I E E E Transactions on Electrical Insulation Vol. 25 N o .

5, October 1990 989

REVIEW

Water Treeing in Polyethylene Cables

E. F. S t e e n n i s
N. V. KEMA, Arnhem, The Netherlands

F. H. K r e u g e r
Technical University Delft, The Netherlands

ABSTRACT
This review discusses water tree growth in polyethylene cable
insulations. The characteristics of water trees, the effect of
aging parameters on water tree growth and the possible mech-
anisms of growth are considered, emphasizing vented tree de-
velopment in polyethylene insulating materials. Moreover, test
methods and measures to reduce water treeing are discussed.

1. INTRODUCTION a fan. It is generally accepted that water trees reduce the

electric breakdown stress level of a n insulating material.
1.1 HISTORY
Miyashita observed water trees in the stator windings
of submersible motors of which the wires were coated with
H E introduction of extruded cables in the second half
T of this century was accompanied by the idea that
water or water vapor would not affect the electrical prop-
P E . Subsequent publications of Tabata et al. [152] and
Vahlstrom [162] described the early experience with water
trees in low-density polyethylene (LDPE) and crosslinked
erties of the cable. Therefore, the cables were not provid-
polyethylene (XLPE) insulated cables. After 5 yr of ser-
ed with water or water-vapor impervious outer sheaths,
vice experience, these cables showed failures that could
often a PVC (polyvinyl chloride) outer sheath over the
be related to water treeing.
ground screen was chosen, while a metal sheath was in
most cases absent. Unlike cable practice in Japan and
Europe, in the United States extensive lengths of extrud- Continued publications of different authors describe
ed cables have been installed without any plastic outer many aspects of water treeing in polymeric insulating ma-
sheath. terials. Review-type publications are given in Table l. It
was found for instance, that water treeing, or more or less
The discovery of degradation of polyethylene ( P E ) by similar tree creation, could occur in all polyolefins. There
the combined action of water and electric stress was pub- was much discussion about the water tree susceptibility
lished by Miyashita and presented a t the Electrical Insu- of different types of insulating materials. Filled insulat-
lation Conference in Boston 1969 [ l o l l . Soon this kind ing materials such as ethylene propylene rubber (EPR)
of degradation was called ‘water treeing’ [152]. A water have the advantage of opacity, but these materials are
tree is defined as a diffuse structure in a dielectric insu- suspected as well, since close examination showed water
lating material with a n appearance resembling a bush or trees also in them.

0018-9307/90/1000-989$1.00 @ 1990 IEEE

990 Steennis et al.: Water Treeing

In 1984 Shaw e t al. [136] estimated the total cable LIST OF SYMBOLS
length harboring water trees to be about 300000 km.
A area m2
c concentration
1.2 PE INSULATION C, specific heat
c' osmolarity
D dielectric displacement
LDPE was often applied as an insulating material for E electric stress
extruded cables. Nowadays, most polyethylene extruded E , electric stress in the original unaffected
insulation
cables have XLPE insulating materials. Other types of
E, electric stress in the vented tree
P E are medium-density polyethylene (MDPE) and high- E.t axial electric stress in the tip of the vented
density polyethylene (HDPE) used for sheath production tree
in modern cable technology. EPt axial electric stress i n the polyethylene
adjacent t o the t i p of the vented tree
E p d radial electric stress in the tip of the v/m
Polyethylenes, (CHz-CHa), , are very long macromole- vented tree
cules. The CH2 groups are strongly joined by bonds of E,, Young's modulus Pa
the shared electron valence type. The ends of the differ- F,, force N
ent chains contain methyl (-CH3) or vinyl (-CH=CH2) h parameter of length, length of vented tree m
groups. k' thermal conductivity J/Ksm
unit-normal vector
The mechanical properties of the different polyethylenes pressure N/m2
are mainly determined by the density of these materials. heat dissipation W/m3
The density in turn is strongly related to the molecu- charge C
lar length and the number and length of side chains per parameter of place, radius m
macro molecule. This is illustrated in Table 2. gas constant 8.31 J/Kmole
time S

P E is a thermoplastic: the upper operational temper- temperature K

ature is limited to % 70°C. By crosslinking of the macro- Tg glass transition temperature K
molecules the operational temperature is increased to % T, melting temperature K
90 'C. In XLPE the macro molecules are incorporated in To temperature of t h e original unaffected K
insulation material
a network in which the effective molecular weight has be-
T, temperature in the vented tree
come infinite. A schematic representation of XLPE chains
e strain
is presented in Figure 1.
c p permittivity of the polyethylene
Table 1. E. permittivity of t h e vented tree
Review-type publications on water treeing. q5 angle
n relative permittivity
n. relative permittivity of the vented tree
path
year author number of reference number p density k/m3
references in this paper U mechanical stress Pa
U, stress in radial direction Pa
ut stress in tangential direction Pa
1977 Wiersma 32
22 ny yield strength Pa
1978 Eichhorn
U, ultimate strength Pa
1979 Kiss 38
1980 Nunes et al. 89 c, electrical conductivity of the vented tree (nm)-'
1983 Dissado et al. 43 II osmotic pressure N/m2
1984 Bernstein 46
227 The polymer, either crosslinked or not, is semi-crystal-
1984 Shaw et al.
147 line, which means that it is partly crystalline and partly
1989 Steennis
amorphous. The crystalline part shows a folded regime of
1990 Nicholls et al. 130
macromolecules organized into platelets with dimensions
strongly depending on the production process. Usually
IEEE Transactions on Electrical Insulation Vol. 25 No. 5, October 1990 QS1

Table 2. Platelets can be grouped to spherulites, varying in size

Relation between density and number and length from x 3 to x 700 pm. Indeed, etching procedures some-
of side chains for LDPE and HDPE (After [l?]). times revealed a kind of superstructure [lo, 30,97,105,
110, 112,117,163], although the results are suspect since
Polvethvlene LDPE HDPE most of the etching procedures produce artifacts which of-
~ _ _ _
ten are misinterpreted [lo]. Moreover, long macromolecu-
density [g/cm3] 0.91-0.94 0.95-0.965 les and the existence of many branches, as in LDPE, re-
duce or even suppress the superstructure. Examples of
average molecule length 1500-3500 7000-14000 complete suppression of superstructures in LDPE are giv-
en by Patsch et al. [120], Mandelkern et al. [88], Capaccio
number of side chains 20-40 < 5 et al. [28] and Ross et al. [126]. This is illustrated in Fig-
[/lo00chain atoms] * ure 3: an electron micrograph of LDPE cable insulation
is given with a magnification factor of 40000. Crystalline
length of side chains 2-5 < 4 and amorphous regions have been revealed, superstruc-
[number of atoms1 * tures have not been found.

* The mechanical behavior of P E is mainly determined

Occasionally there is a side chain with an
by the amorphous regions. In a thermoplastic such as PE,
average molecule length
the intermolecular Van der Waals forces play an impor-
tant role. Although Young’s modulus E,, is high, x 500
MPa a t room temperature, considerable chain flexibility
H H H H H H H H H H or micro-Brownian motion exists. The actual mechanical
I I I I I I I I I I
- c - c - c - c - c - C - c - c - c - c - behavior of P E as a function of temperature T depends
I I I I I I l l 1 on the values of the glass transition temperature Tq and
H H H H H H H H H
the melting temperature T,. A schematic presentation
H H H H H H H H H is given in Figure 4. Above T,, LDPE will melt if it is
I I I I I I I I I
- c - c - c - c - c -
I I I I I
c - cI - c I - c I- c -I not crosslinked, while XLPE behaves like a rubber-like
H H H H H H H H H H
substance.

Figure 1. On the basis of Figure 5, which shows a typical stress-

Molecular structure of XLPE, bond angles are not strain curve for P E , some important mechanical param-
indicated. eters are discussed. For small stresses, uniaxially applied
to the bulk polymer, the stress-strain relation shows a
Hookean behavior with e = E,,€, where E,, is the
Young’s modulus, U the applied stress and E the strain.
The strain is here defined as the length of the elongation
divided by the initial length.

In the linear region of the stress-strain curve instan-

taneous recovery is possible. The molecule chains are
partially uncoiled and coiled again. At the yield point
slippage of the chains will result in incomplete recovery.
10 nrn The stress at this yield point is called the yield strength
e,. Finally, fracture of the bulk polymer will be reached
a t the ultimate strength U,,. Some typical data for LDPE
and HDPE are given in Table 3.
Figure 2.
The PE structure in the nanometer range (After [12]). As all polymers, P E is rather susceptible to effects of
time. Important time effects are
the thickness is x 10 nm, they are generally several p m
wide and long. Between the platelets are interconnecting 1.creep: a t a constant stress the deformation or strain
chains and chain ends which form part of the amorphous slowly increases,
regions. The structure of the P E on this nm scale is shown 2.’relaxation: a t constant strain the required stress slowly
in Figure 2. decreases,
go2 Steennis et al.: Water Treeing

Table 3. Typical tensile properties for PE.

polymer LDPE HDPE

density’ [g/cm3] ~0.92 ~0.95

crystallinity’ [%I = 55 =90
Young’s modulus’ [MPa] 200-400 600-1500
Yield strength’ [MPa] 10-20 25-50
Ultimate strength‘ [MPa] 15-25 25-55
elonaation to fracture2 1%1 400-700 100-600

’) according to [17]
)’ according to (1291

Figure 3.
3. internal friction: a t dynamic loading mechanical ener- Electron micrograph of LDPE cable insulation.
gy is converted into heat. Crystalline and amorphous regions are visible.
Spherulites have not been found.
Time effects are of importance a t moderate or high stress
levels. At lower stress levels these effects are hardly found radicals produce acetophenone and methyl radicals. The
and therefore complete recovery is possible a t lower stress- methyl radicals in turn give methane and carbon radicals
es. in the polymer chain by removing hydrogen in a limited
number of places. It is also possible that the free radi-
Stress cracking or brittle fracture is a n important phe- cals produce cumyl alcohol and radicals in the polymer
nomenon that may lead to complete degradation of the chains directly. Finally, the carbon radicals of (different)
polymer. It starts as a localized phenomenon a t a lo- chains combine to crosslinks [3]. In some production lines
cation where a critical stress is present. Stress cracking crosslinking occurs with the aid of steam a t temperatures
occurs in the brittle region at low temperatures and/or between 200 and 220°C and a pressure of 1.6 to 2 MPa
under alternating stresses during long times intervals. [120,157]. This process is called steam curing. Towards
the end of the process the cable insulation enters the cool-
Fracture is not found under repeated loading conditions ing section where a fairly rapid temperature reduction is
as long as the imposed stress remains below a certain achieved.
endurance limit. This limit is about 1/5 of the static
ultimate strength. Today most cables are not cured with steam, but by
using hot nitrogen. This suppresses the creation of mi-
A liquid which is capable of dissolving or swelling the crovoids considerably, as will be shown. Such a process
polymer promotes cracking. Measures to reduce stress is called dry-curing. Cooling is performed by using gas
cracking are e.g. or water. Therefore, this process is subdivided into ‘dry-
cured dry-cooled’ and ‘dry-cured wet-cooled’. The meth-
1. increasing the molecular weight (fewer chain ends and od of cooling has a minor influence on the creation of
therefore fewer micro-cracks) microvoids.
2. the use of copolymers (the mechanical properties are
improved by the combined effects of the various poly-. Two of the three major chemical residues of the dicumyl
mers). peroxide-curing process are cumyl alcohol and acetophe-
none. These products will eventually diffuse out of the
1.2.1 CROSSLINKING insulation. The rate of diffusion depends on the tempera-
ture, the temperature gradient and barriers (e.g. sheaths).

There are several methods of crosslinking. Often a An essentially different method of crosslinking is silane-
dicumyl peroxide-curing agent is used. Here the dicumyl curing. In this method, curing does not take place direct-
peroxide, added to the polymer compound, is activated ly after extrusion, but in a separate production step. In
immediately after extrusion in a special tube at high tem- the one-shot silane-curing process a silane compound is
perature and high pressure. By thermal energy the per- grafted onto the PE chains during the extrusion. Af-
oxyde is split into two chemical free radicals. These free ter extrusion the cable is slowly cooled in a water-filled
IEEE Transactions on Electrical Insulation Vol. 25 No. 5, October 1990 993

cooling trough. Curing takes place afterwards by putting be collected in the amorphous regions of the polymer. If
the extruded cable on a reel in a water tank at 85'C. these substances are polar, clustering may take place in
The immersion time for medium-voltage cables ranges be- existing voids. Sometimes, for instance if supersatura-
tween one and five days, depending on the insulation tion occurs, cavities will be created. Smaller voids and
thickness. In this tank the silane groups are coupled cavities have dimensions comparable to the dimensions of
chemically under the influence of the water vapor in the the interlamellar regions, in most of the XLPE insulating
insulation. Methanol is also found as a residue. The materials the largest voids and cavities have diameters
amount of residual products is much smaller than the of x 10 pm. In exceptional cases, however, much larger
amount of these products in the peroxide-curing process- voids are found, with diameters of M 500 pm.
es.
n 10r
a The creation of most of these microvoids is attribut-
v
ed to water vapor or other gases which, during the rapid
k cooling of the melt, are prevented from diffusing out of the
8- insulation. As a result, supersaturation, in particular of
Is,
- water during steam-curing, is inevitable and microvoids,
filled with gases and/or water, will be created. The ap-
6- proximate sizes and densities of the cavities/voids have
been measured for typical production processes and are
T9
given in Table 4 [3,4,57,68]. In this Table the effect of
4 I I
, the method of crosslinking on the presence of microvoids/

. '0"'- .'
the insulation during crosslinking, in combination with
A schematic modulus/temperature curve for
the cooling procedure afterwards.
LDPE and XLPE (after [1331).
40
Supersaturation and void creation will usually occur
in the center region of the insulation. The inner and
outer regions will cool down and solidify first, thereby
suppressing further diffusion of gases out of the center
region of the insulation. Such void concentrations are
usually visible as halos in the insulation of steam-cured
20 HDPE
cables.

LDPE

'"oLv
1.2.3 WATER IN PE

012 Oi4 0:s 018 1:O

strain
1(2 1.14
6 (m/rn)
' Owing to its non-polarity, PE surfaces are hydrophobic
until oxidized. Some water in the P E , however, cannot
Figure 5. be excluded. Polar substances in the insulating material
Typical stress-strain curve for PE (After [77]). are capable of attracting water. It is known [65,148] that
XLPE at a temperature of 20'C absorbs < 100 ppm of
1.2.2 VOIDS AND CAVITIES water. The amount of water absorbed depends strongly
on the temperature, as shown in Figure 6. As was stat-
ed in Section 1.2.2, supersaturation temporarily results
As defined here, voids are filled with gases only, and in high amounts of water, mainly collected in already ex-
cavities are filled with liquids and/or solid materials. isting microvoids, or in newly produced microcavities. It
WAS found in several cases that most of this water has
It is generally assumed that during production, water, diffused out of the insulation after several years of service
impurities and residual products from crosslinking will aging [148].

P
994 Steennis et al.: Water Treeing

Table 4. Concentration, size a n d volume of voids in PE as a function of t h e curing process.

curing method void concentration maximum void size void volume/

PE volume

number/mm3 Pm %

stearn* 105-1o6 30 0.7-7

dry* o4
10~4.1 15 0.OO 7-0.4
silane (one-shot)* =io4 15 ~0.07
uncured** =io5 13 =0.005

* After [19], as observed in 10 cm3 insulationvolume chosen at random, the void volume-PE volume ratio
is based on a void diameter of 5 pm.
** After [57], the void volume-PE volume is based on a void diameter of 1 pm.

e
L
W e
4
0 e
3
e
102: e
e
e
e

1 0 " ' ' ' ' 50

I ' ' ' ' 100
I Figure 7.
I I 150
I '
Vented tree initiated from a void in t h e semicon-
temperature ( " C )
ducting screen; t h e void a n d void content is shown
Figure 6. in Figure 0 .
Saturated water content (in p p m ) for XLPE as a
function of temperature (in "C) (After [65]) considered as dangerous, since these trees can initiate in-
sulation failures.
2. MORPHOLOGY OF WATER
Trees can be made permanently visible using different
TREES dyes. A well known and generally accepted dyeing pro-
cedure is given by Larsen [86] and Shaw and Shaw [136].
2.1 INTRODUCTION
2.2 SHAPE
ATER trees are diffuse structures in polymer insu-
W lating materials resembling a bush or a fan, growing
in many different kinds of polymers under the action of
A distinction must be made between two different types
of.water trees [24,148]. These types are the 'bow-tie tree'
water (vapor) and a n electric field. Water trees weak- and the 'vented tree'. This distinction is based on the
en the insulation and especially large water trees can be location where these trees start growing: vented trees
IEEE Transactions on Electrical Insulation Vol. 25 N o . 5 , October 1990 995

the propagation rate of vented trees is lower than that

of bow-tie trees (in a later stage of growth the opposite
is true). Consequently, the study of vented trees is more
time consuming than the study of bow-tie trees. For these
reasons many publications contain information on bow-
tie trees only. In some other cases a distinction between
both types of water trees has not even been made: only
the description ‘water tree’ has been used. In this review
the generic description ‘water tree’ will be used only if
bow-tie trees as well as vented trees are under considera-
tion.

Figure 8.
Electron micrograph of the void in Figure 7, crys-
tallization of various species.

Figure 9.
Figure 10.
Typical vented tree grown from the semiconduct-
Vented tree grown from the graphite outer screen
ing inner screen into the insulation of a 10 kV
into the insulation of a 10 kV XLPE cable.
XLPE cable.

are initiated a t the insulation surfaces, bow-tie trees are 2.3 VENTED TREES
initiated in the insulation volume. Such a distinction is
important since both types show a completely different
growth behavior. The vented tree is defined as growing from the insulat-
ing material boundaries to the other side of the insulation,
This review is focused on vented trees because these predominantly in the direction of the electric stress.
trees are more dangerous under service aging conditions
than bow-tie trees. The origin of vented tree initiation in many cases is
difficult to find. However, it is sometimes mechanical
For several reasons the study of vented trees is more damage to the cable insulation. Scratching the insulation
difficult than that of bow-tie trees. The concentration for instance may initiate treeing. Another origin of vent-
of vented trees is often low compared to the concentra- ed tree initiation can be an irregularity in the semicon-
tion of bow-tie trees, and at the beginning of growth, ducting screen where it is in contact with the insulation.
QQ6 Steennis et al.: Water Treeing

In a n insulating material which is fairly water tree sus-

ceptible, vented trees can reach the other side of the in-
sulation (in 10 kV cables = 4 mm thick) in about 7 yr.
It has been found that this type of water tree has been
responsible for many cable failures. In Figure 12 a vented
tree is presented; here also part of a breakdown channel
is visible.

I1 00 pm

Figure 11.
v‘ented trees, bent near the insulati on outer sur-
frice. These trees were found in the insulation of
a service aged steam-cured 10 kV X LPE cable.
Figure 12.
An example of a vented tree, initiated a t the boundary Vented tree grown from the semiconducting inner
area of a void, located in the semiconducting inner screen screen into the insulation. A part of the small
against the insulation surface, is shown in Figures 7 and breakdown channel is visible. marked 1.
8 [148]. Crystallization of various species (containing Si,
S and Ca) did occur in the void. It is assumed that these 2.4 BOW-TIE TREES
species were dissolved in the water during aging. The
vented tree was grown in a 10 kV XLPE cable insula-
tion which has been aged under service conditions for 8 The other type of water tree is the bow-tie tree. Bow-
yr. Muller [106], among others, found that in most cas- tie trees are defined as initiating in the insulation volume.
es vented trees at the conductor side in the insulation These trees grow in opposite directions, along the elec-
of full-scale steamcured cables were initiated a t inhomo- tric field lines. Exceptions can be found, in some cases
geneities in the inner semiconducting layer. For the ini- the growth direction has been bent: Karakelle et al. [71]
tiation these examples show that apart from water and assumed that this is a consequence of frozen-in mechan-
electric stress also chemical species in the water (or semi- ical stresses. However, Karakelle could also show that
conducting screens) and/or electric stress enhancements this typical growth direction was apparently not affected
may play an important role. by the flow pattern that existed in the melt a t the time
crosslinking occurred.
When aged under moderate service conditions (< 4
kV/mm), a vented tree grown from the outside of the The initiation spots are often clearly visible using nor-
insulation is sometimes pencil-like. Trees grown from the mal optical microscopy. An example is given in Figure 14.
inside have branches that spread a little further, although It is generally assumed, and it has been proven in some
each distinct branch of a large vented trees is also pencil- cases, that these spots contain impurities [137]. Normal-
like. Typical vented trees are presented in Figures 9 and ly the growth of bow-tie trees is strongly reduced after a
10. In a few cases, vented trees, bent near the insulation certain time. The total length is restricted and therefore
surface, have been observed by the authors. An example this kind of water tree is seldom the origin of cable break-
is shown in Figure 11. down. There are indications that the length of bow-tie
 IEEE Transactions on Electrical Insulation Vol. 25 N o . 5 , O c t o b e r 1990 997

Figure 13.
Typical bow-tie trees grown in the insulation of a
10 kV XLPE cable.

trees is related to the size of the location containing the

impurities.

The transformation of a bow-tie tree, formed close to Figure 14.

the insulation surface, into a vented tree was once report- Bow-tie tree, total length 200 pm. This bow-tie
ed by Naybour [lll].A collection of typical bow-tie trees tree was initiated from one of t h e impurities in
in the XLPE insulation of a 10 kV cable is given in Fig- the insulation.
ure 13. A further enlargement of one of these bow-tie the supposed channels, penetrating throughout the insu-
trees is presented in Figure 14. lation, must be < 1 pm. Microscopic investigations of the
tip of a water tree did not even reveal open channels in
2.5 VOIDS AND CHANNELS IN the n m range [28].
WATER TREES
2.6 WATER CONTENT OF WATER
TREES
Apart from the already existing voids or cavities in the
PE, extra void/cavity creation occurs occasionally in a
water tree. Collections of such voids are found sometimes Water trees contain water. If this water is evaporated
in the trunk of vented trees. Near the tip of the vent- e.g. by heating, it is known that the tree will absorb water
ed tree such microvoids are rarely found [28]. If vented again if the insulation is exposed to water or water vapor
trees have been grown under rather extreme aging con- aft erwards.
ditions, for instance by applying high electric stresses of
- 30 kV/mm, some of these microvoids may become in-
terconnected by micro channels [28]. In a gas permeation
Meyer [99] measured the amount of water in vented
trees grown from water needles. The electric stress near
experiment, Cross et al. [32] showed that the diameter of the water needle tip was very high, a t least 60 kV/mm.
998 Steennis et al.: Water Treeing

Meyer found that near the tip of the water needles, the of the aging time. The largest trees observed are present-
vented tree contained 10% water (of the vented tree vol- ed and not the mean length of the trees, since the largest
ume). He assumed that the water had been collected trees correspond to electrical degradation of the insula-
primarily in the microvoids. Such voids are expected to tion. After 11300 h, breakdowns did occur in cable No. 1
be the result of the high initiation stresses applied. Mey- probably as a consequence of the large vented trees. This
er also found that a t a certain distance from the water cable was withdrawn from further aging.
needle, where the electric stress is much more moderate,
the amount of water was about 1 t o 2% (of the vented This accelerated aging test shows the typical behavior
tree volume). of water tree growth. After an initial rapid growth within
the first 3000 h, the length of bow-tie trees does not seem
Apart from clustering of water in voids, some water is to increase any more. The length of vented trees increases
probably molecularly dispersed in other parts of the tree. continuously for most cable insulating materials. After
Water molecules can be found a t any place where polar approximately 12000 h of aging, the lengths of both types
groups are attached to P E [125,126,127]. of trees correspond well to the general picture of tree
lengths in the insulation of service-aged cables after 6
= loor
?! *1
to 13 yr.

.-
f 3. CHARACTERISTICS OF
5 80
.-
Y 701
WATER TREES
3.1 T H E VENTED TREE: AN
* 2 INS U LATl N G M A T ER IA L

starting the discussion of dielectrical proper-

0
B EFORE
ties, definitions should be given of ‘vented tree’ and
‘vented tree path’. The ‘vented tree’ represents the total
10 - region within which the tree can be observed ( p m to mm
0 @! I
-*I - 2
I scale), using the dyeing procedure described in [86,136].
~~~~

0 3000 6000 12000 24000

time (hours) The ‘vented tree path’ or ‘path’ relates t o the attacked
= vented tree P E only (the nm scale). The ratio of the volume of the
- = bow-tie tree
vented tree paths t o the vented tree certainly is much
Figure 15. smaller than 1, but upper and lower limits cannot be
Maximum length of vented trees and bow-tie trees estimated in general. There are several indications that
for five different cables as a function of the aging both the vented tree and the vented tree paths can be
time. considered as an insulating material.

2.7 TYPICAL GROWTH BEHAVIOR Koo et al. [81] and Cross et al. [32] studied the dielectric
OF WATER TREES properties of a vented tree in a water needle experiment.
The change in capacitance between the water needle and
the opposite electrode was measured during the growth
Many publications show that vented trees and bow-tie of the vented tree a t a frequency of 1500 Hz. The capac-
trees have a completely different growth behavior. This is itance variations were derived from a voltage change in a
illustrated in [147]. Five different medium-voltage cables resistor placed in series with the electrode. Afterwards,
were aged for 24000 h in a water tank. The cables var- in a model, the vented tree was replaced by metal or by
ied in construction and in method of crosslinking. The dielectrics with different permittivities. The resulting ca-
electric stress level a t the outside of the insulation was pacitance variations showed that the dielectric behavior
3.9 kV/mm a t power frequency. The tap water in the of a real vented tree differs strongly from that of a con-
water tank was kept a t 30°C. The water contained small ductor. Moreover, it was found that comparable voltage
amounts of NaCl (0.2 kg/m3) and HC1 (pH = 6). The variations in this model could be obtained if the water
water was admitted under the sheath only. In Figure 15 tree was replaced by an actual material with a relative
the largest vented tree and the largest bow-tie tree as ob- permittivity K. M 6. To explain the measured permittivi-
served in the samples are given for each cable as a function ty, Cross assumed that water was collected in the many
IEEE Transactions on Electrical Insulation Vol. 25 N o . 5 , October 1990 909

microvoids distributed over the vented tree volume. Cross

pointed out that in such a situation the observed increase
in the permittivity of the vented tree volume might be ex-
plained by using the multiphase dielectric mixture theory
(e.g. Tinga et al. [159]).

Boggs et al. [18] studied the micro-movement of vent-

ed trees in a water needle experiment. The trees were
not grown in PE but in silicone rubber under the appli-
cation of an electric stress a t different frequencies. Tree
movement has been observed by applying interferomet- 5
ric holography. Boggs found that the movement of the
trees was related to E' and therefore probably related to
Maxwell stresses. He also concluded that certain out-of- 0
phase movements of the tree in relation to the movement 0 10 20 30 40 50 60 7 0 80 90 100
of the needle tip would provide evidence that the vented t r e e size (% of insulation thickness)
tree is not conductive. - mean value
-- 95 % confidence bound
Recently Ross et al. [126] measured the dielectric prop-
Figure 17
erties of a vented tree which had been grown under nor-
mal aging conditions in a full-scale XLPE insulated ca- Relation between the breakdown stress level and
the water tree size.
ble. The trees were saturated with water preceding to the
measurements. They found that the vented tree has the found that the vented trees did not provide a more conve-
properties of an insulating material with a permittivity nient path t o the electrical trees or the breakdown chan-
IE = 2.26 and a loss-factor of about ~ O X ~ Oat- 50 ~ Hz. nels than unaffected PE.
There are other observations confirming that a vented
tree is an insulating material 3.1.2 D I R E C T I O N OF V E N T E D T R E E

"4 axis of s y m m e t r y
4
P ROPAG A T I O N

Vented trees and the related paths grow in the direction

of the electric field lines. High radial electric stresses a t
the tip of the vented tree path would, however, cause
the vented tree to fan out. This can be observed a t the
tip of a water needle, where the direction of growth near
I the water electrode is perpendicular to the water needle
f y-axis surface. One of the many examples of this vented tree
0 1- growth behavior a t the needle tip is given by Filippini et
Figure 16. al. [46].
Electric stresses near the tip of a vented tree path.
In Steennis [148] various electric stresses near the tree
3.1.1 B R E A K D O W N C H A R A C T E R I S T I C S tip are calculated for different vented tree tip radii. The
electric stresses were calculated for a range of permittiv-
ities and conductivities of the vented tree path. In Fig-
If a vented tree path were a conductor, then such a ure 16 the assumed electric stresses and the vented tree
path, grown to the other side of the insulation, would path configuration are given. High radial electric stresses
initiate thermal breakdown or initiate an electrical tree of 30% of the axial electric stress a t the tip of the vent-
followed by breakdown. However, it was found that in ed tree path (Epd/Ept= 0.3) can be expected for vented
most cases such long vented trees do not cause breakdown tree paths with IC, 2 20, or ~ 2 . lo-*5 ~(am)-'. In such
a t stresses of 2 kV/mm or even higher (see Section 4.1 and a situation it is expected that the tree would fan out.
Figure 17). It is also shown that for IC, < 10 and cv < 1.3 x lo-'
(Gm)-' this ratio Epd/Ept is small. In that case vented
In an experiment by Densley [33] electrical trees were tree growth can be expected mainly in the direction of
initiated in vented trees by inserting a needle. It was the axial electric stress a t the tree tip. This conforms
1000 Steennis et al.: Water Treeing

with the phenomenology where it is found that the direc- Bamji et al. [ll]also examined regions with and with-
tion of propagation of vented trees is mainly determined out water trees using FTIR. Different kinds of test spec-
by the local electric field lines of the original unaffected imen were studied. Absorption was found a t 1160 (and
PE. For moderate permittivities and conductivities, the 600) cm-l. Bamji did not attribute these absorptions
electric stresses become independent of the length of the t o ether groups [55] but t o sulfate anions. The absorp-
vented tree path: the influence is confined to a local area. tions a t 1600 cm-' are consistent with the presence of
carboxylate anions. Also in this particular experiment
3.2 PHYSICAL/CHEMICAL the absence 6f carbonyl absorption is noticeable.
PROPERTIES (LOCAL)
In 1987, a publication from Garton et al. [56] described
the observation of traces of oxidation. Trees have been
3.2.1 INTRODUCTION taken from service-aged steam-cured XLPE insulated ca-
bles. The difference in the carbonyl concentrations a t
1720 cm-' was E 15% of the overall level of oxidation
This Section discusses chemical and physical observa-
in the insulation. The results have been confirmed by
tions inside a water tree compared to observations outside
oxidative stability tests where it was found that the tree-
the tree. Studies are mainly carried out on vented trees
affected areas were much less stable than the areas not
from needle tests, from scratched insulation slabs or from
affected by water trees.
full-scale cables.
Ross et al. [125] were also able to observe traces of ox-
3.2.2 INFRARED idation in the vented trees of an accelerated-aged XLPE
cable insulation. The cable was aged for 24000 h, the
mean electric stress applied a t 50 HZ during aging was 4
Infrared (IR) measurements have been carried out by kV/mm. Differences between the material inside and out-
several authors. Bernstein et al. [15] performed measure- side vented trees were found a t 1150, 1710 and 1720 cm-'
ments on model cables aged for 60 days at 4.6 kHz. After and assigned to hydroxyl groups and carbonyl groups, re-
drying the samples, metal ions from the solution were spec tively.
detected a t 1130 cm-'.
Recently, Ross et al. [127] emphasized that carboxylate
Garton et al. [55] tried to find evidence for oxidation groups, which have an absorption a t 1570 cm-', were
in a water tree with FTIR. The insulation of model cable found in the vented trees of full-scale service-aged XLPE
was studied, aged a t a moderate electric stress level a t cable insulations. These carboxylate groups can be re-
1 kHz for 320 h a t 70°C. Garton was not able t o find garded as end groups of the polymer interacting with wa-
a typical oxidation product such as carbonyl. However, ter. Such an interaction was also established in the 1000
he found absorptions at 600, 1100, 1160, 3550 and 3600 to 1200 cm-' range, probably as a consequence of ionic
cm-', assigned t o ether and alcohol groups. sulfur-oxygen groups.

Abdolall et al. [l] were able to distinguish differences 3.2.3 ENERGY DISPERSIVE
in the IR spectrum between affected (in water tree) and SPECTROMETRY
unaffected P E (outside water tree) in a range between 20
and 350 cm-' at a temperature of 4.2 K. Samples were
taken from a full-scale XLPE insulated cable. The dif- Ross [127] observed with energy dispersive spectrome-
ferences observed were attributed to various possible ef- try sulfur and in some cases also silicon in vented trees
fects such as hydrogen bonding, strain or inhomogeneous grown in full-scale service-aged cables.
broadening, breakage of polymer chains, or effects due t o
crosslinking residual products. 3.2.4 ELECTRON SPIN RESONANCE

Differences in the FTIR spectrum in a range between

500 and 2000 cm-' have been found by Yoshimitsu et al. Electron spin resonance was applied by Dorlanne et al.
[173]. These investigators studied XLPE material with [38] and Crichton et al. [30]. They studied LDPE samples
an imbedded copper wire aged for up to 10 days a t 1 from water needle experiments as well as XLPE samples
kHz. They found that in the tree-affected regions the CH2 from full-scale cables. All samples were aged for up to 90
groups lose their absorbance, but chemical species such days a t frequency levels of up to 4 kHz. Both investigators
as hydroxyl groups and carbonyl groups gain strength detected metal ions from the solution in the water trees,
compared with the undegraded area. even beyond the visible part of the water tree [30].
IEEE Transactions on Electrical Insulation Vol. 25 N o . 5 , October 1990 1001

3.2.5 DIFFERENTIAL SCANNING trees themselves had been completely dissolved. This ex-
CALORIMETRY periment shows that the chemical changes assumed above
do not result in crosslinking of the tree-affected material.
Differential scanning calorimetry was applied by Bamji
et al. [ll].Trees have been studied in samples taken from 3.2.10 STAINING
service-aged XLPE insulated cable. It was not possible
to find a difference in melting endotherms between re-
gions containing water trees and regions without these Abdolall et al. [l]colored (most probably) vented trees
trees. This indicates that heating during or after water from a full-scale XLPE insulated cable after drying of the
tree growth is of minor importance. insulation for 12 h a t 95'C. For the staining experiment
many different solvents were used. The revisibility re-
sults show that, with a few exceptions, only solvents with
3.2.6 X-RAY ANALYZES
an OH-group a t the end of a molecule made the trees
strongly visible again after a few days.
Sletbak et al. [137] detected metal ions and sulfur in
the branches of strongly colored bow-tie trees. These ele- 4. CHARACTERISTICS OF
ments could also be detected as a fraction of the impuri-
ties located a t the initiation site. The bow-tie trees were
THE TREE INFESTED CABLE
grown in a XLPE slab for up to 790 h a t a frequency of DIELECTRIC
50 Hz. Bamji et al. [ll] detected metal ions in a vent-
ed tree using X-ray techniques. The vented trees were
4.1 ELECTRICAL PROPERTIES
taken from a full-scale XLPE insulated cable, aged under
service conditions.
N this Section the electrical properties of the bulk insu-
3.2.7 dc ARC I
lation such as breakdown stress level, loss-factor, resis-
tivity and partial discharges, will be discussed. Studies in
this field have mainly been performed on LDPE or XLPE
dc Arc measurements, performed by Bamji et.al. [ll], insulating materials which have been aged under service
again show metal ions in vented trees. Moreover, Garton conditions in most cases for a few years. In general, wa-
et al. [56] clearly demonstrated the existence of apprecia- ter trees (bow-tie trees as well as vented trees) cause a
ble amounts of sodium, calcium, aluminum and silicon in reduction of the 50 HI, 0.1 Hz, dc or impulse breakdown
the tree-affected regions of the insulation of service-aged stress level. This has been observed by many workers [52,
XLPE insulated cables. 58,60,69,78,79,95,143,144,146,147,155]. In the studies
by Franke et al. [52] and Matsuura [95] et al. (1987) in-
3.2.8 OXYGEN PLASMA ETCHING dications are given of a relation between the size of the
water tree and the 50 Hz breakdown stress level.

A reduced oxidative stability of water trees in rela- In 1986 this relation was demonstrated [146]. More
tion to the unaffected surrounding P E was found by Ross than one hundred medium-voltage cable pieces have been
[127]. subjected to a 50 He breakdown test. Special techniques
allowed the energy dissipated in the breakdown channels
3.2.9 HEAT-TREATMENT to be reduced. In this way the cause of breakdown could
be established for several cases; examples are given in
Figures 12, 18, 19. When a vented or bow-tie tree was
Muller et al. [lo71 carried out some unusual experi- found, its size was measured. The relation between water
ments with LDPE slices in which vented trees were pre- tree size and the corresponding breakdown stress level is
sent. In an initial experiment slices were heated for 160 h presented in Figure 17. The breakdown stress level is de-
a t 135°Cand 20 h a t 190°C.In both cases it was found that fined as the arithmetic mean value of the electric stresses
the optical structure of the water tree was unchanged. during breakdown in the inner and the outer region of
Even after recrystallization of the material by cooling the the insulation. The main conclusion is that vented trees
melt, the structure was not changed. Chemical changes as well as bow-tie trees weaken the insulation. Moreover,
in a water tree are assumed, stabilizing the structure. In it 'is shown that there is a clear relation between the size
a second experiment untreated slices were dissolved in xy- of the water trees and the electric breakdown stress lev-
lol. It appeared that not only the P E but also the water el. Finally, it appears that water trees crossing the entire
1002 Steennis et al.: Water Treeing

The breakdown stress of cable insulation can be re-

stored by drying the insulation: a breakdown stress of
up to a t least 50% of the original level can be obtained
[76]. However, water trees do not actually disappear. As
soon as water or water vapor is present, this water will
be absorbed and consequently the breakdown stress level
reduces again:

4.2 LOSS FACTOR (BULK

IN S U LATlO N )

Several investigators reported an increase in the loss

factor of cable insulation containing water trees: Bahder
et al. [7], Tanaka et al. [155], Franke et al. [52], Karner et
Figure 18. al. [72] and Fukagawa et al. [54]). The study of Fukagawa
Vented tree (marked 1) grown from t h e inside of
showed this increase to be related to insulating materials
cable insulation. During the breakdown test not already having a rather low breakdown stress level; it is
only a breakdown channel (marked 3 ) was creat- possible therefore that these particular insulating mate-
ed, but a n electrical bow-tie tree (marked 2) was rials may contain rather large water trees.
also initiated i n the tip of the water tree.

Other investigators also tried to find a relation between

a change in the loss factor of the bulk insulation and the
presence of water trees. The loss factor has been stud-
ied for power frequencies (Kirkland et al. [79], Naybour
[112], Srinivas [144]) and frequencies between 0.01 and
10 Hz (Swingler et al. [151]). In these publications it is
reported that such a relation was not found. Naybour
tried to explain these different findings. He found that
poor conductivity of the conducting screens of the cables
involved can also be an explanation for an increase of the
loss factor.

4.3 RESISTIVITY (BULK

IN S U LAT I0N )

Water trees do not affect the resistivity of the insulation

of a cable as has been reported by Kirkland et al. [79] and
Karner et al. [72].
Figure 19.
Vented tree (marked 1) grown from t h e inside
of cable insulation. Again, a breakdown channel In contrast, a reduction of resistivity was observed by
(marked 3 ) was created during the test, but also Bahder et al. [7] and Wojtas [168]. In Wojtas’s study the
a n electrical bow-tie tree (marked 2) was initiated insulation was aged with a dc voltage, making the results
in the tip of the water tree . suspect. As far as very large water trees are concerned, a
reduction of the insulation resistance was found by Taba-
insulation do not cause immediate breakdown under st:r- t a et al. [153], Tanaka et al. [155], Srinivas et al. [143]
vice conditions. Such insulating materials often still ha.ve and probably also Fukagawa et al. [54]. These results can
a breakdown stress level above the service stress level of explain the observed increase of the loss factor described
PZ 2 kV/mm. above.
IEEE Transactions on Electrical Insulation Vol. 25 N o . 5 , October 1900 1003

4.4 PARTIAL DISCHARGES A N D

ELECTRO-LUMINESCENCE

The growth of water trees is normally not attended by

detectable partial discharges. This was found by Bahder
et al. [7], Kirkland et al. [ 791, Bamji et al. [9, 111 and
Steennis [148] a t a noise level of approximately 0.1 pC.

Optical experiments were performed by Nitta [114] and

Bamji et al. [ll]. In water needle experiments both in-
vestigators tried to find light emission in front of growing
vented trees. Nitta was able to observe the emission of
light. However, the light phenomena observed in this nee-
dle experiment in all probability originated from mecha-
nisms which are not related to vented tree growth under
moderate aging conditions. This is illustrated by Bamji
in a similar experiment. The electric stress applied a t
the needle tip in the experiment carried out by Bamji
was much lower than in Nitta’s experiment. Bamji did
not succeed in observing any light emission during tree
growth, not even with the aid of a photomultiplier and
lens system. Moreover, he was unable to detect partial
discharges a t a detection level of 0.05 pC. T h e experi-
ment performed by Bamji indicates that neither low mag- Figure 20.
nitude partial discharges nor electro-luminescence (emis- Electrical bow-tie tree (marked 3) initiation in the
sion of light by phosphorescent substances) occurs during branch of a vented tree (marked 1). The inclusion
the growth of vented trees. (marked 2) from which the electrical tree was ini-
tiated is indicated.
However, under certain circumstances the water tree 3. a vented tree path is a ‘poor’ insulating material, the
can initiate an electrical tree, for instance when there breakdown stress level is lower than that of the sur-
are overvoltages. After initiation of electrical treeing, rounding material.
breakdown of the insulation cannot be excluded. Such
a mechanism has been described by Tabata et al. [152] Figure 20 shows an enlargement of the initiation sites of
and Gronefeld et al. [60]. electrical trees, grown from the branches of a vented tree.
It is shown that a large amount of impurities is present.
Steennis [148] showed electrical bow-tie trees grown The photograph indicates that the electrical trees were
from the branches of a vented water tree. Such electri- initiated a t such inclusions.
cal trees were initiated during breakdown voltage tests
a t electric stresses much higher than those applied dur- 5. EFFECT OF OPERATING
ing service conditions. An example is shown in Figures
18 and 19. It is assumed that the cause of initiation of
VA RIA BLES
these electrical trees is perhaps the site where a vented 5.1 EFFECT O F ELECTRIC STRESS
tree path crosses a field-disturbing inclusion, such as an
impurity or a void. This is based on the following:
INT ENSlTY

5.1.1 VENTED TREES

1. the electric stress in the vented tree path is of the same
order of magnitude as in the unaffected surrounding
P E (Section 3.1), The development of vented trees is clearly affected by
2. there is always a great number of inclusions in the P E the electric field strength. An increase of the electric
and consequently also in the vented tree path. En- stress intensity leads t o a n increase of the length of vent-
hancement of the electric stress can be expected a t ed’ trees according to Bernstein et al. [15], Yoshimura et
many of these inclusions, al. [174], Srinivas et al. [143], Filippini et al. [45], Hossam
1004 Steennis et al.: Water Treeing

Eldin e t al. [64], Naybour [112] and Muller [106]). An in- 5.1.2 BOW-TIE TREES
crease of the concentration of vented trees was observed

tric stress intensity. Enhancement of the electric stress

One can measure the time needed for a vented tree was found to increase the rate of propagation in experi-
to reach a certain length if the electric stress is dou- ments carried out by Bulinski et al. [26] and Yoshimura
bled. Most of the studies show that the reduction in et al. [175]. The opposite effect was observed by Sletbak
time is approximately a factor 2 (Filippini, Hossam El- et al. [137].
din, Yoshimura and Muller). These studies all deal with
water needle experiments. Bernstein [ 151 and Srinivas 5.2 EFFECT O F FREQUENCY
[143] tested cables with scratched insulation surfaces a t
moderate electric stresses. Here, a reduction factor be- 5.2.1 VENTED TREES GROWN UNDER ac
tween 2 and 7 was found. As in these cases only a few STRESS CONDITIONS

;= 80-
E related to the frequency of the electric stress. To study
0 this subject, water needles were used by Yoshimura et al.
M
[174], Favrid et al. [41], Filippini et al. [44], Pays et al.
70- 120 days [123]. Yoshimura found a linear relation between growth
L
m and frequency in a range of 200 Hz to 3 kHz. The to-
4
0 tal test time in this case was 5 h, which is rather short.
~ 6 0 - Longer tests were performed by Favrie et al. [41] and
0 Filippini et al. [44]. These tests show an optimum tree
-0
growth between 4 and 8 kHz.

From these data it can be concluded that vented tree

growth increases a t higher frequencies. However, above
4 to 8 kHz the propagation rate appears t o decrease. As
a result of the different growth conditions during these
I I I I
50 85 125 170 tests it is hardly possible to make an accurate predic-
tion of the propagation rate a t a certain frequency level.
applied voltage a t 70 " C (V/mil)
Figure 22 therefore only gives a rough picture of the re-
Figure 21. lation between the propagation rate of vented trees and
Effect of the electric stress on the vented tree frequency.
propagation rate (After [15]). Note: 1 mil = 25.4
pxn and 100 V/mil = 3.94 kV/mm 5.2.2 BOW-TIE TREES GROWN UNDER ac
STRESS CONDITIONS
Figure 21 illustrates the relation between the electric
stress applied and the time required for a vented tree to Bow-tie tree growth in different types of P E was stud-
grow to a certain length [15]. ied by Bahder et a1 [8], Sletbak et al. [137] and Yoshimura
IEEE Transactions on Electrical Insulation Vol. 25 No. 5 , O c t o b e r 1990 1005

Pays et al. [123]. The results were obtained from tests on

model LDPE insulated cables [51] or full-scale XLPE in-
sulated cables [170] under the application of high electric
stresses. Pays et al. [123] studied trees in LDPE plates
and vented trees in a needle experiment. The electric
stresses varied from 10 t o 100 kV/mm for the cable and
plate experiments, while in the needle test the electric
stresses were much higher.

The shape of water trees grown under dc stress con-

ditions differs from that of common water trees grown
under ac stress conditions. It was reported that water
trees grown under the application of a dc stress have a
remarkably narrow structure, while vented trees as well
as bow-tie trees only grow in one direction. Consequently
1 10’ io2 io3 io4 io5 the bow-tie trees have only one plume. Franke observed
frequency (Hz) that during the growth of the water trees under the appli-
Figure 22. cation of a dc stress the negative copper electrode became
Graph showing roughly the relation between the black. It is assumed that this is a n indication of the pres-
vented tree propagation rate and the frequency. ence of an electrochemical process. Pays did not find any
tree growth, with one exception: in the plate experiment
et al. [175]. Bahder used model cables and Sletbak full- once a tree-like structure was observed.
scale cables; in the experiment carried out by Yoshimura
a blunt needle has been applied. The electric stress in all There are reasons t o suspect results of tests with dc
experiments was moderate. At higher frequencies accel- stresses: Very high electric stresses have been applied in
eration of initiation and growth was found. An optimum order to initiate and grow these trees, and the shape of
frequency was not given. the trees differs from that of water trees grown under
ac conditions. Perhaps there are different mechanisms of
5.2.3 WATER TREES GROWN UNDER (initiation and) growth for trees grown under the appli-
SIMULTANEOUS APPLICATION OF ac AND cation of high dc stress and those trees grown under the
dc STRESS application of ac stress.

Pays et al. [123] studied the tree propagation rate in 5.3 EFFECT O F TEMPERATURE
a frequency range of approximately 20 Hz to 10 kHz. In
this experiment water treeing was studied under ac stress
as well as under simultaneous ac and dc stress conditions. The effect of the temperature on water tree growth
The results of ac plus dc stress were the same as if only can be considered as rather complex. It is thought to
ac stress had been applied. The type of water trees found depend on the temperature level, temperature gradient,
in these experiments has not been described. However, in temperature load pattern, water diffusion coefficients and
general these results confirm the results presented by oth- moisture absorption of the insulation as well as the oth-
er investigators: in this frequency range the propagation er cable components. In one of his experiments, Muller
rate increases with increasing frequency. Different types [lo61 found that as a consequence of a high conductor
of P E were studied. The aging time of this experiment temperature the relative humidity level of the insulation
was not given. at the conductor side was low. Moreover, it is well known
that cycling the temperature may lead to oversaturation
5.2.4 WATER TREES GROWN UNDER dc of water in certain regions of the cable insulation.
STRESS CONDITIONS
5.3.1 VENTED TREEING AT CONSTANT
TE M PE RATU RE
If the results described above for moderate electric
stresses are extrapolated to very low frequencies, hard-
ly any tree growth is to be expected. This is confirmed in Several publications discuss vented tree growth at con-
the literature. Water tree growth a t dc voltage has been stant temperature levels, without a temperature gradient
studied by Franke et al. [51], Yamada et al. [170] and over the insulation. Most of the tests were carried out
1 006 Steennis et al.: Water Treeing

in water baths on LDPE or XLPE insulation. The test Muller was not able to find a statistically justified differ-
models applied are needle samples by Fourni6 et al. [50], ence in vented tree growth for steam-cured cables after
sandwich models with inserted wire by Yoshimitsu et al. aging with cycled temperatures, compared to the same
[172,173] or more or less full-scale cables by Tabata et cables after aging a t a constant temperature level. With
al. [153] and Srinivas et al. [143]. From these studies it respect to the concentration of these trees, Bulinski noted
can be concluded that the concentration of vented trees that temperature cycles reduced it strongly.
increases and the length of vented trees decreases if the
temperature of the insulation is 2 50°C. Between 20 and 5.3.3 BOW-TIE TREES
50°C there are conflicting views. Excluding the results
of the needle tests [50], the most favorable temperature
range for water tree growth to be 30 to 50”c [143, The development of bow-tie trees has been studied by
1531. Naybour [lll],Sletbak et al. [139], Bulinski et al. [27],
Fredrich et al. [53] and Marsh et al. [89]. Fredrich showed
5.3.2 VENTED TREEING AT CYCLING that bow-tie tree growth is enhanced at higher temper-
T E M P ERATURE atures. In his experiment the temperature of the water
outside the insulation of the full-scale cable was fixed at
40 or 90°C or cycled from 5 to 9O’C. The conductor con-
Sletbak et al. [139], Bulinski et al. [27], Marsh et al. [89] tained water, but was unheated. In this particular study
and Muller [lo61 studied the effect of cycling the temper- the water at high temperatures was oxygen enriched, but
ature of the conductor and the outer screen. In most it was stated that this could have hardly any influence on
cases there was a temperature gradient over the insula- the tests; the amount of oxygen in the water a t low tem-
tion during the experiment. Sletbak, Marsh and Muller perature would have been even higher. Aging was carried
used full-scale cable and Bulinski model cables. The du- out a t moderate stress levels for < 1000 h.
ration of the different tests was < 1000 h. The outlines
of the different experiments are summarized in Table 5. Naybour observed an increase of the concentration of
Table 5. bow-tie trees if the temperature of the water bath was
Cycling temperature and moisture conditions re- increased from 35 to 80’C. The other aging parameters
garding the study of vented treeing.
-
were chosen fairly moderate. Tests have been carried out
on full-scale dry-cured cables for 2000 h.

temperature temperature The experiments by Marsh, Bulinski, Sletbak and Miil-

conductor [“C] outer Screen [“cl Reference ler already have been described in Section 5.3.2. Test re-
(dry or wet) (wet) sults match the findings of Fredrich and Naybour. Slet-
bak assumed that the increased bow-tie tree growth dur-
90 55 [I391 ing cycling of the temperature is a result of supersatura-
55-90 55 tion of water in the insulation during the unloaded part
20-90 20-55 of the temperature cycles applied.
90 (wet) 55
55-90 (wet) 55 5.4 EFFECT OF MECHANICAL
STRESS
90 50 ~ 9 1
20-90 20-50
20 20 A mechanical stress can be imposed on an insulation.
It can also be present in the insulation after cooling from
20-70 20-70 ~ 7 1 the extruded melt. Tanaka et al. [154] studied the rela-
20 20 tion between internal mechanical stresses and water tree-
ing. Slices were taken from several full-scale (crosslinked)
70-80 20-30 I1 061 LDPE insulated cables aged in different ways. By count-
ing the number of isochromatic lines on a polarization
photograph in a cross-section of the insulation, the me-
In general, one can conclude from these experiments ch,anical stresses were estimated. It was found by Tana-
that the growth of vented trees from the outer screen ka that bow-tie trees and vented trees are concentrated
is hardly affected by these variations in test conditions. in regions under higher mechanical stresses. Mechanical
I E E E Transactions on Electrical Insulation Vol. 25 N o . 5 , October 1990 1007

stresses between 1 t o 8 MPa were observed in the insu- 5.6 EFFECT OF T H E CHEMICAL
lation. This is high, but not destructive as described in NATURE OF T H E FLUID
Section 1.2. Such mechanical stresses may result from
the production process or from cable installation. One 5.6.1 N0N- WAT E RY S 0L U T ION S
year later Prigent et al. [124] found water trees in LDPE
samples concentrated a t locations with high mechanical
stresses. However, Prigent’s evaluation is purely qualita- Bahder et al. [8] studied the growth of vented trees and
tive. bow-tie trees in normal cables with a LDPE or XLPE
insulation. T h e tests were carried out a t a frequency
More recently, results were published by Tu and Kao level of 7.3 kHz and for a duration of 274 days. Flu-
[160]. They exposed a P E sample with a water needle to ids surrounding the insulation were water and water with
a pressurized atmosphere. The pressure on all sides of the CuSO4. Moreover, H o s t o p a p and ethylene glycol, both
sample was varied over a range between 0.1 and 3 MPa. easily penetrating the polymers, were used. In all cases
The test time was short, only 15 h. Tu and Kao found treeing in the insulation had the same appearance. From
that the initiation was faster, but growth of the vented this it can be concluded that water treeing is probably
trees was slower if the pressure was increased. a special case in a broader field of insulation degrada-
tion. Treeing has also been studied in a liquid paraffin
solution, which was chosen because of its extreme small
Recently, Patsch et al. [122] found more evidence for dipole moment. No trees could be observed but it must
the influence of mechanical stresses. A bent full-scale 20 be mentioned that the test duration was only 12 days.
kV XLPE cable was aged for 5000 h at a voltage level of All other studies mentioned below deal with water solu-
24 kV. Patsch found that 90% of all the bow-tie trees were tions. The following subjects will be reviewed: type of
created in the stretched zone (near the conductor where salts, salt concentration, electrode materials, acidity and
the elongation was 10%) in contrast to 10% of the bow- solubility.
tie trees found in the compressed zone. This result shows
the effect of the mechanical stresses on the initiation rate
5.6.2 TYPE OF SALTS
of bow-tie trees, however, does not inform about the ef-
fect of the mechanical stress on vented tree initiation and
especially vented tree growth. An indication that vented tree length is affected by the
type of salts dissolved in the water is given by Bamji et
5.5 EFFECT OF RELATIVE al. [ll]. Bamji used XLPE samples in which water nee-
HUM I D l T Y dles were made. A CuSO4 solution produced the greatest
tree growth, followed by a NaCl solution and finally by a
CaClz solution. No trees were observed in distilled water.
The test duration was 90 h, the frequency 1 kHz.
There are a few studies dealing with the subject of wa-
ter treeing under different water-vapor conditions. Slet- Ross [127] studied vented tree growth in similar test
bak et al. [137,139] and Yoshimitsu et al. [172] showed objects. Aging lasted for x 1200 h. The upper electrodes
that the relative humidity of the air surrounding the spec- consisted of stainless steel, and several 1.71 mmol/l salt
imen and inside the specimen is relevant to water treeing. solutions were used: NaCl, NazS04, CuC12, and CuSO4.
Both found that water treeing becomes rare a t a rela- With the CuSO4 solution the longest trees were obtained,
tive humidity of 65 to 70% or less. Above this level the whereas with NaCl l o x smaller trees were grown. The
concentration of bow-tie trees (Sletbak) or vented trees other solutions yielded intermediate results.
(Yoshimitsu) increases if the relative humidity level is in-
creased. The studies do not give a relation between water
5.6.3 SALT CONCENTRATION
tree length and relative humidity. Also, the effect of liq-
uid water around the specimen compared to water vapor
with a relative humidity of 100% around the specimen Many investigators studied the growth of vented trees
has not been studied. in relation to the concentration of certain salt ions dis-
solved in water. The salt concentration can of course
The experiments were carried out on XLPE samples be translated into the conductivity of the fluid. In most
or cables with electric stresses ranging between 4 and 20 cases NaCl and CuSO4 solutions were applied. It was
kV/mm. found that higher salt quantities enhance the vented tree
1008 Steennis et al.: W a t e r Treeing

propagation rate. Most of these investigators used wa- between the concentration and pH of the solutions. How-
ter needles in order to study vented trees: Ashcraft [4, ever, a slight relation between the concentration and the
51, Yoshimura et al. [174], Filippini et al. [44,45], Hossam standard entropy of hydrated ions was found.
Eldin et al. [64]. Test durations < 120 h were used at
<
frequencies 20 kHz. 5.6.6 SOLUBILITY

The effect was confirmed for vented trees in model or

full-scale cables (Tabata et al. [153], Katz et a]. [75] and A relation between solubility and initiation of water
Srinivas et al. [143]). Test durations in the related studies treeing was found by Kat2 et al. [751. Vented trees have
were much longer, < 1 yr, while the frequency level was been studied, which start a t the surface of scratched (cross-
50 HZ in [153] and in the kHz-range in [75,153]). linked) LDPE cable insulation in a test with a duration
of 13 days a t a frequency level of 7.8 kHz.
Fournid et al. [50] found the opposite effect in a 50
HZ LDPE needle test. Rye et al. [128] tried t o explain 5.7 EFFECT OF INSULATING
Fournid ' s findings. They proposed that oxygen promotes MATERIAL AND ADDITIVES
vented tree growth; moreover, it is known that the sol-
ubility level of oxygen is lower in a very strong solution 5.7.1 VENTED TREEING IN NEEDLE TESTS
of salts. A combination of these facts could explain the
reduced growth of vented trees.
The water tree susceptibility of different types of in-
sulating material has often been studied in water needle
5.6.4 ELECTRODE MATERIAL
experiments. Studies with water needles in this specific
field have been carried out by Ashcraft [4,5], Isshiki et al.
[66], Kat0 et al. [73], Braun [25], McMahon [96] and Saure
Fournid examined the effect of the electrode material on
et al. [130]. Ashcraft examined the growth rate of trees
vented tree growth. The fastest vented tree growth was
a t 8.5 kHz for 24 h in polybutene, polystyrene, ethylene
found if P t or Cu electrodes were used, followed by All
propylene diene terpolymer rubber (EPDM), LDPE and
Fe and Pb. It is again suggested by Rye that the amount
in XLPE. Tree growth in polybutene was extreme. The
of oxygen in the fluid affects the growth of water trees.
slowest tree growth was found in freshly cured XLPE.
The amount of oxygen is lower with Fe or P b electrodes,
This was attributed t o crosslink residual products (main-
than when using P t and Cu electrodes.
ly acetophenone). It was shown that acetophenone added
to P E is able to suppress water treeing in this material.
Filippini [47] carried out similar tests with the same ex-
perience. The role of oxygen in water treeing is considered The retardant effect of crosslink residual products was
here too. Filippini showed that oxygen formation a t the confirmed by Saure in a similar test, but a t a frequency of
electrode/liquid interface is not likely. He suggested that 50 Hz. Moreover, Saure found that an antioxidant, used
the effect on vented tree growth is a consequence of ions for the stabilization of'PE, had the opposite effect in this
in the solution through metal corrosion. Filippini stated test. Saure also found that silane-cured XLPE was less
that the actual mechanism causing the tree growth is not water tree susceptible than peroxide-cured XLPE, which
clear. However, the role of the liquid and the interaction matches the findings of van de Laar [85] and Kreuger et
of the liquid/insulation interface is important. al. [83]. In 1986 Dissado et al. [36] examined vented tree
growth under moderate aging conditions (6 kV/mm, 50
5.6.5 EFFECT O F ELECTROLYTE pH Hz, 30'C) in silane-cured and peroxide-cured (degassed)
XLPE slabs for a few thousand hours. In both materi-
als vented trees with a comparable average length could
The effect of pH has been studied by Ashcraft [4,5]. be observed. In the peroxide-cured XLPE, however, the
Ashcraft found that the vented tree growth is enhanced scatter was much higher. These results give the impres-
a t high pH and decreased a t low pH. Solutions of hy- sion that silane-cured P E is less susceptible to vented tree
drochloric acid in 0.01 N NaCl for water-needle experi- development than peroxide-cured PE. However, none of
ments in P E were used. The test duration was 24 h a t the investigators tried or could explain these findings.
a frequency of 8.5 kHz. Morita et al. [lo41 examined the
concentration of vented trees for a great variety of solu- Braun studied the growth of vented trees in dry-cured
tions. Scratched LDPE films were tested for 7 days a t XLPE, steam-cured XLPE, polystyrene and epoxy resin
a frequency of 3 kHz. Morita could not find a relation a t 60 Hz for 100 days. No treeing was observed in epoxy
IEEE Transactions on Electrical Insulation Vol. 25 N o . 5 , October 1090 1009

resin, the other materials showed about equal vented tree of crosslink residual products in the XLPE cable insula-
development. It cannot be concluded that epoxy resin tion. Henkel and Kalkner [62] tested full-scale LDPE and
is water tree resistant: other investigators also observed XLPE insulated cables for 250 days at 50 Hz. The cables
water tree development in this material, e.g. Yoshimitsu were pre-conditioned, causing evaporation of these residu-
et al. [171]. al products. This evaporation resulted in even more rapid
vented tree growth in the XLPE insulated cables than in
Isshiki et al. [66] carried out research work which was the LDPE insulated cables. Katz tested pre-conditioned
mainly intended to find the electric stress level a t which model cables with a scratched inner surface for 15 days a t
various materials initiate water trees. The rate of prop- 7.8 kHz. He found only a small difference in the growth
agation was also studied. The total aging time for this behavior of vented trees in LDPE and XLPE insulated
experiment was 12 days; the frequency 2 kHz. It was cables.
found that the growth of vented trees is faster in soft ma-
terials. The materials PVC, LDPE, XLPE, ethylene vinyl The effect of crosslink residual products is also con-
acetate copolymer, HDPE and polypropylene all showed firmed by Srinivas. In his experiments, model LDPE and
water treeing, while in polystyrene, polycarbonate and XLPE insulated cables with a scratched inner insulation
nylon some kind of vented treeing was only observed when
high initiation stresses were used. In polystyrene, poly- -
surface were tested for 20 to 150 days a t a frequency of
2 kHz.
carbonate and nylon the trees observed can be considered
as suspect since other aging mechanisms may have been Faremo studied vented tree growth in XLPE and some
introduced as a result of the high initiation electric stress- filled ethylene-propylene rubbers. The materials were
es. preconditioned prior t o testing. T h e experiments were
carried out under rather extreme temperature conditions
Kato et al. [73] suppressed the initiation of vented trees and a t high electric stresses on press molded cups, in-
cluding plastic semiconducting screens. Faremo conclud-
in very short term needle tests on XLPE samples at 1.2
kHz. He uses a mix of different additives: ferrocene, silox- ed that the growth rate of vented trees was about the
same in the different materials. These results are all the
ane oligomer and 8-hydroxy quinoline. The initiation re-
sistance observed was attributed to the combined action more important since it has often been suggested that
water trees would not exist in ethylene propylene rubber.
of a migration of these additives to irregularities and a
deactivation of electrons and metal ions through'traps.
5.7.3 BOW-TIE TREES
Another water tree inhibiting additive was presented by
McMahon [96]. This additive, dodecanol, has been tested Bow-tie tree growth has been studied by Katz et al.
with water needles at different frequencies in P E samples [74], Yoshimitsu et al. [171], Kalkner et al. [69], Marsh et
for 28 days. Moreover, the additive was used in full-scale al. [89] and Faremo [40]. Katz and Faremo did not find
cables, tested for 220 days with a 9O'C conductor tem- any difference in the growth of bow-tie trees between pre-
perature. It was found that the dodecanol concentration conditioned LDPE and XLPE insulated cables [74] or be-
stabilizes; this indicates that the water tree inhibiting ef- tween XLPE and ethylene propylene rubber press molded
fect observed could be effective over longer periods. cups [40]. The experiments are described above in Section
5.7.2. Marsh [89] observed bow-tie trees in silane-cured
5.7.2 VENTED TREEING IN (MODEL) and in steam-cured P E insulated cables. The concentra-
CABLES tion and growth of bow-tie trees in the peroxide-cured ca-
bles was shown to be more significant than in silane-cured
cables. Marsh tested the full-scale cables for many years
The development of vented trees in cables or model ca- under moderate aging conditions. In his experiments dif-
bles has been studied by several investigators: Katz et al. ferent temperature conditions were applied. Yoshimitsu
[75], Bahder et al. [8], Srinivas et al. [142,143], Henkel et [171] reduced the bow-tie initiation rate successfully in
al. [62], Kalkner et al. [69], Marsh et al. [89] and Fare- an epoxy by adding a surface active agent that makes
mo [40]. Bahder investigated the difference in vented tree the hygroscopic substances (initiating bow-tie trees) hy-
growth between LDPE and XLPE insulated cables. He drophobic. Kalkner [69] found a decrease in bow-tie tree
found that LDPE insulated cables were more susceptible growth if a certain, not further described, additive was
to vented trees than XLPE insulated cables. The investi- used. The experiments by Yoshimitsu were carried out
gations were carried out on full-scale cables aged for 8 yr on XLPE or epoxy resin films. A rather high electric
under service conditions. The difference observed can be stress level of 40 kV/mm at 1 kHz for 63 h was applied.
attributed, for instance, to the water tree retardant effect Kalkner tested P E films a t 50 Hz for 130 days.
1010 Steennis et al.: Water neeing

5.7.4 EFFECT OF MORPHOLOGY OF THE Vented trees are more dangerous than bow-tie trees as
I NS U LAT ING M AT E RIA L a result of the difference in growth behavior. Vented trees
are diffuse structures, growing in many kinds of polymers.
The vented tree probably contains about 1% water, be-
A few publications only discuss the growth of water ing about 3 x the amount of water in freshly steam-cured
trees with respect to morphology-related parameters of XLPE insulations and about 100 x the saturation level
the insulation. of water in dry-cured and silane-cured PE. This water
is mainly concentrated in micro cavities and channels.
Morita et al. [lo31 varied the melt index (0.3 to 2 g/10 Apart from Clustering of water in voids, water is proba-
rnin), density (0.920 to 0.927 g/cm3) and related crys- bly molecularly dispersed in the path of the vented tree.
tallinity (72 to 76% from density) of LDPE and some Here water molecules can be found a t places with polar
other materials. He was not able to find a relation be- groups attached to P E branches. Nevertheless the vented
tween the growth of the water trees and these morpho- tree can be considered as an insulating material.
logical parameters. No information is given as to whether
bow-tie trees or vented trees were observed. The insula- A vented tree can be dried and rewetted. To make
tion was tested in a sandwich construction a t a frequency the tree clearly visible different water-soluble dyes can
of 50 Hz for 42 days. be used; decolorization of vented trees after dyeing has
never been reported. A high concentration of microvoids
Saure and Golz [130] studied vented tree growth in a is mainly found in the trunk of vented trees grown a t high
50 Hz water needle experiment. The melt index of the electric stresses. The branches of such vented trees also
LDPE and XLPE samples was varied over a wide range become more pronounced. A vented tree grows mainly
(0.2 to 10 g/10 min). Also no relation could be observed in the direction of the original electric field. The growth
in this study. In the same kind of test, Golz [59] conclud- rate under service aging conditions ranges from M 20 to
ed that annealing the insulation resulted in extra vented M 500 pm per year.
tree growth. The total effect, however, is small. Anneal-
ing was carried out in different ways, often at tempera-
tures > 1OO'C for 15 h. Crosslink residual products had Local measurements on tree-affected regions of the in-
sulation with different chemical and physical detection
been removed by pre-conditioning from the P E samples.
methods show evidence of oxidation. The presence of dif-
Go18 attributed the effect observed to the increase of crys-
ferent species like sulfate, carboxylate anions and metals
tallinity and therefore of free spaces in the insulation after
in the tree-affected regions has been established.
annealing.

Namiki et al. [110] also concluded that heat-treatment Bulk measurements on insulating materials with and
of the material produced extra water tree growth. The without vented trees show:
author studied the development of bow-tie trees in XLPE
sheets aged for 40 h a t a frequency of 50 Hz. Namiki as- 1. There is a relation between the size of vented trees and
cribed this effect to an enhancement of the brittleness the breakdown stress level.
of the insulation due t o higher crystallization. An ex- 2. One cannot exclude an increase of the loss factor and
planation for the enhanced water tree growth after heat- a decrease of the dc resistance of insulating material
treatment [59,110] could be the effect of the (further) containing vented trees. However, results from differ-
evaporation of crosslink residual products. Most results ent investigators were not found to be reproducible.
indicate that there is no distinct relation between the 3. At normal electric stresses the growth of vented trees is
morphological parameters of the insulation such as melt not accompanied by measurable partial discharges. At
index, density or crystallinity and the growth of water higher electric stresses electrical treeing may be initiat-
trees. Only a few studies are informative in this partic- ed from the tip of the vented tree, eventually resulting
ular field and unfortunately none of these studies deals in a breakdown.
with vented trees in full-scale cables.
There are no fundamental contradictions between the re-
6. SUMMARY sults of tests where vented trees have been initiated in
a water needle experiment and vented trees initiated on
the surfaces of scratched or even unscratched insulations.
N summarizing the main characteristics of water trees
I and effects of operating variables regarding vented tree
growth, the following can be concluded.
Most of the information justifies the conclusion that the
rate of propagation of vented trees is proportional to the
electric stress.
IEEE Transactions on Electrical Insulation Vol. 25 N o . 5 , October 1990 1011

There is hardly any vented tree growth under the ap- 3. Coulomb forces,
plication of a dc stress. The vented tree propagation rate 4. dielectrophoresis,
increases a t increasing frequency; however, above 4 to 8 5. thermal degradation,
kHz the tree propagation rate decreases with increasing 6. partial discharges,
frequency.
7. chemical degradation.
Vented tree growth is not really affected by tempera- The main focus will be on the growth of vented trees,
ture cycling. The role of a temperature gradient is im-
because this type of water tree appears to be the most
portant and rather complex. At constant temperature
dangerous one for medium-voltage extruded cable insula-
level the most favorable temperature region for vented tions, as was stated before. The discussion of the mecha-
tree growth appears to be 30 to 50°C. nisms of vented tree growth is preceded by the definition
of aging conditions.
To date, the relation between vented treeing and me-
chanical stresses is unclear. Degradation is usually presented under extreme aging
conditions. It is obvious to expect a certain degradation
At a relative humidity of the air surrounding the insu- if, for instance, very high electric stresses are applied.
lation of < 70%, vented treeing becomes rare. Moreover, it is not difficult to show this degradation in ex-
periments with water needles, in which such high stresses
The chemical nature of the fluid surrounding the poly- are available. These experiments show that high electric
mer affects vented tree growth. There are several indica- stresses should be avoided; however, such experiments are
tions that the type of salt is related to vented tree growth; not conclusive on the cause of vented tree propagation a t
NaCl and CuSO4 solutions are often used to speed up ag- moderate electric stresses of a few kV/mm. The aging
ing processes. Higher salt quantities will usually produce conditions therefore will be emphasized when discussing
more progressive growth. One test was made with fluids the various degradation processes.
other than water. This also resulted in some kind of tree-
ing, showing that water treeing is probably a special case
of treeing in general. 7.2 CAPILLARY ACTION

Vented tree growth is not only related to P E bu't can be

Van der Waals demonstrated that a force, directed in-
found in many different materials: polyolefins and epoxy
ward is exerted on the water molecules of a surface layer.
resins. The growth of these trees is clearly affected by
The related surface energy can be written as a product
additives and crosslink residual products. Additives and
of the surface tension and the surface area. This sur-
residual products are able to evaporate in most cases,
face energy is a representation of the potential energy of
resulting in a reduction of this protective effect.
the molecules in the surface layer. Enlarging this surface
requires energy.
Morphological parameters such as melt index, crys-
tallinity and density probably do not affect the growth
P E insulating materials always contain a number of mi-
of vented trees.
crovoids. As a consequence of the different surface ten-
sions of PE/water, water/air and PE/air interfaces, water
7. POSSIBLE MECHANISMS inside microvoids with pure P E walls will in principle be
OF GROWTH expelled from the PE.

Similar considerations describe the behavior of water

7.1 INTRODUCTION
in a capillary channel. Water will not enter a narrow
channel as long as the walls of these channels consist of
variety of mechanisms of bow-tie tree and vented tree pure PE. However, if the walls of these channels become
A growth has been postulated since water trees were
discovered. This important matter has been discussed
polar, their surface tension will change. Water will en-
ter the channel if the surface tension of the PE/water
in detail a t a workshop meeting of EPRI in 1985 [24]. interface is smaller than the surface tension of the P E /
The most important mechanisms will be discussed in the gas interface. Then the absorption of water into the sys-
course of this Section: tem will cause the surface energy t o decrease. In other
wdrds, a certain minimum amount of polar material on
1. capillary action, the walls is required to get an intrusion of water. It could
2. osmosis, be questioned if capillary action itself is able to damage

r
1012 Steennis et al.: Water Treeing

the polymer system, especially in front of the capillary tangential stress ct becomes = II/2. The various
channel. This is not the case: the reduction of the sur- stresses are given in Figure 23.
face energy and, thus, the supply of water will stop as
soon as the water has reached the unmodified area in the 7.3.1 EXAMPLE
channel. The pressure at the end surface is compensated
by the concavity of the water meniscus. There remains
no pressure capable of damaging the P E [148]. For a typical NaCl solution under saturated conditions
the osmolarity c’ = lo4 mole/m3 at a temperature of 300
7 . 3 OSMOSIS K. As a result, the osmotic pressure II becomes large:
in this example the osmotic pressure is M 25 MPa. The
tangential stress in the surface layer of a spherical void
containing this solution is 12 MPa, which is near the yield
Water-soluble substances that are present in microvoids
strength of LDPE (cyin Section 1.2). Creep, a mechan-
attract water from the environment and osmotic pressure
deformation process, may occur. As a
may occur. Moore [lo21 derives such an osmotic pressure
the void will be enlarged and by further absorption of
by means of a thermodynamic approach. It is stated that
water the solution will be diluted and osmotic pressure
the chemical potential of the water in the void and outside
will decrease. This process will continue until eventually
the P E are equal if an equilibrium is reached. Therefore,
the osmotic pressure falls below the yield strength of the
a decrease of the chemical potential of the water in a void
PE.
due to a solute must be compensated by a n increase of
the hydrostatic pressure. The osmotic pressure derived
7.3.2 OSMOTIC PRESSURE AND SURFACE
becomes TENSION
II M RTc (1)
in which R , the gas constant, 8.31 J/Kmole, T the tem-
perature in K and c the concentration in mole/m3. The The osmotic pressure may be (partly) compensated by
basis of this derivation is a diluted solution; however, for a the surface tension of the solid/liquid interface of the void.
saturated solution Equation (1) is an approximation. For The effect of the surface tension has already been dis-
better results the concentration c has to be replaced by a cussed under ‘capillary action’. The surface energy can
practical parameter, c’ being the osmolarity. Therefore, be high, especially if the walls of the P E are unmodified
so that the voids have small radii. Therefore, the pres-
II M RTc’ (2) sure drop over the liquid surface can reach compensating
pressures regarding osmotic pressure. The radii of such
This osmolarity has been measured for many solutions
voids are calculated to be in the nm region. Consequent-
and can be derived from data given in the Handbook of
ly, compensation of the osmotic pressure by the surface
Chemistry and Physics (D-262 [164]).
energy can only be expected in the interlamellar region

t”
of the polymer. Osmotic pressure itself does not explain
vented tree growth. It could act without an electric field,
while from the phenomenology it is known that an electric
stress is necessary for tree growth.

Sletbak [138], who suggests a n osmotic pressure for

bow-tie tree growth, indicated that the electrically con-
trolled mechanism should be found in a Coulomb action.
Another investigator taking osmosis into account is Fe-
dors [42]. Kats [74] also indicated the ability of water
soluble matter to attract moisture, and included other
mechanisms in their overall picture, for example an elec-
trochemical reaction.
Figure 23.
Stresses in the surface layer of a spherical void. 7.4 COULOMB FORCES
For a spherical void the stress ct in the tangential di-
rection of the void surface is 50% of the stress cT in the Coulomb forces are forces on electric charges, that are
radial direction of the surface [65]. With IF,( = II, the caused by an electric stress. Separated stress terms as
IEEE n a n s a c t i o n s on Electrical Insulation Vol. 25 No. 5 , O c t o b e r 1990 1013

the Maxwell stress and electrostriction can be combined They also found a reduction of the surface tension of the
into a single relation: P E surface through the action of an electric stress. On
the basis of these two facts it was concluded that the
(3) propagation of a vented tree is related t o a process which
is called 'environmental fatigue failure'.
In which fi is the dielectric displacement vector, E' the-
electric stress vector, 5 the unit normal vector, the Cherney [29] starts his calculations with the assump-
permittivity of the polymer, and p the density of the poly- tion that the pressure can be derived from the relation
mer. F, = Q E . In this relation the parameter Fn represents
the force on the medium. The electric stress is E and Q
represents all the electric charges present in the water of
a cavity. This derivation, however, is incorrect because
by no means all charges in the water contribute to the
force on the medium.

Recently Zeller [176] and Steennis [148] calculated that

the Coulomb forces are unable to degrade the polymer.
In this case the electric stress E,, has been chosen rather
moderate, according to service aging conditions. More-
over, Steennis assumed that a vented tree can be repre-
sented by an insulating material. There are several re-
sults from the phenomenological experience supporting
this conclusion:

1. If vented tree growth were related to Coulomb forces,

Figure 24. the propagation of the tree would be proportional to
Pressure at the tip of a vented tree path. E'. However, the results of the phenomenology seem
to indicate that its propagation is proportional to E .
Equation (3) expresses the relation between pressure 2. Cracking of a polymer is normally reduced by the in-
p and the electric stress E . The first term on the right crease of the molecular weight (Section 1.2). Such an
is related to the Maxwell stress and the second term t o effect has not been observed clearly in relation to the
electrostriction. The medium here is P E surrounding the propagation of vented trees in P E (Section 5.7).
vented tree, as is schematically illustrated in Figure 24. 3. Even electron micrographs did not reveal cracking in
the tips of vented tree channels [28]. The voids some-
A number of investigators (e.g.Meyer et al. [98]) esti- times observed in the P E a t a certain distance from the
mated Coulomb forces as serious. They considered var- channel tip probably are related to secondary degra-
ious combinations of- the following: Very high electric dation.
stresses in the virgin insulating material, that a water
tree is a conducting material, and the vibration of the
Coulomb forces in relation to the power frequency. 7.5 DIELECTROPHORESIS

In certain cases Coulomb forces are assumed to act

in combination with another force. With regard to the In a n inhomogeneous electric field, water dipoles tend
growth of bow-tie trees, including the effect of osmotic t o move to the point of highest stress. After some time
pressure, Sletbak and Ildstad [138,140] should be men- the flow of water in the direction of the electric stress
tioned here. A high surface tension of the PE/water concentration will reach an equilibrium with the diffusion
interface may counteract the effect of Coulomb forces, backwards as a result of the water concentration gradient.
especially for voids with small radii. A reduction of this Several investigators describe this effect (Tanaka et al.
surface tension is assumed by Tanaka et al. [156] and Min- [155], Isshiki et al. [66] and Patsch [121]).
nema et al. [loo]. The effect of vibrations is emphasized
by Cherney [29], Isshiki et al. [66], Minnema et al. [loo] The difference in water concentration is found by using
and Morita et al. [104]. a thermodynamic approach. An equilibrium is reached
if'the chemical potential of the water in the position of
Minnema et al. concluded that the vibrating elastic high electric stress is equal t o the chemical potential in the
stresses as a result of the Coulomb forces are essential. Dosition
, ~ ~ ~ .of. low (normal) electric stress. From equations
1014 Steennis et al.: Water Treeing

given in the literature it can be calculated that even under

-
the application of a high electric stress enhancement of
1OOx a t the tip of a water tree, the increase of the water
concentrations a t that specific location is < 3%. The
increase of the water concentration becomes negligible if
a water tree is considered as an insulating material. cable insulation

Nevertheless, several authors mention supersaturation Figure 25.

of water in PE. This supersaturation is assumed to be
Cylindrically shaped volume V surrounded by the
caused by dielectrophoresis. Water droplets will be unaffected PE. T h e volume V has the electrical
formed, which cause small voids to be filled with water. properties of a vented tree path.
This process continues until an equilibrium has been
reached and may therefore lead to pressure building up On the other hand, it is possible to assume that the
and P E being attacked [21,93,94,121,141,167]. vented tree can be considered as an insulating materi-
al (Section 3.1) and that there is a thermal equilibrium:
It can also be stated that there is no supersaturation. a T / & = 0 according to [148]. It can be shown that the
The slightly higher water concentration a t the location temperature in the vented tree hardly differs from the
with a higher electric stress is in equilibrium with its sur- temperature of the surrounding polymer. This is demon-
roundings. Dielectrophoresis is not the cause of water strated for a certain volume V in the PE, cylindrically
tree growth but may contribute to the transportation of shaped, as shown in Figure 25, with a length h, radius
T and with one of its extreme surfaces A on the outer
water to locations capable of absorbing water if stress
enhancement exists near the surface of these locations. surface of the insulation. The properties of the vented
Condensation will occur in these places; in particular wa- tree are attributed to the whole cylinder. For the con-
ter trees can be considered as such dislocations. figuration considered, the solution of Equation (4) with
a T / a t = 0 is according t o Wong [169]:
r2 2h
7.6 THERMAL DEGRADATION Tu = To + qh-2k‘ In( -)
T

in which T, is the temperature of the vented tree, T the

The thermal behavior of any arbitrary volume is given radius of the cylinder, h the length of the cylinder.
by the relation
In relation (6) the heat flow through the surface A is
dT neglected. The temperature in the water tree has a max-
p c , - - v .k’VT = qh (4)
at imum for T = 2h/&, thus
With p the density of the material, C,, the specific heat, c, E,”h2
k’ the thermal conductivity, qh the heat dissipation. The Tu < To +7 ek (7)
value of qh is given by Assuming that E, <E,, E, = 2x106 V/m, c, = 1.3x10-’
(am)-’, k’ = 0.25 J/Ksm, h < 1 mm, while e = 2.72.
qh = %Et (5)
in which q, and E, are the conductivity and the electric These assumptions result in
stress in the vented tree respectively. < +
Tu To 0.08K (8)
In agreement with these findings microscopic examina-
It is possible to assume that a vented tree is a conduct-
tions of vented trees do not reveal any thermal degrada-
ing material (water and salts) and that, moreover, the
tion. Moreover, Bamji et al. [ll]using Differential scan-
remaining electric stress in the tree is still considerable
ning calorimetry were not able t o detect any difference
( M E,). Calculations show that under these conditions
between the thermal history within PE affected by a tree
the temperature can become unstable ( a T / a t # 0). On
and the P E free from attack.
the basis of these assumptions the investigators Isshiki et
al. [66], Tanaka et al. [156], Yoshimura et al. [174] and
Meyer et al. [98] state that an increase of temperature 7.7 PARTIAL DISCHARGES
may play an essential role. Yoshimura suggested that
the water in the tree would even evaporate, with obvious Partial discharges may occur in gas-filled cavities under
consequences for the polymer. high electric stresses. Under the application of an ac-field
IEEE Transactions on Electrical Insulation Vol. 25 N o . 5 , October 1990 1015

a continuous repetition of the process causes degradation are known to either interrupt the electrochemical process
of the surrounding insulating material, so that electrical or stabilize H 2 0 2 , strongly diminish the formation of wa-
treeing may occur eventually. These discharges have been ter trees, supports this supposition.
studied for many years [82,84,90,91].
Steennis [148] states that water enters the amorphous
In Section 4.4 it was concluded that vented tree growth regions of polymers if polar impurities are present. He
a t moderate stress levels is in all likelihood not accompa- gives several reasons for the creation of polar regions
nied by such discharges. The experience with partial dis- (which can be seen as a vented tree initiation process).
charge measurements on medium-voltage cables confirm For instance:
this, even if these cables contained large vented trees.
The noise level during the experiment of Steennis [148] 1. pollution or oxidation of the compound or cable insu-
was FZ 0.1 pC, the mean electric stress level in the insu- lation during production,
lation was < 5 kV/mm. 2. diffusion of impurities from a semiconducting screen,
which was also estabilished by Crine [31] and Johnson
However, there are two different situations in which ~71,
partial discharges can be related to vented treeing. In 3. scratching of the insulation surface during or after ca-
the first situation partial discharges may occur during ble production. This will produce polymer chain ends
initiation. If water needles are applied, the very high which may easily be oxidized.
initiation stress level a t the tip of the water needle may
cause partial discharges. This accounts for the light emis-
sion from the water needle electrodes in a test by Nitta
[114]. Secondly it is possible for electrical treeing to orig-
inate from vented trees, for example during overvoltages.
These electrical trees cause partial discharges. Examples
have already been presented in Figures 18 and 19.

Finally, it is assumed by Dakin [37] that perhaps micro

discharges could play a role in the process of water tree
growth.

7.8 ELECTROCHEMICAL
D EG RA DAT IO N

The concept of chemical action in relation to water tree

growth in P E has been mentioned by several authors. Rye polar group
e t al. [128] showed that in the absence of a n electric field, Figure 26.
warm solutions of some salts (sodium chloride and vari-
Polar groups (.) fixed on the polymer chains at a
ous copper salts) attack the interface of P E chemically, scratched PE surface.
possibly oxidizing it. It is possible that such a process
increases the tendency to water tree growth in PE. It has Figure 7 shows an example of a vented tree initiate(
still t o be explained why a tree structure arises. a t such a location. In the cavity were species contain-
ing silicon, sulfur and calcium (Figure 8). Growth due
Tabata et al. [153] found tree structures in voltage en-' to electrochemical action is illustrated in Figure 26 [148].
ergized cable insulation which had been in a solution of In a cross section of an insulation surface, polar groups
H2S and water. The trees had grown from a copper con- are attached to the polymer chains. In Figure 27 an en-
ductor, and contained CuzS and Cu20. This makes the largement is shown. Water enters the amorphous regions
contribution of a chemical action to water treeing proba- as far as these regions contain polar groups, the front of
ble. Yoshimitsu et al. [171] suppose water tree formation this water intake is represented by line F . In the polar
in epoxy resins t o be accompanied by chemical reactions. region, up to front F , charge transport by ions can take
place, represented by c. Beyond front F the polymer
Henkel et al. [63] supposed a n electrochemical process is *pureand acts as a dielectric that transmits capacitive
a t the P E interface, in which H 2 0 2 is generated in water, currents, represented by the permittivity cp of the poly-
successively attacking the PE. The fact that agents which ethylene in the Figure. At interface F the transport of
1016 Steennis et al.: Water Treeing

t P -
-r
F = t F

limiting void with Iimitin g

diameter tropped solt diameter

b
crystalline at.'. cotions
noter
oOoO anions
amorphous
Figure 27.
LVJ ionic endgroups PE outside watertree
Water intake at a polarized insulation surface
[148] and consequent tree growth by electroly- Figure 28.
sis. Up to front F the degraded polar polymer A model of the structure of water trees (After [127])
is slightly conductive by ion movement. At F the
ions lose their charge and free radicals are formed ed that acidic sulfur-oxygen groups may be the result
that attack the virgin polymer. The ionic current of oxidized antoxidants, which derive their functionality
through cv is maintained by a capacitive current from a sulfur atom; in contact with water these groups
through c p so that the mechanism works at ac
may be converted to salts. It shows that antioxidants
voltage only.
may play a n important role in the process of water tree
growth. Ross observed with oxygen plasma exposition
ions is stopped and charge transfer takes place by elec-
that the oxidative stability is locally reduced by water
trolysis as has been described by Kao et al. [70]. Free
trees, which is another argument that oxidation process-
radicals or oxydizing agents such as H202 are formed.
es are important. According to Ross there is a further
The polymer is attacked, the front shifts further into the
argument. This argument is found in the growth rates
polymer and a polar path is created: the polar path con-
of vented trees in needle tests when different salt solu-
taining water forms a vented tree. The vented tree can be
tions are applied (for a description of the test results,
considered as a very high-resistive electrolyte: the ionic
see Section 5.6). The strongest tree growth is obtained
conductivity is low and a high potential difference re-
with copper sulfate. According t o Ross, sulfates can be
mains over the electrolyte. Because the available voltage
considered as oxidizing agents and copper ions are noto-
remains high, the vented tree path may contain several
rious catalysts in oxidation processes. Ross further in-
of such fronts in series. The vented tree is thus consid-
troduced the selective permeability for different entities
ered as an insulator, although a poor one. As the ionic
locally, seeming to bear a close resemblance to ionomer
current in the water tree in this theory is maintained by
membranes. As a result of the selective permeability (at
capacitive current through the unaffected dielectric, tree
places with limiting diameter in Figure 28), salts may be
growth takes place at ac voltage only which corresponds
trapped once the electric field is removed.
with the phenomenology described in Section 5.2.

In 1990 Ross [127] introduced a n advanced model of

7.9 CONCLUSIONS
the structure of water trees in the nm range, see Fig-
ure 28. In this model the hydrophilic nature of water Osmosis and capillary action are not related to electric
trees is explained by the presence of chemically or phys- stress, which is the reason why they cannot be consid-
ically bonded ionic groups. It is assumed that oxidation ered as the cause of water tree growth. However, both
occurs during water tree growth as carboxylate groups processes may play a secondary role, for instance during
and ionic sulfur-oxygen groups (observed with FTIR) can the initiation of a water tree or by pushing water into a
be regarded as possible oxidation products. It is suggest- polarized channel.
IEEE Transactions on Electrical Insulation Vol. 25 No. 5 , O c t o b e r 1990 1017

Partial discharges have never been detected during the on cables, even for cables with small water trees. Espe-
growth of a vented tree under moderate aging conditions cially dc stresses of several times the rated voltage could
and are thus not considered in the first place as the basis lead to cable insulation damage.
of water tree growth. Nevertheless if such discharges are
involved, the level is below detectability. Alternative test procedures, such as 0.1 Hz [60] and os-
cillating wave [6] are under investigation for instance in
Assuming a water tree represents a conducting mate- Germany and in The Netherlands. It was found recent-
rial, various basic mechanisms of degradation are con- ly (to be published, Steennis, Jicable 1991) that 0.1 Hz
ceivable such as thermal degradation and degradation by easily discriminates between cables with different rates of
Coulomb forces. About the role of dielectrophoresis there degradation.
are different opnions. In any case, dielectrophoresis might
be of assistance by carrying water vapor to the tip of a
A further alternative would be the measurement of the
vented tree.
dc current under dc load conditions. According to Hara-
sawa et al. [61] this measurement technique has the dis-
On the other hand, assuming a vented tree is not a
advantage that only heavily deteriorated cables can be
conducting material but an insulating material, and by
discriminated from other cables. A further disadvantage
selecting aging parameters to comply with the actual use
is that still the application of the suspected dc load is
of medium-voltage cables, most of the mechanisms men-
required.
tioned above become rather improbable. Under these as-
sumptions, however, electrochemical degradation can be
considered as a significant mechanism. A more sophisticated procedure was described earlier
by Oonishi et al. [117]. They found that the dc leakage
current under negative voltage application is larger than
8. TEST METHODS under positive voltage application. As a consequence of
this polarity effect, the ac voltage will reveal a dc charg-
8.1 FIELD TESTING ing current or a dc leakage current. Their examinations
showed a good correlation between the breakdown volt-
age, the size of vented trees and this leakage current.
1. Experience with dc tests on service-aged cables con- Moreover, the measurement of the current can be made
taining water tees (vented trees) shows that many cables, without interruption of the cable usage. The results indi-
after a first ac breakdown and repair, have an extreme- cate that both extensively deteriorated and moderate de-
ly low dc breakdown voltage. Often breakdown voltages teriorated cables can be discriminated from cables having
smaller than the rated voltage of the cable were report- small water trees only. It is not possible to distinguish
ed and further dc breakdowns reveal an ever decreasing cables being deteriorated by bow-tie trees since in this
breakdown voltage level [49,60,150]. case there is no polarity effect. It is the authors’ opin-
ion that this disadvantage is not very important, because
2. To find the effect of the dc field test on the 50 He bow-tie trees are less detrimental. However, cables hav-
breakdown voltage level short cable lengths have been ing vented trees grown from both sides of the insulation
stressed with a dc voltage up t o several times the rated may probably be less sensitive t o this procedure.
voltage after which the 50 Ha breakdown voltage level
was measured. It was found that in these specific cases
there was not a reduction of the 50 Hz breakdown voltage
8.2 ACCELERATED AGING
level, although the cables contained many large vented
trees [20,60,145].
The evaluation of the water tree susceptibility of a cable
Gronefeld [60] explained the apparent contradiction be- insulation and in particular the evaluation of water tree
tween l and 2 as follows: under the application of a dc retardant insulating materials must be made by means of
stress space charges can reduce the stress enhancements accelerated aging tests. It is the object of such a test to
in the vicinity of water trees. Then, traveling waves of op- obtain an indication of the service behavior of the cable
posite polarity, occurring after breakdown, will give rise insulation in a short period of time.
to relatively high stress enhancements a t the tip of the
water tree which may result in degradation. The development of such a procedure is rather difficult,
as’history shows. A material test is still under develop-
Although it is difficult t o prove this assumption, prac- ment [131]. With respect t o full-scale cable tests, the
tical experience gives enough reasons to avoid dc stresses AIEC test [2] and the ‘Accelerated life test’ [87] (ACLT,
1018 Steennis et al.: Water Treeing

introduced by Lyle and Kirkland) are being used espe- 2 cable samples, each a t least 5 m long, subjected to a
cially in the United States. In Europe, the 'Accelerat- step test until breakdown. A breakdown voltage of 10
ed water tree aging test' [134] (ACWT, introduced by kV/mm minimum is required for each of the two cable
Schroth, Kalkner and Fredrich) is under consideration. samples. Moreover visual inspections are carried out on
The effectiveness of these three tests can be questioned. slices taken near the places of breakdown. In the 'Stufen-
It was shown that, probably as a consequence of the tem- test' described by Weck [165] three cable samples of 15 m
perature regimes applied, the tests lead to an extensive are subjected to a step test. The step test is terminated
bow-tie tree growth and a reduction of the sizes of the a t a level of 9 x the rated cable voltage (15 to 20 kV/mm).
more important vented trees. Therefore, the water tree It is assumed that water treeing is of minor importance
growth patterns obtained are not according t o service ex- in cables having breakdown voltages above this level. Vi-
perience [134,148]. It was shown also that a bad cable sual inspections are carried out both on randomly chosen
passed the AEIC test [148]. insulation volumes and near and a t the breakdown sites.

Alternative test procedures are under development, for

instance within CIGRE WG 21-11 [113]. As long as a
few thousand hour test is not available, a 2 yr test with
moderate test conditions (such as 2 . 5 ~the rated volt-
age, room temperature, power frequency) can be applied
according to [148].

8.3 C HA RACT ER lZATl0N TESTS

In order to determine the level of degradation of aged

cables (and if necessary the reference level of unaged ca-
bles) a characterization test is carried out. Such a test
comprises the following three basic elements:

1.supply of several cable samples is necessary for break-

down tests and visual inspections in the laboratory,
2. breakdown voltages (in most cases a step voltage test)
give information on the electrical strength of the cable
insulation,
3. visual inspections give information on the sizes, densi-
ties and types of water trees. 01
0
I
10
'
20
'
30
'
40
'
50
I
60 70
I
80
I '
90
I
100

Breakdown voltages are found by testing cable samples largest woter tree observed in insulotion
[ % of the insulation thickness]
with a length of a t least about 5 m. The mean break-
cable circuit with breokdown(s) during
down voltage level can be calculated by means of Weibull J
service operation
statistics [14, 1581.
Figure 29.
Visual inspections are carried out on dyed slices of the The 63% breakdown stress level and the largest
cable insulation. Microscopic evaluations are necessary water tree found in the insulation plotted for dif-
to find water trees with sizes in the range of 100 p m t o ferent cables in one graph. In this Figure, the ca-
1 mm. The slices can be taken from randomly chosen ble circuits which had failures during service are
places, the vicinity of the places of breakdown and the indicated with a lightning mark [149].
breakdown site.
An extensive but very informative test procedure is the
AEIC [2], Weck [165] and CIGRE [149] recommend so called characterization test of CIGRE [149]. The total
their own characterization test procedures. number of 5 m cable samples subjected to the breakdown
test is 12. In this test (apart from visual inspections on
After aging in the AEIC qualification test [2] the rate randomly chosen small cable samples) visual inspections
of degradation of the cable is characterized by means of are carried out a t the breakdown sites. In order to ob-
the ac breakdown test. The characterization is based on tain better visual inspection results, it is recommended to

r
IEEE Transactions o n Electrical Insulation Vol. 25 N o . 5 , October 1990 1019

adjust the equipment for the breakdown tests in such a Concerning the swelling material under the outer sheath,
way that the energy dissipated during breakdown is min- these materials have the ability t o absorb water vapor dif-
imized. As a consequence the actual causes of breakdown fusing through the plastic outer sheath into the cable. In
are often found which is very informative (examples are this way, by absorption, the swelling material is able to
given in Figures 12, 18 and 19, vented trees were found keep the relative humidity low for a long period of time
as the cause of breakdown). Especially for cables hav- for cables having a HDPE outer sheath (with a relative-
ing water trees exceeding about 30% of the insulation ly low water diffusion coefficient). For example, assuming
thickness this inspection is useful. The largest water tree dry swelling materials right after production, it is expect-
observed in combination with the 63% breakdown stress ed that this period will be a few decades at least. It is
level represent one dot in a graph (see Figure 29). This the authors’ opinion that it would be of interest to prove
graph makes it possible to get a n indication of the level of that with a MDPE outer sheath (having a much higher
degradation of the involved cable (each dot in this graph diffusion coefficient) this period is long enough to avoid
represents one cable circuit under investigation) in com- t o much water in the cable. Such a situation would be
parison t o other cables. The region of bad cables (indicat- reached when in and near the cable insulation the relative
ed with lightning marks) starts with a (63%) breakdown humidity level exceeds a level of x 70% (also see Section
voltage smaller than about 15 kV/mm and with water 5.5).
trees having a length exceeding
thickness.
- 40% of the insulation
9.2 WATER TREE RETARDANTS
9. MEASURES TO REDUCE
WATER TREEING Compound manufacturers and cable manufacturers, of-
ten in close cooperation, put much effort in the develop-
different methods can be applied t o obtain a bet- ment of water tree retardant insulating materials. Pub-
T WO
ter aging performance of cables under wet conditions.
On the one hand it is possible to make water-tight cable
lications show that developments are based on differ-
ent philosophies concerning the mechanism of water tree
constructions. On the other hand it is possible to accept growth. As will be shown, the water tree retardant mate-
water in contact with the insulation and to apply so called rials presented are modified polymers, polymers in which
water tree retardant insulating materials. chemical additives are incorporated or both. In many
cases a modification of the polymer is required to prevent
that these additives become fugitive. The test results as
9.1 WATER TIGHT CONSTRUCTIONS described in the various publications give the impression
that tree retardants can be effective a t least over the lim-
ited time period of the tests.
In many countries today the utilities install cables with
water-tight constructions having longitudinal water block-
ings or a combination of longitudinal and radial water Already in 1980 Soma et al. [141] formulated a bow-tie
blockings. tree inhibiting material. Soma assumed that bow-tie trees
are initiated by condensation of water in a void under the
To avoid any radial water ingress by diffusion utilities action of an electric field. Further condensation will lead
often apply cables having radial water barriers. Such a to a pressure build up and creep of the void surface. The
radial water barrier for medium voltage cables is usually mechanically damaged surface contains polar groups and
a metal foil under the plastic outer sheath. Water vapor water will enter the polymer. The water tree retardant
is not able to permeate through the metal. Examples material (intended t o suppress bow-tie tree development)
of such constructions are given in many publications, for is based on additives having hydrophilic groups absorb-
instance by Bourjot [22], Nagabasami [lo81 and Bow [23]. ing the water in the polymer structure around the voids.
The publication describes different additives and presents
To avoid axial water ingress under the outer sheath af- their effectiveness in a 3000 h test on a full-scale cable.
ter a damage of this sheath, longitudinal water barriers
can be applied. These barriers are situated in between In 1984 Nagasaki et al. [log] presented a water-tree re-
the outer sheath and the cable core and mostly consist of tardant insulating material containing a n ethylene copoly-
swelling tapes or swelling powders [135]. Cable damage mer (EVA) that acts as a barrier against water tree growth.
can be expected also when liquid water from the soil can The cause of this retardant effect has not been clarified.
reach the conductor, but this conductor can be made wa- Life tests up to 30000 h were carried out on medium-
ter tight for instance by means of a solid conductor or by voltage cables showing that the material suppresses the
the application of swelling powders between the strands. growth of both bow-tie trees and vented trees.
1020 Steennis et al.: Water Treeing

One year later Matey et al. [92] presented a modified and a reduction of the amount of impurities in these lay-
base polymer which should intrinsically be more resistant ers would be helpful in retarding the water tree growth,
to water tree degradation. Matey assumes that mechani- especially the vented tree growth. A study of types and
cal fatigue plays an important role in the process of water amounts of impurities is given by Belhadfa et al. [13]. Im-
treeing. The modified polymer system should offer a n en- purities diffusing from the layers into the insulation are
hanced resistance to crack propagation. In addition the described by Crine et al. [31] and Johnson et al. [67]. De-
resistance is enhanced by water treeing inhibitor, being velopment of semiconducting layers with lower contami-
an organometallic compound. With respect to this mate- nants levels have been reported for instance by Nitta [115]
rial, it is assumed by Saure et al. [132] that by migration and Umpleby [161].
the polar additives have the ability to reduce the electric
field stress on places of field enhancement. In this publi- Unfortunately, the above described developments and
cation Saure emphasized that the basic polymer is a P E test results are not informative about the actual degra-
copolymer stabilizing the presence of the additives. Short dation mechanism, since modified polymers as well as
duration tests show that both vented trees and bow-tie additives can be considered as polar material. For all
trees are reduced in number and length. Long-term ag- mechanisms discussed above, polar groups have water-
ing tests up to 10000 h reveal the good performance of tree retardant consequences since polar groups are able
the material both with respect to the breakdown strength to reduce the electric stress locally (effective for all mech-
and the bow-tie tree development. However, no data are anisms) or to reduce the interfacial energy of polar in-
presented concerning the growth behavior of the vented terfaces which reduces the absorption of water in general
trees. (effective especially considering electrochemical degrada-
tion). It is concluded that water-tree retardant insulating
In 1986 Fisher et al. [49] described the development materials or a combination of these materials with smooth
of a water-tree retardant insulating material called HF- and clean semiconducting layers can be considered as se-
DA 4202. The development program was intended to rious candidates for solving the water treeing problem.
find a material without a filler and with additives being However, in many cases long duration tests at least up
nonfugitive. From [119] it is known that the additive is to two years have not yet been performed; such tests are
an organo-metallic compound. Short-term water needle required to come to better evaluations of the character-
tests show the retardant characteristics of this material. istics of these materials on the long term. These tests
No more information about test results of long-term tests should emphasize material stability and fugitivity of the
have been found. additives. Moreover, tree growth of especially the vented
trees should be taken into account.
One year later Fischer et al. [48] described a water--
tree retardant material which is based on a modified base
polymer with nonfugitive additives. The additives are REFERENCES
known to stabilize hydroperoxides by deactivation of the
catalytically active metal ions via adsorption and by a [l] K. Abdolall, H. E. Orton, M. W. Reynolds, B. D.
deactivation of the initiating sites at which the assumed Robert, R. Kennedy and B. P. Clayman, “Some
electrochemical reactions occur. The background of this Physicochemical Aspects of Water Trees”, Annual
electrochemical process during tree growth was published Report Conference on Electrical Insulation and Di-
two years earlier by Henkel and Muller [63]. Fischer per- electric Phenomena, pp. 604-614, 1982.
formed aging tests up to 6000 h and observed a reduction
of both densities and lengths of bow-tie and vented trees. [2] AEIC, “Specifications for Thermoplastic and
Crosslinked Polyethylene Insulated Shielded Power
Recently, Field et al. [43] described the results of 3000 h Cables Rated 5 through 35 kV”,9th ed. AEIC CS5-
tests on full-scale cables. Assuming the water tree growth 87. Association of Edison Illuminating Companies
is related to free water, to ionic contaminants and to ox- (AEIC), Birmingham, Alabama, October, 1987.
idation products, different water tree retardant materi-
[3] D. Mc. Allister (ed), Electric Cables Handbook,
als were tested among which successfully a homopolymer
Granada; London, Toronto, Sydney, New York,
with an additive and two different modified polymers. In-
1982.
dication of tree size reduction of both bow-tie trees and
vented trees was obtained. Further testing to demon- [4] A. C. Ashcraft, “Treeing Update Part 111: Water
strate long-term benefits are in progress. Trees”, Kabelitems 152, Union Carbide Corporation.
Based on “Water Treeing in Polymer Dielectrics”,
Apart from water-tree retardant insulating materials it presented a t World Electrotechnical Congress in
is generally assumed that smooth semiconducting layers MOSCOW, June, 1977.
IEEE Transactions on Electrical Insulation Vol. 25 N o . 5 , O c t o b e r 1990 1021

[5] A. C. Ashcraft, (‘Factors Influencing Treeing Identi- on Electrical Insulation and Dielectric Phenomena,
fied”, Electrical World, December 1, pp. 3 8 4 0 , 1977. pp. 296-302, 1975.
[6] C. Aucourt, W. Boone, W. Kalkner, W. Naybour 161 B. S. Bernstein, “Recent Progress in Understand-
and F. Ombello, “Recommendations for a New Af- ing Water Treeing Phenomena”, Conference record
ter Laying Test method for HV Extruded Cable Sys- IEEE International Symposium on Ins. Electr. Mon-
tems”, 1990 International Conference on Large HV treal, June 11-13, pp. 11-21, 1984.
Electric Systems, CIGRE, Paper 21-105, Paris 1990.
171 F. W. Billmeyer jr, Textbook of Polymer Science,
[7] G. Bahder, G. S. Eager and R. G. Lukac, “Influence 3th edition. Wiley-Interscience Publication. Wiley &
of Electrochemical Trees on the Electrical Proper- Sons. New York, Chichester, Brisbane, Toronto, Sin-
ties of Extruded Polymeric Insulation”, Annual Re- gapore, 1984.
port Conference on Electrical Insulation and Dielec-
[18] S. A. Boggs, and S. Rizzetto, “The Application of In-
tric Phenomena, pp. 289-301, 1974.
terferometric Holography to the Study of Watertree-
[8] G. Bahder, C. Katz, J . Lawson and W. Vahlstrom, ing” , Conference Record of 1986 IEEE Internation-
“Electrical and Electrochemical Treeing Effect in al Symposium on Electrical Insulation, Washington
Polyethylene and Crosslinked Polyethylene Cables”, DC, June 9-11, pp. 140-143, 1986.
IEEE Trans. PAS, Vol. 93, pp. 977-990, 1974.
[19] W. Boone, E. F. Steennis, P. A. C. Bentvelzen and
[9] S. Bamji, A. Bulinski, J . Densley and N. Shimizu, A. M. F. J . van de Laar, “Development and Trial of
“Light Emission from Polyethylene Subjected to EHV XLPE Cables in The Netherlands”, Confkrence
Highly Divergent Fields”, Annual Report Conference Internationale des Grandes Rkseaux Electriques a
on Electrical Insulation and Dielectric Phenomena, Haute Tension (CIGRE), session September 1984.
pp. 592-597, 1982. Paper 21-02, 1984.
[lo] S. Bamji, A. Bulinski, J . Densley and A. Garton, [20] W. Boone, P. V. M. van Nes and E. F. Steennis,
“Etching and the Morphology of Crosslinked Poly- “Water, Onoverwinlijke of Overwinbare Vijand van
ethylene Cable Insulation”, IEEE Trans. EI, Vol. 19 Kunststof Middenspanningskabel”, Elektrotechniek,
No. 1, February, pp. 32-41, 1983. Vol. 68, No. 3, March, pp. 219-225, 1990.
[ll] S. Bamji, A. Bulinski, J . Densley, A. Garton and [21] 0. Bottger, R. Niederhagen and W. Schuchard,
N. Shimizu, “Water Treeing in Polymeric Insu- “Hypothese zur Water Tree Entstehung und deren
lation”, Confkrence Internationale des Grandes Uberprufung”, Dauerverhalten von Hochspannungs-
Rkseaux Electriques Haute Tension (CIGRE), ses- isolierungen, Electrotechnische Gesellschaft Fach-
sion September. Paper 15-07, 1984. berichte 16, VDE-Verlag GmbH, Berlin, pp. 109-
113, 1985.
[12] R. Bartnikas and R. M. Eichhorn (ed), Engineering
Dielectrics Volume IIA. Electrical Properties Solid [22] P. Bourjot, P. Gauthier and A. Pinet, “Compor-
Insulating Materials: Molecular Structure and Elec- tement thermomkcanique et ktanchkitk des cibles
trical Behavior, ASTM Special Technical Publica- moyenne tension a isolant synthktique (Themome-
tions 783. ASTM Publication Philadelphia, March chanical behavior and watertightness of medium-
1983. voltage insulated cables)”, Jicable 84, International
Conference on Polymer Insulated Power Cables,
[13] A. Belhadfa, A. J . Houdayer, P. F. Hinrichsen, G. Paris, March, pp. 63-67, 1984.
Kajrys, J . St-pierre, G. Kennedy, J.-P. Crine and N.
Burns, “Impurities in Semiconductive Compounds [23] K. E. Bow, “The Development of Underjacket Mois-
used a s HV Cable Shields”, IEEE, to be published; ture Barrier Cable as a Counter Measure Against
1989. Treeing”, IEEE Trans. PES, Vol. 5, No. 1, January,
pp. 47-53, 1989.
[14] G. Bernard, “Application of Weibull Distribution to
the Study of Power Cables Insulation”, (On behalf [24] E. L. Brancato and B. S. Bernstein (ed), Water Tree-
of Working Group 21-09 of CIGRE). Electra, Vol. ing Theory, Collection of papers and reports of EPRI
127, pp. 77-83, 1989. Workshop report, Monterey, California, December,
1985.
[15] B. S. Bernstein, N. Srinivas and P. N. Lee, “Electro-
chemical Treeing Studies: Voltage Stress, Tempera- [25] J. M. Braun, “Comparison of Water Treeing Rates in
ture, and Solution Penetration Effects under Accel- Steam and Nitrogen Treated Polyethylenes”, IEEE
erated Test Conditions”, Annual Report Conference Trans. EI, Vol. 15 No. 2, April, pp. 120-123, 1980.
1022 Steennis et al.: Water Treeing

[26] A. Bulinski and R. J . Densley, “The Voltage Break- [37] T . W. Dakin, “Rapporteur’s Report, Session D,
down Characteristics of Miniature XLPE Cables Treeing in Polyethylene”, Annual Report Conference
Containing Water Trees”, IEEE Trans. ET, Vol. 16, on Electrical Insulation and Dielectric Phenomena,
August, pp. 319-326, 1981. pp. 197-206, 1974.
[27] A. T. Bulinski, S. S. Bamji and R. J . Densley, “The [38] 0. Dorlanne, M. R. Wertheimer, A. Yelon and J .
Effect of Frequency and Temperature on Water Tree R. Densley, “ESR Study of Cu2+ in Polyethylene
Degradation of Miniature XLPE Cables”, IEEE Containing Electrical and Water Trees”, Annual Re-
Trans. EI, Vol. 21 No. 4, August, pp. 645-650, 1986. port conference on Electrical Insulation and Dielec-
tric Phenomena, pp. 136-143, 1980.
[28] G. Capaccio, W. Golz and L. J . Rose, “The In-
fluence of Morphology and Polymer Structure on [39] R. M. Eichhorn, “Treeing in Solid Extruded Electri-
the Watertreeing Resistance of Crosslinked Polyeth- cal Insulation”, Electronics & Power, February, pp.
ylene”, Elektrotechnische Gesellschaft Fachberichte 125-131, 1978.

ence Upon the Breakdown Strength of E P R Materi-

[29] E. A. Cherney, “Water Trees in Electrical Insula-
als”, Jicable 87, International Conference on Poly-
tion” , Ontario Hydro Research Quarterly, 3d quar-
mer Insulated Power Cables, Paris, pp. 193-198,
ter, pp. 7-12. 1973. 1987.
[30] B. H. Crichton, M. J . Given, 0. Farish and H. M. [41] E. Favrie and H. Auclair, “Effect of Water on Elec-
Bamford, “Measurement of Water Tree Growth by trical Properties of Extruded Synthetical Insulations
Instrumental Neutron Activation Analysis”, Journal Application on Cables”, IEEE Trans. PAS, Vol. 99,
of Physics, Part D: Applied Physics, Vol. 16, pp. No. 3 May/June, pp. 1225-1234, 1980.
L223-L226, 1983.
[42] R. F. Fedors, “Water-treeing as an Osmotic Phe-
[31] W . P. Crine, S. Pelissou, H. St-Onge, J . St-Piere, nomenon”, Polymer, Vol. 21, August, pp. 863-865,
G. Kennedy, A. Houdayer and P. Hinrichsen, “El- 1980.
emental and Ionic Impurities in Cable Insulation
[43] A. W. Field and V. A. A. Banks, “Development of
and Shields” , Jicable 87, International Conference
Water Tree Retardant Crosslinked Polyethylene Ca-
on Polymer Insulated Power Cables, Paris, pp. 206-
ble”, The 6th BEAMA International Electrical In-
213, 1987.
sulation Conference, Session 7, Brighton, May, pp.
[32] J . D. Cross and J. Y. Koo, “Some Observations on 254-258, 1990.
the Structure of Water Trees”, IEEE Trans. EI, Vol. [44] J . C. Filippini and C. T . Meyer, “Effect of Frequen-
19 NO. 4, August, pp. 303-306, 1984. cy on the Growth of Water Trees in Polyethylene”,
[33] J . Densley, “Water Treeing in HV Cable Insulation”, IEEE Trans. EI, Vol. 17 No. 6, December, pp. 554-
IEEE Canadian Communications and Power Confer- 559, 1982.
ence, pp. 269-270, 1974. [45] J . C. Filippini, C. T. Meyer and M. El-Kahel, “Some
Mechanical Aspects of the Propagation of Water
[34] J . Densley, A. Bulinski and T . Sudarshan, “The Ef-
Trees in Polyethylene” , Annual Report Conference
fects of Water Immersion, Voltage and Frequency on
on Electrical Insulation and Dielectric Phenomena,
the Electric Strength of Miniature XLPE Cables”,
pp. 629-637, 1982.
Annual Report Conference on Electrical Insulation
and Dielectric Phenomena, pp. 469-479, 1979 [46] J . C. Filippini, J . Y. Koo, Y. Poggi, C. Laurent,
C. Mayoux and S. Noel, “Aborescences d’Eau et
[35] L. A. Dissado, S. W . Wolfe and J . C. Fothergill, Aborescences Electriques (Watertreeing and electri-
“A Study of the Factors Influencing Water Tree cal treeing)”, Jicable 84, International Conference on
Growth”, IEEE Trans. EI, Vol. 18 No. 6, December, Polymer Insulated Power Cables, Paris, March, pp.
pp. 565-585, 1983. 87-92, 1984.
[36] L. A. Dissado, S. V. Wolfe, J . C. Fothergill, T. Jones [47] J . C. Filippini, J . Y. Koo and J . L. Chen, “Electrode
and S. M. Rowland, “The Distribution in Length of Influence on the Properties of Water Trees in Poly-
Vented Water Trees”, IEEE Trans. EI, Vol. 21 No. ethylene”, IEEE Trans. EI, Vol. 12 No. 1, February,
4, August, pp. 651-655, 1986. pp. 75-82, 1989.
IEEE Transactions on Electrical Insulation Vol. 25 No. 5 , October 1990 1023

P. Fischer, E. F . Peschke, R. G. Schroth and A. A. [59] W. Golz, “Water-tree Growth in Low-density Pol-
Farkas, “Development and Realization of Watertree yethylene”, Colloid and Polymer Sci, Vol. 263, pp.
Retardant XLPE Compounds and Cables”, CIGRE 286-292, 1985.
Symposium Vienna, 1987.
[60] P. Gronefeld, R. von Olshausen and F. Selle, “Fehler-
E. J . Fisher and B. MC. Clung, “Long-life Insula- erkennung und Isolationsgefahrdung bei der Prufung
tion for Industrial and Utility Cables”, IEEE Trans. water tree-haltiger VPE-Kabel mit Spannungen
IA, Vol. 22, No. 5, September/October, pp. 946-951, unterschiedlicher Form”, Elektrizititswirtschaft, Jg.
1986. 84, Vol. 13, pp. 501-506, 1985.
R. Fourni, J. Perret, P. Recoup and Y. Le Gall, “Wa-
ter Treeing in Polyethylene for HV Cables”, Confer- [61] K. Harasawa, Y. Shimamura, H. Fugagawa, K. Soma
ence Record of 1978 IEEE International Symposium and M. Marumo, “Influence of dc Voltage Applica-
on Electrical Insulation, pp. 110-115, 1978. tion on Dielectric Performances of XLPE Cables”,
Jicable 87, International Conference on Polymer In-
E. A. Franke, J . R. Stauffer and E. Czekaj, “Wa- sulated Power Cables, Paris, pp. 125-132, 1987.
ter Tree Growth in Polyethylene under dc Voltage
Stress”, IEEE Trans. EI, Vol. 1 2 No. 3, June, pp. [62] H. J . Henkel, W. Kalkner and N. Muller, “Elec-
218-223, 1977. trochemical Treeing - Strukturen in Modelkabeliso-
H. Franke, H. Heumann and D. Kaubisch, “Testing lierungen aus Thermoplastischem oder Vernetztem
Possibilities and Results Regarding Water Aging of Polyet hylen” , Siemens Forschungs und Entwicklungs
PE/XLPE Insulated Medium Voltage Cables”, Jica- Bericht, Vol. 10, No. 4, Springer Verlag, pp. 205-214,
ble 84, International Conference on Polymer Insulat- 1981.
ed Power Cables, Paris, March, pp. 113-118, 1984.
[63] H. J . Henkel and N. Muller, “Additive zur Inhibie-
D. Fredrich and W. Kalkner, “On the Examination rung von Wasserbaumchen in Polyolefinen fur Kabel-
of the Watertreeing Behavior of XLPE-insulated Ca- isolierungen”, Electrotechnische Gesellschaft Fach-
bles”, Conference report of the Fifth International berichte 16, VDE-Verlag GmbH, Berlin, pp. 132-
Symposium on HV Engineering. Paper no. 21-04. 134, 1985.
Braunschweig, Federal Republic of Germany, August
24-28, 1987. [64] A. A. Hossam Eldin and Z. El Shazli, “A Study of
Water Tree Frequency Dependence and Growth Rate
H. Fukagawa, T. Okamoto, N. Hozumi and T. Shi-
in Polymer”, Annual Report Conference on Electri-
bata, “Developments of Methods to Estimate the
cal Insulation and Dielectric Phenomena, pp. 638-
Residual Life of XLPE Cables Deteriorated by Wa-
643, 1982.
tertrees”, Jicable 87, International Conference on
Polymer Insulated Power Cables, Paris, pp. 457-462, [65] E. Ildstad, Water Migration and Watertreeing in
1987. crosslinked Polyethylene Cables, Dissertation. Uni-
A. Garton, R. J . Densley and A. Bulinski, “Initial versity of Trondheim, Norway, 1982.
Results of a Study of Water Trees in XLPE by In-
frared Spectroscopy”, IEEE Trans. EI, Vol. 15 No. [66] S. Isshiki, M. Yamamoto, S. Chabata, T. Mizoguchi
6, December, pp. 500-501, 1980. and M. Ono, “Water Tree in Crosslinked Polyethyl-
ene Power Cables”, IEEE Trans. PAS, Vol. 93, pp.
A. Garton, S. Bamji, A. Bulinski and J . Densley, 1419-1429, 1974.
“Oxidation and Water Tree Formation in Service-
aged XLPE Cable Insulation”, IEEE Trans. EI, Vol. [67] J . F. Johnson, J. H. Groeger, M. S. Mashikian, B.
22 NO. 4, August, pp. 405-412, 1987. S. Bernstein and A. R. Cooper, “Sensitive Analyt-
W. S. M. Geurts, “Veroudering van Kabels met Iso- ical Methods for Early Diagnosis of Aging in Solid
latie van Lage-dichtheids Polytheen” , Research re- Insulation under Voltage Stress”, Confkrence Inter-
port. Ref. I11 7722/85, N. v. KEMA, Arnhem, The nationale des Grandes R6seaux Electriques A Haute
Netherlands, (in Dutch) 1985. Tension (CIGRE), session August/September, Paper
15-01, 1988.
T . Gloger and R. von Olshausen, “Die Auswirkungen
von Water Trees auf die Gleich- und Stossspannungs- [68] S. Kageyama, M. Ono and S. Chabata, “Microvoids
festigkeit von Polyethylen- Modelprufkorpern”, in Crosslinked Polyethylene Insulated Cables”, IEEE
Electrotechnische Gesellschaft Fachberichte 16, Trans. PAS, Vol. 94, No. 4, July/August, pp. 1258-
VDE-Verlag GmbH, Berlin, pp. 135-139, 1985. 1263, 1975.
1024 Steennis et al.: W a t e r Treeing

[69] W. Kalkner, U. Muller, E. Peschke, H. J . Henkel and [79] J. W. Kirkland, R. S. Thiede and R. A. Reita, “Eval-
R. von Olshausen, “Watertreeing in P E and XLPE uating the Service Degradation of Insulated Pow-
Insulated HV Cables” , International Conference on er Cables”, Conference paper (81 TD645-1) of the
Large HV Electric Systems (CIGRE). Paper No. 21- IEEE PES, Transmission & Distribution Conference
07, 1982. and Exposition, Minneapolis, Minnesota, Sept. 20-
25, pp. 2128-2136, 1981.
[70] K. C. Kao and W. Hwang, Electrical Transport in
Solids, (with particular reference to organic semi-
[8O] K. D. Kiss, H. C. Doepken and N. Srinivas, “Ag-
conductors) International Series in the Science of the
ing of Polyolefin Electrical Insulation”, In Durability
Solid State, Vol. 14, Pergamon Press, 1981.
of Macromolecular Materials, ed.: R. K. Eby, ACS
[71] M. Karakelle and P. J . Philips, “The Influence of Symp. Ser. 95, pp. 433-466, 1979.
Structure on Water Treeing in Crosslinked Polyeth-
ylene - Tree Morphology and the Influence of Organ- [81] J. Y. KOO,J . D. Cross, M. El-Kahel, C. T . Meyer
ic Contaminants”, IEEE Trans. EI, Vol. 24, No. 6, and J . C. Filippini, “Electrical Behavior and Struc-
Dec., pp. 1101-1108, 1989. ture of Watertrees in Relation to their Propagation”,
Annual Report Conference on Electrical Insulation
[72] H. Karner, U. Stietzel, M. Saure and W. Golz, “De- and Dielectric Phenomena, pp. 301-306, 1983.
termination of Small Water Contents in Solid Or-
ganic Insulating and the Influence Of
[82] F. H. Kreuger, Detection and Location ofDjscharges:
ture on the Dielectric Properties”, Conference Inter- in particular in Plastic-insulated HV Cables, Thesis
nationale des Grandes Reseaux Electriques a Haute Delft University, 1960.
Tension, CIGRE, session September, Paper 15-02,
1984. [83] F. H. Kreuger and A. M. F. J van de Laar, “Silane
[73] H. Kato, N. Maekawa, S. Inoue and H. Fujita, “Ef- Crosslinked Medium Voltage Cable”, Cired 85, Inter-
fect and Mechanism of Some New Voltage Stabiliz- national conference on electricity distribution, Part
ers for Crosslinked Polyethylene Insulation”, Annual I. IEE Conference Publication Nr. 250. IEE, London,
Report Conference on Electrical Insulation and Di- pp. 231-233, 1985.
electric Phenomena, pp. 229-238, 1974.
[84] F. H. Kreuger, Partial Discharge Detection in HV
[74] C. Katz and B. S. Bernstein, “Electrochemical Tree- Equipment, Temple Press, London, 1964 & Butter-
ing a t Contaminants in Polyethylene and Crosslinked worths, London, 1989.
Polyethylene Insulation” , Annual Report Conference
on Electrical Insulation and Dielectric Phenomena, [85] A. M. F. J. van de Laar, “Silane Crosslinked Power
pp. 307-316, 1973. Cable: An Investigation of Cable Properties”, Inter-
[75] C. Katz, N. Srinivas and B. S. Bernstein, “Growth national Conference on Large HV Electric Systems,
Rate of Electrochemical Trees in P E and XLPE In- CIGRE. Paper no. 21-2, 1982.
sulation Systems: Effect of Water Nature”, Annual
Report Conference on Electrical Insulation and Di- [86] P. B. Larsen, “Methodes de Coloration d’utilisdes des
electric Phenomena, pp. 279-288, 1974. Arborescences d’Eau dans les Isolation Extrudees de
C5bles”, Electra, Vol. 86, pp. 53-59, 1983.
[76] C. Katz, G. S. Eager J r , E. R. Leber and F. E. Fis-
cher, “Influence of Water on Dielectric Strength and [87] R. Lyle and J . W. Kirkland, “An Accelerated Life
Rejuvenation of In-service Aged URD Cables” , Jica- Test for Evaluating Power Cable Insulation”, IEEE
ble 84, International Conference on Polymer Insulat- Trans. PAS, Vol. 100, No. 8, August, pp. 3764-3771,
ed Power Cables, Paris, March, pp. 127-134, 1984. 1981.
[77] H. S. Kaufmann and H. J . Falcetta (ed), Introduc-
tion to Polymer Science and Technology, An SPE [88] L. Mandelkern, M. Glotin and R. A. Benson, “Su-
Textbook. LC 76-16838. Society of Plastics Engi- permolecular Structure and Thermodynamic Prop-
erties of Linear and Branched Polyethylenes under
neering Monographs. Wiley, New York, 1977.
Rapid Crystallization Conditions”, Macromolecules,
[78] K. Kawahara, T. Yoshimitsu, Y. Ishikawa, H. Aka- Vol. 14, pp. 22-33, 1981.
hori and H. Yoshida, “The Effect of Temperature on
the Electrical Breakdown Voltage of XLPE in the [89] T. G. Marsh, A. P. Smith and J . Drysale, “Long
Presence of Water Trees”, Conference record of 1984 Term Aging of Water Immersed XLPE Insulations”,
IEEE International Symposium on Electrical Insula- Jicable 87, International Conference on Polymer In-
tion, Montreal, June 11-13, pp. 33-36, 1984. sulated Power Cables, Paris, pp. 181-186, 1987.
IEEE Transactions on Electrical Insulation Vol. 25 No. 5, October 1000 1025

[go] J . H. Mason, “The Deterioration and Breakdown [lo21 W. J . Moore, Physical Chemistry, Fourth Edition,
of Dielectrics Resulting from Internal Discharges”, Inc. Prentice-Hall, Englewood Cliffs, New Jersey,
Proc. Instn. Eng. Electr. Part I, Vol. 98, pp. 44-59, 1972.
1951.
[lo31 M. Morita, M. Hanai and H. Shimanuki, “Some
[91] J. H. Mason, “Breakdown of Insulation by Dis- Preventive Methods for Water Treeing in P E and
charges”, Proc. Instn. Eng. Electr. Part IIa, Vol. XLP Insulation”, Annual Report Conference on
100, Nr. 3, pp. 149, 1953. Electrical Insulation and Dielectric Phenomena, pp.
[92] G. Matey, and J . D. Umpleby, “The Development of 303-312, 1973.
a Water Tree Retardant XLPE Insulation”, Dauer- [lo41 M. Morita, M. Hanai, H. Shimanuki, F. Aida and T.
verhalten von Hochspann ungsisolier ungen, Elec tro- Shiono, “Effects of Hydrated Ions on Watertrees in
technische Gesellschaft Fachberichte 16, VDE-Verlag Polyethylene” , Annual Report Conference on Elec-
GmbH, Berlin, pp. 170-173, 1985. trical Insulation and Dielectric Phenomena, pp. 195-
[93] H. Matsuba and E. Kawai, “Water Tree Mechanism 202, 1980.
in Electrical Insulation”, IEEE Trans. PAS, Vol. 95,
[lo51 J . Muccigrosso and P. J . Phillips, “The Morphology
No. 2, March/April, pp. 660-670, 1976.
of crosslinked Polyethylene Insulation”, IEEE Trans.
[94] H. Matsuba, E. Kawai and K. Sato, “Water EI, Vol. 13, No. 3, June, pp. 172-178, 1978.
Tree Mechanism and its Suppression”, Conference
Records (CH 1496-9/80) International Symposium [lo61 K-B. Muller, “Zum Langzeitverhalten neues
on Electrical Insulation, pp. 224-227, 1976. VPE-Kabel unter Feuchteeinwirkung”, Elektriz-
itatswirtschaft, Jg. 88, Vol. 26, pp. 1861-1871, 1989.
[95] K. Matsuura, H. Ohno, A. Ishibashi, K. Yatsuka,
K. Hirotsu, I. Kajiki and J . Shinagawa, “Investiga- [lo71 N. Muller and H. J. Henkel, “Phanomenologische
tion on Deterioration of 22 to 66 kV XLPE Cables Aspekte der Wasserbaumchem in Polyolefinen fur
Removed from Actual Service Lines”, Jicable 87, In- Kabelisolierungen”, Electrotechnische Gesellschaft
ternational Conference on Polymer Insulated Power Fachberichte 16, VDE-Verlag GmbH, Berlin, pp.
Cables, Paris, pp. 471-476, 1987. 119-122, 1985.

[96] E. J. McMahon, “A Tree Growth Inhibiting Insula- [lo81 T . Nagabasami, K. Sasaki, K. Sugiyama and M.
tion for Power Cable”, IEEE Trans. EI, Vol. 16, No. Ishitobi, “Development of XLPE Cables with New
4, August, pp. 304-318, 1981. Laminated Waterproof Layer”, Conference Record
of 1984 IEEE Symposium on Electrical Insulation,
[97] C. W. Melton, D. Mangaraj and M. Epstein, “Mor- Montreal, June 11- 13, pp. 4 1 4 5 , 1984.
phology of Thermoplastic and crosslinked Polyeth-
ylene Cable Insulation”, Annual Report Conference [lo91 S. Nagasaki, H. Matsubara, S. Yamanouchi, M. Ya-
on Electrical Insulation and Dielectric Phenomena, mada, T. Matsuike and S. Fukanaga, “Development
pp. 299-305, 1981. of Water-tree-retardant XLPE Cables”, IEEE Trans.
[98] C. T. Meyer, J. C. Filippini and N. Felici, “Water PAS, Vol. 103, No. 3, March, pp. 536-544, 1984.
Tree Propagation in Relation to Mechanical Proper- 1101 Y. Namiki, H. Shimanuki, F. Aida and M. Mori-
ties of Polyethylene”, Annual Report Conference on ta, “A Study on Microvoids and Their Filling in
Electrical Insulation and Dielectric Phenomena, pp. Crosslinked Polyethylene Insulated Cables”, IEEE
374-381, 1978. Trans. EI, Vol. 15 No. 6, December, pp. 473-480,
[99] C. T. Meyer, “Water Absorption During Water Tree- 1980.
ing”, IEEE Trans. EI, Vol. 18, No. 1, February, pp.
1111 R. D. Naybour, “The Growth of Water Trees in
28-31, 1983.
Cross Linked Polyethylene at Operating Stresses and
[loo] L. Minnema, H. A. Barneveld and P. D. Rinkel, “An Their Influence on Cable Life”, IEE Conference Pub-
Investigation into the Mechanism of Water Treeing lication nr. 177, pp. 238-241, 1979.
in Polyethylene HV Cables”, IEEE Trans. EI, Vol.
15, No. 6, December, pp. 461-471, 1980. [112] R. D. Naybour, “The Growth of Water Trees in
Dry and Steam Cured Polyethylene a t 1.9 MV/m,
[ 1011 T . Miyashita, “Deterioration of Water-immersed 50 Hz and their Influence on Cable Life in the Stress
Polyethylene Coated Wire by Treeing”, Proceedings Range 15 to 4 MV/m”, Annual Report Conference
1969 IEEE-NEMA Electrical Insulation Conference, on Electrical Insulation and Dielectric Phenomena,
Boston, September, pp. 131-135, 1969. pp. 620-628, 1982.
1026 Steennis et al.: Water Treeing

[I131 A. W. Nicholls and E. F. Steennis, “Water Tree- [124] M. Prigent, J . Bobo and I. Eyraud, “Aging of Pol-
ing, State of the Art - Progress Report of CIGRE yethylene in Water under Electrical Stress”, Confer-
WG 21-11 on Water Treeing in Extruded Cable In- ence Record on Dielectric Materials, Measurements
sulations”, 1990 International Conference on Large and Applications. July 21-25, Cambridge, pp. 32-35,
HV Electric Systems, Joint session paper 15/21-02, 1975.
CIGRE, Paris, 1990.
[125] R. Ross, W. S. M. Geurts and J . J . Smit, “FTIR
[114] Y. Nitta, “A Possible Mechanism for Propaga- Microspectroscopy and Dielectric Analysis of Wa-
tion on Water Trees from Water Electrodes”, IEEE tertrees in XLPE”, IEE Conference publication num-
Trans. EI, Vol. 9, No. 3, September, pp. 109-112. ber 289, Dielectric Materials, Measurements and Ap-
1974. plications, pp. 313-317, 1988.

[115] Y. Nitta and Y. Kawasaki, “Evaluation of a New [126] R. Ross, W. S. M. Geurts and J . J . Smit, J . H.
Material for Semiconducting Layer in XLPE Power van der Maas and E. T. G. Lutz, “Vented Water
Cables”, IEEE Trans. EI, Vol. 15, No. 2, April, pp. Trees - Oxidation and Hydrophilic Behavior”, Kema
140-143, 1980. Scientific Technical Reports, Vol. 7, No. 6, pp 357-
364, 1989.
[116] S. L. Nunes and M. T . Shaw, “Water Treeing in Pol-
[127] R. Ross, W. S. M. Geurts, J . J . Smit, J . H. van
yethylene - A Review of Mechanisms”, IEEE Trans.
der Maas and E. T . G. Lutz, “The Hydrophilic Na-
EI, Vol. 15, No. 6, December, pp. 437-450, 1980.
ture of Watertrees”, Conference record of the 1990
[117] H. Oonishi, F. Urano, K. Soma, K. Kotani and K. IEEE International Symposium on Electrical Insula-
Kamio, “Development of New Diagnostic Method for tion, Toronto, Canada, June 3-6, 1990.
Hot-line XLPE Cables with Water Trees”, Confer- [128] J. Rye, P. M. Brown, G. L. Le Poidevin and W.
ence Record 86 SM 397-4 of the IEEE PES 1986 T . Eeles, “Oxidation of Polyethylene as a Possible
Summer meeting, Mexico City, July, 1986. Prelude to Water Treeing”, IEE Conference Record
[118] H. Orton and K. Abdolall “An Investigation into on Dielectric Materials, Measurements and Applica-
the Morphology of XLPE Cable Insulation”, Annual tions. July 21-25, Cambridge, pp. 36-39, 1975.
Report Conference on Electrical Insulation and Di- [129] J . A. Sauer and K. D. Pae, “Mechanical Proper-
electric Phenomena, pp. 306-313, 1981. ties of High Polymers”, In Introduction to polymer-
[119] M. J . KEOGH, “Patent no. 0057 286 A2, dated De- ic science and technology, H. S. Kaufmann, and J .
cember 21, 1981”, Application number: 81110660.8. J . Falcetta, SPE Textbook. LC 76-16838. Society of
Applicant Union Carbide Corporation, 1981. Plastic Engineering Monographs. Wiley, NewYork,
NY 1977.
[120] R. Patsch, H. Wagner and H. Heumann, “Imhomo-
[130] M. Saure and W. Golz, “Uber den Einfluss von
geneities and their Significance in Single-layer Ex-
Mischungskomponenten auf das Water-Tree Wachs-
truded Polyolefine Insulations for Cables” , Interna-
tum in Polyolefin-Materialen und Prufverfahren zur
tional Conference on Large HV Electric Systems, CI-
Bewertung des Water-Tree Wachstums”, Electro-
GRE. Paper 15-11, 1976.
technische Gesellschaft Fachberichte 16, VDE-Verlag
[121] R. Patsch, “Effects of Moisture and Electrical Field GmbH, Berlin, pp. 127-131, 1985.
Strength on Polymer Insulations”, Colloid Polym. [131] M. Saure and W. Kalkner, “On Water Tree Testing
Sci. Vol. 259, pp. 885-893, 1981. of Materials - Status Report of CIGRE T F 15-06-
[122] R. Patsch and A. Paximadakis, “Water Trees in 05”, CIGRE Symposium 05-87, Vienna, Section 6.2,
Cables - Experimental Findings and Theoretical Ex- Number 620-10, 1987.
planations”, IEEE Conference Record 89 T D 350- [I321 M. Saure, W-D. Schuppe, H. Franke and D.
0 PWRD. Presented a t the IEEE PES 1989 Trans- Kaubisch, “Die Entwicklung des Mittelspan-
mission and Distribution Conference, New Orleans, nungskabels Rheytard mit watertreeretardierender
Louisiana, April, 1989. VPE-Isolierung”, Elektrizitatswirtschaft, Jg. 88,
Vol. 7, pp. 376-381, 1989.
[123] M. Pays, M. Louis, J . Perret, C. Alquie and
J. Lewiner, “Behavior of Extruded HVDC Pow- [I331 B. A. Schipper, “Dynamic Mechanical Thermal
er Transmission Cables: Tests on Materials and Analysis van Kunststoffen”, Research report. Ref.
Cables”, International Conference on Large HV I11 9168/84, N. V. KEMA, Arnhem, The Nether-
Electric Systems, CIGRE. Paper 21-07: 1988. lands, (in Dutch), 1984.
IEEE Transactions on Electrical Insulation Vol. 25 N o . 5 , October 1990 1027

[134] R. G. Schroth, W. Kalkner and D. Fredrich, “Test [145] N. N . Srinivas, B. S. Bernstein and R. A. Deck-
Methods for Evaluating the Water Tree Aging Be- er, Conference Record 89 T D 365-8 PWRD of the
havior of Extruded Cable Insulations”, 1990 Inter- IEEE PES Transmission and Distribution Confer-
national Conference on Large HV Electric Systems, ence, New Orleans, Louisiana, April, 1989.
Joint session paper 15/21-01, CIGRE, Paris, 1990. [146] E. F. Steennis and W. Boone, “Watertreeing in
[I351 W. D. Schuppe, “Progress with XLPE Insulated Service Aged and Accelerated Aged XLPE Cables”,
Medium Voltage Cables in the Federal Republic of Institute of Electrical Engineers (IEE). Conference
Germany”, Jicable 84, International Conference on Power Cables and Accessories 10 to 180 kV, Lon-
Polymer Insulated Power Cables, Paris, March, pp. don, pp. 162-166, 1986.
35-39, 1984. [147] E. F. Steennis, C. C. van den Heuvel and W. Boone,
“Accelerated Aging to Predict Watertree Behavior in
[136] M. T. Shaw and S. H. Shaw, “Water Treeing in Extruded Cables”, Jicable 87, International Confer-
Solid Dielectrics”, IEEE Trans. EI, Vol. 19, No. 5, ence on Polymer Insulated Power Cables, Paris, pp.
October, pp. 419-452, 1984. 161-167, 1987.
[137] J . Sletbak and A. Botne, “A Study of Inception and [148] E. F. Steennis, Water Treeing, the Behavior of Wa-
Growth of Water Trees and Electrochemical Trees ter Trees in Extruded Cable Insulation, Thesis Uni-
in Polyethylene and Crosslinked Polyethylene Insu- versity Delft. ISBN 90-353-1022-5; KEMA, Arn-
lations”, IEEE Trans. EI, Vol. 12, No. 6, December, hem, 1989.
pp. 383-388, 1977.
[149] E. F. Steennis and A. M. F. J. van de Laar, “Char-
[138] J . Sletbak, “A Theory of Water Tree Initiation and acterization Test and Classification Procedure for
Growth”, IEEE Trans. PAS, Vol. 98, No. 4, Ju- Water Tree Aged Medium Voltage Cables”, (On be-
ly/August, pp. 1358-1365, 1979. half of Working Group 21-11 of CIGRE). Electra,
Vol. 125, July, pp. 89- 101, 1989.
[I391 J. Sletbak and E. Ildstad, ‘IThe Effect Of Service
[150] E. F. Steennis, w . B~~~~ and A. ~ ~ ~ uwa- t f ~ ~ ~ t ,
and Test Conditions on Water Tree Growth in XLPE
ter Treeing in Service Aged Cables, Experience and
Cables”, IEEE Trans. PAS, Vol. 102, No. 7, July, pp.
Evaluation Procedure”, IEEE Trans. PES, Vol. 5,
2069-2076, 1983. No. 1, January, pp. 40-46, 1990.
[140] J . Sletbak and E. Ildstad, “The Validity of the Me- [151] S. G. Swingler, R. J. Jackson and J. Drysdale, “The
chanical Damage Theory of Water Treeing Tested Dielectric Response of Polyethylene Cable Contain-
Against Experimental Results”, Conference Record ing Water Trees”, Conference University of Lancast-
of the 1984 IEEE International Symposium on Elec- er, 10-13 September, pp. 183-186, 1984.
trical Insulation, Montreal, June 11-13, pp. 29-32,
11521 T . Tabata, T. Fukuda and Z. Iwata, “Investigations
1984.
of Water Effects on Degradation of Crosslinked Pol-
[141] K. Soma and S. Kuma, “Development of Bow-tie yethylene Insulated Conductors”, Paper 71 TP545-
Tree Inhibitors”, Conference Record of the 1980 PWR, IEEE Summer Meeting and International
IEEE International Symposium on Electrical Insu- Symposium on High Power Testing, Ore. Portland,
lation, pp. 212-215, 1980. July 18-23, pp. 1361-1370, 1971.
[153] T. Tabata, H. Nagai, T. Fukuda and Z. Iwata, “Sul-
[142] N. N. Srinivas, S. M. Allam and H. C. Doepken Jr,
fide Attack and Treeing of Polyethylene Insulated
“The Effect of crosslinking and crosslinking Agent
Cables - Cause and Prevention”, IEEE Trans. PAS,
by Products on Tree Growth in Polyethylene”, An-
Vol. 91, Nr. 4, pp. 1354-1360, 1972.
nual Report Conference on Electrical Insulation a n d
Dielectric Phenomena, pp. 380-385, 1976. [154] T . Tanaka and T. Fukuda, “Residual Strain and
Water Trees in XLPE and P E Cables”, Annual Re-
[143] N. N. Srinivas, H. C. Doepken, A. L. McKean and port Conference on Electrical Insulation and Dielec-
M. C. Biskeborn, “Electrochemical Treeing in Ca- tric Phenomena, pp. 239-249, 1974.
ble”, EPRI Final Report EL-647, 1978.
[155] T . Tanaka, T. Fukuda, S. Suzuki, Y. Nitta, H. Go-
[144] N. N. Srinivas, “Estimation of Life Expectancy of to and K. Kubota, “Water Trees in crosslinked Poly-
Polyethylene-insulated Cables”, EPRI Final Report ethylene Power Cables”, Paper T73497-5. IEEE PES
EL-3154, 1984. Summer Meeting & EHV/UHV Conference, Vancou-
ver, BC Canada, July 15-20, pp. 693-702, 1974.
1028 Steennis et al.: Water Treeing

[156] T. Tanaka, T . Fukuda and S. Suzuki, “Water Tree [167] W. D. Wilkens, “Thermal Gradient Effects in Pow-
Formation and Lifetime Estimation in 3.3 kV and er Cable Insulation in Wet Environments”, Annual
6.6 kV XLPE and P E Power Cables”, IEEE Trans. Report Conference on Electrical Insulation and Di-
PAS, Vol. 95, No. 6, Nov/Dec, pp. 1892-1900, 1976. electric Phenomena, pp. 260-269, 1974.
[157] T. Tanaka and A. Greenwood, Advanced Power Ca- [168] S. Wojtas, “Investigations of Polyethylene Insula-
ble Technology, Volume 11,Present and Future, CRC tion Resistivity of Power Cables”, Jicable 87, Inter-
Press, Inc. Boca Raton, Florida, 1983. national Conference on Polymer Insulated Power Ca-
bles, Paris, pp. 436-440, 1987.
[158] D. R. Thoman, L. J . Bain and C. E. Antle, “In-
ferences on the Parameters of the Weibull Distribu- [169] H. Y. Wong, Heat Transfer for Engineers, Longman
tion”, Technometrics, Vol. 11, No. 3, August, pp. Group Limited, Inc. Longman New York, 1977.
445-460, 1969.
[170] Y. Yamada, S. Yamanouchi and S. Miyamoto,
11591 W. R. Tinga, W. A. G. Voss and D. F. Blossey, “Treeing Phenomena in XLPE Insulation under dc
“Generalized Approach to Multiphase Dielectric Voltage” , Annual Report Conference on Electrical
Mixture Theory”, Journal of Applied Physics, Vol. Insulation and Dielectric Phenomena, pp. 500-510,
44, No. 9, September, pp. 3897-3902, 1973. 1979.
[160] D. M. Tu and K. C. Kao, “Effects of Hydrostatic [171] T. Yoshimitsu and T. Nakakita, “New Findings on
Pressure on Water Treeing Properties of Polyethyl- Water Tree in High Polymer Insulating Materials”,
ene”, Annual Report Conference on Electrical Insu- Conference Record, IEEE International Symposium
lation and Dielectric Phenomena, pp. 307-31 1, 1983. on Electrical Insulation, pp. 116-121, 1978.
[161] J . D. Umpleby, E. P. Marsden, R. J . Turbett and [172] T. Yoshimitsu, H. Mitsui, K. Hishida and H.
A. Mendelsohn, “Approaches to W T G R Crosslinked Yoshida, “Water Treeing Phenomena in Humid
Polyethylene”, IEE, Science, Education and Tech- Air”, IEEE Trans. EI, Vol. 18, No. 4, August, pp.
nology Division. Synopses of the opening remarks. 396-401, 1983.
Discussion meeting on “The Mechanisms of Water
Treeing in Polymeric Insulations”, March 14, 1988. [173] T . Yoshimitsu, H. Mitsui, S. Kenjo and T . Nakaki-
ta, “Some Considerations on ac Watertrees in
[162] W. Vahlstrom J r , “Investigation of Insulation De- Crosslinked Polyethylene”, IEEE Trans. EI, Vol.
terioration in 15 kV and 22 kV Polyethylene Cables 18, No. 1, February, pp. 23-27, 1983.
Removed from Service”, Paper 71 C42-PWR, IEEE
PES Underground Distribution Conference, Mich. [174] N. Yoshimura, F. Noto and K. Kikuchi, “Growth of
Detroit, Sept 27- 30, pp. 1023-1035, 1971. Watertrees in Polyethylene and Silicone Rubber by
Water Electrodes”, IEEE Trans. EI, Vol. 12, No. 6,
[163] H. Wagner, “Pseudo-spherulite Structures in December, pp. 411-416, 1977.
crosslinked Low-density Polyethylene” , IEEE Trans.
EI, Vol. 13 No. 2, April, pp. 81-86, 1978. 1751 N. Yoshimura and F. Noto, “Voltage and Frequency
Dependence of Bow-tie Trees in Crosslinked Polyeth-
[164] R. C. Weast (ed), Handbook of Chemistry and ylene”, IEEE Trans. EI, Vol. 17, No. 4, August, pp.
Physics, 61st edition 1980-1981, CRC Press. Inc. Bo- 363-367, 1982.
ca Raton, Florida 1980.
1761 R. H. Zeller, “Thermodynamics of Watertreeing”,
[165] K-H. Weck, “Stufentest eur Ermittlung des IEEE Trans. EI, Vol. 22, No. 6, December, pp. 677-
Isolationszustands betrieblich vorbeanspruchter 681, 1987.
PE- und VPE-Mittelspannungskabel”, Elektrie-
itatswirtschaft, Jg. 88, Vol. 8, pp. 470-473, 1989. Manuscript was received on 27 Jun 1990, in revised form 1 9
Sep 1990.
[166] J . A. Wiersma, “Watertreeing in Cables with Ex-
truded Insulation”, Electra, Vol. 55, pp. 25-38, 1977.