Fourier Transform Infrared Quantitative Analysis of Sugars and Lignin in Pretreated Softwood Solid Residues

Size: px

Start display at page:

Download "Fourier Transform Infrared Quantitative Analysis of Sugars and Lignin in Pretreated Softwood Solid Residues"

Ira Brown
6 years ago
Views:

1 Copyright FTIR Anlysis 2001 by of Humn Sugrs Press nd Inc. Lignins 51 All rights of ny nture whtsoever reserved /01/91 93/0051/$12.75 Fourier Trnsform Infrred Quntittive Anlysis of Sugrs nd Lignin in Pretreted Softwood Solid Residues Abstrct MELVIN P. TUCKER,*,1 QUANG A. NGUYEN, 1 FANNIE P. EDDY, 1 KIRAN L. KADAM, 1 LYNN M. GEDVILAS, 2 AND JOHN D. WEBB 2 Ntionl Renewble Energy Lbortory, 1 Biotechnology Center for Fuels nd Chemicls nd 2 Center for Mesurements nd Chrcteriztion, 1617 Cole Boulevrd, Golden, CO 80401, E-mil: melvin_tucker@nrel.gov Hydrolystes were obtined from dilute sulfuric cid pretretment of whole-tree softwood forest thinnings nd softwood swdust. Mid-infrred (IR) spectr were obtined on smple sets of wet wshed hydrolystes, nd 45 C vcuum-dried wshed hydrolystes, using Fourier trnsform infrred (FTIR) spectrophotometer equipped with dimond-composite ttenuted totl reflectnce (ATR) cell. Prtil lest squres (PLS) nlysis of spectr from ech smple set ws performed. Regression nlyses for sugr components nd lignin were generted using results obtined from stndrd wet chemicl nd high-performnce liquid chromtogrphy methods. The correltion coefficients of the predicted nd mesured vlues were >0.9. The root men squre stndrd error of the estimte for ech component in the residues ws generlly within 2 wt% of the mesured vlue except where reported in the tbles. The PLS regression nlysis of the wet wshed solids ws similr to the PLS regression nlysis on the 45 C vcuum-dried smple set. The FTIR-ATR technique llows mid-ir spectr to be obtined in few minutes from wet wshed or dried wshed pretreted biomss solids. The prediction of the solids composition of n unknown wshed pretreted solid is very rpid once the PLS method hs been clibrted with known stndrd solid residues. Index Entries: Fourier trnsform infrred; biomss; softwood; dilute-cid pretretment; cid hydrolysis. *Author to whom ll correspondence nd reprint requests should be ddressed. 51

2 52 Tucker et l. Introduction The conversion of renewble lignocellulosic resources to bioethnol requires number of process steps. We hve used dilute-cid pretretment with sulfuric cid s one of the erly process steps to hydrolyze significnt portion of the hemicellulose in softwood feedstocks consisting of wholetree forest thinnings or swdust from swmill wstes. A second pretretment t higher tempertures nd cid concentrtions ws used to hydrolyze the remining hemicellulose nd some of the cellulose from the first-stge solid residues. The extent of the pretretment rections must be controlled nd optimized for ech feedstock to mximize the production of bioethnol. In conversion processes requiring enzymtic hydrolysis of the residul cellulose, the pretretment steps my increse the susceptibility of the remining cellulose to hydrolysis by cellulse enzymes resulting in incresed yields of sugrs or ethnol. First- nd second-stge pretretment of softwood feedstocks with dilute sulfuric cid is usully short process step involving only few minutes t high temperture in the rector. Controlling the rections requires rpid nd ccurte methods of nlysis for monitoring the rector. Conventionl methods such s wet chemicl nlysis nd high-performnce liquid chromtogrphy re time-consuming nd too slow to be useful for the control of the rector. The stndrd wet chemicl method for solids compositionl nlysis (1) my tke severl dys to complete. Although this method is quite ccurte, it is too slow to enble control of the short dilute-cid pretretment step. However, if rpid method to monitor nd control the pretretment rector is vilble, ethnol production cn be incresed. Fourier trnsform infrred (FTIR) spectroscopy is rpid nd quntittive, nd coupled with dvnced high-temperture, high-pressure ttenuted totl reflectnce (ATR) probes the cpbility of monitoring pretretment rectors is possible. FTIR spectroscopic nlysis is rpid nd nondestructive technique for the qulittive nd quntittive identifiction of components in solids in the mid-ir region (2). The usefulness of FTIR spectroscopy of solids hs incresed s smpling device technology hs dvnced. The FTIR nlysis of wet solid residues, such s pretreted whole slurries or wshed solids from those slurries, hs been severely restricted in the pst becuse of the high IR bckground bsorbnce of wter. Extremely short pthlength cells (not prcticl for nlyzing high-solids slurries) were needed to obtin trnsmission spectr (3). However, in the ATR mode (4 7), the high bckground bsorbnce cused by wter in pretreted slurries nd wshed solids cn be prtilly controlled by choosing ATR cells incorporting single or multiple reflections within the crystl. This llows ttenution of the incident rdition by the insoluble fiber to give the mid-ir spectr without the high wter bckground bsorbnce completely obscuring the spectr. Recent developments in ATR cells nd probes hve extended the ppliction of mid-ir spectroscopy to the qulittive nd quntittive nlysis of wet nd brsive solid residue (8).

3 FTIR Anlysis of Sugrs nd Lignins 53 Diffuse reflectnce IR Fourier trnsform spectroscopic methods or trnsmission spectroscopy utilizing KBr pellets previously hs been used to study whole wood, pretreted wood, wood surfces, lignin, nd pulp nd pper substrtes (9 17). However, the use of FTIR-ATR spectroscopy to study solids ws severely limited becuse of the soft ATR crystls (i.e., ZnSe, ZnS, nd KRS-5) then vilble. Recent vilbility of ATR cells utilizing very hrd crystls of silicon nd dimond hs overcome this difficulty. These crystls possess considerble hrdness nd chemicl inertness; thus, the smples cn be pressed onto the ATR crystl t high pressures, llowing more uniform penetrtion by the evnescent rdition (5) nd higher degree of reproducibility. The recent development of dimond-composite ATR cells hs extended FTIR spectroscopy into res of reserch not possible before becuse the dimond is highly chemiclly inert, exhibits high hrdness nd mechnicl strength, nd is opticlly trnsprent in the visible nd most of the mid-ir region. Becuse dimond surfce hs the lowest coefficient of friction of ll the vilble crystl mterils, few mterils stick to this surfce (8). Dimond-composite ATR probes re now vilble tht work t tempertures s high s 250 C nd t 1500 psi (10 MP). In the present study we used dimond-composite ATR cell nd n FTIR spectrometer to obtin mid-ir spectr of wshed pretreted softwood solid residues, nd wshed nd dried pretreted softwood solid residues. The prtil lest squres (PLS) option in the commercilly vilble softwre pckge TQ Anlyst (Nicolet Instruments, Mdison, WI) ws used to regress the spectr into methods cpble of predicting the glucn, mnnn, glctn, xyln, nd lignin compositions of pretreted softwood solid residues. Using clibrted method, FTIR-ATR cn rpidly determine (in few minutes) the solids composition of wshed residues from pretretment rector. Mterils nd Methods Preprtion nd Pretretment of Feedstock Whole-tree forest thinning ws obtined by the Pcific Wood Fuels Compny, Redding, CA, from site ner Quincy, CA, nd prepred s previously reported (18). Pretretment of this whole-tree forest thinnings feedstock for first- nd second-stge experiments ws performed using 4-L stem explosion rector (NREL Digester) s described erlier (19). Conditions rnged from 180 to 215 C, 0.35 to 2.5% (w/w) sulfuric cid, nd 120-s to 9-min residence times. Brrels of typicl swdusts from southestern Alsk were obtined from swmills of Ketchikn Pulp, by Selsk of Juneu, Alsk. The brrels of fresh swdusts were rpidly shipped to NREL to minimize degrdtion in trnsit. A mixture of 62% (dry wt bsis) hemlock, 26% Sitk spruce, nd 12% red cedr swdusts ws prepred by mixing four times on lrge trpulin using the method of coning nd qurtering. The composition of

4 54 Tucker et l. mixed swdusts ws chosen to represent the sttisticl species popultions of softwood trees in the forests of southestern Alsk. Ech brrel of swdust ws obtined from different swmill conducting cmpign utilizing n individul softwood species for prticulr forest product or customer. Smples of this swdust mixture were pretreted in the 4-L stem explosion rector. First-stge pretretment conditions vried from 180 to 190 C, 3 to 4 min, nd 0.7% (w/w) H 2 SO 4. Second-stge pretretment conditions vried from 205 to 215 C, 3- to 4-min residence times, nd 0.7 to 1% (w/w) H 2 SO 4. Pretreted slurry smples were extensively wshed by centrifugtion before obtining mid-ir spectr of the wet solids. Representtive smples (10 g) of the pretreted slurries were suspended in 40 ml of deionizedglss distilled wter nd centrifuged t 6000g for 5 min, nd the superntnts were discrded. The wshing procedure ws repeted minimum of five times. The percentge of moisture of the wshed smples ws obtined by stndrd oven-drying methods t 105 C. Wshed nd dried solids were prepred from the wet wshed solids by drying under vcuum t 45 C for 3 d. Representtive smples of the wet wshed nd dried wshed solid smples were ground using n gte mortr nd pestle before ppliction to the dimond surfce of the ATR cell. Grinding ws miniml (usully <30 s on 0.5-g smple) to prevent the solids from being degrded by the het of grinding. IR Spectroscopy Approximtely mg of the ground smple ws required to fill the smple well of the dimond-composite ATR smple cell. The smple cell ws equipped with spring-loded nvil to reproducibly press the solid smple uniformly nd tightly ginst the dimond surfce. Mid-IR spectr were obtined by verging 512 scns from 4000 to 400 cm 1, t 2-cm 1 resolution using Nicolet Avtr 360 FTIR spectrometer (Nicolet Instrument) nd six-reflection dimond-composite ASI DursmplIR ATR cell (ASI SensIR Technologies, Dnbury, CT). PLS Method PLS clibrtion methods for solid compositions were determined using the PLS regression nlysis option in TQ Anlyst (version 6.0) on 35 wet or 34 dried solid smple spectr. Wet chemicl results for percentge of solids, percentge of sugrs, nd Klson nd cid-soluble lignin for ech smple were entered into the softwre pckge spredsheet before clibrting the FTIR method. The softwre utomticlly clcultes the number of fctors to use in the regression nlysis by clculting the predicted residul error sum of squres (PRESS) before clibrting the method. The number of fctors suggested by the PRESS nlysis ws used in the PLS regression nlysis. The mid-ir regions selected for nlysis of the sugrs nd lignin were 3800 cm 1 to 2400 m 1 nd cm 1. The region between

5 FTIR Anlysis of Sugrs nd Lignins 55 Tble 1 Correltion Coefficients (r), r 2, nd SEE for 34 Wshed nd Vcuum-Dried Pretreted Whole-Tree Forest Thinnings nd Southest Alsk Softwood Residues Determined by PLS Regression Anlysis Component Correltion coefficient (r) r 2 SEE (wt%) Glucose Mnnose Glctose Xylose Arbinose Lignin SEE vlues less thn the 1.5 wt% vribility in wet chemicl solids compositionl nlysis nd 2200 cm 1 ws voided becuse of the very strong dimond IR bsorption. The regression ws performed on the entire set of 35 spectr, except one smple (chosen rndomly by the softwre) ws not used in the clibrtion. This smple ws used insted s vlidtion stndrd for the PLS method. In ddition, the regression nlysis ws cross-vlidted using the leve-one-out regression nlysis option in the softwre pckge. The leve-one-out cross-vlidtion regression generlly gve poorer correltion of predicted vs mesured vlues; however, the correltion coefficients were >0.9 (e.g., the correltion coefficients in Tble 1 for glucose decresed from to nd lignin decresed from to ). The lower correltion for the leve-one-out cross-vlidtion is the result of the limited number of smples for which wet chemicl nlyses re vilble. Results from the first PLS regression nlysis (not the leve-one-out nlysis) were used to plot the FTIR predicted wt% compositionl vlues vs the mesured solid compositionl nlyses. Results nd Discussion Figure 1 shows the PLS regression nlysis for glucose of 34 wshed nd 45 C vcuum-dried first- nd second-stge pretreted whole-tree forest thinnings nd southest Alsk softwood residues. Figure 2 shows the PLS regression nlysis for lignin of the sme 34 wshed nd 45 C vcuumdried smples. Tble 1 lists the correltion coefficients (r), r 2, nd root men squre SEE for 34 wshed nd vcuum-dried pretreted whole-tree forest thinnings nd southest Alsk softwood residues determined by PLS regression nlysis. The correltion coefficients for glctose, xylose, nd rbinose re low (<0.9) becuse the compositionl vlues tht were mesured for ech solid residue were low nd ner the detection limit for our FTIR-ATR instrument nd the wet chemicl methods. In ddition, little vrince in the clibrtion set is expected when the results re less thn the wet chemicl error of ~1.5% for solids compositionl nlysis, resulting in

6 56 Tucker et l. Fig. 1. PLS regression nlysis for glucose of 34 wshed nd 45 C vcuum-dried first- nd second-stge pretreted whole-tree forest thinnings nd southest Alsk softwood smples. Fig. 2. PLS regression nlysis for lignin of 34 wshed nd 45 C vcuum-dried firstnd second-stge pretreted whole-tree forest thinnings nd southest Alsk softwood smples. little correltion of predicted vlues to mesured vlues. The wet chemicl solid compositionl nlysis on the 34 dried pretreted smples rnged from 3.77 to 57.09% for glucose, 0.06 to 2.91% for mnnose, 0 to 2.87% for glctose, 0 to 7.87% for xylose, 0 to 1.94% for rbinose, nd to 95.04% for lignin. The number of fctors used in the PLS regression nlysis shown in Tble 1 is two used for glucose, nine for mnnose, nine for glctose, nine for xylose, one for rbinose, nd five for lignin. The SEE for glucose

7 FTIR Anlysis of Sugrs nd Lignins 57 Fig. 3. PLS regression nlysis for glucose from 35 wet wshed solid residues of pretreted softwood feedstocks. Fig. 4. PLS regression nlysis for lignin from 35 wet wshed solid residues of pretreted softwood feedstocks. reported in Tble 1 is >2 wt% becuse the wet chemicl compositionl nlysis for pretreted softwood residues pretreted under severe conditions is inccurte owing to the lrge mount of extrctives in the feedstock nd degrdtion products produced. Figure 3 shows the PLS regression nlysis for glucose of 35 wet wshed first- nd second-stge pretreted softwood smples (both wholetree forest thinnings nd southest Alsk swdust feedstocks). Figure 4 shows the PLS regression nlysis for lignin of the sme 35 wet wshed smples. Solids compositionl dt for ech smple in the clibrtion set were djusted for moisture before entering the dt into the PLS spredsheet.

8 58 Tucker et l. Tble 2 Correltion Coefficients (r), r 2, nd SEE for 35 Wshed (Wet) Pretreted Whole-Tree Forest Thinnings nd Southest Alsk Softwood Residues Determined by PLS Regression Anlysis of FTIR-ATR Clibrtion Set Component Correltion coefficient (r) r 2 SEE (wt%) Wter Glucose Mnnose Glctose Xylose Arbinose Lignin SEE vlues less thn the 1.5 wt% vribility in wet chemicl solids compositionl nlysis. The percentge of moisture ws determined in triplicte on ech smple by tking liquots of the wet solids used for collecting the FTIR spectr nd drying them in n oven overnight t 105 C. Tble 2 lists the correltion coefficients (r), r 2, nd root men squre stndrd error of the estimte (SEE) for compositionl nlysis of 35 wet wshed pretreted whole-tree forest thinnings nd southest Alsk softwood residues determined by PLS regression nlysis of the FTIR-ATR clibrtion set. The closer the vlue r 2 is to 1, the higher the probbility tht the FTIR predicted vlue (y-xis) is relted to the mesured solids compositionl vlue for tht component (x-xis). The correltion coefficients for glctose, xylose, nd rbinose reported in Tble 2 re <0.9 becuse the solid compositionl vlues mesured for ech solid residue re low nd ner the detection limit for this FTIR-ATR instrument nd wet chemicl solids compositionl nlysis. This is the result of hydrolysis of the hemicellulosic sugrs in the feedstock resulting in low vlues for these three sugrs in the residues. In ddition, dilution by wter in the wet residues further reduces the effective concentrtion of ll components in the residues. The low correltion coefficients (<0.9) for these three sugrs my be the result of little vribility in the wet chemicl solids compositionl nlysis (~1.5 wt%). The wet chemicl compositionl nlysis on the 35 pretreted smples (corrected for moisture) in Tble 2 rnged from 0.02 to 1.28% for mnnose, 0 to 1.07% for glctose, 0 to 2.94% for xylose, nd 0 to 0.89% for rbinose. The fctors clculted from the PRESS nlysis used in the PLS regression nlysis for the individul components in Tble 2 re s follows: one fctor ws used for wter, five for glucose, eight for mnnose, three for glctose, four for xylose, four for rbinose, nd three for lignin. Method vlidtion requires tht n nlyticl method hve precision nd ccurcy. An FTIR spectrometer is expected to hve frequency error of pprox 0.01 cm 1 (2). The intensity (bsorbnce) error using the Nicolet Avtr 360 FTIR-ATR instrument combintion ws determined to hve

9 FTIR Anlysis of Sugrs nd Lignins 59 stndrd devition of bsorbnce units (u) for eight replic spectr obtined from typicl dry pretreted solid residue for mjor bsorbnce bnd t 1032 cm 1. These eight replic spectr were obtined following collection of 512 scns ech. The smple ws not removed from the cell or disturbed in order to test the spectrophotometric nd softwre reproducibility. Ech replicte spectrum ws bseline corrected by drwing bseline from 1850 to 915 cm 1. The bsorbnce for the mjor bsorption bnd t 1032 cm 1 ws mesured using the pek height tool in the Omnic (version 5.2) softwre pckge. In ddition, spectr from three individul liquots of the sme smple were obtined to test the smpling error on pretreted biomss residues. The intensity error found ws u for the mjor bnd t 1032 cm 1 between the three spectr. If n obvious inhomogeneity in the ground smple is chosen (e.g., splinter tht ws not pretreted), n bsorbnce error of u or greter is found. This demonstrtes tht the smple must be ground to homogeneous mixture before plcing smple in the dimond cell. The root-men-squre of the noise of this FTIR-ATR combintion ws determined to be u in the region between 2800 nd 2600 cm 1 using stndrd softwre tool in the Omnic softwre. The six-reflection dimond-composite cell used in this study hs n pprent lower quntittion limit (20) of pprox 2 wt% for glucn with this FTIR-ATR spectrometer combintion if 10:1 signl-to-noise rtio is used. This limit is directly relted to the vrince in the wet chemicl method (~1.5 wt%) used for the clibrtion of the PLS regression method nd cn be improved only if more ccurte method of solid compositionl nlysis is developed. Models of these cells with higher number of reflection re commercilly vilble nd could extend the limits of component detection in the wet solids to lower levels, unless the bckground IR bsorption by wter in the wet wshed solids increses sufficiently to interfere with the nlysis. Conclusion In this study, we used FTIR-ATR to obtin mid-ir spectr on wshed solid residues from first- nd second-stge pretreted whole-tree forest thinnings nd softwood swdusts from southestern Alsk. Spectr from 35 wet wshed pretreted solid smples were subjected to PLS regression nlysis, s reported in Tble 2. The correltion coefficients for the regression nlyses for glucose nd lignin were found to be >0.9, with SEE of 1.4 nd 2.5 wt%. The correltion coefficients for glctose, xylose, nd rbinose were low becuse little vrince in the clibrtion set is expected when composition is less thn the wet chemicl error of ~1.5%, resulting in little correltion of predicted vlues to mesured vlues. The moisture contents of the wshed solid residues in this clibrtion set vried between 50 nd 80%. The lrge mount of moisture in the residues would hve mde obtining mid-ir spectr very difficult becuse of the high bckground

10 60 Tucker et l. bsorbnce cused by wter; however, the FTIR-ATR technique ws ble to overcome this difficulty. The method is cpble of predicting moisture content of unknown wet wshed pretreted softwood residues. The PLS regression nlyses for 34 wshed nd 45 C vcuum-dried solid residues from first- nd second-stge pretreted whole-tree forest thinnings nd softwood swdusts from southestern Alsk re reported in Tble 1. The correltion coefficients for the regression nlyses for glucose, mnnose, glctose, xylose, nd lignin were found to be >0.9, with root-men-squre SEE vrying between 2 nd 4.4 wt%. The moisture contents of the 45 C vcuum-dried solid residue smples in this clibrtion set were mesured t <1%. The wet wshed smples cn be generted rpidly by centrifugtion, wheres the 45 C vcuum-drying step tkes 3 d. Similr results re obtined if 105 C dried residues re used (dt not shown). FTIR spectroscopy cn be ccomplished in couple of minutes, nd solids composition cn be predicted rpidly once the method hs been clibrted. The FTIR-ATR technique llows mid-ir spectr to be rpidly obtined on whole slurries, wshed solids, nd dried wshed solids from pretreted biomss. The pretreted softwood slurries used in this study were obtined from very complex feedstock consisting of chipped whole trees, including limbs, brk, nd needles. PLS methods generted using higher temperture nd pressure probes in rector under process conditions should llow extension of this technology to rpidly monitor nd control pretretment rector to mximize product yield. Lrger dtbses of FTIR spectr nd solids compositionl nlysis will increse the ccurcy of predicting the composition of pretreted solids with this method. The results presented in this study for SEE re close to the 1.5 wt% errors expected for the stndrd wet chemicl method for solids compositionl nlysis. However, methods developed using PLS for one feedstock should be redeveloped when chnging to other feedstocks. Acknowledgments We wish to thnk Peter Huberth of Ketchikn Pulp, Alsk, for generously supplying the hemlock, Sitk spruce, nd red cedr swdust smples from vrious swmills in southestern Alsk. This work ws funded by the Office of Fuels Development, the US Deprtment of Energy. References 1. Moore, W. E. nd Johnson, D. B. (1967), Procedures for the Chemicl Anlysis of Wood nd Wood Products, US Forest Products Lbortory, US Deprtment of Agriculture, Mdison, WI. 2. Griffiths, R. P. (1975), Chemicl Infrred Fourier Trnsform Spectroscopy, John Wiley, New York. 3. Krishnn, K. nd Ferrro, J. R. (1982), in Fourier Trnsform Infrred Spectroscopy, vol. 3, Krishnn, K. nd Ferrrro, J. R., eds., Acdemic, New York, p Doyle, W. M. (1990), Appl. Spectrosc. 44, Hrrick, N. J. (1967), Internl Reflection Spectroscopy, John Wiley, New York.

11 FTIR Anlysis of Sugrs nd Lignins Kuehl, D. nd Crocombe, R. (1984), Appl. Spectrosc. 38(6), Fix, O. (1988), Mikrochim. Act 1(6), Milosevic, M., Sting, D., nd Rein, A. (1995), Spectroscopy 10(4), Schultz, T. P., Templeton, M. C., nd McGinnis, G. D. (1985), Anl. Chem. 57(14), Grndmison, J. L., Thibult, J., Kliguine, S., nd Chntl, P. D. (1987), Anl. Chem. 59(17), Fix, O. nd Bottcher, J. H. (1992), Holz Als Roh-Und Werkstoff 40(6), Pndey, K. K. (1999), J. Appl. Polymer Sci. 71, Zvrin, E., Jones, S. J., nd Cool, L. G. (1990), J. Wood Chem. Technol. 10(4), Fix, O. nd Bottcher, J. H. (1993), Holzforschung 47, Rodrigues, J., Fix, O., nd Pereir, H. (1998), Holzforschung 52, Heitz, M., Rubio, M., Wu, G., nd Khormi, J. (1995), Anles Quimic 91S, Schultz, T. P. nd Glsser, W. G. (1986), Holzforschung 40, Nguyen, Q. A., Tucker, M. P., Keller, F. A., nd Eddy, F. P. (2000), Appl. Biochem. Biotech , Nguyen, Q. A., Tucker, M. P., Boynton, B. L., Keller, F. A., nd Schell, D. J. (1998), Appl. Biochem. Biotechnol , Krull, I. nd Swrtz, M. (1997), LC:GC 15(6),

Chapter URL:

Chapter URL: This PDF is selection from n out-of-print volume from the Ntionl Bureu of Economic Reserch Volume Title: Price nd Quntity Trends in the Foreign Trde of the United Sttes Volume Author/Editor: Robert E.