2007 International Conference on Solid Dielectrics, Winchester, UK, July 8-13, 2007

Time Domain Polarisation Measurements with the Sawyer-Tower Method

B. Martin* and H. Kliem
Saarland University, Institute of Electrical Engineering Physics,
P.O. Box 15 11 50, D-66041 Saarbruecken, Germany
* E-mail: b.martingmx.uni-saarland.de

Abstract: A system for time domain polarisation polarisation P(t) is determined directly [2]. Then the
measurements in a wide time range which is based on polarisation current is obtained as the time derivative of
the well-known Sawyer-Tower method is presented. d
The time dependent polarisation of dielectric samples is the polarisation, i.e. j(t)= For the measured
measured after the application of a voltage step. To this dt
aim the voltages on three reference capacitors in series quantity P(t) the dynamic range is smaller and one
to the sample are measured with different data single measurement is sufficient to receive the complete
acquisition systems. During the measurement the set of data. On the other side P(t) has to be measured
capacitors are switched off one after another to avoid with a high amplitude resolution. This is now possible
self-discharge. In this way the polarisation of the due to the last generations of transient recorders.
dielectric can be determined in a time range from 10-6s
to 104s, i.e. over 10 decades, with one single run. From
these data the polarisation current can be calculated as PRINCIPLE OF MEASUREMENT
time derivative of the polarisation. Furthermore, the The presented measurement technique is based on the
dielectric permittivity in the frequency domain can be Sawyer-Tower method which was introduced to obtain
calculated from the data in the time domain by Fourier hysteresis loops at ferroelectric samples [2]. The sample
transformation. Additionally the system allows the is placed in series with a reference capacitor. This
measurement of the switching behaviour of ferroelectric circuit is connected to a voltage source which generates
samples with a three-pulse method. here a step voltage Uo instead of a triangular
signal (Figure 1). The voltage at the capacitor Uc(t) can
be measured with an electrometer with high input
INTRODUCTION impedance to avoid the discharge of the capacitor. From
In material science the determination of the electrical the voltage Uc(t) the charge Qc(t) on the capacitor can
behaviour of dielectric samples can be of high interest. be calculated. The same charge is also present on the
It is possible to measure this behaviour either in the sample. With the area A of the sample the polarisation
frequency domain or in the time domain. Because the P(t) can be derived from the charge, i.e. P(t) = Qc(t)IA.
frequency- and time domain are linked by the Fourier Hereby the reference capacitor should have at least a
transform, there is no difference in which domain the capacitance 100 times larger than the capacitance of the
measurements are performed when the samples are sample. Then the voltage drop at the capacitor is
linear and time invariant [1]. negligible and the source voltage is nearly the same as
the voltage at the sample.
In both cases special problems occur during the
measurement. In the frequency domain each single
frequency has to be observed separately. Before a CSample
measurement can start the steady state at the chosen
frequency has to be reached. Especially when low
frequencies are looked at the measurements can CReference uc
consume long times. Therefore it can be advantageous
to perform the measurements in the time domain where
in principal one single measurement suffices to receive Figure 1: Modified Sawyer-Tower measurement
the complete set of data. Here the common way to system. The sample is connected in series to the
determine the electrical behaviour is to measure the time reference capacitor and the voltage source. The voltage
dependent polarisation current j(t) after a voltage step. at the capacitor is measured with an electrometer with
This current can have an enormous large dynamic high input impedance to avoid self-discharge.
range. Thus, the measurement of the current over a wide
time range in a single step is not possible. For example Thus, the time dependent polarisation can be
the dynamic range for Kohlrausch-like currents j -a- determined. Especially for long times the data
can cover 10 decades or more. acquisition system plays an important role. The input
impedance of the system together with the capacitance
Another possibility is to employ the well-known of the reference capacitor builds a circuit with a certain
Sawyer-Tower method where the time dependent RC-time. This time has to be much higher than the

1-4244-0750-8/07/$20.00 ©2007 IEEE.

729
measurement time to avoid a self-discharge of the a first negative voltage the sample is switched to the
reference capacitor. However, commercial electrometers negative saturation of the polarisation. A second
with sufficient high input impedances have long rise positive pulse switches the sample in another state. The
times. Due to this a measurement in the short time range measured time dependent charge during the second
is not possible. pulse contains the ferroelectric switching as well as all
side effects like relaxational polarisation and leakage
To solve this problem the short time range can be charges. Now a third pulse with the same voltage as the
measured with two additional capacitors (Cl, C2) and second pulse is applied. Because the ferroelectric
pre-amplifiers with shorter rise times but lower input sample is switched during the second pulse the third
impedances (Figure 2). During the measurement these pulse contains only the side effects. Then the results of
capacitors are switched off one after another at the times the third pulse can be subtracted from the results of the
t1 and t2. After t2 only the device with the highest input second pulse. In this way the side effects are eliminated
impedance is active measuring U3. The results of all and the true ferroelectric switching behaviour is
channels have to be combined to get the complete determined.
polarisation.
CSample RESULTS
__I orecsordi ng
r. Measurements using a dummy structure
-r~~~403: C3 i~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
uo IS-
_
_ q
To confirm the results of the measurement system a
reference measurement using a dummy structure is
0

'2 I2
=-
\~~~~~~~~~~~
2
I
_2 _i
performed (Figure 3). After application of a voltage step
t _ _ Ri U1 of Uo = 20V the time dependent polarisation and
I I-
,
I
i a 6 6 a depolarisation charges are determined. The results are
depicted in Figure 4. In the short time range only the
Figure 2: Modified Sawyer-Tower measurement charge on the single capacitor can be observed. The
system. Three measurement capacitors are implemented high frequency charging is much faster than 10-6s. In the
which are switched off one after another during the time range up to 1Os the exponential-like charging of
measurement. After t2 only the device with the highest the two RC-elements with the corresponding RC-times
input impedance is active measuring U3. The voltages is found. The same behaviour can be observed for the
are measured with pre-amplifiers having different input depolarisation process. The theoretical curves for the
impedances R1, R2, R3, and rise times. polarisation and depolarisation are depicted as well in
Figure 4. It can be seen that the measured curves fit the
With new data acquisition systems a sample rate of up theoretical ones very well. The long time stability of the
to 20 MHz with a resolution of 12 bit can be reached to system is also demonstrated. The measured polarisation
measure the voltages. For the long times the high does not change for times up to 104s. As claimed no
sample rate is not necessary and commercial slower data self-discharging of the capacitors appears.
acquisition systems with higher resolutions can be used.
With the use of different systems the sample rate is
changed during the measurement. Thus, the amount of
data is reduced. In this way a measurement of the time IOMQ IGQ
dependent polarisation from 10-6S to 104s, i.e. over 10 33OpF
decades, is possible with one single run. 335OpF 33OpF
After switching off U0 the depolarisation process can be
measured. A comparison between the transient
polarisation and depolarisation charge yields basic Figure 3: Dummy structure. Two RC-elements with
information about the time invariance of the process. RC-times tRC1 = 4.7ms and tRC2 = 470ms are placed
parallel to a single capacitor.
From the time dependent polarisation the polarisation
current j(t) is calculated by differentiation. The data set From the polarisation and depolarisation the
for the polarisation current can then be transformed in corresponding currents can be calculated. The results
the frequency domain to obtain the complex dielectric are shown in Figure 5. Here the two RC-times with the
permittivity £*(ca) of the sample [1]. exponential behaviour can also be found. A comparison
with the polarisation current calculated from the
An additional feature of the system is the possibility to measured capacitance of the dummy structure in the
measure the time dependent switching of ferroelectric frequency domain is shown. The agreement of these
samples. For this purpose a three-pulse measurement results ensures the functionality of the measurement
method is implemented to eliminate side effects like system as well as the link between the frequency and the
space charge polarisation and charge injection [3]. With time domain.

730
 0
2.0x10o~8

1.5x1 08v
1.0x10

5.0x10-9

-5.0xl0-9

Figure 4:
0.0
O
measured polarisation
theoretical curve

measured depolarisation
theoretical curve
i0-4

Time dependent polarisation and

depolarisation charges of the dummy structure. The
measured curves are nearly identical to the theoretical
t (s)
1o-1 102 0

05
1.84x1 08

E 1.82x1 08

1.80xlo08
D
. ,i r . i, . ,i r . ,
Then between 10-6s and 103s the polarisation changes
only by about zP = 7 10-l'As/cm2. This value
corresponds to a voltage change of AU= 5mV at the
reference capacitor during nine decades in time.
Therefore a high resolution of the data acquisition
system is necessary to detect such small changes of the
signal.

1.86x1 08
polarisation

]1A -
8x1 10

6x1 10

4x1 10

2x1 10

curves. 1.78xlo08
J 0
,,,

-6
10-3 10° 11 03

105
o
t (s)
Figure 6: Time dependent polarisation and
"idep' depolarisation of the polyimide sample.
rcalc
As can be seen with Figure 6 polarisation and
0o-8 r\
depolarisation charge have different transients for
1o-X t < 1044s. Here further investigations on a possible time
variant process are necessary. At long times the
_011 polarisation and depolarisation charges increase which
looks like a battery effect. This can be due to residual
water molecules in the sample.

..,.I .. .,.I .. .,.I .. .. ,.I .. ,.I. .. .....

....,.I. ..
.... ..... ......
.....
....
....
For this sample it is also possible to calculate the
10-6 10-4 10-2 10° polarisation and depolarisation currents (Figure 7).
t (s)
Although the data are noisy because of the small
Figure 5: Polarisation and depolarisation currents of the polarisation changes a t- abehaviour is found over nine
dummy structure. The results found by the measurement decades with oc 1.1. This enormous dynamic range of
in the time domain are compared to the data calculated the polarisation current can be measured with one single
run. The observed results are in good agreement with
by inverse Fourier transform from the data measured in
the frequency domain. earlier experiments by measuring the polarisation
current directly but decade per decade [4].

Measurements polyimide
on
Measurements P(VDF-TrFE)
on
Earlier measurements on films of polyimide show that
the polarisation currents have a time dependence which For measurements of the switching behaviour in
follows a t- abehaviour over several decades in time [4].
ferroelectric materials the copolymer
Therefore the material is well suited to test the polyvinylidenefluoride trifluorethylene P(VDF-TrFE) is
resolution of the measurement system. Onto the material used. Very thin films (d- 80nm) are prepared by a
with a thickness of d= 25pim aluminium electrodes are Langmuir-Blodgett technique on glass substrates
evaporated. Then the sample is inserted into a shielded covered with aluminium [3]. On the top the film is also
coaxial cell with dry atmosphere. After relaxation in a contacted with thermally evaporated aluminium
short circuit a voltage of U0= 180V is applied. The electrodes. The measurement is performed at a
resulting polarisation and depolarisation curves are temperature T = 303K.
presented in Figure 6. A fast jump of the polarisation at To measure the ferroelectric switching the three-pulse
method is used. A sample with 100 monolayers of
times below 10-6s to P = 1.79 10-8As /cm2 appears. P(VDF-TrFE) is polarized to the negative saturation

731
 CONCLUSION & OUTLOOK
1 o-3
A measurement system has been developed to
r~~~~~~~~~~~0I epi
determine the electrical behaviour of dielectric materials
10-6
in the time domain. One single run is sufficient to obtain
the complete set of polarisation data from 10-6s to 104s.
r~~~~~~~~~~~ Additionally, the time dependent switching behaviour of
-.

CM
E
r 1~~~~~~~~~~~~~~~~ferroelectric materials can be measured with a three-
.......9 ....9 ......9
r 1 0-9 1~~~"
pulse method. In the future the use of faster data
acquisition systems with higher resolutions should be
advantageous. In this way the measurement of shorter
1 o-12 r
times can be performed with less noise.
1~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

1 o-1 5
1C)-6 1o-3 100 1 03
REFERENCES
t (s) [1] N. Farag, S. Holten, A. Wagner, and H. Kliem,
"Numerical transformations of wide-range time-
Figure 7: Polarisation and depolarisation currents of and frequency domain relaxational spectra," IEE
the polyimide sample. The tf-Ubehaviour over nine Proc. - Science, Measurement and Technology,
decades in time is found with oc 1.1. - vol. 150, pp. 65-74, 2003.
[2] C. B. Sawyer and C. H. Tower, "Rochelle salt as a
with a voltage Uo =-1OV for t= 100s. Then two pulses dielectric," Phys. Rev., vol. 35, pp. 269-273, 1930.
with the same positive voltage Uo = 1OV are applied for
a time of 100s. The first polarisation curve contains the
[3] H. Kliem and R. Tadros-Morgane, "Extrinsic
ferrolectric switching as well as all side effects (Figure versus intrinsic ferroelectric switching:
7). The second curve contains only the side effects and experimental investigations using ultrathin PVDF
no switching can be observed. Then the difference
Langmuir-Blodgett films," J Phys. D, vol. 38, pp.
between the two curves is calculated. In this way the 1860-1868, 2005.
true ferroelectric switching behaviour is obtained. [4] H. Kliem, B. Schumacher, and G. Arlt, "Transient
Similar results have also been found in earlier current measurements in polyimide" 1986 CEIDP
measurements [3]. Ann. Rep., pp. 168-173, 1986.
lop

8p

6p

a-
4p
2p,

O
-
1o06 10-3 100

t (s)
Figure 7: Polarisation curves of a P(VDF-TrFE) film
with 100 monolayers. Pulse 1 contains the ferroelectric
switching and the side effects whereas pulse 2 contains
only the side effects. The difference between them gives
the true ferroelectric switching behaviour.

732