Effect of Humidity on the Time to Breakdown

of Polymers Subject to Partial Discharges

F. GUASTAVINO

Dipartimento di Ingegneria Elettrica

Universita di Genova
161 45 Genova, Italy
and

P. TIEMBIX)*

Instituto de Ciencia y Tecmlogia de Polimeros (C.S.I.Ci

28006,Madrid, Spain

Three polymers, polyethylene terephthalate, a polyimide, and polyamide-6, were

subjected to surface partial discharges up to electrical breakdown under varying
ambient humidity conditions. It was found that increasing humidity increases
lifetime of the samples until a critical value of relative humidity is attained, after
which the polymers' lifetime dramatically decreases. This critical value depends on
the hydrophylicity of the polymer and is lower as the ability of the polymer to absorb
water increases. The evolution of the electrical time to breakdown with the relative
humidity is explained in terms of changes in surface conductivity following varia-
tions in the ambient humidity.

INTRODUCTION phenomenon such as inception voltage, frequency re-

currence, and amplitude of the discharges, and the
ial discharges are electrical discharges that de-
p""
velop in the structural defects of insulation sys- phase of the supply voltage at the instant of each
partial discharge (when working under AC condi-
tems when subjected to an electric field (1, 2). These
tions). The introduction of the phase resolved partial
discharges do not immediately lead to electrical break-
discharge measurement technique (PRPD) has con-
down, but as their effect is cumulative and irrevers-
ible, they cause the degradation of the insulating ma- siderably improved the detection and measurement of
PD, but as the output data available through these
terial and of its properties. If lasting for a long period
methods is enormous, a statistical approach is neces-
they may end causing electrical breakdown. As a mat-
ter of fact, partial discharges (PD)are one of the main sary.
As for the chemical modifications of the materials,
causes of degradation of the insulation systems in
surface chemical analyses have been performed, and
electrical equipment under normal use conditions, es-
pecially on materials derived from organic chemistry, much information about the fmal state of the material
is available (6-9). Experimental conditions usually re-
as is the case of polymers. Still, in spite of its obvious
produce PD under normal use conditions of insulating
technological importance, little is known about the
systems. A discharge in a gas involves the generation
basis of the degradation mechanism.
of various species, all of them highly reactive; when
Research concerning this phenomenon has focused
done in air, atomic oxygen, ozone, nitrogen oxides,
mainly on two areas: i) detecting and characterizing
etc., are found. Chemistry on the polymer's surface is
progressively better the basic quantities involved in
therefore anything but controlled and leads to the
PD and ii) studying a posteriori the chemical transfor-
formation of various oxidation products, branching,
mations suffered by the insulating material in the
chain breaking, unsaturation, etc. For example, in the
course of degradation. The first of these (3-5) has been
case of polyolehs, crystals of oxalic acid are found
accomplished by designing measurement cells and
after extensive PD degradation. The changes in the
sophisticated data acquisition systems that provide
morphology of the surface are usually studied using
information about characteristic quantities of the PD
scanning electron microscopy, while chemical trans-
formations and by-products are followed by IR spec-
T o whom correspondence should be addressed troscopy (IR microspectroscopy).
POLYMER ENGINEERING AND SCIENCE, JANUARY 1998, Vol. 38, No. 1 119
 F. Guastavino and P. Tiemblo

In the last few years, research has been more and TEST CELL SET-UP
more concerned with the finding of a relationship be-
tween characteristic quantities of the PD phenomenon
and the dynamics of the degradation process in insu-
lating systems (9).In a previous paper (10)we found
two characteristic quantities of the surface PD aging
on polymeric thin films that are related to the dynam-
ics of the process, i.e., magnitudes whose evolution
with degradation time characterized the degradation
process. These are the time derivative of the average
charge with respect to the acquisition time (1) and the
time derivative of the external energy (w)of the dis-
charges, both of which will be defined in the Experi- lower electrode
vucuum holes
mental section. These quantities were found (10)to be
related to the time to breakdown of the polymer sub-
jected to PD in the following way:
log tb = log a - Plog rn MEASUREMENT CIRCUIT
(1)
where tb stands for time to breakdown, and rn = ( d / d t l HV

-
or rn = ( d W / d t ) ,that is, the time derivative of PD
resistive
average charge or of the external energy. It has been divider
shown that parameters a and b are common to all
samples tested. This relationship shows that time to
breakdown depends on the increase of the average
charge (electrical current) and on the increase of the
6filter

electrical energy (electrical power) during the degra-

dation and not on the amount of the average charge and oscilloscope
of the electrical energy themselves. This relationship
has proved to be general as far as possible variations
in the experimental conditions of our setup are con- phase
reference
cerned, i.e., it is independent of the nature of the
Fig. 1 . Partial discharge test cell and measurement circuit.
polymer film, of the film thickness, of the humidity
conditions, and of the amplitude of the test voltage.
The fact that the relationship is general for the poly-
mers tested in our previous work (10) (polypropylene,
polyether ether ketone, a polyimide, a polyamide, and take place between the metallic hemispherical elec-
polyethylene terephthalate, all of them commercially trode and the polymer film placed on the lower metal-
available) and for varying ambient humidity values lic plane electrode. A graphite conductive paint layer is
does not mean that the electrical time to breakdown applied on the lower film surface, which is in contact
under PD conditions is independent of the polymer’s with the lower electrode to avoid the presence of PD
structure or of relative humidity, but that changes in between the film and the lower plane electrode. A
them affect time to breakdown through modifications moderate vacuum is applied through tiny holes drilled
in d / d t and dW/dt. As a matter of fact, the relative in the lower electrode to ensure good adhesion be-
humidity at which experiments were performed tween the film and the plane electrode, in order to
greatly affected the time to breakdown of the different have good mechanical precision of the air gap. The
polymers tested. We have therefore undertaken a configuration of the cell has been constructed accord-
study on the effect of relative humidity on the time to ing to calculations that make it possible to obtain a
breakdown of three chemically different polymers, slightly divergent geometric field with force lines as
polyamide-6, polyethylene terephthalate, and a poly- normal as possible to the f ilm surface. Nevertheless,
imide. on the film’s surface, a weak tangential electric field is
present (3).The configuration of the test cell mini-
EXPERIMENTAL mizes the effects of pure electrical aging, provided
setup that films thicker than 20 pm are used (17).
PD signals are monitored by a digital oscilloscope
The test cell, shown in Fig. 1 , has a sphere-plane and are acquired by means of a bipolar PFWDA (Phase
electrode configuration. The air gap between the film Resolved PD analyzer) system with 64 amplitude
surface and the upper electrode is of 0.1 mm and can channels and 128 phase channels at fixed time inter-
be regulated with a precision of 0.01 mm. The sphere vals. The measurement of the AC test voltage ampli-
electrode has a diameter of 6 mm. Over a certain tude is done using a resistive divider; this signal is
voltage, called inception voltage, partial discharges also used as phase reference for the PRPDA. In this

120 POLYMER ENGINEERING AND SCIENCE, JANUARY 1998, Vol. 38,No. 1

 Effect of Humidity on Time to Breakdown

way the so-called PD pattern is obtained (Fig. 2). The where Qiis the apparent charge (related to the i-th
information contained in the PD pattern is the num- amplitude window) and is the instantaneous value
ber N, amplitude Q and the phase of occurrence cp with of voltage (related to thej-th phase window) applied to
respect of the applied voltage, of positive and negative the electrodes at which the discharge takes place. Niis
polarity discharges. As can be seen in Fig. 2, phase and the number of discharges of amplitude i, and & is the
amplitude are recorded in the x and y axes, respec- set acquisition time. This acquisition time is set at the
tively, while the number of discharges are presented beginning of the experiment. In the case of tests where
using a color code. In this Figure (and in Fig. 61, in- the involved phenomena are time variant, the acqui-
stead of color code, a scale of gray is used. sition time must be set short enough to allow a n ap-
The required voltage is applied to the cell, and the proximation to a time invariant system during the
test proceeds continuously until breakdown of the acquisition stage, but it must be sufficiently long to
sample occurs. In the tests presented in this work, the obtain correct information from the PRPDA. In our
applied voltage has been 2.4 KV, which is about three case it has been set at 40 s. An example of the type of
times the PD inception voltage of the samples. PD results obtained appears in Fig. 3, where the evolution
patterns are collected a t fixed time intervals, so their with degradation time of the positive discharges' aver-
evolution with the degradation time may be studied. age charge for the three polymers is shown. Each
Starting from several PD patterns, acquired as the test point in the graph represents the average charge of the
time progresses, it is possible to obtain a wealth of positive polarity discharges that have taken place
statistical information and to calculate several derived during the acquisition time.
quantities, for example, inception phase and maxi-
mum charge, and their evolution with the degradation Materials
(test) time. The quantities relevant to this work, be-
A set of commercially available polymers was used
sides number and amplitude of the discharges, are
in this study: A 25-km-thick film of a polyimide (PI)
average charge of positive I+ and negative r polarity
from DuPont (Kapton),which is currently used in elec-
discharges and external energy W+ and W , defined in
trical applications as an insulator; a 23-pm-thick PET
the following way:
supplied by Polifibra (Mylar);and a 25-km-thick film
of a polyamide (PA) from C&aro (nylon 6).The poly-
mers have been chosen according to two criteria: i)
their common use in electrical insulation and/or ii)
their different chemical structure and ability to absorb

Amplitude

Phase
m.2. Partial discharge ap'pattem"recorded during 40 s.
POLYMER ENGINEERING AND SCIENCE, JANUARY 1998, Vol. 38, No. 1 121
 F. Guastauino and P. Tiemblo

4000000

3000000
. .
Fn
3&000000
+
H

1000000

0
0 10000 20000 30000 40000 50000

Flg. 3. EUoUion of the average charge of the positiue discharges during the degradation time.

water. As for the first one, it is well known that poly- the experiment was canied out. At each relaive iu-
amide-6 readily absorbs water, and so does polyimide midity value, at least three lifetime measurements
though to a lesser extent. PET absorbs little water were made, the average of which is presented in the
compared with these two polymers. In Table 1, water Figure. The low dispersion of the lifetimes obtained
absorption ( A m method D570) of the three poly- ensures the accuracy of the data (4). In Fig. 4 two main
mers is shown. features must be noted. First, in all three polymers
there is a relative humidity value after which lifetime
Test Conditions
quickly decreases and achieves a value that does not
The three polymers were submitted to PD until change much with a further increase of humidity. We
breakdown at relative humidities ranging from lo?!to call this value the critical relative humidity value
83%. In order to control the relative humidity, the cell (RHJ. Second, RH, depends on the ability of the poly-
shown in Fig. 1 was housed in a PMMA closed box mer to absorb water, being lower as the hydrophylicity
(about 4 liter volume) in which a constant flow of fkesh of the polymer increases: it is about 75% for PET, 52%
air with controlled relative humidity was circulated. for PI, and around 17% for PA. As shown in Table 1,
The relative humidity was controlled via a demineral- the polyamide tested absorbs three times more water
ized water bubbler system suppIied with dry com- than the polyimide and over twenty times more than
pressed air. Before the test began, samples were held PET.
at the required humidity for about half an hour. Figure 5 shows the time to breakdown of the sam-
ples tested in this work as a function of the time
RESULTS derivative of the average charge of the positive dis-
Figure 4 shows the lifetime of the tested samples as charges. The time derivative of the average charge was
a function of the relative humidity (RH)under which obtained for each of the samples from plots shown in
Fig. 3. From Fig. 5 it can be seen that all these sam-
Table 1. Water Absorption for PET, PI, and PA ples’ lifetimes follow Equation 1: as the time derivative
(ASTM Method D570). of the average charge increases, the lifetime dimin-
Polymer Water Absorption (Weight %) ishes as described by Eq 1. The implication is that any
changes in lifetime induced by variations in ambient
PA humidity are due to a modification in the time deriv-
PI
PET
ative of the average charge [or of the external energy) of
the discharges according to Eq 1. Now, as defined

122 POLYMER ENGINEERING AND SCIENCE, JANUARY l#S8, Vol. 38, No. 1
 Effect of Humidity on Time to Breakdown

.
5 m I I I I

RHC

n
4 m
m
v

0
.cI,
0,:.

,B loo00
0
0

I
REIC
...""'A
%..
I
KMXAMDE6
*-.._
......I....
..
..........

........................&..................
I I I I

20 40 60 80 100

RH (Yo)
m. 4. Time to breakdown as afunction of relatiue humidity for PI m), PET (O), and PA (A). Error bars represent minimum and
maximum values.

previously, the average charge is I = O;NiQi/Q and between the f ilm and the sphere electrode, and a di-
the external energy W = ZlNiQi31.Therefore, changes minishing of their amplitude. In addition, it is not
in ambient humidity alter the PD process in a way that possible to find clear differences between the patterns
is reflected in the time derivative of these products, of the three polymers if compared all under or over
and thus in lifetime. RH,. It should be noted that the three patterns pre-
The two common quantities involved in the prod- sented for PA, PI, and PET under RH, have been taken
ucts are Ni (number of discharges) and Qi (amplitude of at RH that are, respectively, about lo%, 30%, and
the discharges); their evolution with RH appears in 60%, while those taken over RH, correspond to hu-
Fig. 6. In this Figure the PD patterns for PA, PI, and midity values about 20%. 55%, and 75%. respectively.
PET before and after RH, are shown. In the case of PA Thus, it seems that the characteristics of the PD phe-
the PD patterns correspond to RH values very near to nomenon do not depend on the tested polymer or on
RH,,for it is not possible to obtain patterns in PA for the relative humidity value in itself.
high FW values owing to the very short duration of the
tests. For the three polymers, PD patterns before RH,
DISCUSSION
show features similar to those over RH,:at RJ3 values
under RH,, the number N of the discharges is compar- The results contained in Figs. 3 to 6 can be sum-
atively small and the amplitude (QJlarge. On the con- marized as follows: i) Over RH,a n abrupt decrease in
trary, at RH values over RH,, patterns change, dis- lifetime is observed, which is related to a sudden
charge repetition rate is much higher and amplitude change in the PD features, and ii) RH, decreases with
smaller. In the case of PI, for example, an increase in the increasing ability of the polymer to absorb water.
RH from a 48% to 52% induces a n increase in the The reason for this behavior must depend on the
number of discharges from about 5000 positive or effect that the water absorbed on the polymer's sur-
negative discharges during the acquisition time (40s) face has on the PD phenomenon. As humidity in-
to about 40,000 in the same period. In fact, the num- creases, the quantity of water molecules adsorbed on
ber of discharges increases about 8 times, while the the surface also increases, more quickly the more hy-
amplitude of the discharges diminishes about a drophilic the substrate. These molecules are strongly
fourth. bound to the surface because of the interaction with
Small changes in humidity around RH, produce a it, especially when the polymer contains strongly polar
drastic increase in the number of PD taking place groups. On the other hand, after a certain number of

POLYMER ENGINEERING AND SCIENCE, JANUARY 1998, Vol. 38, No. 1 123
 F. Guastavino and P. Tiemblo

dldt
Flg. 5. Time to breakdown as afundion of the time deriuatiue of thepositive discharges m a g e chargefor the tested samples.

water layers are completed, mobility is attained. In the value RH, is the value of the relative humidity at
fact, proton NMR studies (11) on linewidths have which the surface of the polymer becomes conductive
shown that there is a transition at a certain moisture enough to evacuate the charges deposited on it. Obvi-
level at which the adsorbed water molecules change ously the higher the water absorption of the polymer,
their nature from layer absorbed water to less inter- the sooner this value is attained. It is also well known
acting water, when the number of layers, n, is over a that adsorbed polar layers encourage neutralization of
certain The effect of these mobile water layers on surface charge (12);as regards our results, this may
surface conductivity has been studied (12-16),espe- also happen, for the result is the same, i.e., scarce or
cially for inorganic solids such as quartz. There is a no field shielding.
remarkable increase in surface conductivity after the On the other hand, at the voltages used in this kind
water layer n,- is completed, owing to the mobility of of test, and taking into account the nature of the
the charged particles present on the surface (15, 16). species in the discharge, it is not surprising that the
The implication of this surface conductivity in- chemical or thermal stability of the polymer is of little
crease in the studied phenomenon is the following: importance, for degradation is liable to occur when-
when charges are deposited on the surface after a ever these species interact with the f ilm surface.
discharge they are quickly evacuated from the attack
area (area of the film under the sphere electrode) if the
CONCLUSIONS
surface is conductive, owing to the tangential electric
field on the polymer's surface. This means that the The degradation models presented in our previous
electric field between the electrodes is not greatly paper (10) allowed quantitative prediction on time to
changed and discharge recurrence will be high. On the breakdown of any polymeric sample provided that the
contrary, if the surface is not conductive, charges do slope of the variation with degradation time of average
not have the possibility of being evacuated, and there- charge and external energy could be obtained. The
fore they remain on the attack area. This charge ex- results presented here allow further understanding of
cess of opposite polarity deposited on the film's sur- the nature of the phenomenon, as they have shown
face shields the electric field between the electrodes, that the behavior of the tested polymers when submit-
which therefore diminishes. The diminishing of the ted to PD at varying humidity conditions depends on
electric field in the gap naturally induces a decrease in the characteristics of the surface as regards water
the number of discharges, and ultimately a decrease absorption and surface conductivity. In addition,
in the variation with degradation time of the average some practical conclusions can be drawn from these
charge of the PD, which, as explained before, is re- results. It is important to keep in mind that there is a
sponsible for the duration of the material. Therefore critical value of relative humidity at which the electri-

124 POLYMER ENGINEERING AND SCIENCE, JANUARY 1998, Vol. 38,No. 1

 Effect of Humidity on Time to Breakdown

Amplitude (pC)

Phase
Ffg. 6. PD Patterns before and ajter RHcfor PET, PI, and PA. The number of dischargesN is shown in di@erent shades of gray: black
N > 512, dark gray 512 > N > 64, medium gray 8 < N < 32, and hghtgray < 5.

cal performance of the material subjected to partial REFERENCES

discharges changes. Even if under normal use the
1. J. H. Mason, E E E ?fans. Electr. Ins., 13,21 1 (1978).
service conditions can strongly affect the absorbed 2. R. Bartnikas and E. J. McMahon, eds.. Engineering Di-
humidity, it appears of interest to add the value of RH,, electrics, Volume 1 , Corona Measurement and Interpre-
obtained with the described method, to the properties tation, p. 669, ASTM Special Technical Publication
characterizing the polymers currently used in electri- (1979).
cal applications. 3. G. Coletti and F. Guastavino, Nordic Insulation Sympo-
sium, NORD-IS 92, 5.2:1,Vasteras (1992).
ACKNOWLEDGIYIENTS 4. R.Bozzo, L. Centurioni. and F. Guastavino, IEEE Trans.
Ekctr. Ins., 28. 1050 (1993).
We would like to acknowledge financial support 5. G. Coletti, B. Fruth. F. Guastavino, G. Guerra, Roc. 4th
from European ARPEGE Project. ICSD, 160, Sestri Levante, Italy (1992).

POLYMER ENGINEERING AND SCIENCE, JANUARY 1998, Yo/.38, No. 1 125

 F. Guastavino and P. Tiernblo

6. P. Dejean. N. Foulon, M. Goldman, and H. Dejean. Roc. 13. N. Inagald, S.Tasaka. and K. Hibi, J. P o l y n Sci: Part A:
4 t h International Con$ Conduction and Breakdown in P o l y n C h e m , 30, 1425 (1992).
SofidDiekctrics, 181 (1992). 14.W. J. Feast, Po&~mersurfaces, D. T. Clark, ed., Chapt. 4,
7. Y. 1. Mitchenko, V. A. Fenin. and S. Chegolya. Polymer "The movement of electric charge along polymer sur-
Science USSR, 406 (1989). faces," 65,John Wiley & Sons (1978).
8. J. J. Xu and S. A. Boggs. Trends in Pdymer Science, 9. 15. Y. Awakuni and J. H. Calderwood, J. Phys. D: Appl.
234 (1995). Phys., 8 , 1038 (1972).
9. N. Izeki, ETG Conference, Long term pe$ormance of ?fV 16.G. Sawa, J. H. Calderwood, J. Phys. C: Solid St. Phys.,
isulating systems,Berlin (1985). 4. 2313 (1971).
10. F. Guastavino and P. Tiemblo, IEEE Trans. on Elect. 17. F. Guastavino and G. Guerra, Roc. 4 t h International
Ins., 4,189 (1997). Con$ Conduction and Breakdown in Solid Dielectrics,
11. M. Miyagawa, K. Kohata, A. Takaoka, and H. Kawai, 193.Sestxi Levante, Italy (1992).
Sen'i Gakkaishi, 43, 57 (1987).
12.T. J. Lewis, hoc. 5thInterr~Symp. Electrets. 75,Heidel- Received Nov. 26, 1996
berg (1985). Revised March 1997

126 POLYMER ENGINEERING AND SCIENCE, JANUARY 1998, Vol. 38,No. 1