AN AMERICAN CHEMICAL SOCIETY JOURNAL

VOLUME 1, NUMBER 2

MARCH/APRIL 1987

@Copyright 1987 by the American Chemical Society

Art i c 1es
Molecular Characterization of the Pyrolysis of Biomass. 1.
Fundamentals
Robert J. Evans and Thomas A. Milne*
Solar Energy Research Institute, Golden, Colorado 80401
Received-September 9, 1986. Revised Manuscript Received November 3,1986
The technique of molecular-beam, mass spectrometric (MBMS) sampling is applied to the elucidation of the molecular pathways in the fast pyrolysis of wood and its principal isolated constituents.
The goal is the optimization of high-value fuel products by thermal and catalytic means. The
positive-ion mass spectra shown are obtained from real-time, direct sampling of light gases, reactive
intermediates, and condensible vapors simultaneously. The cellulose, lignin, and hemicellulose (e.g.,
xylan) components of wood pyrolyze largely to monomer and monomer-related fragments and give
characteristic mass spectral signatures. Whole wood appears to behave as the s u m of its constituents,
with few if any vapor species derived from interaction of the main polymer constituents. An important
interaction, however, is the influence of mineral matter in the wood on the carbohydrate pyrolysis
pathways. Vapor phase cracking of the primary products proceeds through a stage of light hydrocarbons and oxygenates to the ultimate formation of aromatic tars and H2, CO, COz,and H20. These
steps are illustrated and discussed. Consistent with these observations, a relatively simple pyrolysis
reaction scheme is proposed.

Introduction
The pyrolysis of lignocellulosics has been studied for
decades, much of the earlier work being motivated by
concern for fire suppression and interest in chemicals from
~ 0 o d . lMore
~ recently, the desire to optimize processes
for conversion of biomass and wastes to heat and to gaseous and liquid fuels has led to many more studies of
pyrolysis, often using new techniques and instrumentation.
Interest continues in fire research and most recently, in
"nuclear winter" related smoke generation. Several current
comprehensive reviews of the pyrolysis of wood and its
major constituents have been published,&" and two spe(1) Roberta, A. F. Combust. Flame 1970, 14, 261.
(2) Beall, F. C.; Eickner, H.W. Res. Pap. FPL-For. Rod. Lab. (U.S.)
1970, FPL 130.
(3) Shafiideh, F. Adu. Carbohydr. Chem. 1968,23,419-474.
(4) Gooe,A. W. In Wood Chemistry, 2nd ed.;Wise, L. E., Jahn, E. C.,
Ede.;ACS Monograph 97; American Chemical Society Washington, DC,
1952; Vol. 2, p 826.
(5) Broido, A.; Kilzer, F. J. Fire Res. Abstr. Rev. 1963, 5, 157-161.
(6) W e , T. A. In A Survey of Biomass Gmification;Reed, T. B.; Ed.;
Noyea Data Corp.: Park Ridge, NJ, 1981; Vol. 2, Chapter 5.
(7) Soltee, E. J.; Elder, T. J. In Organic Chemicals from Eiomuss;
Goldatein, I. S., Ed.; CRC Boca Raton, FL; 1981 pp 64-99.

cialists meetings covered major aspects of pyrolysis12J3so

no attempt will be made in this paper to summarize the
whole field. (Excluded from this discussion are the complex, condensednphase, direct liquefaction processes that
combine pyrolysis with catalysts, high pressure, reactive
environments, and long residence times; see ref 14-17 for
reviews of this chemistry.)
(8) Shafizadeh, F. J. Aml. Appl. Pyrolysis 1982, 3, 283-305.
(9) Antal, M. J. Jr. Adu. Solar Energy 1982,1, 61-112.
(10) Antal, M. J. Jr. Adu. Solar Energy 1986,2, 175-255.
(11) Graham, R. G.; Bergougnou, M. A.; Overend, R. J. Anal. Appl
Pyrolysis, 1984, 6, 95-135.
(12) Diebold, J. P. Proceedings of Specialists Workshop on Fast Pyrolysis, SERI/CP-622-1096; Solar Energy Research Institute: Golden,
c o , 1980.
(13) Overend, R. P., Milne, T. A., Mudge, L. K., Eds. Fundamentals
of Thermochemical Biomass Conuersion; Elsevier Applied Science: New
York, 1985.
(14) Molton, P. M.; Demmitt, T. F. Eattelle Pac. Northwest Lab.,
[Rep.] BNWL 1977, BNWL-2297.
(15) Chomet, E.; Overend, R. P. Bioenergy 84 1985, 3, 276296.
(16) Chomet, E.; Overend, R. P. In Fundamentals of Thermochemical
Biomass Conuersion; Overend, R. P., Milne, T. A., Mudge, L. K., Eds.;
Elsevier Applied Science: New York, 1985; pp 967-1002.
(17) Moffatt, J. M.; Overend, R. P. Biomass 1986, 7, 99-123.

0887-0624/87/2501-0123$01.50/00 1987 American Chemical Society

Evans and Milne

124 Energy & Fuels, Vol. 1, No. 2, 1987

Many techniques have been used to study biomass pyrolysis under conditions chosen to yield both practical and
scientific insight into kinetics and product yields. A large
number of past studies have measured weight loss by TGA
and related techniques1*or have collected final gaseous,
liquid and solid products for analysis and to close mass
balances.lg Conditions have ranged from close simulation
of the particle size and environments of practical devices,
to the study of milligram, or even microgram, samples in
vacuum (as in analytical pyrolysis mass spectrometry20).
The behavior and properties of the solid residue have been
probed in a variety of ways.21
In only a few studies have the intermediate steps in gas
phase pyrolysis been o b ~ e r v e d . ~ *These
~ ~ - ~studies
~
concentrated mainly on the light-gas end products of primary
vapor cracking. The focus of the work described here is
on the complex, organic, vapor product slate in both primary and vapor phase pyrolysis of lignocellulosics, as observed by a unique, direct-sampling, mass spectrometric
technique.
The goal is to extract, in real time, samples of gases and
vapors directly from pyrolyzing systems at ambient pressure, rapidly quench these samples and, introduce them
into the ion source of a mass spectrometer (as a universal
detector), without allowing wall collisons or condensation
of vapors. What is sought are the molecular pathways the
biomass vapors follow, as a function of temperature, residence time, environment,and other variables of practical
interest. Since the emphasis of our work has been on the
optimum conditions for the production of high yields of
condensed oxygenates or hydrocarbon gases, the conditions
chosen have been those of moderate-to-fast heating (degrees per second rather than degrees per minute) where
char, water, and C02 yields are minimized.
This paper attempts to summarize the insight we have
gained in several years of study of the pyrolysis of wood
and its main components under fast pyrolysis conditions,
with references given to more detailed discussions in other
publications.

Experimental Section
The experimental approach to achieving the sampling goal
stated above is to couple extractive sampling (as a free-jet through
a small orifice) with rapid transition to molecular flow, collimation

of a molecular beam, and line-of-sight introduction of the molecular beam into the ion source of a mass spectrometer. The
apparatus to accomplish this is shown in Figure 1. Vapor from
a pyrolysis reactor flows over and into a sampling orifice. The
initial expansion is nearly adiabatic and is isentropic, with the
consequence that extreme collisional and internal energy state
cooling can occur.25 With a low enough pressure in the first
vacuum stage, and proper placement of the second slit (skimmer),
the supersonic flow enters the skimmer without shock formation.
Rapid quenching and formation of an intense molecular beam
(18) Antal, M. J., Jr.; Friedman, H. C.; Rogers, F. E. Combust. Sci.
Technol. 1980,21,141-52.
(19) Nunn, T. R.; Howard, J. B.; Longwell, J. P.; Peters, W. A. Ind.
Eng. Chem. Process Des. Dev. 1985,24, 836-844.
(20) Meuzelaar, H. L. C.; Haverkamp, J.; Hileman, F. D. Pyrolysis
Mass Spectrometry of Recent and Fossil BiomateriaIs; Compendium and
Atlas; Elsevier: Amsterdam; 1982.
(21) Sekiguchi, Y.; Shafizadeh, F. J. Appl. Polym. Sci. 1984, 29,

1267-1286.
(22) Ekstrom, C.;Rensfelt, E. In Proceedings of Specialists Workshop
on Fast Pyrolysis, SERI/CP-622-1096; Solar Energy Research Institute:
Golden, CO, 1980; p 303.
(23) Diebold, J. P. M.S. Thesis T-3007, Colorado School of Mines,
Golden, CO.
(24) Steinberg, M.; Fallon, P. T.; Sundaram, M. S. Proceedings of the
1985 Biomass Thermochemical Conversion Contractors Meeting, PNL-

SA-13571, CONF-8510167; NTIS: Springfield, VA, 1986; p 15.

(25) Milne, T. A.; Green, F. T. J. Chem. Phys. 1972,56, 3007.

Quadrupole
MS
Electron
beam

Stage 3
Pump
4

Molecular
beam

Skimmer
Secondary
pyrolysis
zone
Primary
pyrolysis
zone

Heater

Gas introduction

Electrical
connection

Figure 1. Schematic of pyrolysis vapor generator coupled to a

molecular-beam mass spectrometer sampling system.
are achieved. The molecular beam intercepts a low-energy electron
beam (15-22.5 eV) of a quadrupole mass spectrometer, yielding
a positive ion mass spectrum. The details of such sampling have
been discussed sin several papers and review^.^^^' The main
features for our sampling applications are as follows: collisions
are minimized as the gas cools so that even highly reactive and
condensible species can be preserved; rotational and vibrational
cooling of molecules occurs, leading to less ionization fragmentation in the mass spectra; light gases and heavy vapors are
sampled simultaneously and in real time; the mass spectrum
provides a fmgerprintthat gives a rapid, semiquantitative picture
of the complex vapor; sampling can occur from ambient atmosphere pressures and temperatures that closely simulate the environment of pyrolyzers, gasifiers, or combustors. The limitations
of this technique will be discussed as results are presented.
A number of reactor systems for pyrolyzing small samples of
biomass have been interfaced with the free-jet, molecular-beam
mass spectrometer (MBMS) sampling system. These are placed
in the desired proximity of the sampling orifice, as illustrated in
Figure 1for a two-section furnace source. In early work an open,
premixed, H2-02-He or -Ar flat-flame burner was used to create
a laminar flow of hot steam-He or -Ar into which the sample to
be pyrolyzed was placed. By varying the placement of the burner
and sample relative to the sampling orifice, one could quickly
survey both the primary species and the secondary reactions of
primary vapor species in the gas phase. Results from this system
have been p ~ b l i s h e d . ~ ~ * ~ ~ - ~ ~
(26) Milne, T. A.; Soltys, M . N. J. Anal. Appl. Pyrolysis 1983, 5,
93-110,111-131 (93-131).
(27) Milne, T. A.; Beachey, J. E. Combust. Sci. Technol. 1977, 16,
123-38,139-52.
(28)Knuth, E. L. Appl. Mech. Rev. 1964,17, 751.
(29) Anderson, J. B.; Andres, R. P.; Fenn, J. B. In Molecular Beams;
Ross, J., Ed.; Wiley: NY 1966;pp 275-317.
(30) Ashkenas, H.; Sherman, E. S. Rarefied Gas Dyn. 1966,2,84-105.
(31) Steams, C. A.; Kohl,F. J.; Fryburg, C. G.; Miller, R. A. NBS Spec.
Publ. (US.)1978,No. 561, 303.
(32) Milne, T. A.; Soltys, M. N. In Fundamentals of Thermochemical
Biomass Conversion; Overend, R. P., Milne, T. A., Mudge, L. K., Eds.;
Elsevier Applied Science: New York, 1985; pp 361-383.
(33) Evans, R. J.; Milne, T. A.; Soltys, M. N.; Schulten, H. R. J. Anal.
Appl. Pyrolysis 1984,6, 273-283.

Pyrolysis of Biomass

Energy & Fuels, Vol. 1, No. 2, 1987 125

olution from total ion current curves such as that shown in Figure

t
O

2.

/
/

Io

15

29

25

$0

3s

40

is

5'0

55

60

Time (s)

r
.-c
v)

a,

.-c

a,

>
.-c

-m

0,

LT

Time (s)
loor

43

6P

m/z

Figure 2. Example of data from a typical batch pyrolysis of a

small sample of pine wood (30 mg): (a) total ion count to time
(8); (b) selected ion profiles from extractives (m/z 226), lignin (m/z
180 and 138), hemicellulose ( m / z 114 and part of m / z 43), and
cellulose ( m / z 43); (c) average spectrum with background correction over the whole pyrolysis interval.
In order to have a better defined temperature history for the
secondary cracking and to operate at pyrolysis temperatures lower
than about 800 "C, the burner system has been replaced by single
or two-section resistance-heated flow reactors as shown in Figure
1. This simple system allows preheating of gas to the desired
temperature in the lower section, insertion of a sample quickly
into the convective and radiant heating environment in the middle
of the reactor, and control of the vapor temperature and residence
time in the upper section where secondary gas phase reactions
occur. Antal was perhaps the first to use separated zones to isolate
primary from secondary reactions? In work to date, we have only
used captive pieces (or samples in boats) as the source of pyrolysis
vapors, in order to clearly separate the primary pyrolysis region
from the secondary region by preventing the solid from entering
the secondary region as in entrained flow pyrolysis. To simulate
entrained flow pyrolyzers and to achieve the extraordinary heating
rates experienced by micrometer-sized powders injected into hot
gases (>lo3 "C/s), powder-laden gas flows could be sampled as
a function of distance from injection, as was done in pulverized-coal flames in past work.n For primary pyrolysis work, the
upper section of the reactor is removed so that the orifice can be
placed quite close to the pyrolyzing surface, where temperatures
are low due to the endothermic nature of fast pyrolysis and
secondary reactions are minimized.
No direct measurement of sample heating rate or sample
temperature was made. Only the temperature of the flowing gas
into which the sample was placed was measured. The effect of
surrounding gas temperatures, in the range 400-500 "C used t o
observe primary spectra, was negligible. Where heating rates are
quoted they are estimated from the rise-time to peak vapor ev(34) Evans, R.J.; Milne, T. A.; Soltys, M. N. J.Anal. Appl Pyrolysis
1986,9, 207-236.

Temperature profiles in the upper, vapor phase cracking region

were measured by using doubly shielded thermocouples to establish rough constancy to f10 OC. This was felt to be more than
adequate for the qualitative deductions reported here. Residence
times were simply calculated from bulk flow a t the temperatures
used and the tube cross section.
Cooling effects at the upper end of the reactor were minimized
by inserting a small-diameter extractor tube about 5 cm into the
reactor to minimize residence time in that region. For the secondary cracking experiments, both the lower and upper section
temperatures were set at the same value to increase the isothermality of the upper section. Past work with open-flame heat
sources has shown that the product slate emerging from samples
placed in hot gases is little changed over the surrounding gas
temperature range of 400-900 "C. This is attributed to the
endothermic nature of primary pyrolysis.
In addition to the pyrolysis of small inserted samples, small
conversion devices (gasifiers, combustors) can be directly interfaced with the sampling orifice. Larger devices, such as wood
stoves, would require the use of heated probes connecting the
sampling orifice to the region to be sampled, due to the geometric
constraints of the rather open sampling stages needed for undisturbed, free-jet expansion to molecular flow.
Data acquisition is continuous, through digitization of electron-multiplier signals from the arrival of positive ions and
programmed storage in an IBM PC computer. The data system
used was made by Teknivent= primarily for GC-MS applications,
but it is well suited for our purposes. We repetitively scan
(typically one 300 amu scan/s) during the evolution of a pyrolysis
wave from samples of 10-1000 mg inserted into the hot environment. The stored spectra can be manipulated to give average
spectra, subtracted spectra, or time evolution of different masses.
In addition, the individual ion data for groups of samples can be
processed by using available data analysis programs that allow
flexible graphical display as well as multivariate analysis of the
data.
Examples of the kinds of data that emerge in a batxh pyrolysis
experiment are given in Figure 2. The upper curve shows the total
mass spectral ion current, which profiles the duration of the
pyrolysis wave during a typical fast pyrolysis (>30 "C/s) experiment. The spectrum is the average of 50 consecutive 1-s scans
during the fast pyrolysis of the sample, with the background signal
due to electronic noise and ion source gases subtracted. The
middle curve shows single-ion monitoring of six selected masses
during pyrolysis.
A word of caution is necessary a t this point as the reader
interprets the data to be shown below. A number of variables
can be manipulated to alter the appearance of the positive ion
mass spectra. Foremost are the electron energy, which affects
fragmentation and relative parent-ion sensitivities, and the tuning
of the q ~ a d r u p o l e where
, ~ ~ tradeoffs between mass resolution,
intensity, and high and low mass emphasis are routinely made.
In addition, the free-jet sampling process causes a "massseparation" effect, which favors heavier molecules by approximately the first power of the molecular eight.^' Also the low
electron energy, used in most of our work to minimize parent-ion
fragmentation, leads to overemphasis of molecules with low
ionization energies (e.g., aromatics and olefins) vis-a-vis permanent
gases and saturated hydrocarbons (e.g., CHI, HsO).
Without tedious calibrations using standard additions of known
compounds under actual pyrolysis-free-jet conditions, the results
of most significance are the kinds of species observed and their
relative change with pyrolysis conditions, rather than their absolute values. Thus the approach taken here is complementary
to other more easily quantifiable techniques such as GC, GC-MS,
FTIR, condensation of products followed by chemical analysis,
etc. Its outstanding advantages are the rapidity of data acqui(35) Teknivent (1984) Model 1050 mass spectrometer data system,
Teknivent Corp., St. Louis, MO 63146.
(36) Extranuclear (1979) quadrupole mass filter Model 4-162-8, Extranuclear Laboratories, Inc., Pittsburgh, PA 15238.
Brewer, J.; Milne, T. A. J. Chem. Phys. 1964, 40,
(37) Greene, F.T.;
1488-1495.

126 Energy & Fuels, Vol. 1, No. 2, 1987

Evans and Milne

sition, the fact that all volatile species can be directly sampled
pyrolysis takes place and the varieties of biomass. Thus
from their reactive environment,and the nearly universal detection
various mixtures of primary and secondary organics are
capability of mass spectrometry. Because no prior separation is
obtained, depending on the thermal history and environmade, as in GC-MS,isomers can seldom be distinguished and
ment of the pyrolysis.
severe fragmentation can obscure parent molecules. FragmenMass spectra of primary pyrolysis products from a
tation can be used to distinguish compounds with the same
number
of samples of whole biomass and its major comnominal molecular ion. Ethylene, for example, has a fragment
ponents are shown in Figure 3 and a listing of inferred
ion at mf z 27, which can be used to determine the relative conchemical species is given in Table I. By primary spectra
tribution to mlz 28 of ethylene and CO. These disadvantages
we mean observed spectra that are substantially free of
are in part offset by the ability to observe 100% of pyrolysis vapors
that by most techniques are often only 30-70% chromatosecondary gas phase cracking products as inferred from
graphable when the condensed vapors are r e v a p o r i ~ e d . ~ ~ ~ ~ the secondary cracking studies discussed below. Cracking
Finally, mass spectral fingerprints,particularly if fragmenof primary products within the pyrolyzing particle can also
tation is minimized, are more intuitively interpretable than gas
occur during their escape to the surface. Over the dichromatograms, since the displayed axis is molecular weight and
mensions of particles typically pyrolyzed (1-6 mm narresolution is quantized and roughly constant. In general the m
rowest dimension) only small changes in relative mass
spectra, vis-a-vis gas chromatograms, are simplified by the
spectral peak heights occur (less than a factor of 1.5 for
overlapping of parent ions of isomers, but are complicated by
almost all peak ratios). Such changes have no effect on
virtue of fragment ions. Ultimately it is probably not affordable,
the qualitative discussion to follow. For large particles,
or desirable, to attempt to routinely quantify the hundreds of
of industrial interest, internal cracking would become imminor species known to be involved in pyrolysis. What will be
more important is the behavior of major products or functional
portant and should be quantified.
classes of compounds, which affect product properties and define
In our tabulation in Table I, we have restricted the list
mechanistic pathways. The rapidly emerging techniques of
to what we believe are primary pyrolysis products from
multivariate statistical analysis, such as factor analysis,41combined
wood and its separated components, as revealed by parent,
with the ready acquisition of pyrolysis fingerprintsmay be the
or prominent fragment ions from fast-pyrolysis mass
preferred strategy for exploratory research.
spectra such as shown in Figure 3, and from recent analyses
of primary products from collected pyrolysis oi1s.38939v40*45
Results and Discussion
Many confirming reports of these species exist in the literature7J4s46,47No species are observed from wood that
A number of reviewers of biomass pyrolysis have postulated more or less complex reaction schemes to account
appear to derive from a chemical interaction of the organic
for slow and fast pyrolysis and primary and secondary gas
components, though proportions may change due to matrix
phase pathways. These range from simple s c h e m e ~ ~ * ~ J ~ ~ -effects and inorganic interactions. This is in agreement
to complex, multipath s c h e m e ~ . ~ * ~ J ~
with Goos observation^.^ The references in Table I cite
It has been a major goal of our work, and of this paper,
independent, chemical confirmation of the compounds
to add molecular detail to the substances variously labeled
listed.
as volatiles, monomers, tar, anhydrosugars,
Almost all of the even-mass ions observed in our mass
primary tars, low molecular weight liquids,
spectra can be related to observed chemical species, though
oxygenates, transient oxygenated fragments,
ambiguities are inherent in our mass spectra due to fragsecondary tar, vapor-phase-derived tar, refractory
mentation, as the case of levoglucosantypifies.% A number
tars, refractory condensibles, secondary gases, and
of odd-mass ions (due to fragmentation since nitrogen is
pitch and the pathways for their formation. One can
virtually absent in wood) are prominent and are useful for
appreciate the need for a standardizations of nomenclature
pattern recognition correlations. Of special note are m/z
for pyrolysis products. The simple use of oil and tar
43 (CH30+),a common fragment from aldehydes, ketones,
is ambiguous for biomass-derived oxygenates, however, due
and other carbonyl compounds, and m/z 57 and 73, which
to the traditional use of those terms in the fossil literature.
are common fragment ions of carbohydrates like levoNature of Primary Products from Biomass Pyroglucosan.
lysis. The number of chemical species that have been
Through spectra in this paper are shown only to 300
reported from wood pyrolysis and wood distillation is
m u , we have carried out mass spectral scans to 1400 m u .
enormous. As an example Goos4lists 231 compounds that
With the quadrupole adjusted to greatly over represent
have been found in the liquid products from the dehigh masses (as judged by the standard fragmentation
structive distillation of wood. Soltes and Elder in a recent
pattern of perfluorotributylamine) levoglucosan oligomers
review, tabulate about 230 compounds identified in
up to the octamer and lignin polymers beyond the trimer
pyroligneousacid and tar products from softwood and
were seen from cellulose and lignin, respectively. On the
hardwood pyrolysis. Part of the reason for these large
basis of calibrations at lower mass and comparisons of our
numbers is the variety of conditions under which wood
spectra for levoglucosan and lignin with field ionization
r e ~ u l t s ,the
~ ~ higher
, ~ ~ molecular weight species are estimated to constitute less than a few percent of the vapor.
(38) Elliott, D. C. Final Report on IEA Co-Operative Project D1,
The main processes leading to the primary pyrolysis
Biomass Liquefaction Test Facility Project; Pacific Northwest Laboraspectra for carbohydrates and lignin have been extensively
toires: Richland, WA, 1983; Vol. 4.
(39) Beaumont, 0. Wood Fiber Sci. 1985,17, 228-239.
discussed in two specialized report^^^.^ as has the special
(40) Menard, H.; Belanger, D.; Chauvette, G.; Gaboury, A,; Khorami,
A synopsis of these mechanisms
problem of levogluco~an.~~
J.; Grise, N.; Martel, A.; Potvin, E.; Roy, C.; Langlois, R. Fifth Canadian
follows.
Bioenergy R&D Seminar; Hasnain, S., Ed.; Elsevier Applied Science:
New York, 1985; pp 418-439.
Primary Pyrolysis Pathways. There have been two
(41) Wjndig, W.; Meuzelaar, H. L. C. Anal. Chem. 1984,56,2297-2303.
main theories of cellulose pyrolysis: free radical and
(42) Kilzer, F. J.; Broido, A. Pyrodynamics 1966, 2, 151-163.
heterolytic depolymerization. Both of these theories have
(43) Bradbury, A. G. W.; Sakai, Y.; Shafizadeh, F. J. Appl. Polym. Sci.
1979,23, 3271-3280.
(44) Hajaligol, M. R.; Howard, J. B.; Longwell, J. P.; Peters, W. J. Ind.
Eng. Chem. Process Des. Deu. 1982,21, 457-465.
(45) Piskorz, J.; Radlein, D.; Scott; D. S. J.Anal. Appl. Pyrolysis 1986,
9, 121-137.
(46) Fullerton, T. J., Franich, R. A. Holtforschung 1983,37,267-269.

(47) Obst, J. R. J. Wood Chem. Technol. 1983, 3, 377.

(48) Evans, R. J. Direct Mass Spectrometric Studies of the Primary
Pyrolysis of Carbohydrates, SERI/TR-234-3061; Solar Energy Research
Institute: Golden, CO, 1986.

Pyrolysis of Biomass
100

Energy & Fuels, Vol. 1, No. 2, 1987 127

60

43

Cellulose

60

Birch
c

looI

(1

Cellulose + 0.1% K and K & 0 3

Douglas Fir

43

l o r

6.73

85

1 9,8

lZ6137

>

c
.-

.-c

v)

c
5

v)

0)

.-C

C
._
100

43

60

Bagasse

180

Sweet gum
Milled wood
Lignin

0
0

m/z
m/z
Figure 3. Characteristic mass spectral patterns of primary pyrolysis products for several whole biomass samples and for separated
constituents of biomass. Approximately 30 mg of powdered sample was suspended in a steel boat (0.05 mm thickness) in flowing 550
O C helium, giving an estimated heating rate of 30-50 OC/s. The particle size ranged from 50 to 250 pm. The gas phase residence time
before free-jet sampling was approximately 75 ms. The actual solid temperature is unknown and less than that of the surrounding
550 "C helium due to the endothermic nature of the pyrolysis. Product evolution occurred over 15-25 s depending on the sample.

some experimental support, but the evidence is inconclusive. Discussions of this debate can be found in recent
articles by AntallO and Evans48and are only briefly summarized here. A free-radical mechanism has been proposed
by Kislitzyn et al.49on the basis of experiments with dinaphthylphenylenediamine as a free-radical scavenger.
When cellulose was pyrolyzed with 5 mol % of this material, the levoglucosan yield was reduced from 30 to 5 wt
%. Shdizadeha has proposed a heterolytic mechanism on
the basis of the work with phenylglucosides where the
nature of the substituents on the phenyl ring dramatically
affected transglycosylation. The influence of electron
withdrawal or donation led to the hypothesis that the
reaction is not a free-radical process but rather involves
heterolytic cleavage of existing glycosidic acid bonds via
a transglycosylation process involving nucleophilic displacement of the glycosidic groups by one of the free hy(49) Kislitayn, A. N.; Rodionova, Z. M.; Savinylch, V. I.; Guseva, A. V.
Zh. Prikl. Khim. (Leningrad) 1971,44, 2518-2524.

droxyl groups. The results of the molecular-beam mass

spectrometry studies have been reported in depth48and
favor a concerted displacement reaction that is similar to
Shdizadeh's proposed mechanism; however, unambiguous
proof of intermediates is still lacking.
The primary pyrolysis spectrum of pure cellulose is
shown in Figure 3. The major peaks a t m / z 162 and 144
(largely an ionization fragment ion) are due to levoglucosan
(XXIX; Chart I), which can be obtained from pure cellulose in yields of up to 60%.s Work with chemical ionizat i ~ has
n ~shown
~ that a major part of mlz 57,60,73, and
98 ions are also E1 fragment ions from levoglucosan. The
dominance of this one product is due to the structural
homogeneity of cellulose and the possibility of its undergoing intramolecular condensation in the sequential depolymerization of the glycosidic units (Scheme I). This
mechanism allows complete devolatilization, with char
yields of less than 1% observed in fast pyrolysis in pure
cellulose. The importance of cellulose structural features
such as crystallinity and degree of polymerization (DP) has

Evans and Milne

128 Energy & Fuels, Vol. 1, No. 2, 1987

Table I. Major Ions in Mass Spectra of Wood and Compounds Reported in Primary Oils and Gases

ion,
m/z
16

18
28

32
44
46
58
60

68
72
74

76
82

84
86
88

94
96
98
100

102
108
110
112

114
116
120
122

124
126
132
138
146
150
152
154
162
164
166
168
178
180

formula

structure

likely product

methane
water
carbon monoxide
co
ethene
CH4H2
methanol
MeOH
carbon dioxide
OCO
acetaldehyde
OCHMe
formic acid
HCOOH
2-propanone (acetone)
MeCOMe
acetic acid
H02CMe
hydroxyacetaldehyde (glycoaldehyde)
HOCHzCHO
methyl formate
MeOCHO
furan
I
2-butanone
EtCOMe
acrylic acid
CH2-CHCOOH
1-hydroxy-2-propanone (acetol)
HOCH2COMe
propanoic acid
HO2CEt
hydroxypropanal
HOCHzCHzCHO
methyl acetate
CH3COOCH3
glycolic acid
HOCH2COOH
2-methylfuran
I1
cyclopentenone
I11
cyclopentanone
IV
2(5H)-furanone
V
2,3-butanedione
MeCOCOMe
crotonic acid
MeCH=CHCOOH
butyrolactone
VI
butanoic acid
H02CEtMe
1-hydroxy-2-butanone
MeCh2COCH20H
phenol
PhOH
dimethylcyclopentene
VI1
furfural
VI11
2-methyl-2-cyclopentene-1-one
IX
furfuryl alcohol
X
5-methyl-2(3H)-furanone (a-angelicalactone)
XI
3-methyl-2(3H)-furanone
XI1
valerolactone
XI11
2,3-pentanedione
MeCOCOEt
pentanoic acid (valeric/isovaleric)
CH,(CH*)&OOH
o-cresol
CH~C~HIOH
m/p-cresol
5-methylfurfural
XIV
dihydroxybenzenes (catechol, hydroquinone, resorcinol)
2-hydroxy-3-methyl-2-cyclopentene-l-one
dimethylcyclopentanone
3-hydroxy-2-penteno-1,5-lactone
hexanoic acid
1-acetyloxy-2-propanone
vinylphenol
dimethylphenols
ethylphenol
benzoic acid
2-methoxyphenol (guaiacol)
trimethylcyclopentenone
5-(hydroxymethyl)-2-furfural
2-methyl-3-hydroxy-4-pyrone
1-hydroxy-2-propanone acetate
4-methylguaiacol
XXII
MeCOO(CH2)2(CO)CHzOH 1-hydroxy-2-butanone acetate
p-vinylguaiacol
XXIV
coumaryl alcohol
xxv
4-ethylguaiacol
XXVI
vanillin
XXVII
2,6-dimethoxyphenol (syringol)
XXVIII
levoglucosan
XXIX
isoeugenenol
xxx
eugenol
1-(4-hydroxy-3-methoxyphenyl)ethanone
XXXI
4-methy1-2,~i-dimethoxyphenol
XXXII
coniferyl aldehyde
XXXIII
coniferyl alcohol
XXXIV
vinylsyringol
xxxv
XLV
a-D-glucose

MeH
HOH

ref to
chem
anal. of
oils

major
ionh

in
product
at >0.1%'

X
X

X
X
X

f
f

X
X
X
X

X
X
X
X
X
X

a, b, d , e

X
X
X
X
X
X
X
X
X
X
X
X
X
X
X

a, b, e

e
e
c-e
c, e
e

X
X
X
X

d, e
e
c-e
a, d

a
c , d, f
c, d

X
X

a
c, e, f
d, e
d, e
a, e

b, c, e

c, d
a

g
X

c, d

X
X
X

X
X
X
X

a-e
a

g
d
e
a, b, e

e
i?

g
X

g
X
X

a, b, d , e

c , d, f
g

X
X
X

X
X
X

a-d

X
X

X
X
X
X
X
X

c-e
a, b

a, b

e
a-f

e
b

X
X

a-f
a-f
a
g
a, c-e

X
X
X
X
X
X

b, e
b-e
e

X
X

a-f

d
a
a, e
a

g
X

g
d

Pyrolysis of Biomass

Energy & Fuels, Vol. 1, No. 2, 1987 129

Table I (Continued)

ion,
mlz

182
194

formula
C9H1004

cl&140s
C11H1403
C10H1004

196
208

210
272

c181204
C11H1204
C11H1404

C16H16O4

structure
XXXVI
XXXVII
XKXVIII
XXXIX
XL
xL1
XLII
XLIII

major
ion"

likely product
syringaldehyde
4-ethylsyringol
4-propenylsyringol
ferulic acid
(4-hydroxy-3,5-dimethoxyphenyl)ethanone
sinapyl aldehyde
sinapyl alcohol

ref to
chem
anal. of
oils

in
product
at >0.1%'

a-d
a
a

X
X

X
X

k-

'Reference 38. GC/MS and GC of a vacuum pyrolysis oil from the University of Sherbrooke (poplar). *Reference 38. A flash pyrolysis oil from
the University of Waterloo fluidized bed (poplar). 'Reference 40. A variety of analytical techniques applied to a University of Sherbrooke vacuum
pyrolysis oil (poplar). dReference 40. A flash pyrolysis oil from the University of Waterloo (poplar). eReference 39. GC/MS of oils from a 1-atm
pyrolysis of powdered beechwood. /Reference 45. GC analyses of hybrid poplar, pyrolyzed at 504 O C in a fluidized bed. #Identified in other studies.
"Under arbitrary tuning of mass spectrometer, as used in Figure 2. 'As percent of water-free, organic oil for ref a-d and percent of dry wood for ref
e and f.

Scheme I

Scheme I1

R-

R-

R = CELLULOSE CHAIN
HOC H,

RR-

k:

OH

0
11

HOCH&H

c=c

,CH -CH
11

R-0'

%H

"9
R-

@OH

/
been discussed in past work,1 but in general these effects
are not large32in fast pyrolysis compared to the effect of
inorganic impurities, as discussed below.
Several workers have shown, by kinetic analysis of
thermal gravimetry and by differential scanning calorimetry experiments, the existence of a low DP stage42or
"active
stage, which corresponds to a decrease
in DP from 100&2000 to 200 at temperatures between 220
and 250 "C. Thermogravimetric studies (TGA) have
shown that cellulose undergoes a rapid weight loss beginning around 280 "C with an inflection point around 350
0C.8 At this temperature cellulose undergoes transglycosylation to form levoglucosan in high yield~.~7&
There
are several possible transformations that could contribute
to the rate-limiting step: the decrease in DP, the breaking
of inter- and intramolecular hydrogen bonding to form a
less rigid structure (i.e., a glass transition zone), the formation of a reactive intermediate by the cleavage of the
glycosidic linkage (i.e., a free radical or ionic species), or
the interconversion from the chair to the twist conformation so that the carbon in the 6-position will be in the
axial position as it is in levoglucosan.50 It is possible that
(50) Cemy, M.;Stanek, J., Jr. A d a Carbohydr. Chem. Biochem. 1977,
34, 23-177.

CROSS-LINKING

CARBONYL
COMPOUNDS

BETWEEN CHAINS

more than one of these transitions are important in the

rate limiting step and that they are all enhanced at higher
temperatures.
In the pyrolysis of most biomass materials, however, the
levoglucosan yield is low despite cellulose concentrations
of approximately 50% .a The treatment of cellulose with
small amounts of alkali material can drastically change the
product slate as shown in Figure 3. The addition of 0.1
w t ?& potassium or sodium as the carbonate or hydroxide
inhibits the formation of levoglucosan and leads to a different product state composed of furfural derivatives (mlz
126, 110,96), low molecular weight carbonyl compounds
represented by mlz 43, a common fragment ion of carbonyl
compounds such as acetaldehyde and acetol, high molecular weight condensed species (in relatively low abundance), and an increase in char yield from 1to 10%. Our
work over the past several years has led to the speculative
mechanism shown in Scheme 11for the alkali-metal-catalyzed pathway. With the disruption of intramolecular
condensation (transglycosylation), glycosidic rupture becomes more important with dehydration to form carbonyl
groups, double bonds, and substituted furans. Piskorz et

Evans and Milne

130 Energy & Fuels, Vol. 1, No. 2, 1987

Chart I

b
II

IV

111

VI1

0
& C H .

pn

\ I
Vlll

n3cfY

IX

XIV

XI1

XI

q qn2
0 hMe

@a;

QOMe

xv

XVll

XVI

XVlll

cn=cn,

"OH"QH

XIX

cn=cH-cn,on

XXI

HrCH1

Me@Me

@Me
no

Me

no

XXll

&

QOMe

me

xxx

XXlX

xxv

XXlV

XXlll
cn=cn-cn,

XXXl

XXVlll

XXVl
0
II

cn=cn-cH

J+
Me

Me

@Me
no

Me

XXXll

xxxv

XXXIV

XXXlll

Me

no

HO

0
II

Me0

cn=cn-cH,
Me

M e 0QMe

no

XXXVI

XXXVlll

II

cn=cn-coon

00.
no

XXXIX

cn=w-cn

Me

XL

M e 0@Me
no

XLI

CH=CH-CHPH

no

XLll

en &

noc n

w&n,

ns

XLlll

-%

XLlV

XLV

XLVl

al.45proposed that the monomer unit of cellulose decomposes to two-carbon and four-carbon fragments with the
two-carbon fragment rearranging to give high yields of
hydroxyacetaldehyde (up to 17% yield). Hydroxyacetaldehyde may be the major product of the glycosidic
rupture pathway accounting for m/z 60 and fragment ions
a t m / z 31 and 32 in the spectra. Piskorz et al. did not
address the effect of alkali-metal catalysis in their work.
The biomass samples shown in Figure 3 show the distribution of the two major carbohydrate mechanisms. The
distribution of the carbohydrate peaks in these whole
biomass samples shows that the levoglucosan yield is low,
as judged by the relative amounts of m/z 43 and 126 from
the glycosidic rupture mechanism and m / z 60 and 144
from the transglycosylation mechanism. Only the douglas
fir has a major peak a t m / z 144, showing the relatively
minor importance of levoglucosan in the other biomass
samples relative to ash-free cellulose. However, even with
yields of levoglucosan less than 5%, it is still one of the
major single pyrolysis products.

These two major mechanisms for cellulose are also of

importance for the other carbohydrates in the hemicellulose fraction of biomass. The pyrolysis of larch wood
xylan is shown in Figure 3. Xylan is a major hemicellulose
constituent in hardwoods that is a copolymer with glucuronic acid and is commonly acetylated in the 0-2position.
In separating the xylan from wood, the acetyl group is
cleaved and as a result the pyrolysis spectrum for xylan
shows little acetic acid (mlz 60). The distribution of
products is similar to cellulose with a series analogous to
the m/z 162,144,and 126 series in cellulose present at m / z
132,114,and 96-a difference of 30 m u due to the missing
hydroxy methyl group in xylan. This shows that transglycosylation can occur without the primary hydroxyl
group on the number 6 carbon, and one possible structure
for mlz 132 would be 1,4-anhydroxylanopyranose.Previous
has shown that the m / z 144 abundance for
pure cellulose pyrolysis is largely due to ionization fragmentation and the species at mlz 114 may also be in part
due to the ionization fragmentation of the mlz 132 species.

Pyrolysis of Biomass
However, Ohnishi et aLsl in a study of xylan pyrolysis,
found a product a t m / z 114 and identified it as 3hydroxy-2-penteno-l,5-lactone
(XVII). This compound
was also identified by Van der Kaaden et a1.52from the
pyrolysis of amylose. Schulten et
in field ionization
mass spectrometry (FIMS) studies of xylan pyrolysis in
vacuum, also found high abundances of m/z 132,114,and
96. The m / z 132 peak was about equal in height to the
m/z 114 peak in FIMS studies, indicating that significant
pyrolysis products a t m / z 114 are probably present since
field ionization minimizes ionization fragmentation. In the
pyrolysis of most biomass samples, m / z 114 is a significant
peak, especially in grass species and hardwoods. The
bagasse and birch pyrolysis spectra in Figure 3 show that
m/z 114 is among the major peaks.
The primary pyrolysis of lignins separated by a variety
of methods has been previously reported.34 Among the
common methods of separating lignin from wood, only
milled-wood lignin (MWL) shows pyrolysis behavior that
resembles the pyrolysis of lignin in whole biomass. Other
methods of separating lignin, such as steam explosion,
kraft, or organic solvent extraction, change the nature of
the lignin and affect the pyrolysis product distribution.
Even ball milling causes notable changes in the pyrolysis
product distribution. In Figure 3, sweet gum MWL and
whole sweet gum pyrolysis spectra are shown and there is
a greater relative abundance of m / z 180 and 210 relative
to the other lignin products (e.g., mlz 194) in the whole
wood than in the milled-wood lignin spectrum.
As noted for the carbohydrates, the lignin fraction of
biomass also undergoes primary pyrolysis by structurally
controlled depolymerization. The sweet gum MWL in
Figure 3 shows the predominance of the precursor monomers of hardwood lignin a t m / z 180 (coniferyl alcohol
(XXXIX)) and 210 (sinapyl alcohol (XLII)). These
products not only form in high abundance, but also form
early. The evolution profile of several major products from
pine are shown in Figure 2. Coniferyl alcohol is the earliest
pyrolysis product to form (the species a t m/z 226 is an
extractive that is volatilized). The next product to form
is derived from hemicellulose, m/z 114. This is followed
by the cellulose derived species a t mlz 43. The last
products to form are derived from lignin again, shown in
Figure 2 by m / z 138 (methylguaiacol). Actually, lignin
peaks are sequentially evolved over the time needed for
complete devolatilization of the particle. The formation
of the double bond in the alkyl side chain allows devolatilization with minimum need for transferable hydrogen.
The other early predominant products also show the formation of a double bond in conjugation with the aromatic
ring: isoeugenol (mlz 164),vanillin (mlz 152), and vinylguaiacol (mlz 150). The higher molecular weight series
of peaks also show this mechanism and a possible structure
for m/z 272 is an enol ether dimer34(XLIII). The whole
sweet gum wood sample in Figure 3 shows that the generation of the precursor monomers is even more predominant in whole biomass than in the separated milled wood
lignin. The other class of lignin primary products are the
lower molecular weight primary products, such as methylguaiacol ( m / z 138),guaiacol (mlz 124),and catechol
( m / z 110). These products tend to reach their maximum
evolution rate a t the end of pyrolysis as shown by m/z 138
in Figure 2. This is consistent with the observation in
(51) Ohnishi, A.; Kato, K.; Takage, E. Carbohydr. Res. 1977,58,387.
(52) Van der Kaaden, A.; Haverkamp, J.; Boon, J. J.; de Leeuw, J. J .
Anal. Appl. Pyrolysis 1983, 5 , 199-220.
(53) Schulten, H.-R.; Bahr, U.; Gortz, W. J . Anal. Appl. Pyrolysis
1981/1982,3, 229-241.

Energy & Fuels, Vol. 1, No. 2, 1987 131

thermogravimetric analysis that cellulose pyrolyzes over

a relatively narrow temperature range while lignin pyrolyzes over a much wider range.
The lignin from grass species has three monomers:
coniferyl, sinapyl, and coumaryl alcohols. Coumaryl alcohol (m/z 150),which has no methoxy groups, is probably
reponsible for mlz 150 in the bagasse sample in Figure 3.
Vinylphenol (mlz 120) is a major low molecular weight
lignin peak from grasses, which not only forms in high
abundance from virtually every grass species but, unlike
other low molecular weight lignin products, also forms early
with an evolution profile similar to the lignin precursor
monomers. One explanation for the late evolution of most
low molecular weight lignin products is that during the
early evolution of the alkenyl aromatic species (i.e., coniferyl alcohol), the bulk of the lignin is becoming more
deficient in hydrogen. A portion of the bonds that are
cleaved in the hydrogen-deficient matrix are likely to undergo condensation reactions, leading to a more refractory
solid as pyrolysis progresses that can only undergo further
devolatilization by hydrogen transfer to form stable aromatic species such as guaiacol. This means that lignin
pyrolysis covers a wide range from the most labile to the
most refractory species. At the latter stage of product
evolution the methoxy groups are cleaved and there is the
simultaneous evolution of dihydroxybenzene and methane.54 This is the only source of methane from the primary pyrolysis of biomass observed in this work. (The
detection conditions used in the experiments reported here
do not allow observation of methane in these spectra due
to mass discrimination and low ionization sensitivity.)
In summary, primary pyrolysis vapors are rather low in
molecular weight, representing monomers and fragments
of monomers of the biopolymers of biomass, no chemical
interactions are observed among the organic components
of biomass, mineral matter is suspected of altering the
carbohydrate pathways substantially in wood in contrast
to pure cellulose, and the patterns for the separate components of biomass are rather characteristic.
Nature of Vapor Phase Cracking Products from
Biomass Pyrolysis. When cracking of primary vapors
is observed by increasing the temperature of the upper
column of the reactor shown in Figure 1 beyond 500 "C,
systematic changes are seen in the behavior of the primary
pyrolysis products of wood and its components at residence
times of the order of tenths of seconds. Examples of
spectra observed for cellulose and lignin are shown in
Figure 4 and for xylan (a representative hemicellulose) and
wood in Figure 5. Tables I1 and I11 show prominent
secondary cracking peaks that grow a t the expense of
primaries. Many of these have been seen in liquids subjected to high cracking severity (see below ands5). One
can distinguish a "primary oxygenates" zone from 400 "C
(same spectrum as 500 "C) to about 700 O C , a "hydrocarbon* or secondary zone from 700 to 850 "C and a largely
"aromatic" or tertiary zone above 850-1000 "C. A full
unraveling of this complex chemistry will require the study
of a number of pure primary pyrolysis products, since even
a single compound such as coniferyl alcohol gives a dozen
or so major secondary products.56
(54) Evans, R. J.; Soltys, M. N.; Milne, T. A. Annual Report, Oct 1,
1983 to Dec 30, 1984, SERI/PR-234-2701; Solar Energy Research Institute, Golden, CO.
(55) Elliott, D. C. In Proceedings of the 2985 Biomass Thermochemical Conuersion Contractors' Meeting, PNL-SA-13571,CONF-8510167;
NTIS: Springfield, VA, 1986; p 361-382.
(56) Evans, R. J.; Milne, T. A. Proceedings of the 1986 Biomass
Thermochemical Conversion Contractors' Meeting, PNL-SA-13571,
CONF-8510167;NTIS: Springfield, VA, 1986; p 57-79.

Evans and Milne

132 Energy &Fuels, Vol. 1, No. 2, 1987

100

28

78

c
.v)
C

Cellulose 95O0c

Sweet gum lignin 9 5 0 " ~

Cellulose 7OOOc

Sweet gum lignin 7OOOc

Q)

.-C

181

I Sweet gum lignin

Cellulose 500" c

500"c
151

loo[

m/z

137

m/z

Figure 4. Vapor phase cracking of the primary products derived from cellulose and milled-wood lignin from sweetgum wood at 500,
700, and 950 O C and residence times of about 750 ms. The primary pyrolysis was like that used in Figure 3 except the gas temperature
of the lower reactor was increased to match the secondary zone. This change in primary pyrolysis surrounding gas temperature had
little effect on the composition of the primary products; see Experimental Section for details.
Table 11. Prominent Intermediate Ion Peaks in the
Secondary Cracking Regime of Wood
ion, m / z
formula
Dossible compd
methane
16
CH4
ethene
28
CZH,
ethane
30
CZH6
40
C3H4
ProPYne
propene
42
C3H6
carbon dioxide
44
COZ
butadienes, butyne
54
C4H6
butenes
56
C4H8
cyclopentadiene
66
C5H6
toluene
C7H8
92
phenol
94
C6H60
styrene
104
C8H8
xylene
106
C8HIO
cresol
108
C7H8O
dihydroxybenzene
110
C6H602
benzofuran(?)
118
C8H60
vinylphenol
120
C8H80
benzodioxole (?)
122
C7H60Z
dihydroxytoluene, guaiacol
124
C&Oz
130
methylbenzofuran (?)
132
C9H80
methylbenzodioxole (?)
136
CBH8OZ
142
144
vinylbenzodioxole (?)
148
CSH8OZ
160
168
182

The s p e c t r a at 500 "C for the four materials shown in

Figure 4 and 5 are dominated b y peaks that are all due to
primary products; however, there is some decrease i n the

Table 111. Prominent Growth Ion Peaks in the Tertiary

Cracking Regime of Wood
ion, m / t formula
possible compd
16
CH,
methane
18
water
26
acetylene
28
carbon monoxide
44
carbon dioxide
52
vinylacetylene (?)
cyclopentadiene
66
78
benzene
to1uene
92
94
phenol
104
styrene
116
indene
128
naphthalene
152
acenaphthylene
166
fluorene, benzindenes
178
anthracene, phenanthrene
202
pyrene, fluoranthene, benzacenaphthylene
216
methylpyrene
226
benzo[ghi]fluoranthene
228
chrysene, benz[a]anthracene, triphenylene,
benzo[clphenanthrene
240
methylbenzo[ghi]fluoranthene
252
benzofluoranthenes, benzo[ae]pyrene,
perylene
276
anthanthrene et al.
relative abundance of the heavier species. No major
change i n composition occurs until approximately 550 "C
where the first species to crack are the alkenyl aromatics,
such as coniferyl alcohol. A significant decrease i n mlz
180 and 210 is already observed at 500 "C for sweet g u m
MWL i n Figure 4.

Pyrolysis of Biomass

Energy & Fuels, Vol. 1, No. 2, 1987 133

100

Xylan 950"

Pine 950"

128

'

is','

292

178

152

ins

iA

'

2A

'

2;s

" 2 i 0,'

Xylan 700"

c
.-

ln

"'3b

e,

Pine 700'

c
.-

ln
C

e,

.-c

>
.-c
m
-

.-c
e,
>
.c
m
-

e,

e,

'"r L

137

Xylan 500"

Pine 500'

272

m/z

m/z
Figure 5. Vapor phase cracking of the primary products derived from xylan and pine wood at 500, 700, and 950 "C and residence
times of about 750 ms. The primary pyrolysis was like that used in Figure 3 except the gas temperature of the lower reactor was increased
to match the secondary zone. This change in primary pyrolysis surrounding gas temperature had little effect on the composition of

the primary products; see Experimental Section for details.

Cellulose products are essentially unchanged a t the

lowest temperature when that spectrum is compared to the
primary spectrum in Figure 4. Levoglucosan is known to
be thermally labile so the lack of change in product distribution between the primary slate (750-ms gas phase
residence time in Figure 4) and the 750-ms slate at 500 "C
(in Figure 5 ) is surprising. Shafizadeh and Lai5' studied
the thermal degradation of levoglucosan at 600 "C but with
an extreme residence time of 8 min. They found cracking
to carbon dioxide, carbon monoxide, and carbonyl compounds. No increase in the relative abundance in m/z 43
is observed due to increased gas phase residence time,
indicating only minimal carbonyl compound formation.
This is surprising since retroaldolization is a likely pathway
from the anhydropyranose sugar. The spectra for xylan
is also largely unchanged by the increase in gas phase
residence time from approximately 75 to 750 ms.
Within the secondary cracking zone, sliown in Figures
4 and 5 by the spectra at 700 "C and 750 ms, major changes
for all products are observed. Cellulose is dominated by
C 0 2 and CO with low molecular weight alkenes such as
ethylene, propylene, and butene. Aromatic species have
also formed such as furan, benzene, toluene, and phenol.
The levoglucosan has been completely decomposed. The
same secondary product slate is also observed from xylan
even though it is a five-carbon sugar and has a primary
slate diffferent from that of cellulose. The xylan primary
producta are essentially 30 amu less than those from cellulose (96,114, and 132 vs. 126, 144, and 162, respectively).
(57) Shafizadeh, F.; Lai, Y. Z. J. Org. Chem. 1972, 37, 278-284.

These products are not likely to be more aromatic in nature than the primary products from cellulose. This indicates that benzene, toluene and phenol from carbohydrates are probably due to gas phase polymerization of
unsaturated species such as propylene, butadiene, and
butene.
The lignin secondary products have appreciable yields
of low molecular weight aromatics at 700 "C, however, and
this is undoubtedly due to cracking of the higher molecular
weight methoxy phenols (e.g., m / z 150, 164, 180, etc.),
which are no longer present in the lignin and pine wood
spectra at 700 "C. The peaks at m/z 94,110, and 124 are
probably due to the initial phenolic secondary products
from lignin cracking (the peak at m/z 124 is probably due
to both methylcatechol and guaiacol, which is a secondary
as well as a primary product). The peak at m / z 136 is a
major intermediate product, and although this may also
be a phenol, a possible structure is methylbenzodioxole
(XLVI), which could be formed by the intramolecular
condensation of the methoxyl and hydroxyl groups. A
homologous series is possible with benzodioxole at m/z 122
serving as the basis for the peaks at m / z 136 (methyl-),
148 (vinyl-), 162 (l-propenyl-),and 178 (l-propen-3-al).
Past
on the cracking of vanillin (mlz 152) has
shown a major peak at m l z 150, and the cracking of isoeugenol (mlz 164) gave a peak at mlz 162 in addition to
other products. The cracking of the MWL is very complicated, especially at the higher masses, in contrast to the
other three samples, although the pine wood does have
some high molecular weight peaks. Despite these high
molecular weight species there is no evidence of condensed
aromatics, such as polynuclear aromatics, in this inter-

Evans and Milne

134 Energy &Fuels, Vol. 1, No. 2, 1987

mediate zone. Although it is impossible to distinguish

isomers, the polynuclear aromatics are thought not to be
present in the secondary regime (700 "C epectrum) since
plots of intensity vs temperature show these species increasing from low values a t 700 "C to very high values at
950 "C. Therefore, it is not likely that they are present
at 700 "C and simply survive at the higher temperature
relative to the other species present a t 700 "C.
The tertiary products are shown a t 950 "C and 750-ms
residence time. The product distribution from the carbohydrates is still dominated by CO (with ethylene also
contributing to m/z 28) and COz, but the formation of
benzene and the heavier aromatics such as naphthalene
( m / z 128) and anthracene (mlz 178) have increased and
those aromatics containing oxygen, such as furan and
phenol, have been cracked. The olefins are decreasing
within this zone and since the higher aromatics grow in
absolute terms from zero, they are probably due to the
continued polymerization of the unsaturated species. The
lignin and wood samples have a higher proportion of high
molecular weight aromatics indicating the importance of
the light aromatics that are directly formed from the lignin
primary materials. The building from the lignin aromatics
is not a necessary step for polynuclear aromatics, however,
since the same types of aromatic products are obtained
from cellulose and xylan where no aromatic structures exist
as primary products. The growth of the polynuclear aromatics occurs beyond the temperature where the primary
products survive, which indicates that the condensed
aromatics are derived from the polymerization of low
molecular weight hydrocarbons along with benzene and
toluene formed directly from the gas phase cracking of
lignin primary products. The chemistry of the cracking
of biomass in this kinetic regime is dominated by hydrocarbon cracking reactions.
The generation of olefins has been a major goal of
fast-pyrolysis research.23 Early work by Antalg on the
pyrolysis of cellulose-derived primary products showed that
methane and ethylene were produced by a common path
and that all gas-producing pathways generated roughly
equal amounts of CO. However, as mentioned previously,
the cellulose pyrolysis in most biomass does not occur via
the transglycosylation mechanism but via the glycosidic
cleavage mechanism, which is catalyzed by the alkali-metal
material present in the biomass matrix. This primary
product slate has secondary cracking products different
from those of pure cellulose. The yields of benzene, phenol,
and cyclopentadiene are significantly higher from the alkali-metal-treated cellulose,Mand the rate of formation of
ethylene and other olefins may be sufficiently different
from the rate of CO formation that significant improvement in ethylene yield could be obtained by pretreatment
with alkali-metal salts.
Multivariate Analysis of Vapor Phase Transformations. As mentioned above, the recent work by Windig
et a1.41358959in multivariate analysis of pyrolysis mass
spectrometric data provides a means of simplifying the
data and observing trends. The interested reader should
consult the work of Windig et al. for an in-depth description of this technique. In brief, data reduction is
accomplished by finding correlated masses that can be
expressed by a new single variable, the factor. In a typical
pyrolysis data set, greater than 95% of the variation in the
100 most important masses in the data set is contained in
~~

(58) Windig, W.; Kistemaker, P. G.; Haverkamp, J. J. Anal. Appl.

Pyrolysis, 1981/1982, 3, 199-212.
(59) Windig, W.; Haverkamp, J.; Kistemaker, P. G. Anal. Chem. 1983,
55,81-88.

only 5-10 factors. The frst factor is the linear combination

of masses that accounta for more of the variance in the data
than any other combination of variables. The subsequent
factors are similarly extracted on the basis of the residual
variance after the effect of the previous factors has been
removed from the data. If the masses are completely
uncorrelated, then there would be as many factors as
masses and no data reduction would occur. However, as
shown in the primary and secondary pyrolysis results, the
masses are highly correlated so significant data reduction
is possible. This data reduction makes possible graphical
display of the data that not only shows trends but also can
provide chemical insight into the transformations.
Each mass has a correlation coefficient, "the loading",
aij,with each of the factors. The masses with high loadings
on a particular factor are correlated, and that factor represents that group of masses and is quantified for each
sample by the factor score, Fj. The factor score for a
sample is a linear combination of the masses based on the
loading of a mass for a factor and the intensity of that
mass, Z, for the sample:

By use of the factor loadings for the masses and factor

scores for the samples, factor analysis can reflect the
presence of chemical components or trends in the data set.
After the preparation of the correlation matrix between
masses and the extraction of the initial factors, a third
operation that is generally performed is the rotation of the
mathematically derived factors to coincide with real
chemical components. Certain real chemical components
will have high loading on more than one initial factor.
Windig and Meuzelaar41developed variance diagrams that
graphically show which linear combination of factors best
describe the component of interest. A description of the
methodology is beyond the scope of this work and again
the interested reader is directed to the work of Windig et
a1.41,58959

The secondary cracking of pine wood pyrolysis products

was studied a t three residence times from 500 to 1000 "C.
Factor analysis was performed on the 100 masses in the
data set with the most variance by using the SIGMA program developed by Windig et a1.60 Ten factors were extracted from the data, which together accounted for 98.2%
of the original variance. The plot of factor score 1vs factor
score 2 is shown in Figure 6a. The factor score plot shows
the well-behaved transition of sample composition as a
function of temperature and residence time. The trend
shown graphically by factors 1and 2, due to variation in
vapor phase residence time and temperature, can be given
chemical meaning by studying the major masses which
contribute to the two factor scores. In Figure 6b is shown
the variance diagram described above that shows the location of the major components expressed as degrees from
factor 1. A major maximum is seen at 340, and from a
comparison of the variance diagram to the factor score plot,
it can be seen that this location corresponds to the temperatures between 500 and 600 "C. The representation
of the loading as a factor spectrum in Figure 7a for this
linear combination of factors 1and 2 (340"rotation) shows
that this represents the primary products, and it is logical
that the low-temperature samples will be located in this
(60) Windig, W.; Chakravarty, T.; Richards, J. M.; Nguyen, V. T.;
Dedes, A.; Meuzelaar, H. L. C. "SIGMA, an IBM 9OOO computer software
system for interactive, graphics-oriented multivariate analysis"; on loan
from the Biomaterials Profiling Center, the University of Utah,Salt Lake
City, UT, 1986.

Pyrolysis of Biomass
a)

Energy & Fuels, Vol. 1, No. 2, 1987 135

43

150ms

a) 340"
0 750 ms

kooo1t

500
Temperature

("C)

F2

110

-2
.-

0.8'
1

94

m
2
O.S

911
44

28

I/

'I

b) 90"

I'

128

270"

Figure 6. Fador analysis of mass spectrometric data for the effect

of gas phase temperature and residence time on product distribution: (a) plot of factor score 1vs. factor score 2; (b) variance
diagram showing maxima at 340,90,and 210' rotation where the

main chemical component vectors lie for primary, secondary, and

tertiary products, respectively. The maxima indicate the linear
combinations of Fland F2 that best describe the chemical components. The factor spectra for these components are shown in
Figure 7.

region. This component axis a t 340 in Figure 6b can be

used to quantitate the cracking of the primary products
as a function of time and temperature.
The variance diagram shows three intermediate maxima
at 40,90, and 120 rotation (the maximum a t 90 corresponds to F2). Those three maxima correspond to degrees
of secondary cracking. Only the maximum at 90 will be
discussed here with the maximum at 40' displaying less
cracking severity and the maximum at 120 displaying
more cracking severity (i.e., more phenols a t 120O). The
factor spectra at 90 (Figure 7b) shows that this factor
represents the secondary products with high abundances
of low molecular weight hydrocarbons (i.e. ethane, mlz 30;
propylene, m / z 42; butadiene, m / z 54), the phenols (m/z
94, phenol; m / z 108, resol; mlz 110,catechol; m / z 120,
vinylphenol) as well as other products which have not been
identified yet. The third maximum in the variance diagram is a t 210 rotation, and the factor scale plot shows
that this is the region of high temperature. The factor
spectrum for 210' rotation shows that this group contains

m/z

Figure 7. Factor spectra of the component axes (as located by

the maxima of the variance diagram in Figure 6)that explain the
main chemical trends in vapor phase pyrolysis of wood-derived
volatiles: (a) the factor spectrum at 340' rotation showing the
primary products; (b) the factor spectrum at 90' showing the
secondary products; (c) the factor spectrum at 210' showing the
tertiary products.

the tertiary products CO and C 0 2 ,benzene, and the condensed aromatics. These two variables can thus be used
to describe the major transitions in biomass vapor phase
chemistry and graphically show the relative importance
of temperature and residence time within the major
cracking regions.
Summary of Pyrolysis Pathways. We summarize our
observations of the pyrolysis pathways for whole biomass
(and its components) in the simple diagram shown in
Figure 8. The intent is to show the origin of the major
products in both high-pressure and low-pressure pyrolysis,

Evans and Milne

136 Energy &Fuels, Vol. 1, No. 2, 1987

Primary processes

Secondary processes

Primary

Vapor

Light hydrocarbons,
aromatics and
oxygenates

phase

Primary
liquids

\
&

Condensed
011s
(phenols, aromatics)

I
Solid
phase

I1

Biomass

Olefins,
Aromatics
CO. COz, H2, HzC

High pressure

Liquid
phase

Tertiary processes

1
soot

Charcoal

Pyrolysis severity

__t

This box is meant to include solid biomass but also its plastic or rubbery forms prior
to major rupture of covalent bonds to produce low molecular weight vapors or liquids

Only first-formed products are shown All of these products could be condensed
to liquids and solids

Figure 8. Pyrolysis pathways that are consistent with observation reported in this paper.

with an emphasis on the latter. The solid products can

be distinguished by their origins: charcoal retaining the
morphology of the original lignocellulosic;coke arising from
continued thermolysis after the deposition of liquids and
organic vapors; soot from homogeneous nucleation of
high-temperature decomposition products of hydrocarbons
from the vapor phase.
The direct production of liquids is postulated to occur
mainly at pressures above atmospheric. Diebold has shown
that under direct-contact, fast-pyrolysis conditions, wood
exhibits many properties of a molten plastic state.23
IrvineG1has shown that wood has a glass transition temperature, a t a low temperature of 60-90 C, that is highly
dependent on the amount of water present as well as derived essentially from the lignin component. According
to Irvine, this glass transition is associated with the onset
of rubbery flow caused by slippage of the biopolymers.
Under high-pressure conditions, the direct formation of
a liquid product slate due to the pyrolysis of the plastic
wood is postulated to occur. The primary, true liquids
formed in the rapid pyrolysis of biomass under pressure
(but in the absence of catalysts or reactive atmospheres)
appear to have the same composition as the vapor obtained
directly from low-pressure pyrolysis.62 The further,
high-pressure conversion of the primary liquid is practiced
under a great variety of conditions, often with catalysts
and added H2 or CO. The discussion of the condensed oils
is beyond the range of this article, but has recently been
reviewed by Beckman and Elliott.63 Pyrolysis mass
spectra of several condensed liquids are given in Milne et
a1.64
The upper, vapor-phase, pathway in Figure 8 is the
(61) Irvine, M. G. TAPPI J. 1984,67, 118.
(62)Evans, R.J.; Milne, T. A,, submitted for publication in Energy
Fuels.
(63) Beckman, D.; Elliott, D. C. Can. J. Chem. Eng. 1985,63, 99.
(64)Milne, T.A.; Evans, R. J.; Soltys, M. N. In Energy from Biomass
and Wastes VIII; Institute of Gas Technology: Chicago, IL, 1984;

1371-1374.

major one elucidated by our work. At atmospheric pressure or below it is not clear whether a liquid phase exists,
after breaking of the main polymer covalent bonds, prior
to volatilization of the main components of wood. As noted
lignin is known to soften a t rather low temperatures, and the charcoal, though retaining structural features of the wood, does shrink, which could indicate a
plastic state where pyrolysis products pass directly into
a liquid state before devolatilization. Our experimental
system does not provide any insight into the condensedphase transitions that occur in pyrolysis, and this interesting hypothesis should be further explored by other
experimental techniques. The direct formation of gaseous
species from primary pyrolysis reactions, the prompt gases,
as reported by many workersB are primarily COz,H20,and
CO. These are largely associated with the char forming
reaction^.^ The primary vapors are very similar under
high- and low-pressure conditions. In fact, our primary
spectra from fractions of a gram at one atmosphere are
qualitatively similar to Curie-point and field ionization
pyrolysis results from micrograms volatilized in hard
v a c u ~ m . Figure
~ ~ , ~3~illustrates the range of spectra observed for the primary vapors.
The sequential transformation of the primary products
in the vapor phase is shown in Figure 8 as going through
three stages. Between 500 and 600 OC slight cracking
reactions occur (on a time scale of about one second) before
substantial conversion to permanent gases occurs. (This
stage is not shown in Figure 4 and 5 but is the subject of
current research activity. The factor analysis results in
Figures 6 and 7 indicate potential combinations of temperature and residence time to achieve slight cracking
intermediate between the product slates shown in Figure
7a,b.) The higher molecular weight lignin products are
cracked to lighter aromatics and oxygenates. This stage
may be of practical interest in conversion techniques since
several desirable transformations occur without substantial
loss of liquid yield: the cracking of coniferyl alcohol, which
likely polymerizes in the condensed, acidic phase, hence

Pyrolysis of Biomass

raising the molecular weight of the products; organic acid

decarboxylation, increasing the pH from the initial value
of approximately 2.
The second stage is the formation of secondary products
characterized by CO, light olefins, and the formation of
aromatics, even from the carbohydrates. The spectra a t
700 "C in Figures 4 and 5 demonstrate the products in this
class. This regime is of interest since high-value olefins
and light aromatics are a desirable product slate and have
been studied in depth.12
The third stage leads to the tertiary products characterized by the polynuclear aromatics as shown in Figure
7c and listed in Table 111. These products generally form
only in high-temperature conversion processes such as
gasification and combustion and generally in low yield.
These classes of products are logical generalizations from
our mass spectrometric studies and provide a means of
relating this work to practical methods of thermochemical
conversion, such as primary pyrolysis liquid production,
catalytic conversion of the primary products, secondary
cracking of the primary products to hydrocarbon gases, the
nature of the residual tars produced in gasification, and
source of the types of products found in emissions from
wood combustion. These applied topics will be covered
in the next paper in this
which focuses on this
classification scheme and MBMS studies of applied
thermochemical conversion systems.
Acknowledgment. ,The authors gratefully acknowledge
the support of the DOE Biofuels program under Simon
Friedrich and Gary Schiefelbein and the technical monitoring of Donald Stevens. Collaborations and discussions
with James Diebold, Tom Reed, Helena Chum, William
Peters, Douglas Elliott, and Michael Anta1 were particularly valuable. Crucial experimental and design help with
the MBMS was provided by Michael Soltys. The SIGMA
software for multivariate analysis was provided by Willem
Windig and Henk Meuzelaar of the Biomaterials Profiling

Energy &Fuels, Vol. 1, No. 2, 1987 137

Center, University of Utah.

Registry No. I, 110-00-9; 11, 534-22-5; 111, 14320-37-7; IV,
120-92-3;V, 497-23-4; VI, 96-48-0; VII, 2453-00-1; VIII, 98-01-1;
IX, 1120-73-6;X, 98-00-0; XI, 591-12-8; XII, 25414-24-8; XIII,
108-29-2; XIV, 620-02-0; XV, 80-71-7; XVI, 4041-09-2; XVII,
55100-07-7; XVIII, 2628-17-3; XIX, 90-05-1; XX, 67-47-0; XXI,
118-71-8; XXII, 93-51-6; XXIII, 68732-99-0; XXIV, 7786-61-0;
XXV, 3690-05-9; XXVI, 2785-89-9; XXVII, 121-33-5; XXVIII,
91-10-1; XXVIX, 498-07-7; XXX, 97-54-1; XXXI, 498-02-2;
XxXII,663805-7; XxXIII,458-36-6; XXXIV,458-35-5;XXXV,
28343-22-8; XXXVI, 134-96-3;XXXVII,14059-92-8; XXXVIII,
6635-22-9; XXXIX, 1135-24-6; XL, 2478-38-8; XLI, 4206-58-0;
XLII, 537-33-7; XLN, 106544-450;XLV, 492-62-6; MeH, 74-82-8;
CO, 630-08-0; C2H4, 74-85-1; MeOH, 67-56-1; OCO, 124-38-9;
OCHMe, 75-07-0; HCOOH, 64-186; MeCOMe, 67-64-1; H02CMe,
64-19-7; HOCH2CH0, 141-46-8;MeOCHO, 107-31-3;HOCH2CH,CHO, 2134-29-4;EtCOMe, 78-93-3; CH,=CHCOOH, 79-10-7;
HOCH2COMe,116-09-6; HO,CEt, 79-09-4; CH3COOCH3,79-20-9;
HOCH,COOH, 79-14-1; MeCOCOMe, 431-03-8; MeCH=
CHCOOH, 372465-0; HO,CEtMe, 107-92-6;MeCH2COCH20H,
5077-67-8; PhOH, 108-95-2; MeCOCOEt, 600-14-6; CH3(CH2),COOH, 109-52-4;o-CH~C&OH, 95-487; m-CH3CsH,OH, 108-39-4;
P-CH&H~OH, 106-44-5;CeH,(OH)2, 12385-08-9;CH3(CHz),COzH, 142-62-1; CH&O&H2(CO)CH3,592-20-1; C G H ~ ( C H ~ ) ~ O H ,
1300-71-6;C ~ H I ( C H ~ C H [ ~ ) O
25429-37-2;
H,
C&&,CO2H, 65-85-0;
C&, 74-84-0; C3H4,74-99-7;C3H6,115-07-1; CzHz, 74-86-2; C&3,
71-43-2; K2C03,584-08-7; K, 7440-09-7; 1-hydroxy-2-propanone
acetate, 106544-46-1; 1-hydroxy-2-butanone acetate, 1575-57-1;
eugenol, 97-53-0; butadiene, 106-99-0;butyne, 25339-57-5;butene,
25167-67-3; cyclopentadiene, 542-92-7;toluene, 108-88-3; styrene,
100-42-5;xylene, 1330-20-7; indene, 95-13-6; naphthalene, 91-20-3;
acenaphthylene, 20896-8; fluorene, 86-73-7; anthracene, 120-12-7;
phenanthrene, 85-01-8; pyrene, 129-00-0; fluoranthene, 206-44-0;
benzacenaphthylene, 76774-50-0; methylpyrene, 27577-90-8;
benzo[ghi]fluoranthene, 203-12-3; chrysene, 218-01-9; benz[a]anthracene, 56-55-3; triphenylene, 217-59-4; be"[
clphenanthrene,
195-19-7;methylbenzo[ghi]fluoranthene,51001-44-6; benzo[ae]pyrene, 73467-76-2; perylene, 198-55-0; anthanthrene, 191.26-4;
cellulose, 9004-34-6; lignin, 9005-53-2; hemicellulose, 9034-32-6;
xylan, 9014-63-5.