Vacuum pyrolysis of cellulose: Fourier transform infrared characterization of solid residues, product distribution and correlations

Journal of Analytical and AppIied Pyrolysis, 19 (1991) 81-104 Elsevier Science Publishers B.V., Amsterdam

81

Vacuum pyrolysis of cellulose: Fourier transform infrared characterization of solid residues, product distribution and correlations

S. Julien **a, E. Chornet a, P.K. Tiwari a and R.P. Overend b a Department of Chemical Engineering, UniversitP de Sherbrooke, Sherbrooke, Qutbec, JIK 2R1 (Canada) b National Research Council of Canada, Division of Biological Sciences, Ottawa, Ontario, KIA OR6 (Canada) (Received July 12, 1990; accepted in final form December 12, 1990)

ABSTRACT Vacuum pyrolysis of purified cellulose in constant rate of temperature increase experiments has been used, in conjunction with X-ray crystallography and FTIR of the solid residues, to probe the mechanism of conversion of cellulose into liquid products. The high yield of levoglucosan in the absence of cationic impurities, and the high yield of hydroxyacetaldehyde in the alternate pathway of impure or metal ion doped cellulose has been confirmed. The decomposition pathways correlate well with the yields of carbon oxides as has been found previously in different types of apparatus. The FTIR information shows that the residual solid is slowly transformed from a cellulose, i.e. pyranose ring structure, to a solid with little or no pyranose structure having considerable olefinic and carbonyl character at a conversion of about 50% of the solid to gaseous and liquid products. This occurs at about 325°C for the impure high DP cellulose and at about 3OO’C for the purified but lower DP cellulose. Increasing temperature continues to result in weight loss and increasing aromatization of the residual solid. The transition from a cellulosic to a dehydrated structure is much more clean cut in the case of the purified cellulose and the yield of levoglucosan (i.e. cyclised anhydro pyranose rings) is greater. An experiment conducted in the hydroxyacetaldehyde regime has demonstrated that the choice of pathways is dictated not only by the presence of the metal ions, but also is a function of the DP even at relatively high DPs. Carbon oxide correlations; cellulose; infrared spectroscopy; levoglucosan; pyrolysis; vacuum pyrolysis.

INTRODUCTION

The pyrolysis of biomass and its components such as cellulose has been studied for a long time and gradually a comprehensive picture of the processes involved is emerging. The effects of pressure, temperature and 0165-2370/91/$03.50

0 1991 Elsevier Science Publishers B.V.

82

time have been reviewed recently by Antal [l] and the effects of these on the products and their presumed pathways have been identified. The overall features of cellulose pyrolysis may be well described; however, the product distribution is found to be a function of the isolation and pretreatment of the cellulose, with two major and possibly mutually exclusive pathways having been identified. Relatively impure cellulose produces a significant amount of hydroxyacetaldehyde, while cellulose purified by an acid pretreatment leads to levoglucosan as a major product pathway. Irrespective of the pathway the overall progress of the reaction is well correlated with the production of even smaller product molecules, the carbon oxides CO and CO, [23]. This effect was presumed to be independent of the pyrolysis device, the final temperature and the rate of heating, though this has not been demonstrated for vacuum pyrolysis. Scott and his co-workers [9] have demonstrated that the removal of inorganic ions from the cellulose by an acid pretreatment (which also reduces the DP of the cellulose polymer) increases the yield of levoglucosan in a flash pyrolysis apparatus. Richards [lo] contends that both levoglucosan and hydroxyacetaldehyde are produced in parallel reactions and has proposed a mechanism for the formation of hydroxyacetaldehyde directly from the pyranose ring of the glucose units of the cellulose. He was unable to establish if the levoglucosan pathway was solely a function of inorganic ion removal or some concomitant change in the fine structure of the cellulose such as DP. Golova [ll], and Basch and Lewin [12] suggested that DP reduction to about 200 units is a precursor to an unzipping reaction of the interlinkage p-0 1-4 linkages that starts at the free end of the molecules. The Waterloo group [9] suggested that the free ends could be capped by the highly polar alkaline cations to prevent this heterolytic scission that liberates free pyranose rings that cyclise to levoglucosan. One technique that is capable of providing access to the primary rather than the secondary reactions of cellulose and its thermal pyrolysis products is vacuum pyrolysis [2--Q which is able to produce extraordinary high yields of liquid products as a result of the primary decomposition products being swept from the pyrolysis zone and hence trapped. We have utilized this technique in conjunction with a detailed examination of the pyrolysed cellulose by X-ray diffractometry and FTIR to follow the evolution of the different product pathways. The evolution of the carbon oxides as a marker of reaction extent was also utilized as a convenient description of the extent of reaction. A statistical experiment with specially prepared celluloses having different DPs and levels of inorganic ion doping was tried in what was primarily the hydroxyacetaldehyde pathway in order to test the conflicting hypotheses of inorganic ion or DP as controlling the choice of reaction pathway.

83 EXPERIMENTAL

Apparatus

A complete description of the laboratory bench scale vacuum pyrolysis appears in Roy et al. [4] and only a brief description is given here. A schematic diagram of the experimental pyrolysis apparatus is depicted in Fig. 1. Experiments were conducted in a batch mode with a cylindrical stainless steel retort equipped with an inner quartz liner of 50 mm in diameter. The cellulose sample weighed 40 g and occupied a volume of about 300 cm3 in the quartz liner. The retort was installed in a Lindberg vertical three-zone electric furnace equipped with a temperature profaner to control the final decomposition temperature and the heating rates at desired levels. The final temperature and heating rates were varied between 210 and 525°C 2 and 16”C/rnin, respectively for the experiments herein reported. The total degradation time for all of the experiments was 180 min. The average pressure was 130 Pa in the retort. Condensables were recovered in a series of cold water and dry ice-acetone traps. The non-condensable gases were pumped into an evacuated steel tank of 44.15 1. The

unit

1. Recorder 2. Thermoccuple meter 3. Themlocoupte 4. Retort 5. Thrae zone electric furnace 6. Sample 7. Temperature programmer

6. Coolinc tracs flee ai w&r) 9. Cooling traps (dry Ice and acetone) 10. Carder gas 11. Gas sampling valve 12. Sampling loop

Fig. 1. Diagram of bench scale vacuum pyrolysis unit.

13. Electronic manometer head 14. Digital raadout 15. Vacuum pump 16. Gas ccllecting bottle 17. Thermometer

84

pressure changes in the tank were recorded, the average molecular mass is calculated from the gas analysis and the total mass of gas was calculated assuming ideal gas behavior. The solid residue remaining in the retort was collected and weighed. The material balances closed between 97 and 100% of the initial charge for all the experiments herein reported. Losses were assumed to be gases. Materials used The starting material was Sigmacell cr-cellulose from Sigma Chemical Co. with a purity of 99.6%. Its elemental composition was 43.65% C, 6.55% H and 49.80% 0 (by difference). The degree of polymerization measured by high performance size exclusion chromatography (HPSEC) was 1100. The crystallinity index was 0.83. The initial ash content was 0.2%. Residual moisture of the air-dry sample varied between 1.5 and 6%. Cellulose pretreatment Acid pretreatment of cellulosic samples was performed by stirring cellulose with 10 times its weight of 1% sulfuric acid (Anachemia, practical grade 95-98% H,SO,). Pretreatments were conducted at room temperature for between 0.5 and 32 h. The samples were washed to neutrality (pH = 7) with distilled water. The ash content and the degree of polymerization are summarized in Table 1. Ion-exchange treatments were carried out by soaking the acid pretreated cellulose in 0.01 to 0.16 N solutions of potassium carbonate (J.T. Baker Chemical Co. with an assay 100% K&O,) for 2 h at room temperature. A solids consistency of 10% weight was used. After each treatment the samples were washed with 41 of distilled water. The potassium contents are shown in Table 2.

TABLE 1 Ash content Soaking time

and degree of polymerization

0.5

washing

times

DP

Inorganic ions (W wt./wt.)

1105 1010 785 530 460

0.030 0.025 0.022 0.026 0.024

(h) 2 8 16 32

at different

TABLE 2 Ash content of acid-washed cellulose for different solutions of potassium carbonate DP

Solution of K&O, (N)

Inorganic ions (K+ ) (% wt./wt.)

785 530,lOlO 460,785,1105 530,lOlO 785

0.01 0.02 0.04 0.08 0.16

0.2040 0.2154 0.2389 0.2661 0.2750

Infrared spectroscopy

The cellulose Sigma and residual chars FTIR spectra were obtained with a Nicolet 5-DXB system equipped with a TGDS detector at a resolution of 4 cm-‘. The dry samples were milled with IR grade KBr (Aldrich Co.) at a concentration of 1% (wt.) and 250 mg was used to form the potassium bromide disk for analysis. Four hundred scans of the samples were collected using the diffuse reflectance accessory and baseline correction was applied in the 1800-400 cm-’ area in order to eliminate the ramping effect of the diffuse reflectance [14]. Pure KBr background spectrum was subtracted from the sample spectrum. The 400-4600 cm-’ region was investigated. Elemental

analysis

Carbon and hydrogen were determined from air-dried samples using a Perkin-Elmer 240C elemental analyzer. The oxygen content was calculated by a difference method. The qualitative analysis of the inorganic ion content of original cellulose Sigma was carried out by X-ray scattering using a Link AN 10000 instrument. Thermogravimetric

analysis (TGA)

Thermogravimetric analysis was performed using a Netzsch Model 409 Thermobalance and Model 413 Controller [14]. The sample (25 mg) was placed in an alumina crucible and was heated from room temperature to 600°C using a heating rate of S’C/min. A gas flow of nitrogen (100 ml/mm) was used. Crystallinity

Crystallinity measurements were carried out by X-ray diffraction using a Rigaku (Gergerflex D/max Series) equipped with a 2 kW X-ray generator

86 TABLE 3 Characteristics of HPLC columns used to determine major components

Column Eluent Flow rate Temperature Detector Standard

in organic liquids

Monosaccharide

Organic acid

Biorad HPX-97P Distilled water (HPLC grade) 0.6 ml/min 85°C Refractometer Gilson No. 131 internal (mannitol)

Biorad HPX-87H aqueous 0.01 N H,SO, 0.6 ml/min 25°C Refractometer Gilson No. 131 external (calibration curve)

and a copper source (KIY at 1.5405 A, operated at 40 kV and 30 mA) for the diffractometer traces. High performance liquid chromatography (HPLC)

Condensable products were diluted with water (1: 50 in weight) and filtered on a micropore silicone filter (0.45 pm), prior to analysis using different columns to identify monosaccharides and organic acids. This was accomplished by matching the product retention times with those of an internal standard (mannitol) on the monosaccharide column and by means of external standards for the organic acids column. Hydroxyacetaldehyde was analysed by HPLC (Biorad HPX-97P column at 85OC, eluent distilled water (HPLC grade), eluent flow-rate 0.6 ml,knin). The details of the HPLC technique used in our laboratory [14] for analysis of the water soluble fractions are given in Table 3. Water analysis

Water in the pyroligneous liquor was determined by the Karl Fisher technique using an automatic amperometric titremeter (Model KF132, Fisher Scientific Co.). High performance size exclusion chromatography (HPSEC)

The degree of polymerization of cellulose was determined on a Varian model 5000 liquid c~omato~aph equipped with a variable wavelength UV-100 detector (280 nm) (141. The carbanylated cellulose [15] was dissolved at a concentration of 0.1% (wt./wt.) in THF. lo-$ samples were injected in a series of two columns of styrene-divinylbenzene gel (Polymer Labs., 300 X 7.5 mm with a 5 pm particle size for column A and 10 pm

87 100

PYROLYTIC OILS

A 10

u-0

I

0

200

300

400

I

1

!500

TEMPERATURE ( C)

Fig. 2. Thermal decomposition of cellulose Sigma under vacuum (lO’C/min). Product yields as function of final temperatures.

at low heating

rate

particle size for column B) Column A consists of two columns with porosities of 100 and 1000 A respectively while the porosities of column B are lo4 and lo5 A. The elution flow rate was of 1.0 ml/min.

RESULTS

AND DISCUSSION

Untreated cellulose Product yields The solid residue consists of unreacted substrate and char. The pyrolytic oils refer to the total condensibles collected at - 80°C. The weight loss, liquid yields and FTIR data all show that a major transformation of the cellulose takes place between 300 and 35OOC. Our discussion will focus at this transformation and will use 325°C as the characteristic temperature showing the highest rate of weight loss. Characterization of the char The yields and composition of the charred residue are given in Fig. 2 and Table 4 respectively for different final temperatures. A van Krevelen di-

88 TABLE 4 Yields and composition of the charred residue obtained final temperatures noted (heating rate lO”C/min) Cellulose Sigma Temperature (“C) Heating rate (“C/mm) Cellulose (g) Cellulose (m.a.f. basis) (g) Moisture (W wt./wt.) Ash (W wt./wt.) DP % H,SO, (% wt./wt.)

Untreated 525

by pyrolysis

Pretreated

of cellulose

Sigma at the

Limits of error%

525

10 40.1 39.48 6.16 0.20 1100 0.00

10 40.2 39.58 1.52 0.02 500 1.00

5.00 0.25 0.25 1.00 5.00 1.00

3.73 11.13 0.25 11.84 0.65 1.63 4.40 3.09 17.57 54.29 21.89 76.18 10.13 13.69 100.00

33.09 0.00 2.03 6.94 0.20 0.80 2.44 4.04 23.10 72.64 14.50 87.14 8.56 4.30 100.00

1.15 1.05 2.05 1.50 2.15 1.95 1.45 1.30 1.35 0.25 0.50 0.75 0.35 0.40 1.50

Yields, S of m.a.f. feed Levoglucosan Hydroxyacetaldehyde Glucose Acetic acid Hydroxymethylfurfural Formic acid Levulinic acid Propionic acid Others (tar) Total organic liquid Water Total liquid Char Gas Total

agram [16] illustrating the changes in elemental composition of the charred residue with increasing reaction temperature is shown in Fig. 3. Along with the theoretical lines formed by the dehydration, decarboxylation and decarbonylation routes, the evolution of the curve suggests that three different trends are predominant during the char formation. First, the decomposition is mainly due to dehydration reactions between 210 and 325°C. In the range between 325 and 425”C, the second stage shows an increased loss of oxygen due to decarbonylation (CO) and decarboxylation (CO,) reactions. Finally, an increased loss of hydrogen due to dehydrogenation and demethanation reactions seems to occur above 425°C. This increased loss of hydrogen is mainly due to decomposition of the aliphatic groups and condensation to polycyclic aromatic compounds [17]. It is possible to illustrate the sequence of thermal decomposition of cellulose if we compare the same bands in the spectra of different residues

89

1.6

0.6

0.2

0.0

0.0

0.2

0.4 ATOMIC

0.6 RATIO

0.6

1.0

O/C

Fig. 3. Van Krevelen diagram illustrating the relationship between H/C and O/C atomic ratio of chars derived from the pyrolysis of untreated cellulose Sigma at temperatures noted.

obtained at different final temperatures. The spectra of cellulose Sigma and the charred residue obtained with increasing reaction temperatures are shown in Fig. 4. Probable band assignments [18] for each significant peak of infrared spectra of cellulose and residual materials are also indicated. At a final temperature < 325°C the evolution of spectrum for the char shows a decrease in the intensities of absorption bands due to hydroxyl stretching (3300-3400 cm-‘) and bending (1300-1340-1360 cm-‘). Also the spectrum shows the emergence of bands at 1735 and 1620 cm-’ attributed respectively to carbonyl C=O stretching and unsaturated carbon-carbon C=C stretching. These are the changes expected for dehydration reactions leading to the formation of anhydrocellulose [19]. Also at T 2 240°C we observed a gradual decomposition of the glycopyranose ring by a decrease of bands at 2900 (-C-H stretching), 1160 (antisym. bridge C-O-C stretching), 1125 and both 1060 and 1035 cm-’ (skeletal vibration involving C-O stretching). The spectrum, however, showed that the structure of the material is predominantly still that. of cellulose. At T c 325°C X-ray diffraction (Fig. 5) shows a decrease in the crystalline regions of the cellulose (peaks at 20 = 15.5” and 22.5 “). The low temperature breakdown of cellulose chains evidently occurs in the crystalline regions. Halpem and Patai [24] have suggested that the crystalline regions of the cellulose were the major source of levoglucosan.

90

91

,.___,____,_.__,____,.___,._..r 10

15

20

25

30

35

40

D’E~RLE (28)

Fig. 5. X-ray ~ffracto~a~es temperatures noted.

of char obtained by vacuum pyrolysis of cellulose Sigma at

At T > 325”C, we observed a complete destruction of the signal corresponding to the glycopyranose ring with a shift of the band from 1620 (non conjugated) [16] to 1600 (conjugate) carbon-carbon C==C s~etc~ng 1171. The material is completely amorphous (by X-ray diffraction) at a temperature > 325*C (Fig. 5). We suggest that the band at 1620 cm-l may be assigned to unsaturated carbon-carbon C=C and not to OH absorption water because we observed the growth of this band in the subsequent samples. This is in agreement with Morterra [20] who degassed the char samples at 190°C and concluded that the band at 1620 carmot be attributed to the OH stretch of absorbed water. The spectrum shows the emergence of three major absorption bands at temperatures > 325OC. There is a complex absorption pattern between 1500 and 1300 cm-’ with a maximum near 1440 cm-’ (-CH, bending). Also there is an absorption band between 1300 and 1100 with a maximum near 1260 cm-‘. These bands, corresponding to aliphatic C-C and C-O stretching, appeared when the cellulose structure is destroyed and are at a maximum of 375°C and decline at higher temperature. Also, the aliphatic C-H stretch bands between 3000 and 2800 cm-’ assigned to CH, and CH groups

Fig. 4(a). FTIR spectra (between 400 and 2000 cm-‘) of char obtained by vacuum pyrolysis of cellulose Sigma at different final temperatures and a constant heating rate (lOT/min). Fig. 4(b). FTIR spectra (between 2600 and 3800 cm-‘) of char obtained by vacuum pyrolysis of cellulose Sigma at different final temperatures and a constant heating rate (lO”C/min).

92

behave similarly as the bands in the 1500-1100 cm-’ assigned to aliphatic structure. A broad absorption of 900-700 cm-’ grows in intensity and becomes a trio of well resolved bands at 880,820 and 760 cm-’ at higher temperatures (aromatic C-H out of plane bending mode). These bands grow in intensity up to 525OC and then decline at higher temperatures. Morterra and Low [20] have noted that the first appearance of an aromatic network is better monitored by observing the C-H wagging bands (900-700 cm-‘) than the stretchings at higher wavenumbers (3000-3100 cm-‘) because of the low photoacoustic response in the C-H stretching region. At 525°C the broad

J J / 2ooo

1800

.,

1600

1400

WAVENUMBERS

1200

[CM-l)

Fig. 6(a). FTIR spectra (between 400 and 2000 cm-‘) of char obtained by vacuum pyrolysis of cellulose Sigma at different heating rates and a constant final temperatures (325T).

93

16 CIMIN i

; 10 C/MN

3800

3600

3400

3200

WAVENUMBERS

3000

2800

2600

(CM-11

Fig. 6(b). FTIR spectra (between 2600 and 3800 cm-‘) of char obtained by vacuum pyrolysis of cellulose Sigma at different heating rates and a constant final temperatures (325°C).

absorption at 1600 cm-’ is assigned to aromatic C=C stretching mode [20]. The results indicate that the samples change from a mixture having mostly aliphatic C-H (325-425”(Z) groups to one having mostly aromatic C-H groups (425525°C). The influence of heating rate The influence of heating rate has been studied at the optimum decomposition temperature of near 325°C (Fig. 6(a) and Fig. 6(b)). The slow heating rate (3 2”C/min) suggests that the remaining carbon polymers are stabilized by the dehydration of cellulose. The spectrum (Fig. 6(a)) for the char shows that the polymeric structure is mainly retained. The dehydrated cellulose is probably less accessible to cleavage than the original polycellobiose because the elimination of the hydroxyl groups results in carbon-

carbon C=C and carbonyl C=O groups and subsequently there is a reduced volatilization at higher temperature. Slow heating rates provide more time for dehydration (Fig. 6(b)) and as a result this is complete before the onset of cleavage and depolymerization reactions at higher temperature. Product correlations

We have observed that carbon monoxide, levoglucosan, glucose, hydroxyacetaldehyde, acetic acid, propionic acid and hydroxymethylfurfural seemed to exhibit the same behavior with a change in the severity of the treatment (280°C and 525°C). For example the Fig. 7 and Fig. 8 cross-plot respectively the yields of levoglucosan and hydroxyacetaldehyde as weight percentage of original cellulose Sigma against the carbon monoxide yield. Table 5 shows that there exists a correlation between the yield of levoglucosan, glucose, hydroxyacetaldehyde, acetic acid, propionic acid, hydroxymethylfurfural and the yield of carbon monoxide. In the same way, there is also a good correlation (Table 5) between the yield of char, pyrolytic water, formic acid, levulinic acid and carbon dioxide. Correlation coefficients generally greater than 0.95 for the cellulose Sigma data suggest that the correlation is significant. Examples of cross-plot yields of char and pyrolytic water as a weight of original cellulose Sigma against the carbon dioxide are shown

4

”

I

--

I

0

I

I :

;

I ;

I ;

CO YIELD, WT % OF SAMPLE (mat)

Fig. 7. Levoglucosan

yield vs. CO yield for cellulose Sigma pyrolysed between (280-525°C).

CO YIELD, WT % OF SAMPLE (m.a.f.)

8. Hydroxyacetaldehyde (280-525”(I).

yield vs. CO2 yield for cellulose Sigma pyrolysed between

respectively in Fig. 9 and Fig. 10. These linear correlations have similarities with the ones developed by Funazukuri et al. [23] for cellulose flash pyrolysis.

TABLE 5 Results of product correlations Temp “C

ccl 210 240 280 300 325 375 425 525

Yields %of m.a.f. feed 97.20 93.02 78.22 45.37 20.47 13.55 12.08 9.96

Composition (wt.%) C H 0

43.65 43.40 44.36 48.40 60.69 75.16 76.34 80.65 87.55

6.55 6.45 6.37 6.01 4.91 3.62 3.84 3.44 2.34

49.80 50.15 49.27 45.58 34.40 21.23 19.83 15.91 10.11

Chemical formula (ref. to G)

GH1O.aO05.13

GH,,,,4., W-&0.354., f-%H8.9404.24 f-%HS.8302.5S C6H3.4601.27 f%H3.~20L17 f%H3.0,00.89 (%Hl.9200.52

Atomic ratio H/C

O/C

1.801 1.783 1.724 1.490 0.971 0.577 0.604 0.512 0.321

0.856 0.867 0.833 0.706 0.425 0.212 0.195 0.148 0.087

96

00

ae

5_4o 9 c 30 : 5 20

10

0 0

2

4

6

6

CO2 YIELD, WT % OF SAMPLE (m.a.t)

Fig. 9. Char yield vs. CO* yield for cellulose Sigma pyrolysed between (280-525°C).

22 20

0 0

2

4

6

6

CO2 YIELD, WT% OF SAMPLE (m.a.f.)

Fig. 10. Water yield vs. CO, yield for cellulose Sigma pyrolysed between (280-525°C).

10

97

The effect of acid-washed pretreatment Results from 525°C pyrolysis of untreated (DP 1100) and acid-washed (DP 500) cellulose Sigma are given in Table 6. The acid-washed cellulose (1% H,SO, wt./wt.) reduced the ash content of the samples from 0.22 to 0.02%. The original untreated cellulose contained principally the following inorganic ions: potassium (K), calcium (Ca) and silicon (Si) (Fig. 11) and all of them were removed (about 90%) by the acid-washed pretreatment. A fine gold (Au) and palladium (Pd) coating is used to avoid charging of the samples (Fig. ll), hence the presence of the corresponding peaks. The total organic liquid yield (i.e. the moisture free liquid yield) was increased from 54.3 to 72.6% when the inorganic cations are reduced. Char, gas and water yields are greatly reduced in comparison to values for untreated cellulose. These results suggest that decomposition reactions yielding liquid products have been favoured at the expense of decomposition reactions yielding char, gases and water. The very low quantities of acids and aldehydes obtained when acid-washed treated cellulose was pyrolysed indicates that the decomposition of the glucose unit from cellulose have been greatly reduced. The pretreated cellulose gave a much higher yield of levoglucosan (33.1% compared with 3.7% for the untreated cellulose) and the amount of hydroxyacetaldehyde (low molecular weight aldehyde) was drastically reduced from TABLE 6 Pyrolysis products from untreated and pretreated Sigma cellulose Products

Ai

Bi

‘co

co*

N

Levoglucosan Glucose Hydroxyacetaldehyde Acetic acid Propionoic acid Hydroxymethylfurfur~

- 0.215 0.008 - 0.282 0.563 1.119 0.088

0.973 0.063 2.597 2.504 0531 0.135

0.989 0.968 0.965 0.940 0.950 0.982

0.11-4.01 0.11-4.01 0.11-0.41 0.11-0.41 0.11-0.41 0.11-0.41

10 10 10 10 10 10

Products

Ci

Di

‘co,

co,

N

- 9.353 1.716 0.174 0.452

0.98 0.957 0.950 0.940

0.17-8.85 0.17-0.85 0.17-0.85 0.17-0.85

Char Pyrolytic water Formic Acid Levulinic acid Y, =

A, + ~YCO

Y[ = C; + QYCO, I =

UiCO/U,~oico

92.117 5.477 0.133 0.479

*

10 10 10 10

0) (2)

98 0 - ;?O ke'v'

LTV:

100s 132s

R&l

i

:SA9 MEMl

:CELLULOSE

8K

Praost:

29%

1005 Rtmai

Dead

6.000

ch

SIEMAIUNTIIEATEDI

krU 310=

ni ng:

Or

11.1 > 391 cts

Fig. 11. The qualitative analysis of inorganic ions content of original cellulose Sigma by X-ray scattering.

11.1 to 0%. Levoglucosan is thus formed at the expense of hydroxyacetaldehyde by removal of inorganic ions from original cellulose. Shafizadeh et al. [22] have explained the formation of hydroxyacetaldehyde by secondary decomposition of levoglucosan; the following reaction scheme was proposed: cellulose + levoglucosan + glucose + hydroxyacetaldehyde + C4 This reaction scheme has been rejected by Piskorz et al. [9] and Richards [lo]. They suggest that hydroxyacetaldehyde is formed by a primary ring fragmentation of cellulose. Richards et al. [lo] have proposed a homolytic mechanism for the formation of hydroxyacetaldehyde from cellulose pyrolysis, while the formation of levoglucosan results from a heterolytic mechanism [21]. Our results are in agreement with Richards’ view. We have established the negative influence of ionic impurities on levoglucosan yield during vacuum pyrolysis. Golova [ll] assumed that inorganic cations (K and Ca) can act as catalyst for complete decomposition of glycosidic units by a homolytic mechanism. The absence of inorganic ions favoured the direct depolymerization of celluliose to levoglucosan. A series of infrared spectra of char residues obtained by pyrolysis of acid-washed cellulose at different reaction temperatures are shown in Fig. 12. The collapse of the acid-washed cellulose structure is observed at 300°C instead of 325°C as in the untreated samples (Fig. 4(a)). The inorganic ions seem to stabilize thermally the glycosidic units. The thermogravimetric analysis of various celluloses show a downward shift of the temperature of maximum rate of decomposition in the pretreated cellulose (Fig. 13). Piskorz

99

-I 2000

1800

16W

1400

1200

1000

800

600

400

WA!ENUMBERS[CM-11 Fig. 12. FTIR spectra of char obtained by vacuum pyrolysis of acid-washed cellulose Sigma at different final temperatures and a constant heating rate (lOT/min).

25”

”

300 0

35”

0

6,

TEMPERATURE DE&C

Fig. 13. Thermograms of the untreated and pretreated (1% H,SO,)

cellulose Sigma.

100 x2

4

5

I

(0,1.414)

1

Fig. 14. Locus of the central composite design.

et al. [9] have concluded that a clear link exists between increased levoglucosan yields and the decrease of the temperature of maximum decomposition rate of 310°C to 300°C for flash pyrolysis. Our results also confirm the existence of this relationship even if under vacuum the volatilization temperatures are somewhat lower. The acid washing of cellulose (l%H,SO, wt./wt.) reduced also the degree of polymerization (DP) from 1100 to 500. Basch and Lewin [12] consider that the rate of levoglucosan formation will depend also on the initial degree of polymerization (DP). To confirm this hypothesis, we have used a statistical analysis design to study the effect of inorganic ions and the degree of polymerization (DP) on various response variables (products yields). In order to reduce the number of experiments, a central composite design was selected [25]. This design is visualized in Fig. 14. The experiments (Nos. 1- 11, Table 7) were performed in a random manner. An empirical quadratic model was obtained:

Then a second model with k = 2 variables is given by: The variables Xi and X, are dimensionless (Table 8). The first four experiments belong to a 22 factorial design; the &l coded value are obtained by calculating:

x, = (x1 - x,)/d x2 =

(x2 - X2)/d

The independant variables xi and x2 are respectively the degree of polymerization and the content of inorganic ions K+(% wt./wt.). The level

101

TABLE 7 Groups of experiments of the orthogonal central composite design (Product yields at 525°C and a heating rate of lO’C/min) Exp. X1 No. 1

1 2 3 4 5 6 7 8 9 10 11

1 1

Tar % wt./wt. 4 0.25%

X,

1

48.42 50.06 -1 52.92 -1 1 50.92 0 1.414 49.18 1.414 0 48.67 - 1.414 52.56 0 -1.414 0 52.87 0 0 51.25 0 0 51.22 0 0 51.30 -1 -1

Water %wt./wt. * 0.50%

Char 4%wt./wt f0.35%

Gas % wt./wt. f0.40%

Levoghlcosan % wt./wt. * 1.15%

21.00 20.73 19.88 20.19 20.31 21.09 19.79 20.11 20.11 20.14 20.04

11.26 10.48 9.82 10.34 11.26 11.37 9.85 9.75 10.51 10.63 10.52

19.32 18.72 17.39 18.55 19.24 18.87 17.80 17.28 18.13 18.00 18.14

3.25 3.49 3.74 3.55 3.43 3.37 3.65 3.73 3.58 3.62 3.54

Hydroxy acetaldehyde % wt./wt. f 1.05% 10.90 10.45 9.98 10.31 10.78 10.90 10.18 9.92 10.50 10.55 10.60

values of natural variables are summarized in Table 8. In this system of coordinates (Fig. 14) these 4 experiments are represented by the four vertices of a square. The next four experiments (Table 7) are represented on Fig. 14 by 2 pairs of points on the 2 axes at a distance (2k)“4 (k = 2 variables, a) from the center. Three replications at the center point have been made to evaluate the experimental error. From the statistical analysis of data (Table 9), the coefficients & @i, &, LL &Y rB12are obtained for the yields of char, tar, gas, water, levoglucosan and hydroxyacetaldehyde respectively. According to the results shown in Table 9, it appears that the effects of the main variables, the degree of polymerization (Xi) and the content of inorganic ion Kf ( X,) are significant at a 95% level of confidence for the I; distribution. TABLE 8 Facto&

levels used for the multiple regression analysis

Variables

Levels

Coded variables

Xi, X, *

Natural variables

Degree of polymerization (DP) xi Inorganic ions tK+) x2 (W wt./wt.)

-1.414 1105

-1

0

1

1010

785

530

0.2040

* Xi = (x, - 778)/ - 232; X, = (x2 - 0.2399)/0.0262. ues.)

0.2154

0.2389

1.414

0.2661

(Xi and X, are normalized

460 0.2750

level val-

102 TABLE 9 Results

of multiple

regression

and statistical

Y,

w

I%

Tar Water Char Gas Levoglucosan

51.26 20.10 10.56 18.09 3.58

- 1.4120 0.3815 0.4849 0.5456 -0.1324

Hydroxyacetal dehyde

10.55

* Significant

analysis

a * * * * *

- 1.0511 0.1637 0.4117 0.4757 - 0.0927

0.3058 *

&

82

&2

r

-0.3047 * 0.2828 * -0.0191 0.0410 -0.0245

-0.2547 0.7721 - 0.0191 0.2661 * - 0.0295

0.0900 - 0.0125 0.0633 - 0.1408 - 0.0125

0.987 0.974 0.956 0.978 0.964

-0.0789

- 0.0439

0.0300

0.980

* * * * *

0.2036 *

at a level of confidence

higher than 95%.

It also appears that the rate of levoglucosan formation depend on the initial degree of polymerization (DP). Basch and Lewin [12] have proposed that the formation of levoglucosan proceeds through free chain ends of the cellulose polymer. It is conceivable that the number of free chain ends is higher for a low DP. Then the increase in free chain ends (low DP) favours the levoglucosan formation. The low levoglucosan yields can be explained by the action of cation K+ associated with the cellulose polymer which might retard the unzipping reaction at the terminal free chain. This mechanism has been explained by Golova [ll]. Through the statistical analysis, we can conclude that there exists an interrelationship between the DP as well as the

$HO Decarbonylation

-CO

Decarboxylation-CO Dehydratation

- HP

FRAGMENTATION (Homolytic) High heating rate

CELLULOSE

D

(High DP)

CELL”LOSE

-1

(Low DP) High heating rate Low (K,*Ca’*) Low temperature

(300°C)

Low heating rate

DEPOLYMERlZATlON

CH,OH 2

r

:

HYDROXYACETALDEHYDE

Acetic acid

b

L

Propionic

acid

Formic acid

OH LEVOGLUCOSAN

(Heterolytic) Glucose

High (K,*C$) Other hexosan

I.-

)

Char, CO, CQ. H20

Fig. 15. Proposed mechanism for the pyrolysis of cellulose (temperature, heating rate and inorganic ion content).

Sigma under different

parameters

103

content in inorganic ion (K+) on the formation of levoglucosan and hydroxyacetaldehyde. All our results on the pyrolysis of cellulose Sigma can be summarized by a generalized decomposition mechanism (Fig. 15) [9]. Levoglucosan is produced at a temperature higher than 300°C a heating rate higher than 7 2”C/min and is favoured by a low DP and a low inorganic ion content (0.02% wt./wt.) (Fig. 15). The relative importance of the major pathways will be determined by the final temperature, the heating rate, the characteristics of the cellulose (DP) and the inorganic ion content.

ACKNOWLEDGEMENTS

The collaboration and technical assistance of P. Vidal and M. Trottier been greatly appreciated. The authors acknowledge the Natural Sciences Engineering Research Council of Canada and le Fonds pour la formation Chercheurs et l’Aide a la Recherche (F.C.A.R.) for financial support for project.

has and de this

LIST OF SYMBOLS

A, B, C, D d

DP k N r

x, ** Xl x2

Xl

x2

*i, YC

Yi Yco Yco,

*j

product-specific parameters of Eqns. (1) and (2) unit of variation of x from X degree of polymerization obtained by HPSEC the number of independent variables number of data used to evaluate the product-specific parameters correlation coefficient coded variable for the degree of polymerization coded variable for the content of inorganic ion K+ (wt.% of sample) natural variable for the degree of polymerization (DP) natural variable for the content of inorganic ions K+ (wt.% of sample) mean of natural variable expressing the degree of polymerization (DP) mean of natural variable expressing the content of inorganic ions K+ (wt.% of sample) the independent variables response for the cth experiment product yields (wt.% of sample) carbon monoxide yield (wt.% of sample) carbon dioxide yield (wt.% of sample)

104

Pi YCO

f?S,,

mean of product yields (wt.% of sample) mean of carbon monoxide yield (wt.% of sample) mean of carbon dioxide yield (wt.% of sample) ,, least square parameter estimates obtained from multiple Bigi;,B” regression

Greek letters u

C

standard deviation experimental error

REFERENCES 1 M.J. Antal Jr., in: K.W. Boer and J.A. Duffie (Editors), Advances in Solar Energy, American Solar Energy Society, Golden, CO, 1983, 61. 2 F. Shafizadeh, R.H. Fumeaux, T.G. Cochran, J.P. Scholl and Y. Sakai, J. Appl. Polym. Sci., 23 (1979) 3525. 3 F. Shafizadeh and T.T. Stevenson, J. Appl. Polym. Sci., 27 (1982) 4577. 4 C. Roy, B. de Caumia, P. Plante and H. Menard, in: R.P. Overend, T.A. Milne and L.K. Mudge (Editors), Fundamentals of Thermochemical Biomass Conversion, Elsevier, London, 1985, 237. R.J. Evans, T.A. Milne, and M.N. Soltys, J. Anal. Appl. Pyrolysis, 6 (1984) 273. R.K. Agrawal and R.J. McCluskey, J. Am. Chem. Sot., 28 (5) (1984) 301. R.K. Agrawal and R.J. McCluskey, J. Anal. Appl. Pyrolysis, 6 (1984) 325. R.K. Agrawal and R.J. McCluskey, Fuel, 64 (1985) 1502. J. Piskorz, A.G. Radlein, D.S. Scott and S. Czernik, J. Anal. Appl. Pyrolysis, 16 (1989) 127. 10 G.N. Richards, J. Anal. Appl. Pyrolysis, 10 (1987) 251. 11 O.P. Golova, Russian Chem. Rev., 44 (8) (1975) 687. 12 A. Basch and M. Lewin, J. Polym. Sci., 11 (1973) 3071. 13 T. Funazukuri, R.R. Hudgins, and P.L. Silveston, Ind. Eng. Chem. Proc. Des. Devel., 25 (1986) 172. 14 J. Bouchard., E. Chomet and R.P. Overend, Report of contract file No. 24ST 23216-6-6168, Renewable Energy Division; Energy, Mines and Resources Canada, Ottawa, Ontario, 1988,128 pp. 15 D.M. Hall and J.R. Home, J. Appl. Polym. Sci., 17 (1973) 3727. 16 M.M. Tang and R. Bacon, Carbon, 2 (1964) 211. 17 Y. Sekiguchi and F. Shafizadeh, J. Appl. Polym. Sci., 29 (1984) 1267. Consultants Bureau, 18 R.G. Zhbankov, Infrared Spectra of Cellulose and its Derivatives, New York, 1966. 19 A.D. Pouwels, G.B. Eijkel and J.J. Boon, J. Anal. Appl. Pyrolysis, 14 (1989) 237. 20 C. Morterra and M.J.D. Low, Carbon, 21 (3) (1983) 283. 21 F. Shafizadeh and Y.L. Yu, Carbohydr. Res., 29 (1973) 113. 22 F. Shafizadeh, J. Anal. Appl. Pyrolysis, 3 (1982) 283. 23 T. Funazukuri, R.R. Hudgins and P.L. Silveston, J. Anal. Appl. Pyrolysis, 13 (1988) 103. 24 Y. Halpem and S. Patai, Israel J. Chem., 7 (1969) 673. 25 DC. Montgomery, Design and Analysis of experiments, 2nd ed., Wiley, New York, 1984.