Experimental investigation of the degradation mechanism of silicone rubber exposed to heat and gamma rays

: Cable-grade silicone rubber was aged thermally or by combining heat and gamma-ray radiation, and resultant changes in chemical, thermal, mechanical and electrical properties were examined. The experimental results obtained in these analyses are clearly consistent with the mechanism that silicone rubber is degraded by forming cross-linked structures via formation of abundant siloxane bonds. With further progress of degradation, decomposition becomes dominant. Reflecting these mechanisms, mechanical properties deteriorate dramatically by losing elasticity. Both the real and imaginary parts of complex permittivity decrease, which is a contrastive difference from typical ageing behaviour of organic insulating polymers. In addition, both the elongation at break and indenter modulus are good indicators of degradation of silicone rubber.


Introduction
For electric cables in nuclear power plants (NPPs), degradation of their insulation by heat and radio-active radiation is a matter of serious concern [1-8]. In such cables, silicone rubber (SiR), ethylene-propylene-diene copolymer and cross-linked polyethylene are mainly used for electrical insulation [2, [4][5][6][7]. If the cable insulation is degraded, it may cause various malfunctions such as loss of controllability of machines, which may lead to dangerous accidents [1,8]. For realising truly reliable maintenance of cables, thorough understanding of degradation mechanisms of such insulating materials is important [8][9][10][11].
Elongation at break (EAB) is often used as an index of degradation and to verify the integrity of cable insulation in NPPs [11,12]. The reason for this is that mechanical properties of insulating materials deteriorate severely with the progress of degradation. In this sense, mechanical properties are important design factors compared with electrical properties [13].
In this research, the degradation mechanism of SiR is studied by various instrumental analyses such as Fourier-transform infrared (FTIR) spectroscopy, solid-state nuclear magnetic resonance (NMR), gel-fraction measurement, scanning electron microscopy (SEM), differential scanning calorimetry (DSC) and thermogravimetry/differential thermal analysis (TG/DTA). Mechanical properties were examined by scanning probe microscopy (SPM) and by measuring indenter modulus, EAB and tensile strength (TS). In addition, electrical properties were examined by measuring complex permittivity. By taking the structural changes clarified by these instrumental analyses into account, reasons for the changes in electrical and mechanical properties induced by the ageing are discussed.

Experiments
We used SiR sheets, bought from two providers (B and C) of cables to NPPs, as samples for this research. They are of the shape of a sheet with a thickness of about 0.50 mm. Some sheets were aged by heat in air at temperatures of 220, 250 and 280°C for 500, 900, 1350 and 1500 h. Some other sheets were aged concurrently by heat and radiation of gamma rays in air at 125, 145 and 185°C at a dose rate of 150 Gy/h for 500, 900, 1350 and 1500 h. The ageing temperatures were chosen to give a similar degree of degradation to all the SiR samples. We determined each ageing condition, namely the dose rate and total dose of gamma rays and the temperature and duration of ageing, based on the experience and report of a national project conducted under the supervision of one of the present authors [1]. As a result, the temperatures of thermal ageing were determined to be higher than those of the concurrent ageing in this research. In order to confirm the reproducibility, samples provided from the two providers were used for most of the experiments. For convenience, the samples provided by each company are distinguished by putting the letters B and C like SiR-B or by referring to as 'samples provided by company C'.
First, in order to confirm significant changes in chemical structure induced by the ageing, FTIR spectra were measured in an attenuated total reflection (ATR) mode using a spectrometer (FT/ IR-4200, JASCO), before and after the degradation. We paid attention mainly to the intensities of vibrations due to Si-CH 3 and Si-O-Si. Furthermore, solid-state 29 Si NMR spectra were observed by a spectrometer (JNM-CM400, JEOL).
For the purpose of obtaining information on the crosslinking degree, a Soxhlet extraction method was employed to estimate the gel fraction in mass. First, the sample was washed in boiling toluene at 110°C in an extractor and filtered over and over again continuously for 24 h. After it was dried, the mass of the residue remaining behind was weighed. On the other hand, for measuring the swelling ratio (SR), the sheet sample was first cut into chips with an adequate size. The chips were then put in a closed container filled with toluene at room temperature for 24 h and the change in their total volume was measured by employing the Archimedes drainage method.
To examine possible changes in surface properties induced by the ageing, each sample was examined by field emission SEM (S-4500, Hitachi). The sample surface and cross-section were coated with gold and then observed at an acceleration voltage of 15 kV. The sample surface was also observed by SPM (SPM-9700, Shimadzu). Although we can obtain various pieces of information by SPM, we used the delay angle as a main evaluation parameter. It is the delay in phase angle of a needle-like probe of SPM, with an effective diameter of 14 nm, against a sinusoidal oscillating force. If the surface is soft and sticky, the delay angle becomes large, while the delay becomes small if the surface is hard [3,14]. In this study, the delay angles were measured at 256 × 256 (= 65,536) points in an area of 5 × 5 μm 2 by scanning the probe. For each sample surface, three different places were examined. The average of the medians in the delay angle distributions obtained at the three places was chosen as an effective datum for each sample.
To examine the surface mechanical properties, we also measured the indenter modulus using an indenter modulus measurement device (IM-INSS III, Tobusa Systems), by inserting an indenter tip at a speed of 0.08 mm/s. Here, the modulus is defined as a derivative of the force by depth needed to push the tip into a sample, which becomes higher if the surface is harder. We used two IEEE-standardised indenter tips with diameters of 0.79 and 0.56 mmϕ [15], which are called Φ0.79 and Φ0.56, respectively. The average of indenter moduli measured at six different points was chosen as an effective datum for each sample.
For the purpose of examining mechanical properties of the sample bulk, we measured EAB and TS for a dumbbell-shaped test piece, which is defined as type 2 in the ISO Standard 37 [16], using a universal testing machine (5565, Instron) at a tensile speed of 500 mm/min. The averages of EAB and TS measured for five different test pieces were chosen as effective data for each sample.
Furthermore, we obtained DSC spectra using a calorimeter (DSC 3100, Netzsch Japan) in dry nitrogen in a range from −140 to 200°C by raising the temperature at a rate of 10°C/min. We also estimated thermal decomposition temperature T D by checking TG/DTA spectra obtained using an analyser (6200, Seiko Instruments) while raising the temperature from 30 to 600°C in dry nitrogen at a rate of 20°C/min.
We also obtained two components of complex permittivity, namely its real part ε r ′ and imaginary part ε r ′′, in a frequency range from 10 −2 to 10 5 Hz. The spectra were obtained at temperatures from 20 to 350°C in nitrogen of one atmospheric pressure using an impedance analyser (SI126096, Solartron) by applying an ac voltage of 3 V rms . For the measurement of complex permittivity, aluminium foil electrodes with a thickness of 11 μm and a diameter of 20 mm had been pasted with silicone oil on both sides of a sample. For any material, its primary structure is the most important basic factor that determines its various properties. From this viewpoint, FTIR and NMR spectra were compared before and after the ageing. As shown in Fig. 1, the FTIR spectrum observed in a 0.55 mm thick unaged SiR-B sheet agrees with that of polydimethylsiloxane [17][18][19]. The peaks at 1256 and 2963 cm −1 are, respectively, assigned to symmetric bending and stretching vibration of C-H in CH 3 . The broad band at 1064 cm −1 is assigned to the stretching of Si-O-C, which is seemingly in a terminal group or cross-linked part. Two large sharp peaks at 785 and 1006 cm −1 are assigned to Si-CH 3 stretching and Si-O-Si stretching, respectively, reflecting the abundance of constituent groups of SiR [17,18]. Fig. 1 also shows that no obvious changes are induced in SiR-B by the ageing. Although not shown in this paper, the spectrum of unaged SiR-C is almost the same as that of the unaged SiR-B. Fig. 2 shows the changes in absorbance of peaks at 785 and 1006 cm −1 , induced in the SiR-B sheets by the two types of ageing for 1500 h. As mentioned above, the two peaks are assigned to Si-CH 3 and Si-O-Si. It is clear that the number of Si-O-Si increases in marked contrast to that of Si-CH 3 . Both the decrease in the number of Si-CH 3 and the increase in number of Si-O-Si become obvious when the ageing temperature reaches and exceeds 250°C in the thermal ageing. They seem to be more significant in the concurrent ageing than in the pure thermal ageing if the ageing temperature is the same, although we could not confirm this in our experiment. Fig. 3a shows 29 Si NMR spectra obtained for SiR-B before and after the thermal or concurrent ageing for 1500 h. Here, letters D, T and Q, which appear at about −20, −65 and −120 ppm, represent that two, three and four oxygen atoms form chemical bonds with one Si atom, respectively [17,20,21]. No signals are seen at 5 ppm in the spectra. This means that neither the unaged nor the aged samples have Si atoms forming one bond with oxygen. Fig. 3b shows the abundance ratios of D, T and Q in SiR-B before and after the thermal or concurrent ageing for 1500 h. It is natural that the unaged sample exhibits a strong signal of D. However, signal Q is seen in the unaged samples. The reason for this is most likely that the signal Q appears due to the presence of talc, an additive commonly used in polymers [22,23]. On the other hand, signal T appears and becomes stronger with the progress of ageing.

Results
To examine structural changes further, the gel fraction and SR were measured according to the procedures mentioned above. Figs. 4 and 5 show the results. In general, if a polymer is crosslinked, it becomes insoluble into a chemical reagent. Therefore, the gel fraction in mass can be measured by the Soxhlet extraction  method as the ratio of the insoluble weight to the original weight. The estimated gel fraction is around 96-98% in the unaged sample and it increases to 99-100% by the ageing as shown in Fig. 4. However, there is a risk that the correct gel fraction cannot be estimated by this method. That is, even if cross-linking occurs only at the sample surface, the entire volume would become insoluble in toluene. This leads to the misunderstanding that the entire volume was cross-linked. Therefore, we measured SR or volume expansion of the sample after it had been immersed into toluene for 24 h at room temperature. As shown in Fig. 5, SR decreases significantly by the concurrent ageing at the three temperatures or by the thermal ageing at 280°C. This clearly indicates that the sample becomes more densely cross-linked to the inside with the progress of the ageing under severe conditions. Fig. 6 shows high-resolution SEM images observed before and after the ageing at different conditions. The two leftmost photos are the images of the unaged sample. Both the surface and crosssection are flat and uniform. The surface becomes very rough and grainy if the sample was aged by heat at 280°C for 1500 h, as shown in Fig. 6a3, whereas Fig. 6a2 indicates that the ageing by the concurrent heat and gamma-ray radiation at a relatively low temperature like 185°C seems not so severe. Furthermore, it has become clear that tiny holes appear in cross-sections as shown in Figs. 6b2 and b3. Fig. 7 shows results obtained by SPM. Each symbol represents the average of medians of 65,536 data of delay angles measured at three different points on each sample surface as mentioned above, while the error bar shows their first and third quartiles. Since the experimental data differ between SiR-B and SiR-C, we made two graphs. Although the delay angle does not show a monotonic change in both samples, it appears evident that the delay angle becomes small if SiR is aged severely. The small delay angle means that the surface of SiR is hard as mentioned above. Fig. 8 shows results obtained by the indenters. Each symbol and each error bar represents the average and standard deviation of the indenter moduli measured six times for each sample. The standard deviation is always narrower than the width of the corresponding symbol and thus the error bars for solid symbols cannot be seen. Except for the thermal ageing at 220°C, the modulus is increased by all other ageing processes. This indicates that the surface becomes hard. Here, the moduli measured with Φ0.79 are higher than those measured with Φ0.56 when compared with the data taken under the same ageing condition. Fig. 9 shows the average and standard deviation of EAB measured five times for each sample. As is the case with Fig. 8, the error bar that represents the standard deviation is small in most cases, which makes it difficult to find the error bars for several solid symbols. Under all the ageing processes, EAB decreases with the increase in ageing time in both SiR-B and SiR-C. Fig. 10 shows the average and standard deviation of TS measured five times for each sample. Compared with the small error bars shown in Figs. 8 and 9, those in this figure are relatively large. This means that the values of TS change comparatively widely even under the same ageing condition. Nevertheless, the overall trend shows that TS decreases with the increase in ageing time. Both the results of EAB and TS measurements indicate that the SiR sheets become hard and brittle when they are aged. Fig. 11 shows enthalpy of fusion E F as a function of ageing time when the sample were aged thermally or concurrently by heat and radiation, while Fig. 12 shows the melting point of crystalline parts T PM . Both E F and T PM of SiR-B, shown in Figs. 11a and 12a, hardly change when the ageing temperature is 220 or 250°C for the ageing only by heat. However, by the thermal ageing at 280°C and the simultaneous thermal and radiation ageing at the three temperatures, both E F and T PM are decreased significantly in SiR-B. This is also the case in SiR-C, although no data are shown for T PM measured after the thermal ageing at 280°C. The reason for not showing these data of T PM for SiR-C is that the corresponding endothermic peak could not be seen in the DSC spectrum of this sample. This indicates that most of the parts which crystallise at low temperatures in this sample had been decomposed by the ageing. Fig. 13 shows T D estimated by the measurements of TG/DTA. For both SiR-B and SiR-C, T D increases first up to the ageing time of 900 h, which is followed by a sharp decrease, if the samples were aged only by heat. On the other hand, T D decreases at an early stage of ageing mainly up to 500 h and then it increases slightly in the case of the concurrent ageing by heat and gamma radiation. Fig. 14 shows spectra of ε r ′ and ε r ′′ measured for SiR-B as a function of frequency and temperature. Here, the measurement was conducted at 20 and 50°C and at every 50°C from 100 to 350°C in a wide frequency range from 10 −2 to 10 5 Hz. Fig. 15 shows the values of ε r ′ and ε r ′′ measured at 63.1 Hz. The reason for selecting 63.1 Hz is that it is the frequency closest to the power frequency. The relative permittivity of the unaged SiR-B decreases significantly with the increase in temperature. That is, at 63.1 Hz, the value of about 2.8 at 20°C becomes about 2.4 at 350°C. It is true that various factors such as charge transport and space charge accumulation exert strong influences on complex permittivity, especially at low frequencies. In this regard, effects of gamma radiation at 145°C with a dose rate of 150 Gy/h on the real part of complex permittivity are shown in Fig. 16. It is clearly shown in Figs. 14-16 that the permittivity decreases in SiR regardless of frequency if severe degradation occurs.

Discussion
It has been known that the degradation of SiR is induced by the formation of cross-linked structures, which is caused by oxidation [4,[24][25][26]. That is, if Si-C bonds break during the ageing, formation of Si-O bonds or oxidation follows. If oxidation occurs between two adjacent siloxane chains, it forms a cross-linked structure. In addition to the decrease in the IR absorption intensity due to Si-CH 3 and the increase in intensity due to Si-O-Si shown in Figs. 1 and 2, the NMR results shown in Fig. 3 clearly support the above-mentioned degradation process. The increase in gel fraction is not so obviously demonstrated by the Soxhlet extraction method shown in Fig. 4, since its value already reached a very high  value of 96 before the ageing. However, as shown in Fig. 5, SR decreases nearly monotonically, depending on the ageing time. If cross-linking occurs inside a polymer, swelling should become difficult. Therefore, the results shown in Fig. 5 indicate that crosslinking proceeds more in SiR as the degradation proceeds.
The formation of cross-links should shrink the volume, which would be the main reason for the rough surface appearing in the surface SEM image shown in Fig. 6a3. The holes appearing in Figs. 6b2 and b3 seem to be induced by the degassing due to the detachment of pendant methyl groups.
Furthermore, the increase in cross-linking degree should harden SiR [4,27]. Namely, the increase in cross-linking density inevitably results in loss of flexibility or elasticity. As a result of this hardening of SiR, the delay angle of SPM decreases as shown in Fig. 7 and the indenter modulus increases as shown in Fig. 8. However, the results obtained by SPM do not correlate well with the results obtained by the indenter. While SPM is an adequate tool to examine surface conditions of a sample microscopically, indenter modulus reflects its more macroscopic conditions. It is assumed that SPM may reflect properties of additives bled out on the surface. In such cases, SPM does not reflect the correct mechanical properties of sample surface correctly. Therefore, in order to examine macroscopic mechanical properties, the indenter is superior to SPM.
As shown in Fig. 9, EAB decreases with the increase in ageing time. This seems to be due to the decomposition caused as a result of the bond breaks at Si-O bonds. This decomposition also decreases TS inevitably as shown in Fig. 10.  As mentioned above, the endothermic peak at around −40°C appearing in DSC spectra (not shown) is due to the melt of crystalline parts [28]. Therefore, the decrease in E F and the decline in T PM shown in Figs. 11 and 12 should reflect the collapse and decrease in amount of crystalline parts caused as a result of radiation-induced cross-linking and breaking of chemical bonds [29]. The increase in T D at the relatively early stage of thermal degradation shown in Fig. 13 seems to reflect the formation of cross-linked structures. On the other hand, the following sharp decrease in T D seems to be caused by the decomposition of SiR. Furthermore, the dependence of T D on ageing time is quite different between the pure thermal ageing and the concurrent ageing. This difference in thermal behaviour indicates clearly that the heat and the radiation exert different influences on SiR. As we already confirmed, SiR degrades by forming a crosslinked structure, which makes SiR hard and brittle. From the complex permittivity spectra shown in Figs. 14-16, it is clear that both the real and imaginary parts of complex permittivity decrease as the sample degrades more. It is quite natural that polarisation becomes difficult if SiR becomes hard. In contrast, both real and imaginary permittivity increase significantly in many organic polymers if they are degraded, since polar carbonyl groups are formed abundantly by the oxidation of organic polymers [30,31]. These contrastive differences in dielectric behaviour between SiR and organic polymers remind us of the fact that SiR is an inorganic polymer with siloxane backbones.

Fig. 8 Changes in average (symbol) and standard deviation (top and bottom of the vertical bar) of indenter moduli, as a function of ageing time
For many rubbers and polymers, EAB is used as an index of mechanical degradation [11,12]. Fig. 17 shows the relation between EAB and TS, measured for SiR-B and SiR-C. Here, in order to show the relation between EAB and TS clearly, experimental data obtained for the same sample are represented using the same symbol regardless of the difference in ageing temperature and time. Only the difference in ageing condition, namely, the thermal ageing only by heat or the concurrent ageing with heat and gamma radiation, is distinguished using different symbols. The data of the samples degraded by heat only and those degraded concurrently with heat and radiation can be fitted to respective curves. The coefficient of determination R 2 is 0.959 for SiR-B degraded only by heat, 0.877 for SiR-B degraded concurrently by heat and radiation, 0.995 for SiR-C degraded only by heat and 0.859 for SiR-C degraded concurrently by heat and radiation. The data of SiR-C heated at 280°C were discarded, because this condition hardened SiR-C too much. All the values of R 2 are close to 1, indicating that EAB correlates well to TS. Here, TS of SiR degraded only by heat is lower than that of SiR degraded concurrently by heat and radiation, even when EAB is the same. This fact indicates that different mechanical properties are induced between the two degradation processes given by heat only and by simultaneous exposure to heat and radiation. Fig. 18 shows the relation between EAB and E F represented in a similar manner to Fig. 17. The two sets of data obtained for SiR samples degraded only by heat or concurrently with heat and radiation are fitted to respective exponential equations. Here, R 2 is 0.960 for SiR-B degraded only by heat, 0.957 for SiR-B degraded concurrently by heat and radiation, 0.814 for SiR-C degraded only by heat and 0.844 for SiR-C degraded concurrently by heat and radiation. All the values of R 2 are close to 1, indicating that EAB correlates to E F in each case where the degradation is induced only by heat or concurrently with heat and radiation. The value of E F measured for the SiR sample degraded only by heat is higher than that of the SiR sample degraded concurrently by heat and radiation, if EAB shows a similar value. This fact seems to indicate that the structural changes in SiR induced during the degradation by concurrently-given heat and radiation are more significant than those induced by the purely thermal degradation. The relation between EAB and TS or the relation between EAB and E F shown in Figs. 17 or 18 indicates that the heat and the gamma radiation   exert different influences on SiR and that EAB can be a good indicator of the changes in mechanical and thermal properties. Fig. 19 shows the relation between EAB and indenter modulus measured for various sample sheets degraded only by heat or concurrently by heat and radiation. The measurements were conducted with two indenters with different tips, Φ0.79 and Φ0.56. The moduli measured by Φ0.79 are larger than those measured by Φ0.56 as was already shown in Fig. 8. This seems simply attributable to the fact that a bigger force is needed to insert an indenter with a larger tip. In addition, data obtained for the samples provided by different companies B and C are on respective curves. If we draw exponential approximation curves as shown in Fig. 19 56. These values are fairly close to 1, indicating that the indenter modulus and EAB correlate to each other well. Needless to say, indenter modulus is an indicator of the material's mechanical property at the surface, while EAB reflects strongly its bulk property. Therefore, the fact that the good correlation between the indenter modulus and EAB is seen in each kind of sample, independently of the ageing condition, means that both indenter modulus and EAB are reliable monitoring tools of degradation of SiR, especially in NPPs where thermal and radiation conditions differ significantly for each cable. Furthermore, the scatter of data of indenter moduli is very small as mentioned above. Taking all these into account, use of indenter modulus for monitoring the degradation of SiR is commendable. The indenter modulus measurement can also be applied to a sheath material as well as to an insulating material. Therefore, the integrity of an in-service cable with a SiR sheath can also be evaluated by the modulus.

Conclusions
Sheets of SiR were aged either thermally or concurrently by heat and gamma radiation and the resultant changes in various properties were systematically analysed from various aspects. As a result, the following important findings have been obtained.
i. All the experimental results are consistent with the ageing mechanism of SiR reported in literature. That is, SiR is degraded by forming cross-linked structures via oxidation of Si atoms that follows the loss of pendant CH 3 groups. ii. The above-mentioned cross-linked structure is essentially nonpolar. Therefore, the degradation behaviour of SiR is significantly different from that of many organic polymers where oxidation of C atoms induces polar carbonyl groups abundantly. iii. When the degradation proceeds to a very high level, decomposition becomes dominant. iv. Heat and gamma radiation give different influences on SiR.
That is, pure thermal ageing causes mainly cross-linking at an early stage of ageing, which is followed by decomposition. On the other hand, decomposition starts much earlier in concurrent ageing by heat and gamma radiation. This difference in degradation behaviour causes significant differences in many properties such as mechanical and thermal behaviour. v. The fact that EAB is a good indicator of possible changes in mechanical properties of SiR has been reconfirmed. vi. Indenter modulus is a reliable indicator for monitoring the degree of degradation of SiR and its use is commendable.

Acknowledgment
This research was partly supported by the Nuclear Regulation Authority in Japan through Ageing Management Technical Evaluation Enhancement Program.