Role of Semiconducting Compounds in Water

Treeing of XLPE Cable Insulation

S.A. Boggs and M.S. Mashikian, Electrical Insulation Research Center, University of Connecticut

Introduction remains dry. This facilitates a direct comparison of the

semicon-dielectric interface when maintained under
Water trees are known to be associated with reduc-
wet and dry conditions. Ageing was carried out over
tion of dielectric strength and eventual failure of poly-
long periods of time (typically 3500 and 7000 hours) at
ethylene-insulated cable which does not incorporate an
moderate stress (3.4 kV/mm or 85 V/mil and 2.6
effective moisture barrier [1,2]. However, past studies
kV/mm or 65 V/mil). Direct comparison of trees
of this phenomenon often correlated poorly with data
grown in this cell configuration and at these stresses
gathered from field-aged cable [3,4]. In particular,
with trees taken from field-aged cables indicated very
chemical analysis of water treed regions of laboratory-
similar tree chemical characteristics.
aged polyethylene demonstrated substantial differ-
ences from field-aged cable. Such results cast doubt on Experiments were carried out using four commercial
the usefulness of typical laboratory studies. semiconducting insulation shield materials (C1 to C4)
and four ‘‘model’’ shield materials (M1 to M4) formu-
A 1983 survey of solid dielectric cable technology
lated to vary such properties as the amount of sulfur in
conducted on behalf of EPRI indicated that several ca-
the shield. The dielectric used in all cases was a com-
ble manufacturers in Europe and Japan placed great
mercial cross-linkable polyethylene resin in the form of
emphasis on the properties of the semiconducting
pellets which had been subjected to 100% optical in-
shields used to manufacture power distribution and
spection.
transmission cables. The carbon blacks used in North
America differed substantially from those used by
these offshore manufacturers. As a result of this find-
Raw Material Impurities
ing, the Department of Energy initiated a project titled Analysis of the cell water showed that in all cases, the
‘‘Interfacial Ageing Phenomena in Power Cable Insula- ion concentration of the distilled, deionized water in-
tion Systems.’’ The findings of this project have been creased during ageing, and after as little as 100 hours of
reported in several technical publications [4-7] which ageing, many ion species could be identified which
form the basis for the present summary. were not initially in the water. This effect was inde-

Experimental Test Program

Table I [4]
The test program was based on a sample configura-
tion which incorporates the interface between the semi-
conducting compound and the polyethylene dielectric. Contaminant Content of Raw Material (weight %)
The sample cell used in this project (Fig. 1) is a modifi-
cation of a sample configuration originated by Union Sample Surface Bulk Cleaned Surface
Carbide. This cell provides a well which allows one C1 .0513 .0746 .0087
semiconducting layer to be maintained in a controlled,
wet environment while the other semiconducting layer C2 .0306 .1038 .0096
C3 .0994 .1072 .0100
C4 .3992 .2970 .0110
C5 .2484 .0286 .0010
M1 .0818 .0355 .0090
M2 .0180 .0860 .0013
M3 .3892 .1430 .0101
M4 .0494 .0581 .0091
Fig. 1. Test cell design which provides wet and dry semicon-dielec-
tric interfaces [4].
Fig. 2. Contaminants on the surface of Compound C5 (a pellet of Fig. 3. Detail of the interface between two pellets after melting. The
polyethylene resin). ‘‘defect’’ in the interface caused by contamination on the surface of
the pellets is obvious.

pendent of electrical stress. Various analytical tech- Results

niques were used to evaluate contaminants in and on
Fig. 4 shows vented water tree length vs. ageing time
the raw material pellets. The results of this analysis are
for an ‘‘old technology’’ commercial semiconductive
shown in Table I. In addition, the pellets were cleaned
shielding compound while Fig. 5 shows a similar graph
through ‘‘bathing’’ in hot, distilled, deionized water.
for a model shielding compound formulated with
As seen in Table I, the surface contamination gener- acetylene black. Note the difference in vertical scales.
ally differs from the bulk contamination. As well, the Obviously the cleaner model compound results in the
surface contamination can be reduced substantially growth of substantially fewer and smaller water trees.
through cleaning in hot, distilled water, which indicates Evidently, tree growth is saturated by 7000 hours. As
that the contaminants are water soluble. The surface the tree growth was thought to be related to the supply
contamination is believed to come from processing of ions, tree growth might be limited as the supply of
water used during the pelletizing process. As seen in ions is depleted. While this could easily occur in the test
Fig. 2, the surface contamination is substantial and as geometry used during this project, it is less likely to oc-
shown in Fig. 3, it can interfere with the quality of the cur in the field, as installed cables are subjected to
pellet-to-pellet interface. This effect may be much less ground water containing an ample supply of ions and
pronounced in extruded cable; however, the ionic con- which can pick up more ions from the cable conductor
taminants must still end up dispersed throughout the which is not included in this study. To test this hy-
dielectric. pothesis, water containing ions extracted from ‘‘soil’’

Fig. 4. Vented water tree length distribution for several ageing

times. The semiconducting compound was an ‘‘old technology’’ Fig. 5. Vented water tree length distribution for several ageing
commercial shielding compound. The cell was aged at 3.4 kV/mm times. The semiconducting compound was formulated with acety-
(85 V/mil) [7]. lene black, and the cell was aged at 3.4 kV/mm (85 V/mil) [7].

2
 Fig. 7. Comparison of vented water tree length distributions for
Fig. 6. The importance of ions for tree initiation is demonstrated cells fabricated with cleaned and as-received ‘‘old technology’’ com-
through comparison of the number of trees vs. time for cells filled mercial shielding compound [7].
with deionized, distilled water and cells filled with water contain-
ing ions extracted from soil and semiconducting shield material. In
the case of distilled, deionized water, the ions from the sample material after cleaning to remove water-extractable
shield become depleted after about 7000 hours and tree initiation contamination. The effect of cleaning is dramatic.
stops. Tree initiation continues in the other samples, which have an
ample supply of ions. Ageing was carried out at 3.4 kV/mm (85 Previous literature indicates that surfactants (deter-
V/mil) [7].
gent-like materials which reduce surface tension) in-
crease tree initiation and growth. However, the surfac-
and from semiconducting compound was added to a
tants used in previous studies were ionic. Fig. 8 shows
cell after 7000 hours of ageing. As seen in Fig. 6, the
a comparison of tree growth for cells in which the semi-
number of vented water trees increased at a fairly
con-dielectric surface was treated with an ionic surfac-
steady rate for the samples which had an adequate sup-
tant, a nonionic surfactant, and no surfactant. As is
ply of ions, whereas the number of trees leveled off for
clear from the figure, the effect on water treeing is the
the sample which did not. Likewise, the effect of water-
result of added ions, not the added surfactant.
extractable contaminants on tree growth could be dem-
onstrated. Fig. 7 shows a comparison of the vented During the extensive tree counting required to as-
water tree length distribution for cells manufactured semble the data presented in the above figures, a num-
with as-received ‘‘old technology’’ semiconducting ber of effects were noted. For example, asperities at the
shield and cells made with the same semiconducting semicon-dielectric interface did not generally result in

Fig. 9. Silicon ion profile as a function of position from the semi-

conducting compound into the dielectric. The transition is sharp for
Fig. 8. Influence of ionic and nonionic surfactants on the number
the dry electrode but is smeared out by ion migration from the wet
of vented water trees after 7000 hours of ageing at 3.4 kV/mm (85
semicon to the wet dielectric. The sample has been aged for 7000
V/mil). Evidently, water treeing is stimulated by ions rather than
hours at 2.6 kV/mm (65 V/mil) [4].
surfactants [7].

3
Fig. 10. Silicon ion profile as a function of position from a second
semiconducting compound into the dielectric. The transition is Fig. 11. Change in chemical potential with change in ion concentra-
sharp for the dry electrode but is smeared out by ion migration from tion as calculated by Zeller for two water-filled cavities connected
the wet semicon to the wet dielectric. The sample has been aged for by a thin channel [9]. A change in chemical potential over about 1
7000 hours at 2.6 kV/mm (65 V/mil) [4]. eV is sufficient to ‘‘drive’’ an electrochemical process such as water
treeing.
the growth of water trees unless they were accompa- 3. The growth of water trees does not depend solely on
nied by some form of contamination. Thus such asperi- ions originating from the semiconducting shields.
ties appear to have much less influence on water tree Ground water will generally carry substantial ionic
initiation than does contamination. contamination.
Clearly, the above results suggest that ions are mi- 4. The tree length and size distribution differs signifi-
grating from the semiconducting material into the di- cantly for the semiconducting compounds tested,
electric and water tree. Several studies within the pro- with the newer technology products which incorpo-
ject demonstrated beyond any doubt that this can occur. rate the cleanest carbon black showing a trend to-
For example, Figs. 9 and 10 show the profile of silicon ward reduced propensity for the growth of vented
(ion) concentration from within the semicon to within water trees.
the dielectric for the wet and dry semicon-dielectric in-
terfaces of the same samples. Clearly the transition is 5. While the propensity toward treeing appears to de-
very abrupt on the dry side but has become smeared as pend on the degree of surface and volume contami-
a result of ion mobility on the wet side. Virtually no nation of the shield material pellets, the role of each
vented water trees grew at the dry semicon-dielectric impurity constituent has yet to be determined.
interface, while many vented water trees grew on the 6. Surface contamination of the insulation pellets re-
wet interface. mains segregated at the interpellet interfaces and
stimulates the initiation of bowtie trees.
Conclusions
1. As noted above, detailed comparisons of the chemis- Technology Transfer
try of trees grown during this project with trees from Various aspects of the technology developed during
field-aged cable demonstrated excellent correlation this Department of Energy-supported work have been
in all parameters which could be measured. Thus the transferred to Cablec Polymers, Inc., Dow Corning, Inc.,
test protocol employed in this study appears to Union Carbide, Elastimold Corporation, Cabot Corpo-
mimic field ageing accurately. ration, Uniroyal, AT&T, Exxon, and BP Polymers. This
research has resulted in substantial changes to semicon-
2. As a result of the extensive testing carried out during
ducting shield technology and compounds which have
this project, we can conclude that the initiation of
improved the reliability of solid dielectric cable.
water trees at a semicon-polyethylene interface re-
quires ions and that the required ions are generally
present in and on the semiconducting material. The
Theoretical Context (Speculation)
concentration of water soluble ions on the surface of A full theory for water treeing is still lacking. How-
semiconducting material can be reduced through ever, the outlines of a theory are available. H.R. Zeller
simple water extraction, and such reduction has a has published two seminal papers in the field [8,9]. One
substantial effect on the number and length of water paper outlines the basis for condensation of water
trees. within a polymer. Using basic thermodynamics, Zeller
points out that water dispersed in a hydrophobic mate-
rial which contains hydrophilic regions will condense

4
into those hydrophilic regions [8]. In the second paper, Matthew Mashikian was graduated from the Ameri-
Zeller provides a credible basis for the ‘‘driving func- can University of Beirut and received his Dr. of Engi-
tion’’ of water treeing [9]. He points out that the change neering degree from the University of Detroit. Matt
in chemical potential with ion concentration is suffi- worked for ASEA from 1958 to 1962 as an application
cient to drive water treeing. Fig. 11 shows the change engineer for lightning arresters. From 1963 to 1979, he
in chemical potential (in electron volts) which accompa- was in the Engineering Research Department of Detroit
nies the change in ion concentration from zero to the Edison where he rose to the position of Supervisor of
stated value for two small cavities connected by a small Electrical Equipment and instrumentation. After early
channel. Anything over 1 eV can drive a chemical reac- retirement from this position in 1979, he started his own
tion. Thus this mechanism has the potential to drive consulting company, Mashikian & Associates, Inc.
Since 1983, he has been Director of the Electrical Insula-
water treeing over a wide range of concentration.
tion Research Center at the University of Connecticut.
However, the driving function drops at sufficiently
Matt was elected a Fellow of the IEEE for his contribu-
high and low concentrations. These computed data can
tions to the technology of solid dielectric power cables.
be taken as qualitative and suggestive; they are not de- Matt is presently Secretary of the PES Insulated Con-
finitive computations of the ‘‘driving force’’ for a realis- ductors Committee and Chairperson of the DEIS Edu-
tic geometry. Recently published experimental data in- cation Committee.
dicate that water trees consist of cavities connected by
‘‘channels’’ of altered material [10]. References
Taken together, these results suggest that water tree- 1. Lawson, J.H. and W.A. Thue. ‘‘Summary of Service Fail-
ing results from a localized chemical reaction which ures of High Voltage Extruded Dielectric Cables in the
changes the polymer from hydrophobic to hydrophilic. United States’’. 1980 IEEE International Symposium on
Water and ions can travel along and condense into Electrical Insulation, pp. 100-104.
these hydrophilic paths from cavity to cavity. The com- 2. Lawson, J.H. and W. Vahlstrom. ‘‘Investigation of Insula-
bination of water, ions, and (possibly) cavities results in tion Deterioration in 15 kV and 22 kV Polyethylene Cables
a chemical reaction at the channel tip which converts Removed from Service - Part II’’. IEEE Trans. PAS-92, pp.
material from hydrophobic to hydrophilic and propa- 824-835 (1973).
gates the water tree as a myriad of such channels. In- 3. Patsch, R. ‘‘Water Treeing in Cable Insulation - Are Labo-
itially, the channels are probably not tunnels but rather ratory Tests Meaningful?’’ 1988 IEE Conference on Solid
paths of polymer in which the material has been reacted Dielectrics, pp. 242-245.
to change from hydrophobic to hydrophilic. Unfortu- 4. Mashikian, M.S., J.H. Groeger, S. Dale, and E. Ildstadt.
nately, the channels tend to be from factions of a mi- ‘‘Role of Semiconducting Compounds in the Premature
crometre to a micrometre in diameter, so analysis of the Ageing of XLPE Cable Insulation’’. 1988 IEEE International
Symposium on Electrical Insulation, pp. 314-320, IEEE Pub-
material in a channel to determine the precise chemical
lication 88CH2594-0.
reactions therein poses a substantial challenge which
has yet to be overcome. 5. Groeger, J.H., J. Henry, and A. Garton. ‘‘Location and Con-
centration of Ionic Impurities in Polymeric Cable Insula-
tion’’. 1988 IEEE International Symposium on Electrical In-
Acknowledgements sulation, pp. 300-305, IEEE Publication 88CH2594-0.
The experimental research described in this article 6. Mashikian, M.S. and J.H. Groeger. ‘‘Ionic Impurities in Ex-
was carried out by Matthew Mashikian, Joseph truded Cable Insulation: Analytical Detection Techniques,
Groeger, and others at the Electrical Insulation Re- Sources, Nature, and Effects’’. Proceedings of the Jicable
search Center (EIRC), University of Connecticut. The Conference, Versailles, 1987.
article was written by Steven Boggs, who recently 7. Mashikian, M.S. and J.H. Groeger. ‘‘Ageing Phenomena at
joined EIRC. Cable-Shield Interfaces Affecting the Propensity of the In-
sulation to Develop Water Trees’’. Proceedings of the Jica-
Steven Boggs was graduated from Reed College and ble Conference, 1991.
received his Ph.D. and M.B.A. degrees from the Univer-
8. Zeller, H.R. ‘‘Thermodynamics of Water Treeing’’. IEEE
sity of Toronto. He conducted research for 12 years Trans. EI-22, p. 677 - 681, 1987.
with Ontario Hydro, primarily in the areas of SF6 gas-
insulated substations, solid dielectrics, and partial dis- 9. Zeller, H.R. ‘‘Noninsulating Properties of Insulating Mate-
rials’’. 1991 Annual Report of the Conference on Electrical
charge detection. Prior to joining the Electrical Insula-
Insulation and Dielectric Phenomena. p. 19 - 47.
tion Research Center as Associate Director, he was Di-
rector of Engineering and Research at Underground 10. Moreau, E., C. Mayoux, C. Laurent, and A. Boudet. ‘‘The
Systems, Inc. (Armonk, NY). Steve was elected a Fel- Structural Characteristics of Water Trees in Power Cables
low of the IEEE for his contributions to gas-insulated and Laboratory Specimens’’. IEEE Trans. EI-28, No. 1, Feb-
ruary 1993. pp. 54-64.
substation technology.