IEEE Transactions on Electrical Insulation, Vol. 23 No. 2, April 1988 319

Dielectric Study at Microwave Frequencies of Halos in

Crosslinked Polyethylene Cable

T.K. Bose and M. Merabet Groupe de recherche sur les dielectriques

Departement de physique UniversitC du Quebec a Trois-Flivikres Trois-Rivikres Quebec, Canada

J.P. Crine and S. P6lissou Institut de recherche d’Hydro-Quebec

Varennes, Quibec, Canada

ABSTRACT We have studied the dielectric properties of halos in crosslinked polyethylene cable by using the method of time domain reflec- tometry (TDR) between 1 MHz and 8 GHz. Although the halo is predominantly composed of free water, there also seems to be present a low relaxation frequency which is probably due to bound water or impurities. The water contents deduced from the TDR method and a Mitsubishi Moisturemeter are comparable.

INTRODUCTION he presence of a halo in the insulation of steam- T cured crosslinked polyethylene (XLPE) cables is a

well-documented fact [l]. It has been shown that halos are essentially water clusters filling the microvoids gen- erated during the curing [l]. Water in cable insulation will increase its dielectric losses and conductivity and will reduce its dielectric strength. But the worst influ- ence of water on cable insulation is to induce the degra- dation phenomenon known as water-treeing [2]. The exact amount of water required to generate water trees in P E or XLPE is still a matter of debate [2,3] but it is obvious that knowledge of the water content in ca- ble insulation, and especially in tree regions, is of prime importance. There exist several techniques that could measure the water content in dielectrics [4] but few allow

a distinction to be made between free and bound wa- ter. It was shown recently [5] that water in P E could be bonded to various impurities and oxidation by-products. It has sometimes been suggested that the nature of the chemical bond between water and P E (or XLPE) could affect treeing [2]. Dielectric property measurements in the region of microwaves and radio-frequencies may en- able one t o distinguish between the two states of water and also allow one to measure a t the same time the water content. We have performed dielectric relaxation mea- surements a t microwave frequencies on the halo region of a steam-cured XLPE cable. Measurements have been carried out using the method of time-domain reflectome- try (TDR). Water contents thus deduced were compared with measurements performed with a Mitsubishi Mois- turemeter.

Time-domain reflectometry has been used with suc-

0018-9307/88/0400-319¶1.00 @ 1988 IEEE

320Bose et al.: Dielectric study at microwave frequencies of halos in crosslinked polyethylene cable

cess in the determination of dielectric properties of sim- ple polar liquids [6,7], biological substances [8], liquid crystals [9] and emulsions [lo], The earlier results were more qualitative [11,12] than quantitative, but recent developments [13-161 in data acquisition, processing, and reduction of system errors have improved the TDR technique to the point where its accuracy is now com- parable to the traditional frequency-domain methods. Although the experimental measurements in a TDR sys- tem are obviously obtained in the time domain, an anal- ysis of the results is simpler if they are converted to the frequency domain [16].

EX P ER1 M EN TAL PRO C E D U RE he present experimental measurements have been T carried out using a total reflection configuration

with an open termination. For a thin dielectric sample with an open termination, the complex permittivity is given by:

true content in the halo. The water content of the cable sample was evaluated with a Mitsubishi Moisturemeter, model CA-02. In this instrument, water is extracted by vaporization a t 150 C and it is injected into a modified Karl-Fisher titration cell where the electrolytic current is directly proportional to the water content. The in- strument requires relatively large samples for reliable results, which is difficult to achieve with the specific shape of halos.

Sompler

pulse Coupler Channel A Dirlectric 8 Termination

Channel B -Terminollon

Trigger

where E' = E' - i ~ " , dand E " , are the real and imagi- nary parts of E - , Vo(s ) and R ( s ) are the Laplace trans- forms of the incident signal Vo(t ) and the reflected signal r ( t ) in the time domain, respectively. X = wdc*1/2/c, c = 1/(LcCc)1/2 is the speed of light, C,, L, are, re- spectively, the geometric capacitance and inductance per unit length of the line, d is the length of the di- electric sample, w = 2 ~ f is the angular frequency, and f is the frequency in Hertz. Measurements between 1 MHz and 8 GHz were taken using a Hewlett-Packard 181 TDR system and the experimental setup is shown in Figure 1. The correction of various experimental er- rors such as slow asymmetric drift of the incident pulse, noise, jitter and unwanted reflections arising from dis- continuities have been discussed in detail elsewhere [6]. The sample cell is a short section of a 50 n, 2 mm coax- ial line. The length of the sample (0.15 rnm) is deter- mined by the electrical length of the inner conductor of the coaxial cell. Since the external diameter of the cell is quite small, it was possible to investigate locally the halo region by introducing the cell in small holes drilled at various locations and depths in the sample, which was cut from a 138 kV steam-cured XLPE cable (Figure 2). The dielectric measurements were taken immediately af- ter drilling the hole, which was done very slowly in order to reduce temperature gradients and water evaporation in the sample. This means that some water may have been lost during drilling, and the water content deter- mined by T D R could thus possibly be lower than the

1 -

Figure 1.

Block diagram of the experimental setup.

FRONT VIEW SIDE VIEW

halo

expo ee d

t o air -

Figure 2. Photograph of the cable studied showing the halo (the dark ring area) and the location and depth of holes drilled in one half of the insulation for the TDR measurements. The other half of the cable was used for the measurements with the Mitsubishi Moisturemeter.

IEEE Transactions on Electrical Insulation, Vol. 23 No. 2, April 1088 321

0.4

I I I I

- - 1 I I

E. I

I Y

I / 0, I

I I I 2000i 5 10 15

Distance from surface (mm)

Figure 3. Water content in halo as a function of distance from surface exposed to air.

RESULTS AND DISCUSSION

he dielectric spectra, obtained a t different depths in the haloed region of the cable insulation, show an

absorption beginning a t 300 MHa and increasing contin- uously up to 8 GHa. Such absorption is essentially due to water which forms halos in the cable sample. From the static dielectric permittivity, the water content C, of the investigated region was calculated according to [18]:

E A e o a = cwE: + (1 - Cw)E:, (2)

where dw = 80 and E ’ , = 2.25 are the static dielec- tric constants of water and of polyethylene, respectively. This simple linear relation is valid for water content not exceeding a few percent. Because of its experimental limitations, the T D R method is not sensitive to changes in E’ less than - 0.04, i.e. for water content below 500 ppm. The sensitivity limit of E” is some orders of mag- nitude smaller. The results thus obtained are compared in Table 1 with those determined with the Mitstibishi Moisturemeter; the water contents determined by the two methods are seen to be comparable. The observed small discrepancies could be due to the fact that mois- turemeter samples contain a part of the adjacent non- halo (i.e. dryer) region of the cable insulation. A plot of water content determined by the TDR method as a function of depth from the sample surface (Figure 3),

suggests that some water has diffused out of the sam- ple: the water content rapidly increases away from the air exposed surface to reach a ’bulk,’ value of - 3600 ppm a t 12 to 15 mm from the surface. Water contents of - 7000 ppm have already been measured in halos [I].

The analysis of the frequency dependent dielectric permitivity is made according to the method of Cole [22] and of Brot [23] by plotting f . e’’ against E ’ . The plot yields a straight line for the case of a single relax- ation process. If however, two relaxation mechanisms are involved, the plot will show two slopes corresponding to relaxation times TI and ~2 provided these relaxation times are sufficiently far apart. Such a plot, correspond- ing to the deepest sample investigated in the XLPE is given in Figure 4; it shows the presence of two relax- ation processes. One can deduce from Table 2 that the

.-l--l-- -

T = 8.5 x d 2 s f R 18.8 GHz -

7.0712

5.3034 T

Y

* t w 3.5356 c a ,,76781 , L- fR 4.33 GHZ

2.2 $3 2.4 2.5 0 2.0 2.t

E Figure 4.

Plot of fc” vs E’ at 10 mm depth from sample sur face.

2.8r-] zd . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . , , . . . . . . .

Figure 5. Frequency dependent dielectric constant and losses of dry polyethylene cable (outside the halo).

322Bose et al.: Dielectric study at microwave frequencies of halos in crosslinked polyethylene cable

Method

TDR

Wltqubishl m l s t u r c -

meter

Table 1. Water contents in halo evaluated from TDR and compared with those measured with Mitsubishi Moisturemeter

Position w.cer C0"tC"t

LO 3413 ppm Dqpth In mm from 4 2958 ppo

2 2958 ppm 0 2315 ppo

alr-free a u c f s c e

Var1""s 1 O C B t i O " l l w i c h l n the hnlo 2255 f 1020

Table 2. Relaxation frequency in( GHz) a t different depths from air-free surface in halo

two relaxation frequencies at different locations within the sample are independent of depth. The higher relax- ation frequency of 18.8 GHz can be identified with that of free (Le. liquid) water [20], while the origin of the lower frequency, 1.33 GHz, may either be due to bound water or impurities. It is known that bound water can relax in the frequency domain ranging from kHz to GHz, depending on the nature of the binding [21].

The possible polar impurities present in XLPE ca- ble are cumyl alcohol and acetophenone, curing by-pro- ducts generated by the decomposition of dicumyl perox- ide, the crosslinking agent [2]. We carried out dielectric absorption measurements on these liquids and found the relaxation frequency to be 1 GHz for cumyl alcohol and 4 GHz for acetophenone. The lower frequency relax- ation in halos, therefore, could be attributed either to bound water or to the presence of cumyl alcohol. How- ever, in the study of a dry XLPE sample we did not observe any relaxation mechanism at 1 GHz (Figure 5) which tends to suggest that the lower frequency relax- ation should be assigned to bound water. We feel that a more detailed study is needed to identify the origin of the low frequency relaxation mechanism, and intend to carry out such a systematic study with varying amounts

of alcohol in polyethylene in an attempt to confirm the origin of the low frequency relaxation in halos.

CONCLUSION t has been shown that the water content in halos in I steam-cured XLPE cable can be reliably measured by

TDR. The study of the relaxation mechanism has also shown that water in the halo is essentially free liquid. A smaller relaxation of unknown origin was detected around 1 GHz. It is suggested that it may be associated with either bound water or with cumyl alcohol.

ACKNOWLEDGEMENTS his work was supported by the Natural Sciences T and Engineering Research Council of Canada, the

government of QuCbec through FCAR funds, and the Institut de recherche d'Hydro-QuCbec.

REFERENCES

[l] J . Tanaka and R. Luther, "Analysis of Cables with Visible Halos", 1982 IEEE Int. Symp. Elec. Insul., IEEE Publ. NO 82 CH1780-6-E1 pp. 292-295, 1982.

[2] S. L. Nunes and M. I. Shaw, "Water Treeing in Polyethylene - A Review of Mechanisms", IEEE Trans. Elec. Insul. Vol. 15 pp. 437-450, 1980.

[3] R. A. Weiss, S. H. Shaw and M. T. Shaw, "The char- acterization of Water in Treed Crosslinked Polyethy- lene films", Comm. Ann. Meet. Am. Chem. Soc., Seattle, March 1983.

[4] J . W. Pyper, "The Determination of Moisture in Solids", Analytica Chim. Acta Vol. 170, pp. 159- 175, 1985.

[5] D. W. McCall, D. C. Douglass, L. L. Blyler, G. E. Johnson, L. W. Jelinski and H. E. Bair, "Solubil- ity and Diffusion of Water in Low-Density Polyethy- lene", Macromolecules Vol. 17, pp. 1644-1649, 1984.

[6] R. Chahine and T . K. Bose,"Comparitive Studies of Various Methods in Time Domain Spectroscopy", J . Chem. Phys., Vol. 72, pp. 808-815, 1980.

[7] R. H. Cole, S. Mashimo, and P. Winsor IV, "Eval- uation of dielectric behavior by the domain spec- troscopy. 3. Precision difference methods", J . Phys., Chem., Vol. 84, pp. 786-793, 1980.

IEEE Transactions on Electrical Insulation, Vol. 23 No. 2, April 1988 323

[8] T. K. Bose, A, M. Bottreau, and R. Chahine, “Devel- opment of a dipole probe for the study of dielectric properties of biological substances in radiofrequency and microwave region with time domain reflectome- try”, IEEE Trans. Inst. & Meas, Vol. 35, pp. 56-60, 1986.

[21] J . B. Hasted, “The dielectric properties of water”, Progress in dielectrics, J.B. Birks Ed., Vol. 3, pp. 101-149, Wiley 196 1.

[22] R. H. Cole, “Analysis of dielectric relaxation meas- turements”, J . Chem. Phys., Vol. 23, pp. 493-499, 1955. [9] T. K. Bose, R. Chahine, M. Merabet, and J . Thoen,

“Dielectric study of the liquid crystal compound oc-

copy”, J . Physique, Vol. 45, pp. 1329-1336, 1984.

[231 C. B ~ ~ ~ , ~ t ~ ~ ~ ~ h i ~ ~ l determination of the param-

ation”, Comptes rend. Acad. Sci. Paris, Vol. 248, tY1cYanobiPhenY1 (8CB) using time domain ‘Pectros- eters of a Debye-type dispersion absorption relax-

pp. 397-399, 1959. [10]T. K. Bose and R. Chahine, “Measurement of the dielectric properties of water-oil emulsions by time domain reflectometry”, IEEE Trans. on Elec. Insul., Manuscript was received on 16 Dec 1986, in final form 15

June 1987. Vol. 20, pp. 935-938, 1985.

[ll] H. Fellner-Feldegg and E. F. Barnett, “Reflection of a voltage step from a section of transmission line filled with a polar dielectric”, J . Phys. Chem., Vol. 74, pp. 1962-1965, 1970.

[12] B. E. Springett and T . K. Bose, “Thin sample time domain reflectometry for nonideal dielectrics”, Can. J . Phys., Vol. 52, pp. 2463-2468, 1974.

[13] W. L. Gans and J. R. Andrews, “Time domain auto- matic network analyzer”, Nat. Bur. Stand., Boulder, CO., Tech. Note 672, p. 176, 1975.

[I41 B. J . Elliot, “High-sensitivity picosecond time do- main reflectometry”, IEEE Trans. Instrum. Meas., Vol. 25, pp. 376-379, 1976.

[15] R. H. Cole, “Time domain reflectometry”, Ann. Rev. Phys. Chem., Vol. 28, pp. 283-300, 1977.

[16] R. Chahine and T . K. Bose, “Drift reduction of the incident signal in time domain reflectometry”, Rev. Sci. Instrum,, Vol. 54, pp. 1243-1246, 1983.

[17] M. J . C. Van Gemert, “Time domain reflectometry as a method for the examination of dielectric re- laxation phenomena in polar liquids”, Thesis, Uni- versity of Leiden, Leiden, The Netherlands, p. 115, 1972.

[18] C. J . F. Bottcher and A. Bordewijk, Theory of elec- tric polarization, Elsevier Amsterdam, 1978.

[19] R. H. Cole, “Evaluation of dielectric behavior by the time-domain spectroscopy. 11. Complex permit- tivity”, J. Phys. Chem., Vol79, pp. 1469-1474, 1975.

[20] U . Kaatze and V. Uhlendorf, “Dielectric properties of water at microwave frequencies”, J. Phys. Chem., Vol. 126, pp. 151-165, 1981.