A MATHEMATICAL MODEL OF A SIMPLE AMPLIFIER

USING A FERROELECTRIC TRANSISTOR

Rana Sayyah1, Mitchell Hunt1, Todd C. MacLeod2, and Fat D. Ho1
1

The University of Alabama in Huntsville, Department of Electrical and Computer

Engineering, Huntsville, Alabama, 35899, USA
2
National Aeronautics and Space Administration, Marshall Space Flight Center,
Huntsville, Alabama, 35812, USA

ABSTRACT
This paper presents a mathematical model characterizing the behavior of a simple
amplifier using a FeFET. The model is based on empirical data and incorporates several
variables that affect the output, including frequency, load resistance, and gate-to-source
voltage. Since the amplifier is the basis of many circuit configurations, a mathematical
model that describes the behavior of a FeFET-based amplifier will help in the integration
of FeFETs into many other circuits.
Keywords: FeFET; FFET; ferroelectric transistor; analog amplifier; model

INTRODUCTION
The use of ferroelectric field-effect transistors (FeFETs) to create simple amplifiers has
not been extensively studied. Moreover, the FeFETs unique characteristics of hysteresis
and nonlinearity result in an amplifier that displays properties that are different from
those of a MOSFET-based amplifier [1]. The FeFET amplifiers behavior has never been
modeled. Thus, a mathematical-based model that describes a FeFET-based amplifier was
created. The model is based on empirical data and incorporates several variables that
affect the output, including frequency, load resistance, and gate-to-source voltage. The
amplifier is the basis of many circuit configurations, and creating an easy-to-use model

that illustrates the behavior of the FeFET amplifier will greatly aid in the integration of
FeFETs into many other circuits.

FeFET AMPLIFIER CIRCUIT CONFIGURATION

A FeFET simple amplifier is built using a FeFET and a resistor. Figure 1 illustrates the
FeFET amplifier circuit configuration. As can be seen from the figure, the resistor, R, is
connected to the source of the FeFET. The input voltage, Vin, is provided from the gate
of the transistor to ground. Thus, Vin includes both VGS and the voltage across the
resistor. VDD is the voltage from the drain of the FeFET to ground, thereby including
both VDS and the voltage across the resistor. The output voltage, Vout, is measured at the
source of the transistor and above the resistor.

VVGS
GS
in

VDS
VDD
DS

Vout
R

Figure 1 FeFET amplifier circuit configuration

DESCRIPTION OF THE MODEL

The model was created in Microsoft Excel and is based on the equations [2] for the drain
current, ID, that are derived from the Fermi-Dirac equation. When the gate voltage is on,

.
(1)
When the gate voltage is off,

.
(2)
In the two previous equations, ID is the drain current, IDSAT is the drain current when the
FeFET is in saturation, B is the drain current decay coefficient, Vgs is the gate-to-source
voltage VGS, Vp is a polarization voltage, which is the gate voltage at which half the
saturation current is reached, k is a constant defining the rate of change of the function,
and t1 is the elapsed time in seconds since the last poling. More needs to be said about
the polarization voltage, Vp. Each FeFET has two different Vps, one is for the
negatively-poled drain current, while the other is for the positively-poled current. This
provides the hysteresis characteristic of the FeFETs I-V relationship.
The model allows the user to enter the desired value for several of the parameters
that affect the output. Thus, the user can define values for VDD, load resistance (Rload),
input frequency, amplitude of Vin (Vin,amp), offset of Vin (Vin,offset), phase shift, and
frequency coefficient. Vin,offset was included in order for the model to be able to account
for offsets in the input voltage. The phase shifting parameter phase is not the precise
angular phase shift of the signal; rather it is a tunable parameter that is used to
figuratively represent the amount of phase shift desired. Therefore, it is not a measurable
parameter; it is an adjustable, user-specified parameter. The frequency coefficient is used

to regulate the effect of frequency. Other parameters incorporated into the model include
positive and negative Vp (Vp,pos and Vp,neg, respectively), k, which is the constant given in
Equations (1) and (2), and maximum ID,. IDmax was needed for the calculation of ID. The
parameters Vp,pos, Vp,neg, k, and IDmax are determined through experimentation.
EQUATIONS OF VARIABLES
Besides the parameters defined by the user and those obtained by experimentation, the
model also includes variables that are determined using simple mathematical equations.
These parameters are Vin, Vin,phased, Polarization1, Polarization2, IDpos, IDneg, ID, ID,FA, and
Vout. Vin is the value of the input voltage that takes into account both the amplitude of the
input voltage and an offset, if given. The value of the input voltage that includes Vin,amp,
Vin,offset, and the phase shift is Vin,phased. Polarization1 and Polarization2 are parameters
that reflect the polarization property of ferroelectric material. IDpos is the drain current
found using positive Vp, Vp,pos, when the gate voltage is on, while IDneg is the drain current
found using negative Vp, Vp,neg, when the gate voltage is on. ID is the final value of the
drain current. ID,FA is the frequency adjusted value of ID. Thus, ID,FA adjusts ID based on
the input frequency and the frequency coefficient. Finally, Vout is the output voltage. The
equation for each parameter is given in the following set of equations.

Vin = Vin,amp*sin(radian value) + Vin,offset

(3)

Vin,phased = Vin,amp*sin(radian value + phase shift) + Vin,offset

(4)

At radian value 0,
Polarization1 = 0

(5)

At all other test points,

Polarization1
1 if Vin>Vp,pos
=

(6)

the value of Polarization1 at the preceding test point if Vin<Vp,pos

(7)

At radian value 0,
Polarization2 = 0

(8)

At all other test points,

Polarization2
0 if Vin<Vp,neg
=

(9)

the value of Polarization1 at this test point if Vin>Vp,neg

(10)

IDpos = IDmax / (ek*(Vp,pos Vin,phased) + 1)

(11)

IDneg = IDmax / (ek*(Vp,neg Vin,phased) + 1)

(12)

ID
=

IDpos if Polarization2 = 1

(13)

IDneg if Polarization2 = 0

(14)

ID,FA = ID*frequency coefficient*log10frequency

(15)

Vout = ID,FA*Rload

(16)

The Vin and Vin,phased equations are based on the sinusoidal wave equation [3]
given by
x(t) = A*sin(t + ) + D,

(17)

where A is the amplitude of the signal, is the radian frequency in radians per second, t
is time in seconds, is the phase angle in radians, and D is the DC offset. In the equation

for Vin, Vin,amp is the amplitude, radian value is t, and Vin,offset is the DC offset. There is
no phase shift in Vin; however, this variable exists in the equation for Vin,phased. In order to
fully explain the term radian value in the equations for Vin and Vin,phased, the method of
measurement used in the model must first be discussed.

The model measures the

parameters at specific test points. The value of these test points are the nonnegative
numbers. The degree equivalent of each test point was chosen to be (360/30) times the
value of the test point. This relationship is given as follows
degree value = test point * (360/30)

(18)

The radian equivalent of the degree value of each point is utilized throughout all the
equations for the variables. Since t is equivalent to 2ft, where f is frequency, and t in
this model is a multiple of the period 1/f (t=n/f, where n is a nonnegative number), then
t can be reduced to 2n. However, degree value can be rewritten as 360*n/30, whose
radian equivalent is given by 2n/30. Thus, t in Equation (17) is replaced with 2n/30
in the equations for Vin and Vin,phased, and the term 2n/30 is denoted by radian value.
The equation for each of the other variables is derived from the basic properties of
ferroelectric material. The polarization variables, Polarization1 and Polarization2, are
initially set to zero, then altered based on the value of Vin at the current test point. If Vin is
greater than Vp,pos, then Polarization1 is set to 1, otherwise it is set to its value at the
preceding test point. Polarization2 is set to 0 if Vin is less than Vp,neg. If Vin is greater
than Vp,neg, Polarization2 takes the value of Polarization1 at the current test point. The
equations for IDpos and IDneg are derived from the Fermi-Dirac-based equation for ID when
the gate voltage is on, where Vp in Equation (1) is replaced with Vp,pos for IDpos and Vp,neg
for IDneg, IDSAT is replaced with IDmax, and Vgs is replaced with Vin,phased.

The term

B*log(t1) in Equation (1) is ignored in this model with little loss of accuracy. ID equals
IDpos if Polarization2 is 1, but if Polarization2 is 0, then ID equals IDneg.

Vout is

determined simply using Ohms Law with the load resistance and the frequency-adjusted
value of ID. This set of equations along with the user-defined and experimentation-based
parameters provide a very good approximation of the output of a FeFET simple amplifier.
MODELED AND MEASURED DATA ANALYSIS
The accuracy and efficiency of the model created was examined by comparing the output
of the model with empirical data for specific test cases. It is important to note that this
model is the first mathematical-based computer model to characterize the behavior of a
FeFET simple amplifier. Thus, this model is a first attempt and can be considered a
preliminary version of more sophisticated models that can be created in the future.
Given the user-defined and experimental-based parameters, the model outputs a
plot of Vin and Vout with respect to time. As was aforementioned, time is given as a
multiple of the period. Several representative test cases were examined. For each test
case, two sets of parameter values are included: one set leads to a model output where
the amplitude of Vou, Vout,amp, is as close to the oscilloscope value as the model allows,
while the other set results in a waveform for Vout that is very similar in shape to that of the
oscilloscope. For each set of parameter values, a table of the values and a plot of the
modeled output are given. The modeled plots are compared to the oscilloscope plot. For
all the test cases, the sum of Vin,amp and Vin,offset never exceeds 8V to ensure that the
FeFET is not burned.

Test Case I
For the first test case, VDD was set to 0.709V, the amplitude of Vin was 6V, the frequency
was 100Hz, and the load resistance was selected to be 35k. No offset was included in
the input signal. The following table lists the values of the user-defined parameters that
resulted in Vout,amp that is greater than that of the oscilloscope, but as small as the model
allowed.
Table 1 Initial parameter values for Test Case I
Parameter
Value
VDD
0.709V
Rload
35k
Frequency
100Hz
Vin,amp
6V
Vin,offset
0V
Phase
1.5
IDmax
6.00E-04A
Vp,pos
2.5V
Vp,neg
-0.5V
k
0.75
Frequency Coefficient
0.007
The parameter phase was set to 1.5 in order to reproduce the phase shift seen in
the oscilloscope output of the empirical measurement, which is shown in Figure 2. It was
observed that a positive value for phase caused the output signal to be shifted to the left,
while a negative value for phase led to a rightward shift of the output signal. Thus, in
this test case, the output signal leads the input signal. The value of the parameter k was
selected after experimenting and noting the effect of changing that parameter. For this
test case, decreasing k below 1 removed the clipping effect at the peaks of the output
signal, and further decreasing k below 0.8 removed the clipping effect at the troughs of
the output signal. A very small value for frequency coefficient was chosen to ensure that

Vout,amp was as small as possible. Decreasing frequency coefficient decreases ID,FA, as

seen in Equation (15), which leads in turn to a decrease in Vout. The values of IDmax,
Vp,pos, and Vp,neg were determined from previous experiments and will remain unchanged
throughout all the test cases. Figures 2 and 3 show the oscilloscope output and the
modeled output of the signals, respectively.

Vin
Vout

Figure 2 Oscilloscope output for Test Case I

FeFET Amplifier Model
8
6
4

Volts

2
Vin

0
0

0.2

0.4

0.6

-2
-4
-6
-8
Time

0.8

Vout

Figure 3 Modeled output for Test Case I based on initial parameter values

In this and all upcoming oscilloscope plots, it must be noted that the input and
output waveforms are not plotted on the same coordinate system, despite their sharing the
same axes. For example, in Figure 2, the coordinate system of Vin is partitioned into
divisions that are 2V each, whereas Vouts coordinate system is divided into 5mV per
division blocks. Therefore, Vin,amp is 6V, while Vout,amp is only 11.8mV.
The maximum value of Vout as measured by the oscilloscope is 11.8mV, which is
significantly larger than the amplitude of the models Vout, as seen in Figure 3. However,
this is the smallest value of Vout,amp that still displays the shape of the signal. Comparing
Figures 2 and 3, one might not believe that the model accurately illustrates the behavior
of the FeFET amplifier, but increasing frequency coefficient to 0.15 and decreasing k to
0.3 produces the plot shown in Figure 4, which closely resembles the oscilloscope output.
The given value of the frequency coefficient was found to produce the desired amount of
frequency effect. At 0.3, k was small enough to model the jumps in the peaks and
troughs visible in the oscilloscope output of this test case.
parameter values.

10

Table 2 lists the new

FeFET Amplifier Model

8
6
4

Volts

2
Vin

0
0

0.2

0.4

0.6

0.8

Vout

-2
-4
-6
-8
Time

Figure 4 Modeled output for Test Case I based on new parameter values
Table 2 New parameter values for Test Case I
Parameter

Value
VDD
0.709V
Rload
35k
Frequency
100Hz
Vin,amp
6V
Vin,offset
0V
Phase
1.5
IDmax
6.00E-04A
Vp,pos
2.5V
Vp,neg
-0.5V
k
0.3
Frequency Coefficient
0.15
Comparing the modeled waveforms of Figure 4 with the measured waveforms
shows that, in fact, significant similarities exist between the two plots. The modeled
output signal displays the same jumps in its peaks and troughs as those of the
oscilloscope output signal. Moreover, the input and output waveforms of the model

11

intersect at almost the same locations as the measured waveforms.

Despite these

parallels, one particular divergence existing between the two output signals is the fact that
Vout of the model never takes on both positive and negative values i.e., Vout is either
greater than or equal to zero or less than or equal to zero, depending on the chosen values
of the parameters. This fact will be noted in all the test cases. Further improvements in
the model may resolve this issue. Even though the amplitude of Vout is not correct since
frequency coefficient has been significantly increased for the purpose of depicting the
details of Vout, overall, Vout of the model has the same shape as the oscilloscopes Vout and
nearly the same values when frequency coefficient is small enough.

Test Case II
For the second test case, VDD was again set to 0.709V, the amplitude of Vin was decreased
to 2V, Vin,offset was set to 2V, the frequency was 1MHz, and the load resistance was
selected to be 600k. The following table lists the values of the user-defined parameters.
Table 3 Initial parameter values for Test Case II
Parameter
Value
VDD
0.709V
Rload
600k
Frequency
1MHz
Vin,amp
2V
Vin,offset
2V
Phase
-0.75
IDmax
6.00E-04A
Vp,pos
2.5V
Vp,neg
-0.5V
k
1
Frequency Coefficient
0.0002
In this test case, a negative value for phase was chosen since the output signal
lags the input signal. The magnitude of phase is much smaller for this test case than for

12

the previous test case since, as can be seen in the oscilloscope output of Figure 5, the
input and output signals are out of phase by a small amount. The removal of the clipping
effect was again noted when the value of k was decreased. However, decreasing k below
1 resulted in more rounded peaks and troughs than desired, so k was set to 1. The value
of 0.0002 was selected for frequency coefficient because it led to the smallest Vout,amp
while maintaining the desired shape of Vout. Figure 6 shows the modeled output.

Vin
Vout

Figure 5 Oscilloscope output for Test Case II

FeFET Amplifier Model
4.5
4
3.5

Volts

3
Vin

2.5

Vout

2
1.5
1
0.5
0
0

0.00002

0.00004

0.00006

0.00008

0.0001

Time

Figure 6 Modeled output for Test Case II based on initial parameter values
13

The modeled output displays the same phase shift present in the oscilloscope
output.

Moreover, Vout,amp of the model is as close as possible to Vout,amp of the

oscilloscope output, which is 74mV, while displaying the same trends as the empirical
output signal. To further verify the accuracy of the model, plots were obtained for this
test case when frequency coefficient was increased to 0.002. Figure 7 shows the modeled
output, and the table below summarizes the parameter values.
Table 4 New parameter values for Test Case II
Parameter

Value
VDD
0.709V
Rload
600k
Frequency
1MHz
Vin,amp
2V
Vin,offset
2V
Phase
-0.75
IDmax
6.00E-04A
Vp,pos
2.5V
Vp,neg
-0.5V
k
1
Frequency Coefficient
0.002

14

FeFET Amplifier Model

4.5
4
3.5

Volts

3
Vin

2.5

Vout

2
1.5
1
0.5
0
0

0.00002

0.00004

0.00006

0.00008

0.0001

Time

Figure 7 Modeled output for Test Case II based on new parameter values

The plots of Figure 7 look very similar to those generated by the oscilloscope.
The modeled plots display the same shape and intersect at almost the same points as the
oscilloscope plots. It can be concluded that the plots of Figure 6 have the correct,
expected shape but were too small to be effectively compared to the measured results.

Test Case III

For the final test case, VDD remained 0.709V, the amplitude of Vin was further decreased
to 1V, 3V of offset were applied to Vin, the input frequency was 100kHz, and Rload was
560k. These values and the other user-defined parameters are summarized in Table 5.

15

Table 5 Initial parameter values for Test Case III

Parameter

Value
VDD
0.709V
Rload
560k
Frequency
100kHz
Vin,amp
1V
Vin,offset
3V
Phase
1.5
IDmax
6.00E-04A
Vp,pos
2.5V
Vp,neg
-0.5V
k
2
Frequency Coefficient
-0.00015
This list of parameter values is different from those presented previously in
several aspects. Here, frequency coefficient is a negative value. It was determined that a
negative frequency coefficient most accurately represents the fact that the maximum
value of Vout on the oscilloscope is 75mV, and its minimum value is -50mV, as seen in
Figure 8. In this figure, the output signal is plotted on a coordinate system with 50mV
per division. Another point of interest is the fact that the measured output signal is
lagging the input signal, but a positive value of phase was selected. This can be
attributed to the fact that the negative frequency coefficient flipped the plot of Vout across
the x-axis, thereby making Vout lead Vin. This accounts for the positive value of phase,
since a leading signal is given by positive phase. The parameter k was set to 2 because
larger values for this parameter resulted in the clipping of the output signal. Thus,
comparing the modeled plots of Figure 9 with the measured plots of Figure 8, it can be
seen that the modeled Vout is close in value to the measured Vout.

16

Vout
Vin

Figure 8 Oscilloscope output for Test Case III

FeFET Amplifier Model

4.5
4
3.5
3
Volts

2.5
Vin

Vout

1.5
1
0.5
0
-0.5

0.0002

0.0004

0.0006

0.0008

0.001

Time

Figure 9 Modeled output for Test Case III based on initial parameter values

17

To obtain plots of Vout and Vin that closely resemble the oscilloscope plots, several
changes were made to the parameter values. First, frequency coefficient was changed to a
positive value, and its magnitude was increased to 0.0035. Since frequency coefficient
was then positive, phase was changed to a negative value so that Vout lagged Vin as shown
in the oscilloscope plots. A value of 1.3 was chosen for k because higher values resulted
in more rounded peaks and troughs for Vout than desired. The following table lists the
new parameter values. The modeled plots are shown in Figure 10.
Table 6 New parameter values for Test Case III

Parameter

Value
VDD
0.709V
Rload
560k
Frequency
100kHz
Vin,amp
1V
Vin,offset
3V
Phase
-1.5
IDmax
6.00E-04A
Vp,pos
2.5V
Vp,neg
-0.5V
k
1.3
Frequency Coefficient
0.0035

18

FeFET Amplifier Model

6
5

Volts

4
Vin

Vout

2
1
0
0

0.0002

0.0004

0.0006

0.0008

0.001

Time

Figure 10 Modeled output for Test Case III based on new parameter values

These plots for Vin and Vout closely resemble the empirically-derived plots. As
with the previous test cases, these plots verify the correctness of the model. It is clear
that the desired results are obtainable using this model.
CONCLUSION OF DATA ANALYSIS
From these test cases, it can be concluded that the models results are not only
comparable but similar to the oscilloscope outputs. The models output plots have the
same shape and exhibit the same trends as those produced by the oscilloscope.
Examining the mathematical equations behind the model brings to light their simplicity
and the models ease of use. Therefore, this model is accurate enough to be used now
and efficient enough to be easily improved for further study.

19

REFERENCES
1. T. C. MacLeod, T. A. Phillips, and F. D. Ho, Characterizing an Analog Amplifier
Utilizing a Ferroelectric Transistor, Integrated Ferroelectrics 104, 40-47 (2008).
2. T. C. MacLeod and F. D. Ho, Modeling of Metal-Ferroelectric-Semiconductor Field
Effect Transistors, Integrated Ferroelectrics, vol. 21, pp. 127-143, 1998.
3. S. Haykin and B. Van Veen, Signals and Systems, 2nd ed. New York: John Wiley
and Sons, 2003.

20