Cellulose (2007) 14:295–309 DOI 10.1007/s10570-007-9124-1 Identification of a chemical indicator of the rupture of 1,4-b-glycosidic bonds of cellulose in an oil-impregnated insulating paper system Jocelyn Jalbert Æ Roland Gilbert Æ Pierre Tétreault Æ Brigitte Morin Æ Denise Lessard-Déziel Received: 20 February 2007 / Accepted: 24 April 2007 / Published online: 21 June 2007
Springer Science+Business Media B.V. 2007 Abstract In this study, headspace gas chromatography/mass spectrometry has been used to assess the volatile by-products generated by the ageing of oilimpregnated paper insulation of power transformers. Sealed-glass ampoules were used to age under oxidative conditions 0.5-g specimens of insulating paper in 9 mL of inhibited mineral oil in a temperature range of 60–120 8C and moisture of 0.5, 1 and 2% (w/w). A linear relationship between one of the oil-soluble degradation by-products, i.e. methanol, and the number of ruptured 1,4-b-glycosidic bonds of cellulose, regardless of the type of paper (ordinary Kraft or thermally-upgraded (TU) Kraft paper), was established for the first time in this field. Ageing at 130 8C of model compounds of the Kraft paper constituents (a-cellulose, hemicellulose and lignin) and two cellulosic breakdown byproducts (D-(+)-glucose and 1,6-anhydro-b-D-glucopyranose) confirmed that the a-cellulose degradation was mostly responsible for the presence of this molecule in the system. Furthermore, additional 130 8C-tests with six different papers and pressboard samples under a tight control of initial moisture indicated that at least one molecule of methanol is formed for each rupture of 1,4-b-glucosidic bond J. Jalbert
R. Gilbert (&)
P. Tétreault
B. Morin
D. Lessard-Déziel Institut de recherche d’Hydro-Québec, 1800, boulevard Lionel-Boulet, Varennes, QC, Canada J3X 1S1 e-mail: gilbert.roland@ireq.ca of the molecular chains. Stability tests showed that the ageing indicator is stable under the oxygen and temperature conditions of open-breathing transformers. The presence of methanol was detected in 94% of oil samples collected from over than 900 in-service pieces of equipment, confirming the potential for this application. Lastly, the tests have shown that oiloxidation by-products and TU-nitrogenous agents modify the methanol partitioning coefficients in the paper/oil/air system, which makes their study essential over a range of field conditions encountered by power transformers. Results are presented and discussed in comparison with 2-furfuraldehyde, which is the current reference in the domain. Keywords Ageing indicator
Cellulose insulation
Damage monitoring
Degree of polymerization
Glycosidic bond scissions
Kraft paper
Methanol
Mineral oil
Remaining life
Thermally upgraded paper
Transformer
Volatile degradation by-products
2-Furfuraldehyde Introduction Cellulose, an unbranched homopolysaccharide composed of b-D-glucopyranose rings joined together by 1,4-b-glucosidic bonds, has been used for over 100 years as electrical insulation in oil-filled transformers (a typical power transformer contains about 12000 kg of cellulose and 40000 kg of oil). Its 123 296 inherent good mechanical and electrical properties, its ease of use in the manufacturing process, and its abundance (obtained from the delignification of wood pulp by the Kraft process) have made it a virtually universal choice. Moreover, this material is generally recognized as the most significant limiting factor in the operating temperature and thermal life of transformers. The combined action of temperature, oxygen and moisture cause the insulation to lose mechanical strength and become weak and brittle. The transformer is then at the mercy of the first short-circuit whose longitudinal electromagnetic stresses crush the paper, in spite of all precautions that may have been taken. Owing to the worldwide growth of large-scale utility systems in the 1950s and 1960s, a large number of transformers now operate at an age beyond the nominal design life. Then, some means of condition monitoring has become essential to promote satisfactory programs of planned maintenance or replacement. From the standpoint of molecular structure, the ageing of the cellulose fibers is related to the crosswise break of the inter- and intra-molecular hydrogen bonds and the lengthwise break of the molecular chains through the rupture of the 1,4-b-glucosidic bonds. Specifically, the average length of the cellulose chains is the parameter that governs the mechanical strength of the insulating papers. Such a relationship makes the viscometric degree of polymerization (DPv) an appropriate measurement for directly assessing the progress of paper ageing (20% of tensile strength corresponds to DPv & 200). This widely used parameter by electric power utilities gives information on the average number of b-D-glucopyranose rings per cellulose molecule. However, the DPv measurement is impractical in the field due to the need to de-energize the transformer to extract representative paper specimens. To overcome the difficulty, it would be especially useful to identify a relationship between the cellulose DPv and a specific product dissolved in the insulating oil that is resulting from ageing. Once established, such a relationship would permit oil analyses to be used to assess the insulating paper condition of transformers in operation. As early as 1981, Tamura et al. reported a relationship between the amount of carbon oxides (CO and CO2) in the oil and the degree of polymerization of insulation papers. However, its applicability was found to be limited considering that these 123 Cellulose (2007) 14:295–309 indicators could arise not only from the degradation of paper but also the decomposition of oil during long-term oxidation. Three years later, Burton et al. (1984) suggested the use of a family of furan compounds that could be directly extracted from the oil to characterize the thermal decomposition of insulation papers. The advantage of such compounds over carbon oxides is that they arise more specifically from the breakdown of the paper insulation. The cumulative amount of the most abundant furan found in transformer oil, 2-furfuraldehyde (2-FAL), was very early related directly to the reduction of the degree of polymerization of the cellulose, both in the laboratory and from in-service transformers (Schroff and Stannett 1985; Burton et al. 1988). However, considering that 2-FAL could be generated not only from the cellulose degradation but also from hemicellulose (5-membered ring polysaccharides known to possess the lowest stability of all wood-pulp materials), this decreases its value as a chemical indicator (Emsley and Stevens 1994). Moreover, the recent literature is conclusive on the existence of significant differences in the relationship between 2FAL and DPv during ageing with the type of paper (ordinary Kraft versus thermally upgraded (TU) Kraft paper), and varying contents of water (Oommen et al. 1993; Soares et al. 2001; Lundgaard et al. 2004). Despite the numerous attempts to correlate 2-FAL with cellulose damage, no satisfactory relationship has yet been established to cover the various inservice transformer conditions. Since the first use of high-performance liquid chromatography (HPLC) by the electric power industry (Burton et al. 1984), the research that followed on ageing was for the most part subject to HPLC standard test methods (ASTM D5837 or IEC 61198) for assessing oil samples (giving access to a limited number of molecules of a specific class). This use of an unidirectional analytical approach is at the very least surprising considering that studies in the 1960s in connection with the thermal degradation of pure cellulose indicated for the by-products a large spectrum of low molecular weight compounds starting from small molecules (H2, CH4, CO2) through intermediate-size hydrocarbons, alcohols and carbonyl compounds to 1,6-anhydro-b-D-glucopyranose and other anhydrides equal in weight to the cellulose monomer (Schwenker and Beck 1963; Glassner and Pierce 1965; Shafizadeh 1968; Kilzer 1971). During Cellulose (2007) 14:295–309 this period, the total number of volatile by-products isolated and identified by gas chromatography was well over 50 and included formaldehyde, acetaldehyde, methanol, acetone, ethanol, glyoxal, 2-butanone and furan. Very little attention in the field has been paid to these molecules over the last 20 years. To our knowledge, only two papers reported the use of such by-products for the assessment of paper insulation, one based on the use of a semiconductortype sensor in the oil headspace and the other on the determination of acetone in oil (Abe et al. 1994; Awata et al. 1997). In this investigation, the limits of the HPLC test methods are overcome by using headspace gas chromatography/mass spectrometry (HSGC/MS) to assess the volatile by-products generated by the ageing of oil-impregnated cellulosic materials. The study was carried out with the aim of identifying an oil-soluble chemical indicator that is specific to the rupture of the 1,4-b-glucosidic bonds of cellulose and present in the oil regardless of the type of paper (ordinary Kraft or TU Kraft). The origin of the identified indicator was confirmed by studying model compounds of the different components of paper insulation (cellulose, hemicellulose and lignin). The viscometric degree of polymerization of the paper specimens was used to establish the existence of a relationship between the amount of indicator in the oil and the fraction of glycosidic bonds ruptured, and to demonstrate the no significance of the amount of moisture on the number of molecules formed by ruptured bond. Tests were also carried out to evaluate the stability of the ageing indicator under the operational conditions of transformers. Finally, our capacity to detect the indicator in the field was measured by assessing oil samples collected from in-service electrical equipment. The results are presented and discussed in this paper in comparison with 2-FAL, which is the current reference in the domain. Experimental section Materials The test specimens were obtained from five insulating papers of different manufacture: one ordinary Kraft 297 paper known as Clupak HD75 (wood cellulose containing up to 7% (w/w) of lignin and hemicellulose), one ordinary Kraft paper that underwent additional purification at the pulp stage (elimination of the residual lignin and hemicellulose) known as Munksjö Thermo-70, and three thermally upgraded papers (obtained by incorporating variable amounts of stabilizing nitrogenous agents, particularly dicyandiamide), known as CE Rotherm and Manning 220 Mannitherm D. Specimens were also obtained from a low density calendered transformerboard. Some characteristics of the materials are given in Table 1. All the tests were carried out with specimens immersed in Naphthenic Nynas 10 CX insulating oil (Nynas Naphthenics) that contains about 3000 mg/ kg of 2,6-di-tert-butyl-p-cresol (oil antioxidant). To trace the origin of the indicators, the following model compounds were studied: microcrystalline a-cellulose (Aldrich #31,069-7), Whatman No. 41 paper (considered to contain more than 98% of a-cellulose, Fisher #1441-866), xylan isolated from birch (Fluka #95588), mannan isolated from Saccharomyces cerevisiae (Sigma #M7504), and alkali Kraft lignin (Aldrich #37,095-9). Two intermediate by-products of the degradation of a-cellulose were also tested: D(+)-glucose (Sigma #G7528) and 1,6-anhydro-b-Dglucopyranose (levoglucosan) (Sigma #G7528). Moisture conditioning A glove box Model DL-001-SP equipped with an automatic pressure control, Model HE-63P from Vacuum/Atmosphere Co. (Hawthorne, CA), and two internal fans for providing uniform humidity was use for moisture conditioning. The glove box is supplied with air freed from water vapor by passing compressed air through a Balston unit Model 75–60, Lexington, MA (dew point at 73 8C). The atmosphere of the glove box is monitored by using a microprocessor humidity controller (0–100% RH with 0.1% RH resolution, Model 37700-03 from ETS Electro-Tech Systems, PA) equipped with a capacitive film sensor and an RTD temp sensor. The signal of the humidity sensor is used to meet a setpoint value by activating a dehumidifier, Drierite desiccant Model 37700-50 or an ultrasonic humidifier distillation system, Model 37700-60 (also from ETS Electro-Tech Systems). 123 298 Cellulose (2007) 14:295–309 Table 1 Some characteristics of the insulation papers and pressboard studied Munksjö Thermo-70 Munksjö Paper AB Sheets Sweden 63 ± 3 61–67 0.75–0.85 0.013 Clupak HD75 Weidman Whitely Limited Tullis Russel Co., Scotland, UK Sheets 80 ± 2 N.A.b 1.00–1.15 0.073 Wrapped on copper conductor 83 ± 3 N.A. N.A. 0.91 Roll 75 ± 2 58 0.90 ± 0.05 1.15 Manning 220 Mannitherm D Lydall Inc., USA (mixture of manilla hemp and Manning Div. Kraft wood pulp)c Sheets 84 ± 1 48 0.63 3.9 Hi-Val Kraft transformerboard Sheets 1798 ± 9 N.A. Tullis Russel Co., Scotland, UK a EHV-Weidmann Thickness Nominal grammage (mm) (g/m2) Nitrogen content (%)a(w/w) Manufacturer CE Rotherm (Insuldur type) Format Apparent density (g/cm3) Type of paper and pressboard 0.90–1.05 < 0.03 Total nitrogen by the Kjeldahl method b N.A.: Non Available c The wood pulp does not exceed 50% Ageing cells for stability tests The stability of the chemical indicators was assessed using a solution prepared by adding 5 mL of methanol (Fisher #AC32695) and 2.5 mL of acetone (Aldrich #27,072-5), ethanol (Les Alcools de Commerce Inc.) and 1-butanol (Fisher #A399) in 0.5 L of insulating oil. Seventy-five 20-mL glass ampoules were volumetrically filled with 3 mL of this solution and 6 mL of oil for a final concentration of about 1500 mg/kg for each compound, except for methanol with 3000 mg/kg. These ampoules were then sealed in open air and introduced in forced-air ovens (Salvis, Sweeden) thermostatically maintained at 70, 90, 110 and 130 8C. Three ampoules were analyzed at the time of preparation and then at 41, 113, 185, 449, 785 and 1505 h of ageing. To verify a possible contribution of oil ageing in the formation of these compounds, an equal number of ampoules containing 9 mL of insulating oil (blank samples) were aged under identical conditions. Ageing cells for the 168-h tests at 130 8C Five strips of 20 · 2.5 cm (depending on the paper grammage) weighing 0.5 g were cut from each sheet 123 of insulating paper and rolled loosely. The same number of strips with equivalent weight was also prepared with the transformerboard. The specimens were first vacuum-preconditioned in the glove-box hatch to attenuate the humidity disparity noted through the various materials. They were then introduced in the glove box along with an equal number of empty 20-mL glass ampoules (VWR #12010L-20) and a 1-L bottle of oil. Prior to this, the ampoules were serially cleaned using an ultrasonic bath, rinsed three times with demineralized water and dried at 130 8C for 24 h in a vacuum oven. When the target humidity content was achieved in the specimens (1.2–1.6% H2O (w/w)), each strip was inserted into a 20-mL pre-weighed glass ampoule. The ampoules were filled volumetrically with 9 mL of oil (giving an oil-paper volume-weight ratio of 18:1), temporarily closed and withdrawn from the glove box to be sealed in open air. The sealed ampoules were placed in a forced-air oven maintained at 130 8C for 7 days. After the thermal treatment, the ampoules were allowed to cool for 3 h in a low temperature incubator (Fisher Scientific, model 146D) maintained at 20 8C, after which oil aliquots were transferred into a 10-mL headspace vial (Supelco #27295) and 2-mL amber screw-top glass vial (Agilent #5182-0716) for Cellulose (2007) 14:295–309 299 analysis. Each paper specimen was then kept in the dark until it could be analyzed for the degree of polymerization. To ensure that the degradation products observed at the end of the tests mostly resulted from the papers and not the oil, an equal number of cells containing only the equilibrated oil were aged under identical conditions. Some ageing cells were also prepared by introducing 0.1 g of each model compound and cellulose degradation by-products in five ampoules with 9 mL of oil. Contrary to the previous cells, the humidity of the starting materials (constituents and oil) was not equilibrated prior to testing. Ageing cells for tests with varying temperature and moisture conditions The conditioning of the cell components (paper and oil) was done as for the tests at 130 8C, except that the specimens were not subjected to vacuum preconditioning and the RH % of the glove box was successively changed to obtain specimens at three moisture levels: 0.5%, 1.0% and 2.0% (w/w). After conditioning, the sealed-test cells were placed in forced-air ovens maintained at 60, 70, 80, 90, 100, 110 and 120 8C. They were withdrawn from the ovens after varied lengths of time depending upon the ageing temperature (maximum time of 14 000 h for T = 60 8C). As in the case of the previous tests, the withdrawn cells were allowed to cool for 3 h in the low temperature incubator, after which the ampoules were broken and oil aliquots collected for the analysis. The paper specimens were set apart for the determination of the degree of polymerization. Apparatus and methods A G1888 static headspace sampler coupled with a 6890N gas chromatograph equipped with a 5973N mass selective detector at 70 eV ionization energy in the electron impact mode (all from Agilent Technologies) was used to assess the volatile degradation byproducts of the cellulose. The instrument interface was maintained at 250 8C and a mass range, m/ z = 10–300 amu, in a 0.21-s cycle, was scanned in total ion count mode (TIC). The separation was performed with a 60-m-long Stabilwax (Restek) polar column, 0.25 mm in diameter and with 0.5 mm film thickness, under the instrumental conditions given in Table 2. The signal was calibrated by injecting a series of dilutions prepared from a mother solution of methanol, acetone, ethanol and 1-butanol in oil, each at a concentration of 5 ppm (v/v), except for methanol with 10 ppm (v/v) (6-point calibration Table 2 Instrumental conditions for the HSGC/MS analysis of the volatile degradation by-products of cellulosic insulating material Headspace sampler parameters Temperature: Sample 60 8C Transfer line 120 8C Gas sampling valve and injection loop 150 8C Pressure: Vial over-pressure 70 kPa Times: Equilibration at 60 8C with shaking 60 min Shaking: Pressurization 0.2 min Loop fill 0.2 min Loop equilibration 0.2 min Injection 0.9 min Power Maximum level Gas chromatograph parameters He carrier gas flow: Oven: 2 mL min-1 20 8C for 12 min 20–110 8C at 5 8C min-1 110–240 8C at 24 8C min-1 240 8C for 12 min 123 300 (a) Clupak HD75 specimen 30 15 22 10 20 23 24 20 5 0 33 34 35 10 2 6 (b) (c) 3 4 1 0 56 7 9 8 21 25 26 15 19 17 29 30 13 14 10 12 Clupak HD75 specimen + oil 30 27 28 18 16 11 Abundance of the TIC signal x 10 curves). Quantification was done in selected ion monitoring (SIM) mode. 2-FAL was assessed using a high-performance liquid chromatograph (Agilent Technologies, 1100 Series) based on a method adapted from Lessard et al. (1995). The signal was calibrated (6-point calibration curve) by injecting a series of dilutions prepared from a mother solution of 2-furfuraldehyde (Aldrich #31,991-0) in oil at a concentration of 2500 ppm (w/w). The determination of the average viscometric degree of polymerization was performed after shredding the samples using a water-cooled shredder (Janke&Kunkel, IKA-Werk) and achieving dissolution in an aqueous solution of bis(ethylenediamine)copper(II) hydroxide (Anachemia #29072-360). The procedure used was based on ASTM D4243. Prior to measurement, the oilimpregnated samples were degreased in a Soxhlet (Soxtec Avanti 2050) using fresh distilled hexane. The moisture content of the specimens was determined by titration with a Karl Fischer 756 KF Coulometer (Brinkmann) according to Method C of ASTM D1348. Cellulose (2007) 14:295–309 28 22 20 21 19 16 10 2 1 0 9 3 5 A 13 14 17 7 Oil 30 28 Results and discussion Identification of molecules with potential diagnostic significance Figure 1a shows a typical TIC chromatogram of the volatile degradation by-products obtained from the injection of the headspace of a paper specimen heated at 120 8C for 912 h (*0.5 g of Clupak HD75 in presence of air in a 20-mL headspace vial). The components identified by comparing the mass spectrum recorded at the maximum of each peak with those from the NIST MS 2002 library are listed in Table 3. As seen in this table, the peaks eluted in the time range of 5–25 min are associated with lowmolecular-weight volatile carbonyl compounds, with the exception of peak 2, which is identified as carbon disulfide. Most of these molecules have already been reported in the pyrolysates of pure cellulose (Schwenker and Beck 1963; Glassner and Pierce 1965; Shafizadeh 1968; Kilzer 1971). The proposed mechanism to account for their presence involves the formation by chain scission of carbonium ions that may decompose irreversibly to form unsaturated products containing aldehyde and enol groups (Byrne 123 20 29 10 2 Unaged oil 12 10 5 0 5 10 7 15 15 16 17 A 20 25 30 35 40 45 Retention time (min) Fig. 1 Typical TIC HSGC/MS chromatograms of volatile degradation by-products recorded after 912 h of ageing at 120 8C in presence of air et al. 1966). These products may yield volatile carbonyl compounds such as those compiled in Table 3, or alternately, they may undergo aldol-type condensation with the elimination of water to form ethylenic crosslinks between carbon chains and thus ultimately form a carbon-rich char (paper with a darkened aspect). Over 25–35 min, the comparison of the mass spectra is indicative of higher molecular weight carbonyl compounds (including 2-FAL), with the exception of peak 18, which is matched with Cellulose (2007) 14:295–309 301 Table 3 Identification of the volatile degradation by-products in the TIC-chromatogram recorded from the headspace sampling of a Clupak HD75 specimen after 912 h of ageing at 120 8Ca Peak numbera Retention time (min) Compounds identified MW % match 1 6.34 Acetaldehyde 44 2 6.90 Carbon disulfide 76 9 3 8.08 Methyl formate 60 78 4 9.59 Furan 68 91 5 10.43 Acetone 58 86 6 11.31 Acetic acid, methyl ester 74 91 7 14.34 Butanal 72 94 8 15.84 2-Butanone 72 91 9 16.28 Methanol 32 43 10 18.58 Ethanol 46 83 11 19.01 Furan, 2,5-dimethyl 96 94 12 19.69 2-ethylacrolein 84 95 13 20.35 2,3-butanedione 86 90 14 24.01 2,3-pentadione 100 83 15 16 24.24 26.75 1,4-dioxane Water 88 18 94 1 17 27.97 1-butanol 74 91 18 30.41 Cyclohexene, 3,5,5-trimethyl 19 31.87 2-propanone, 1-hydroxy 20 33.11 Ethanol, 2-butoxy 118 91 21 33.40 Acetic acid 60 94 22 33.37 2-furfuraldehyde 96 95 23 33.98 Formic acid 46 90 24 34.26 Propanoic acid 74 95 25 35.75 2-Butyl-3,4,5,6-tetrahydropyridine 139 47 26 36.63 Formamide, N, N-dibutyl- 157 97 27 37.07 1-piperidinecarboxyaldehyde 113 97 28 37.65 2,6-di-tert-butyl-p-cresol 220 95 29 38.46 Ethanol, 2,20 -oxybis- 106 90 30 39.09 Benzothiazole 135 94 a 91 124 38 74 91 The numbers in the table show the peak numbers in Figure 1 3,5,5-trimethyl-cyclohexene. The peak detected at 26.75 min shows that a great amount of water was present in the vial at the end of the test. Over the 35– 45 min, the peak identification becomes more uncertain because of the gradual loss of resolution; no attempt was made to improve this resolution. Nevertheless, the mass spectra comparison pointed the presence of 2,6-di-tert-butyl-p-cresol (peak 28) and a compound containing a sulfur atom (peak 30). Figure 1b shows the TIC chromatogram of the volatile by-products found in the headspace of a similar aged specimen except that 9 mL of insulating oil was present in the vial. It is interesting to see how the compositional profile of the oil components may interfere with the elution of the paper degradation byproducts. The 2,6-di-tert-butyl-p-cresol, which is present in substantial amounts in unaged oil, appeared as a well-shaped peak in the overlap on the right side of the profile. Under the present HSGC/ MS conditions, it is evident that the degradation byproducts eluted after 30 min could hardly be retained as indicators, thereby eliminating 2-FAL. On the other hand, Fig. 1c shows a TIC chromatogram of a sample collected in the headspace of an oil sample 123 302 Cellulose (2007) 14:295–309 Fig. 2 Stability with time of the molecules of potential diagnostic significance under conditions prevailing for open-breathing equipment T = 70 °C Concentration (mg/kg) 12 Methanol Acetone Ethanol 1-butanol (a) T = 90 °C 12 10 10 8 8 6 6 4 4 2 2 0 (b) 0 0 400 800 1200 1600 0 T = 110 °C Concentration (mg/kg) 12 800 1200 1600 800 1200 1600 T = 130 °C 12 (c) 10 10 8 8 6 6 4 4 2 2 0 (d) 0 0 400 800 Time (h) aged under identical conditions with in overlay the signal recorded for unaged oil. It is evident from these chromatograms that oil oxidation may contribute to a certain extent to the signal of some of the paper degradation by-products detected at retention times under 30 min (peaks 5, 7, 10, 12 and 17). A peak showing a 90% match with 2-methyl-2-propanol was seen to be associated only with the ageing of oil (totally absent in Fig. 1a and identified as A in Figs. 1b and 1c). After examining the mass spectrum for co-elution and confirming peak identity by injecting pure compounds, acetone (peak 5), methanol (peak 9), ethanol (peak 10) and 1-butanol (peak 17) were retained for further testing (all detected in preliminary oil samples collected from in-service transformers). The interference of oil ageing over the 123 400 1200 1600 0 400 Time (h) use of these molecules for estimating paper damage would therefore have to be evaluated. Stability of the indicators under some operating conditions The stability of the molecules with time was assessed in a range of temperatures that could be found in open-breathing transformers. The oxidative conditions prevailing in such units (oil in contact to the atmosphere through a conservator), were modeled by filling with air the headspace over the 9-mL oil solution of the four molecules. The results are depicted in Fig. 2. As seen in this figure, within the experimental variability measured with triplicate cells, the concentration of the four molecules Cellulose (2007) 14:295–309 remained reasonably stable over the 1600-h test duration when subjected to 70 and 90 8C. However, at 110 8C, acetone is dramatically unstable after an induction period of about 450 h, showing an increase of 2.1 mg/kg of oil/h. It is even worse at 130 8C where an increase of 11.6 mg/kg of oil/h is measured after a very short induction period (<100 h), which is indicative of a rapid exhaustion of the 2,6-di-tertbutyl-p-cresol, followed by a drop after about 800 h of testing. As it could be deduced from tests carried out with the blank samples, the appearance of acetone could essentially be attributed to the oxidizing deterioration of the oil components. Indeed, an increase rate of 2.2 and 11.4 mg/kg of oil/h were measured at 110 and 130 8C, respectively, which totally accounts for the values noted for the solution. Contrary to what was reported by Awata et al. (1997) for breathing transformers, oil oxidation contributes significantly under our conditions to the uptake of acetone, making this molecule inappropriate for indicating paper damage. On the other hand, a measurable decrease of CH3OH is noted when the solution is tested at 110 and 130 8C (note that the latter temperature is about ten degrees over what is normally experienced in the hotter parts of a fully loaded transformer). Such a disappearance could be attributed to a modification of the matrix polarity by the presence of oil-oxidation by-products. A reduction of the gas-oil partitioning coefficient of methanol with time would have the effect of reducing the amount of molecules accessible to the HSGC/MS analysis as reported for CO (polar compound) in a similar case (Jalbert et al. 2003). As for the two other candidates (CH3CH2OH and CH3CH2CH2CH2OH), this apparent loss could easily be corrected by using the appropriate gas-oil partitioning coefficients. In comparison, Unsworth and Mitchell (1990) reported an induction period of 800–1000 h at 110 8C for the 2-FAL in solution in uninhibited oil under oxidative conditions, after which significant losses were observed. Origin of the indicators For this application, it is essential to verify that the indicator is being generated by the damage of acellulose rather than by secondary paper constituents. A typical Kraft paper, which is found in most of the equipment still in operation, is composed of about 303 95–97% of a-cellulose, up to 4% of hemicellulose and up to 3% of lignin, along with varying trace amounts of inorganic salts (Clark 1962). Moreover, there is since the mid 1960s a growing trend toward using TU Kraft papers for oil-filled transformer applications. In the case of the TU products studied in this paper (Table 1), variable amounts of stabilizing nitrogenous agents were added to the Kraft components, and in addition for the Manning 220 Mannitherm D, medium hemp fibers were mixed with the modified Kraft fibers for improving the mechanical strength. Note that these two types of papers are often combined in the same piece of equipment. While a-cellulose is known to possess a crystalline/ amorphous character, the hemicelluloses are for their part amorphous polymers that are preferentially hydrolyzed (Shafizadeh 1982). Acetyl-4-O-methylglucuronoxylans (xylan) constitute the main hemicelluloses of hardwoods, while glucomannans (mannan) are found in softwoods. Lignin is a randomly linked, amorphous, high-molecular-weight phenolic compound that is more abundant and polymeric in softwoods than in hardwoods. It is generally considered to possess the greatest stability of all wood constituents in thermal treatment below 200 8C (Kollmann and Fengel 1965). To establish the contribution of the above constituents, model compounds were then aged in oil at 130 8C for 168 h under oxidative conditions. The results are presented in Table 4 together with those of a major by-product of the cellulose hydrolysis, D-(+)-glucose, and of cellulose pyrolysis, 1,6-anhydro-b-D-glucopyranose. It is interesting to note that all the indicators are generated by the model a-cellulose compounds tested, though the ethanol generation would significantly be masked by a concurrent formation from the oxidation of the oil components (last line in Table 4). The microcrystals of a-cellulose were obtained from cotton linters from which the amorphous regions linking the naturally occurring crystals were removed by acid hydrolysis (DPv of 157 ± 4 in this case). This well-structured cellulose yielded a higher amount of CH3OH when compared to Whatman paper, which suggests that the surface area of the tested material (powder of 20 mm particle size) was largely compensating for the lesser penetrability of the chemical reagents into the structure. In contrast, the microcrystals of the Whatman paper (also from cotton linters with a DPv of 281 ± 4) are still embedded by 123 304 Cellulose (2007) 14:295–309 Table 4 Contribution of the Kraft paper components to the formation of the molecules of potential diagnostic significance as established with model compoundsa Model compounds studied CH3OH CH3CH2 OH CH3CH2CH2CH2OH mg/kg of oil/g of component 2-FAL Components of Kraft paper a-cellulose Microcrystals from cotton linters 8940 640 430 29330 Whatman paper No. 41 1730 1080 590 14470 lignin Alkali Kraft lignin 153200 <D.L. <D.L. 1000 hemicelluloses from softwoods Mannan isolated from Saccharomyces cerevisiae <D.L.b 5390 12630 560 Hemicelluloses from hardwoods Xylan isolated from birch 227500 14770 790 674350 Major by-product from a-cellulose hydrolysis D-(+)-glucose <D.L. 620 460 29990 Major by-product from a-cellulose pyrolysis 1,6-anhydro-b-D-glucopyranose 1750 38630 <D.L. 2900 440 430 48 <D.L. Blank oil a Based on five replicates (each data = amount from model compound in oil heated at 130 8C––amount from oil heated at 130 8C ) b D.L.: Detection Limit less stable amorphous regions (7.4 ± 0.4% amorphous (Nelson 1960)), which means that a reduction of the stability of the material was expected. For the alkali lignin, the results showed a major yield of CH3OH with no detectable amount for the two other candidates. The fact that lignin constitutes less than 3% of the paper weight and has the greatest stability of all paper constituents could largely attenuate the impact of this finding for the present application. In regard to hemicelluloses, no detectable amount of CH3OH is found for mannan while a large contribution is associated to xylan. This observation is of interest considering that the insulating papers are generally manufactured from softwood species (mostly Black Spruce). The higher content of acetyl and methoxy groups in xylan may explain why this material has generated such a large amount of methanol. On the other hand, the test performed with D-(+)-glucose showed no detectable amount of CH3OH, although moderate yields are noted for the two other candidates. In the case of 1,6-anhydro-b-D-glucopyranose, a moderate contribution to methanol was measured together with a major production of ethanol. In addition, the results in Table 4 indicate that 1-butanol is mostly associated with the softwood hemicelluloses and not with a-cellulose, which makes this molecule of lesser interest as an indicator of paper damage. Finally, it is interesting to note that the 123 relative yields measured for 2-FAL (last column in Table 4) are in conformity with data reported by Scheirs et al. (1998), showing the highest amount from the hardwood hemicelluloses (xylan), followed by a moderate contribution from a-cellulose and a weak contribution from lignin. It is also significant that 2-FAL is in large part produced from D-(+)glucose whereas this is not the case for CH3OH, though both appear to be formed in relatively small amounts from 1,6-anhydro-b-D-glucopyranose. To further demonstrate that methanol originates from a-cellulose and more specifically from the rupture of its 1,4-b-glucosidic bonds, additional ageing tests were carried out at 130 8C for 168 h with five strips of each insulating paper and pressboard. Contrary to the above test with model compounds, the moisture of the starting material was conditioned at about the same level (1.2–1.6% H2O (w/w)). The results are presented in Table 5. From the initial DPv (DPv, 0 h), an average weight for the molecular cellulose chains was obtained by dividing the gram-formula weight of the average chain length (using the formula suggested by Klemm et al. 1998) by the Avogadro number. From the final DPv (DPv, 168 h), the average statistical number of ruptured 1,4-b-glucosidic bonds (NS) experienced by the molecular chains was obtained by applying the relation NS = DPv, initial/DPv, final 1 (Lundgaard Cellulose (2007) 14:295–309 305 Table 5 Confirmation of the existence of a relationship between methanol and the number of ruptured 1,4-b-glucosidic bonds of the cellulose (data from tests at 130 8C for 168 h with 0.5-g specimen immersed in 9 ml of oil under air atmosphere)a DPv, DPv, CE Rotherm (1.15% N2) Manning 220 Mannitherm D Munksjö Thermo-70 Clupak HD75 Hi-Val CE Rotherm transformerboard (0.91% N2) 0h 1144 ± 9 1168 ± 17 1170 ± 21 1140 ± 18 1099 ± 9 1300 ± 13 168 h 568 ± 11 570 ± 10 523 ± 9 618 ± 20 640 ± 8 690 ± 4 Average weight of a single molecular chain (g) 3.081 · 10 NS 1.016 19 Weight of specimen in ampoule 0.4666 (g) 3.146 · 10 19 3.152 · 10 19 3.071 · 10 19 2.960 · 10 19 3.502 · 10 1.050 1.237 0.844 0.718 0.886 0.5134 0.4516 0.4718 0.5182 0.4776 19 # of molecular chains in ampoule 1.514 · 1018 1.518 · 1018 1.333 · 1018 1.428 · 1018 1.628 · 1018 1.268 · 1018 18 18 18 18 18 1.123 · 1018 # of molecules of indicator according to 1 scission = 1 molecule 1.539 · 10 1.593 · 10 1.649 · 10 1.206 · 10 1.168 · 10 81.9 245.6 84.8 254.3 87.8 263.2 64.2 192.6 62.2 186.5 59.8 179.3 CH3OH 59.1 63.5 73.8 72.0 80.1 100.6 2-FAL 3.7 4.9 7.4 1.6 1.0 0.4 Expected amount in ampoule (mg) CH3OH 2-FAL Measured amount in oil (mg) a Based on five replicates et al. 2004). The number of molecular chains present in each ampoule at the beginning of the test was estimated knowing the weight of both, the paper specimens (after subtracting a contribution of 7% for the lignin and hemicellulose components except for Munksjö Thermo-70) and the average molecular chains (see Table 5). By multiplying the number of molecular chains by the number of ruptured bonds and assuming that cleavage of a 1,4-b-glucosidic bond leads to one molecule of indicator, it is then possible to calculate a microgram amount of CH3OH that should be found in the ampoule at the end of the test. As seen in Table 5, the concept that at least one molecule of CH3OH is generated each time a 1,4-bglucosidic bond is ruptured is well supported by the measured data considering that a certain amount of methanol was subtracted from the analysis due to absorption in the paper. For these tests, the effect of oil ageing on the partitioning of methanol is assumed to be at the same level for all the materials. It is also interesting to see that the nitrogenous agents incorporated in the Kraft papers (expressed in % N2 by dry weight in Table 1) reduce the material’s capacity to retain methanol. Finally, our data are conclusive on the difficulty for 2-FAL to reach a good level of agreement between the expected and measured values. Contrary to what was obtained for the ordinary Kraft specimens, very low amounts of 2FAL were detected in the ampoules of the TU papers. This behavior has also been reported by other authors who formulated the possibility of the 2-FAL destruction through a reaction with dicyandiamide, one of the agents used for TU papers (Morais et al. 1999). Dependence of methanol on temperature and moisture To verify the dependence of methanol on temperature and moisture, ageing tests were carried out with specimens of the five insulating papers for which the moisture content was set at 0.5%, 1% or 2% (w/w) (corresponding to a range of values found for transformer papers). The temperatures in the range 60–120 8C were chosen in order to maintain the system in a single mode of deterioration to which the Arrhenius equations could be applied. The sealed ampoules used for these tests all contained a 0.5 gpaper strip immersed in 9 mL of oil with the headspace filled with air (oxidative conditions). Typical results of the DP change (expressed as 123 306 Cellulose (2007) 14:295–309 (1986), it could be incorrect to classify the fine structure in cellulose into only two straight fractions—amorphous and crystalline—but rather into several fractions, which vary in degree of perfection of lateral arrangement of the chain molecules. On the other hand, it is very revealing to see that the CH3OH versus Time plots displayed in Fig. 3b show patterns very similar to those observed for the depolymerization (Fig. 3a). To the authors’ knowledge, such a good agreement with the DPv of the aged specimens has never been reported in the past (including for 2FAL). By obtaining the rates of the CH3OH formation in the second section of the degradation for the temperatures experienced for in-service transformers, it would then be possible to evaluate the thermal deterioration and the thermal life of the insulating papers over long periods (the details on the kinetics are beyond the scope of this paper). Typical overall results of the dependence of CH3OH in oil on the number of chain scissions are shown in Fig. 4 together with the equivalent relation for 2-FAL. These relations were built up by integrating the data obtained at the seven temperatures and three moisture levels tested (each symbol corresponds to a different initial moisture content). It is very interesting to note that the methanol production is virtually linear with NS regardless of the type of paper assessed (ordinary Kraft or TU Kraft papers). It is obvious that a greater dispersion of the data points is seen when going to NS > 3, corresponding to a DPv fall below 300 (upper value adopted by the electric power utilities as a criterion of the end of service life). The slopes of the linear relations for the TU 1/DPv, t) for all temperatures at one moisture level as a function of ageing times are shown in Fig. 3 for CE Rotherm (case of 1.15% N2) along with the evolution of the CH3OH content of the oil. As usually reported for the heterogeneous hydrolysis of cellulose (Feller et al. 1986), three sections are expected for the curves in Fig. 3a: a first section corresponding to a fast initial rate followed by one of moderate rate, and then by one with an extremely slow rate. In this figure, it is possible to distinguish the fast rate from the moderate rate only for T < 100 8C. Moreover, this section is extended over a wide timescale (0–1000 h), contrary to what is seen for pure hydrolysis, probably due to the presence of oil in the system. The fast rate section is considered to represent an attack on a small fraction of 1,4-b-glucosidic bonds of the cellulosic fibrils that are particularly sensitive to rupture. These bonds, which are designated as weak links in the literature, could be of native origin (Kilzer 1971) or could have been introduced during the delignification of wood pulp. The second section of the curves implies the rupture of the 1,4-b-glucosidic bonds between the normal b-D-glucopyranose units located in the amorphous regions of the cellulosic fibrils. In the case of the curves at T > 100 8C, this latter section appears to be analyzable into more than one straight line if sufficient data at very long time intervals are available. This is indicative of a transition through the third section corresponding to an attack of the microcrystalline regions of the cellulosic fibrils. The patterns of the curves in Fig. 3a do not allow a clear transition to be identified between the moderate and slow rate sections. As discussed by Feller et al. (b) (a) Formation of ageing indicator Paper depolymerization 450 6 400 DP = 200 5 Temperature (°C) 120 110 100 90 80 70 60 350 -3 (µg CH 3OH/g paper) Criterion of end of service life 4 1/DPv, t x 10 Fig. 3 Typical overall results for the relationship between paper depolymerization and formation of indicator with time—case of CE Rotherm 1.15% N2 specimens equilibrated at 2% (w/w) H2O DP = 300 DP = 437 3 DP = 446 DP = 445 2 300 250 200 150 100 1 50 0 0 0 4000 8000 Time (h) 123 12000 16000 0 4000 8000 12000 Time (h) 16000 Cellulose (2007) 14:295–309 Clupak HD75 specimens 25 DPv = 200 initial % (w/w) H2O 1.92 (a) 100 1.11 DPv = 250 DPv, initial = 1225 3 3 80 DPv = 300 60 15 DPv = 200 10 40 DPv, initial = 1225 DPv = 250 5 20 (µ g 2-FAL/kg of oil) x 10 0.47 20 (µ g CH3OH /kg of oil) x 10 Fig. 4 Typical overall results for the dependence of methanol and 2furfuraldehyde with the number of ruptured 1,4b-glucosidic bonds of the cellulose 307 DPv = 300 0 0 0 1 2 3 4 5 6 0 1 2 NS 3 4 5 6 NS Manning 220 Mannitherm D specimens 100 initial % (w/w) H2O 1.99 1.23 0.49 DPv, initial = 1275 80 3 25 (µg CH3OH/kg of oil) x 10 3 30 (b) DPv, initial = 1275 20 60 DPv = 300 15 40 DPv = 450 10 20 (µg 2-FAL/kg of oil) x 10 35 5 DPv = 300 0 0 0 1 2 3 NS Kraft papers are somewhat more pronounced than for the ordinary Kraft, as can be seen in Fig. 4 (also as expected based on the data in Table 5). These variations in the slopes could be attributed to the presence of nitrogenous agents, which modify the partitioning of methanol of this paper/oil system. Furthermore, the moisture content of the starting material seems to have no significance on the number of CH3OH molecules formed for each ruptured 1,4-bglucosidic bond. In comparison, the production of 2FAL with the number of chain scissions differs markedly from CH3OH and also between the two paper types investigated. In the case of Clupak HD75, the concentration is seen to increase very slowly up to a NS of 2, corresponding to a DPv of about 400, and 4 5 6 0 1 2 3 4 5 6 NS then to increase exponentially to a maximum of 100 mg/kg of oil. A turning point at about 400 b-Dglucopyranose rings has already been reported by others under similar conditions (Heywood 1997). This implies that any life-predicting model based on 2-FAL would suffer from a lack of sensitivity in the 1200–400 DPv range. The situation is even worse for the TU papers, as shown for Manning 220 Mannitherm D in Fig. 4b, where 2-FAL is found to increase and then to decrease with all the values below 0.1 mg/kg of oil. Such a discrepancy between Kraft and TU Kraft papers for 2-FAL with the number of chain scissions was recently reported for tests performed under similar conditions (Lundgaard et al. 2004). The data for the Clupak HD75 specimens 123 308 Cellulose (2007) 14:295–309 (see Fig. 4) did not allowed a distinction to be made between the three-moisture levels on the production of 2-FAL. For this specific case, about half of the HPLC analyses were performed on oil samples that were in contact with the paper strip after the test for a longer period than the mandatory 3 h, which is seen as a source of additional dispersion. Occurrence of methanol in field equipment For a final view of the possibility of using methanol as an indicator of paper damage, its occurrence in the field was measured in over than 900 pieces of equipment of the actual transmission network of Hydro-Québec (including power transformers, autotransformers, shunt reactors and grounding reactors with the oldest units operating for over 80 years). The analytical results are presented in Fig. 5 together with the situation experienced for 2-FAL. Under a detection limit of about 10 mg/kg for CH3OH and 2 mg/kg for 2-FAL, the results showed that the former is seen in 94% of the cases with a reduction to 56% for the latter. Moreover, the 2-FAL concentrations were seen to decrease markedly in the 1980s corresponding to the introduction at Hydro-Québec of the TU-Kraft papers. CH 3OH 22 20 % of detection = 94 % (n = 943) (a) 3 Concentration (mg/kg of oil) 2 1 0 2-FAL 12 % of detection = 56% (n = 943) (b) 11 6 4 2 0 0 20 40 60 80 100 Age (years) Fig. 5 Occurrence of methanol and 2-furfuraldehyde in oil samples collected from equipment of the Hydro-Québec’s transmission network 123 Conclusion Notwithstanding the fact that this study was carried out without a precise understanding of the chemical reactions involved in the system, sufficient evidence was gained to link the evolution of methanol in oil to the depolymerization of a-cellulose (rupture of 1,4-bglucosidic bonds). In this respect, the molecule was found to possess the best potential among all the HSGC/MS investigated cellulose degradation byproducts. This finding introduces the possibility of developing a sensitive ageing model for assessing the condition of both ordinary Kraft and TU-Kraft papers in power transformers. However, to go further with this application, the effect of oil ageing and nitrogenous agents on the partitioning of methanol would have to be clarified (under study in our laboratory). Acknowledgements The authors would like to express their gratitude to M. C. Lessard and B. Noirhomme from IREQ for valuable discussions on this concept. They would also like to thank A. Jolicoeur and P. Gervais from TransÉnergie for supporting the project from the beginning. Finally, a note of appreciation is extended to E. Grenier and M. Dontigny for the initial technical support. References Abe K, Maeda T, Shimonaka T, Umemura T (1994) A sensing system for degradation diagnosis of oil-filled transformers. Conference Record of IEEE International Symposium on Electrical Insulation, p 29 Awata M, Mizuno K, Ueda T, Ohta N, Ishii T, Tsukioka H (1997) Diagnosis by acetone for determination of breathing transformers containing an adsorbent in the insulating oil, Trans. IEE Japan 117-B(5):706 Burton PJ, Graham J, Hall AC, Laver JA, Oliver AJ (1984) Recent developments by CEGB to improve the prediction and monitoring of transformer performance. CIGRE Conference Paper 12-09 Burton PJ, Carballeira M, Duval M, Fuller CW, Graham J, de Pablo A, Samat J, Spicar E (1988) Application of liquid chromatography to the analysis of electrical insulating materials, Proc CIGRE Conf, Paris, Paper 15-08 Byrne GA, Gardiner D, Holmes FH (1966) The pyrolysis of cellulose and the action of flame-retardants. J Appl Chem 16:81 Clark FM (1962) Cellulose and liquid impregnated cellulose insulation. In: Insulating materials for design and engineering practice, John Wiley and Sons, Inc., London, p 251 Emsley AM, Stevens GC (1994) Kinetics and mechanisms of the low-temperature degradation of cellulose. Cellulose 1:26–56 Cellulose (2007) 14:295–309 Feller RL, Lee SB, Bogaard J (1986) The kinetics of cellulose deterioration. J Adv Chem Ser 212:329 Glassner S, Pierce AR (1965) Gas chromatographic analysis of products from controlled application of heat to paper and levoglucosan. Anal Chem 37:525 Heywood RJ (1997) The degradation models of cellulosic transformer insulation, Ph.D. thesis, University of Surrey, Guilford, Surrey Jalbert J, Gilbert R, Tétreault P, El Khakani MA (2003) Matrix effects affecting the indirect calibration of the static headspace-gas chromatographic method used for dissolved gas analysis in dielectric liquids. Anal Chem 75(19):5230 Kilzer FJ (1971) Thermal degradation of cellulose. In: Bikales NM, Segal L (eds) High polymers, vol 5, Pt 5, p 1015 Klemm D, Philipp B, Heinze T, Heinze U, Wagenknecht W (1998) Comprehensive cellulose chemistry, vol 1. WileyVCH, p 10 Kollmann F and Fengel D (1965) Cited in: Thermal degradation of wood components, U.S.D.A. Forest Service Research Paper FPL 130, U.S. Department of Agriculture, Forest Products Laboratory, Madison, Wisconsin, 1970 Lessard M-C, Lamarre C, Gendron A, Massé M (1995) Determination of paper degradation by-products by direct injection on an HPLC column, Annual Report of CEIDP Conference, Session 4B Lundgaard LE, Hansen W, Linhjell D, Painter TJ (2004) Aging of oil-impregnated paper in power transformers. IEEE Trans on Power Delivery 19(1):230 Morais RM, Mannheimer WA, Carballeira M, Noualhaguet JC (1999) Furfural analysis for assessing degradation of thermally upgraded papers in transformer insulation. 309 IEEE Trans on Dielectrics and Electrical Insulation 6(2):159 Nelson ML (1960) Apparent activation energy of hydrolysis of some cellulose materials. J of Polymer Sci 18:351 Oommen TV, Petrie EM, Breemen RB, Haney CA (1993) Analysis of furanic compounds from cellulose aging by GC-MS and attempts to correlate with degree of polymerization, CIGRE Berlin Symposium, Paper 110-02 Scheirs J, Camino G, Avidano M, Tumiatti W (1998) Origin of furanic compounds in thermal degradation of cellulosic insulating paper. J Applied Polymer Sci 69:2541 Schroff DH, Stannett AW (1985) A review of paper aging in power transformers. IEE Proc 132(6):312 Schwenker RF, Beck LR (1963) Study of the pyrolytic decomposition of cellulose by gas chromatography. J Polymer Sci Part C 2:331–340 Shafizadeh F (1968) Pyrolysis and combustion of cellulose materials. In: Wolfrom ML, Tipson RS (eds) Advances carbohydrate chemistry, vol 23. Academic press, New York, p 419 Shafizadeh F (1982) Introduction to pyrolysis of biomass. J of Anal Applied Pyrolysis 3:283 Soares S, Ricardo N, Heatley F, Rodrigues E (2001) Low temperature thermal degradation of cellulosic insulating paper in air and transformer oil. Polymer Int 50:303 Tamura R, Anetai H, Ishii T, Kawamura T (1981) Diagnostic of ageing deterioration of insulating paper. JIEE Proc Pub A 101:30 Unsworth J, Mitchell F (1990) Degradation of electrical insulating paper monitored with high performance liquid chromatography. IEEE Trans on Electrical Insulation 25(4):737 123