.

doi:10.1063/1.2169319

[4] Woodward, J. F., “‘Breakthrough Propulsion’ and the Foundations of Physics,”

Foundations of Physics Letters, vol. 16, 2003, pp. 25–40.

doi:10.1023/A:1024198022814

[5] Woodward, J. F., “Flux Capacitors and the Origin of Inertia,” Foundations of Physics,

vol. 34, 2004, pp. 1475–1514. doi:10.1023/B:FOOP.0000044102.03268.46

[6] Sciama, D. W., “On the Origin of Inertia,” Monthly Notices of the Royal Astronomical

Society, vol. 113, Feb. 1953, pp. 34–42. doi:10.1093/mnras/113.1.34

[7] Barbour, J., and Pfister, H., eds., Mach’s Principle: From Newton’s Bucket to Quantum

Gravity, Boston: Birkhäuser, 1995.

[8] Ciufolini, I., and Pavlis, E. C., “A Confirmation of the General Relativistic Prediction of

the Lense-Thirring Effect,” Nature, vol. 431, 2004, pp. 958–960.

doi:10.1038/nature03007

[9] Everitt, C. W. F., Debra, D. B., Parkinson, B. W., Turneaure, J. P., Conklin, J. W.,

Heifetz, M. I., Keiser, G. M., Silbergleit, A. S., Holmes, T., Kolodziejczak, J., Al-

Meshari, M., Mester, J. C., Muhlfelder, B., Solomonik, V. G., Stahl, K., Worden, P. W.,

Bencze, W., Buchman, S., Clarke, B., Al-Jadaan, A., Al-Jibreen, H., Li, J., Lipa, J. A.,

23
 Lockhart, J. M., Al-Suwaidan, B., Taber, M., and Wang, S., “Gravity Probe B: Final

Results of a Space Experiment to Test General Relativity,” Physical Review Letters, vol.

106, 2011, pp. 1–5. doi:10.1103/PhysRevLett.106.221101

[10] Buldrini, N., “Verification of the Thrust Signature of a Mach Effect Device,”

Proceedings of the Estes Park Advanced Propulsion Workshop, H. Fearn and L.

Williams, eds., Moja: Space Studies Institute Press, 2017, pp. 83–88.

[11] Wanser, K. H., “Center of Mass Acceleration of an Isolated System of Two Particles

with Time Variable Masses Interacting with Each Other via Newton’s Third Law Internal

Forces: Mach Effect Thrust,” Journal of Space Exploration, vol. 2, 2013, pp. 122–130.

[12] Fearn, H., Rossum, N. van, Wanser, K., and Woodward, J. F., “Theory of a Mach Effect

Thruster II,” Journal of Modern Physics, vol. 6, 2015, pp. 1868–1880.

doi:10.4236/jmp.2015.613192

[13] Fearn, H., and Woodward, J., “Recent Results of an Investigation of Mach Effect

Thrusters,” 48th AIAA/ASME/SAE/ASEE Joint Propulsion Conference and Exhibit,

Reston, Virigina: American Institute of Aeronautics and Astronautics, 2012, p. AIAA

2012-3861. doi:10.2514/6.2012-3861

[14] Rodal, J. J. A., “Mach Effect Propulsion, an Exact Electroelasticity Solution,”

Proceedings of the Estes Park Advanced Propulsion Workshop, H. Fearn and L.L.

Williams, eds., Mojave, CA: Space Studies Institute Press, 2017, pp. 127–136.

[15] Fearn, H., Zachar, A., Wanser, K., and Woodward, J., “Theory of a Mach Effect Thruster

I,” Journal of Modern Physics, vol. 6, 2015, pp. 1510–1525.

doi:10.4236/jmp.2015.611155

[16] Tajmar, M., “Revolutionary Propulsion Research at TU Dresden,” Proceedings of the

Estes Park Advanced Propulsion Workshop, H. Fearn and L.L. Williams, eds., Mojave,

CA: Space Studies Institute Press, 2017, pp. 67–82.

[17] Buldrini, N., Tajmar, M., Marhold, K., and Seifert, B., “Experimental Results of the
24
 Woodward Effect on a µN Thrust Balance,” AIAA Paper 2006-4911, 2006.

doi:10.2514/6.2006-4911

[18] Sukesha, Vig, R., and Kumar, N., “Effect of Electric Field and Temperature on Dielectric

Constant and Piezoelectric Coefficient of Piezoelectric Materials: A Review,” Integrated

Ferroelectrics, vol. 167, Nov. 2015, pp. 154–175. doi:10.1080/10584587.2015.1107383

[19] Zhou, D., Kamlah, M., and Munz, D., “Effects of Uniaxial Prestress on the Ferroelectric

Hysteretic Response of Soft PZT,” Journal of the European Ceramic Society, vol. 25,

Apr. 2005, pp. 425–432. doi:10.1016/j.jeurceramsoc.2004.01.016

[20] Piezoelectric Actuators: Components, Technologies, Operation (CAT128E R3),

Lederhose: Physik Instruments (PI) GmbH & Co. KG, 2016.

[21] Zhang, X., Zhang, X., Sessler, G. M., and Gong, X., “Quasi-Static and Dynamic

Piezoelectric Responses of Layered Polytetrafluoroethylene Ferroelectrets,” Journal of

Physics D: Applied Physics, vol. 47, 2014, p. 15501. doi:10.1088/0022-

3727/47/1/015501

[22] Fearn, H., and Wanser, K., “Experimental Tests of the Mach Effect Thruster,” Journal

of Space Exploration, vol. 3, 2014, pp. 197–205.

[23] Woodward, J. F., “Musings on Mach Effects,” Proceedings of the Estes Park Advanced

Propulsion Workshop, H. Fearn and L.L. Williams, eds., Mojave, CA: Space Studies

Institute Press, 2017, pp. 147–162.

[24] Wolff, A., Cramer, D., Heliebrand, H., Schuh, C., Steinkopff, T., and Lubitz, K., “Energy

Considerations of PZT Multilayer Actuators under Dynamic Driving Conditions,” ISAF

’96. Proceedings of the Tenth IEEE International Symposium on Applications of

Ferroelectrics, IEEE, 1996, pp. 317–320. doi:10.1109/ISAF.1996.602758

[25] “An Overview of the Properties of Different Piezoceramic Materials Piezo

Technologies” Available: http://www.piezotechnologies.com/knowledge-

desk/overview-piezo-materials
25
[26] Fearn, H., and Woodward, J. F., “Experimental Null Test of a Mach Effect Thruster,”

arXiv, Jan. 2013

[27] Morin, D., “Normal Modes” Available:

http://www.people.fas.harvard.edu/~djmorin/waves/normalmodes.pdf

[28] “Steiner & Martins, Inc.” Available: https://www.steminc.com/

26
 Brass
End-Cap PZT Electrode (~)
Rubber Pad Clamping
Screw
Aluminum ... Aluminum
End-Cap End-Cap

Thrust x
Epoxy Electrode (Ground)
Support / Thrust Balance
(Ground)

Fig. 1 Schematic Sketch of a Mach Effect Thruster

27
Fig. 2 Typical Mach-Effect Thruster Model (provided by J.F. Woodward)

28
 0.4 5 1.2

4
0.3 0.9
Switch-Off
3
Transient
0.2 0.6
2

Thrust [arb]
Thrust [µN]
0.1 0.3 Steady
1
Thrust [µN]

Thrust
0.0 0 0.0

-1 Switch-On
-0.1 -0.3
Transient
-2
-0.2 -0.6
-3
-0.3 -0.9
-4

-0.4 -5 -1.2
0 3 6 9 12 15 18 21 24 0 4 8 12 16 20 24 28 32 0 1 2 3 4 5 6 7 8
Time [s] Time [s] Time [arb]

a.) Buldrini Measurement b.) Woodward Measurment c.) Thrust Components

(with 200 V p-p) (with 200 V Amplitude)

Fig. 3 Typical Thrust Signature of a similar Mach-Effect Thruster at 36.3 kHz showing On/Off

Transients and a Constant Thrust [10]

29
 0

FThrust [µN] -1
-2
Force

-3
-4
-5
-6
14 * Only Electrostrictive Part
Acceleration
a [m/s ]

7
2

0
*

-7
-14
Mass Fluctuation

0.50
∆mPZT [mg]

0.25

0.00

-0.25

-0.50
200
Applied Voltage

100
V [V]

-100

-200

0 5 10 15 20 25 30
Time [µs]
Fig. 4 Quasi-Static Mach Effect Thruster Energy-Model Outputs for V0=200 V and f=36.3
kHz (Note: Electrostrictive acceleration is shown here because only electrostrictive acceleration
has the right phase to produce a net-time averaged thrust. The piezoelectric acceleration is much
larger and would therefore mask electrostriction in the illustration. This figure intends to show
that the mass fluctuation is pushed- and pulled by electrostriction with the right frequency and
phase. The model thrust equations, however, do not separate between the two effects.)

30
 0

-1

-2
Thrust [µN]

-3

-4

Measured [Fearn et al, 2015]

-5 PZT Model
Energy Model
Dynamic Model
-6
0 50 100 150 200 250

Voltage [V]

Fig. 5 Comparison of Quasi-Static Thruster Model to Measurements [12],[23] (Note: The

measured thrust values were taken by measuring the difference between the switch on-off

transient thrust values – therefore the measured thrust values where divided by a factor of two

to get the single thrust value from the baseline to the peak observed.)

31
 0.0

-0.5

-1.0
Thrust [µN]

-1.5

-2.0

-2.5
Measured [Buldrini et al, 2017]
Dynamic Model
-3.0
0 50 100 150 200 250

Voltage [V]

Fig. 6 Comparison of Combined Quasi-Static and Dynamic Thruster Model to Measurements

[10]

32
 -65
4.3 kHz
-70 92.5 kHz

-75 23.7 kHz

39.2 kHz
-80
Impedance [dB]

58.8 kHz
-85 78.4 kHz

-90

-95 63 kHz
70.8 kHz
-100

-105 81.6 kHz

-110
0 10 20 30 40 50 60 70 80 90 100
Frequency [kHz]

Fig. 7 Resonance Frequencies Measurement obtained with SR780 Spectrum Analyzer (data

supplied by H. Fearn) using Geometry and Material from Tables 1 and 2

33
Table 1 Basic Dimensions and Materials Parameter of Mach-Effect Thruster (some parameters

taken from [14])

Parameter Value

PZT

Number of PZTs in Stack (NPZT) 8

Length of PZT Disc 2 mm

Diameter of PZT 19 mm

Density of PZT (ρPZT) 7900 kg/m³

Young Modulus (Y33) 73 GPa

Screws (Stainless Steel)

Number of Screws (NScrew) 6

Diameter of Screw (dScrew) 2.845 mm

Density of Screw (ρScrew) 7850 kg/m³

Young Modulus (YScrew) 190 GPa

End Cap (Aluminum)

Diameter of Cap (dCap) 28.9 mm

Length of End Cap (lAlu) 4.3 mm

Density of End Cap (ρAlu) 2700 kg/m³

Young Modulus (YAlu) 69 GPa

End Cap (Brass)

Length of End Cap (lBrass) 16 mm

34
 Density of End Cap (ρBrass) 8700 kg/m³

Young Modulus (YBrass) 100 GPa

Electrode Thickness (Brass, lElectrode) 50 µm

Epoxy

Thickness (lExpoy) 5 µm

Young Modulus (YEpoxy) 20 GPa

Table 2 SM-111 Piezo Material Parameters ([14],[28])

Parameter Value

Piezoelectric Constant (d33) 320×10-12 m/V

Electrostrictive Constant (M33) 1.35×10-17 m²/V²

Electromechanical Coupling Coefficient (kp) 0.58

Dissipation Factor @ 1 kHz (tan δ) 0.004

Relative Electric Permittivity (εr33) 1400

35

View publication stats