Mechanocatalysis for biomass-derived chemicals and fuels

Article abstract of DOI:10.1039/b923079c

Heterogeneous catalysis cannot be easily applied to solids such as cellulose. However, by mechanically grinding the correct catalyst and reactant, it is possible to induce solid-solid catalysis or mechanocatalysis. This process allows a wide range of solids to be effectively utilized as feedstock for commercially relevant compounds. Here we show a set of structural and physical parameters important for the implementation of catalysts in mechanocatalytic processes and their application in the catalytic depolymerization of cellulose. Using the best catalysts, which possess high surface acidities and layered structures, up to 84% of the available cellulose can be converted to water-soluble compounds in a single pass. This approach offers significant advantages over current methods - less waste, insensitivity to feedstock, multiple product pathways, and scalability. It can be easily integrated into existing biorefineries - converting them into multi-feedstock and multi-product facilities. This will expand the use of non-food polysaccharide sources such as switch grass.

Full text of DOI:10.1039/b923079c

View Article Online / Journal Homepage / Table of Contents for this issue
PAPER
www.rsc.org/greenchem | Green Chemistry
Mechanocatalysis for biomass-derived chemicals and fuels
Sandra M. Hick, Carolin Griebel, David T. Restrepo, Joshua H. Truitt, Eric J. Buker, Caroline Bylda and
Richard G. Blair*
Received 4th November 2009, Accepted 8th December 2009
First published as an Advance Article on the web 25th January 2010
DOI: 10.1039/b923079c
Heterogeneous catalysis cannot be easily applied to solids such as cellulose. However, by
mechanically grinding the correct catalyst and reactant, it is possible to induce solid–solid
catalysis or mechanocatalysis. This process allows a wide range of solids to be effectively utilized
as feedstock for commercially relevant compounds. Here we show a set of structural and physical
parameters important for the implementation of catalysts in mechanocatalytic processes and their
application in the catalytic depolymerization of cellulose. Using the best catalysts, which possess
high surface acidities and layered structures, up to 84% of the available cellulose can be converted
to water-soluble compounds in a single pass. This approach offers significant advantages over
current methods - less waste, insensitivity to feedstock, multiple product pathways, and scalability.
It can be easily integrated into existing biorefineries - converting them into multi-feedstock and
multi-product facilities. This will expand the use of non-food polysaccharide sources such as
switch grass.
We have found that it is possible to overcome diffusion in
1. Introduction
a solid–solid reaction by using mechanical force without the
Lignocellulosic biomass represents a rich source of feedstock for
addition of solvents. This phenomenon is often referred to as
fuels1 and chemicals; major work in developing this resource is
mechanocatalysis or tribocatalysis. Little work has been done
underway. Biological sources such as switch grass, corn stover,
on this phenomenon.11-14 In fact, recent work has focused on
bagasse, and other agricultural waste are largely underutilized.
using traditional heterogeneous catalysts (such as zeolites) in
Most lignocelluosic biomass is processed in one of three ways:
mechanochemical processes.15-21 This approach ignores the ag-
acid hydrolysis, enzymatic hydrolysis, and pyrolysis.2,3 New
gressive nature of mechanical processing. Effective mechanocat-
biological4,5 and chemical6 approaches seek to circumvent the
alysts must be mechanically robust, and possess sites that are
drawbacks associated with these processes.
physically accessible and chemically active.
Catalytic processing of these materials is an important topic.
Mechanocatalytic processes require no external heat. All
Although efficient catalysts have been developed for a wide
of the energy for the reaction comes from the pressures and
range of heterogeneous systems, they are not suitable for solid-
frictional heating provided by the kinetic energy of milling
solid catalysis. A major difficulty with these solids is mass
media moving in a container. Although there is a large body
transport. High surface area structures suited to liquid and
of work examining mechanochemical syntheses,22-39 to date
gas systems do not work as efficiently in solid systems. Two
no one has examined the properties required for an effective
typical structures, porous solids and supported particles have
mechanocatalyst.
limitations when applied to solid systems. Porous solids have
In a mechanocatalytic system, it is important that intimate
pore sizes too small to accommodate molecules much larger
contact between the catalyst and reactant be maintained.
˚
than 30 A; and supported particle systems still require some
Although most biomass processing facilities utilize some form
of attrition technology, it is usually in the form of hammer mills,
disk mills, or roller mills. These technologies are being actively
researched,40-42 but they do not allow good interaction between a
solid catalyst and the material being treated. Pebble (or rolling)
mills, shaker mills, attrition mills, and planetary mills are a few
examples of mills that effectively “push” the catalyst into contact
with the material being treated (biomass).43
method to overcome the solid-solid diffusion barrier.
Biochemical pathways are also important in processing these
polymers. However, enzymes have limited pressure/temperature
regimes in which they can function and are not as robust as
inorganic solids.
There are two issues that need to be addressed in order
to realize effective catalysis in solid-solid reactions. The solid-
solid diffusion barrier (mass transport) must be overcome and
the catalyst must be structured to allow access to catalytic
sites. Currently, most catalytic depolymerization involves solvent
systems to overcome this diffusion barrier.7-10
Experimental
Reagents
Pure microcrystalline cellulose (Avicel, Brinkmann) was utilized
to investigate the performance of different solid catalysts.
The natural cellulose sources Z. mays indurata (flint corn),
University of Central Florida, Department of Chemistry, P.O. BOX
162366, Orlando, FL, 32816, USA, rblair@mail.ucf.edu; Fax: +1
407-823-2252; Tel: +1 407-823-0639
468 | Green Chem., 2010, 12, 468–474
This journal is
The Royal Society of Chemistry 2010
©

View Article Online
Prunus stone, paper, aspen wood, and mixed biomass were col-
lected from local sources. The grasses: A. gerardii (Big Bluestem),
S. scoparium (Little Bluestem) and P. virgatum (Switchgrass)
were supplied by Agricol Corporation (Madison, WI). All
natural cellulose sources were dried at room temperature to a
moisture content of <10% and cut to 2 cm or smaller pieces.
The materials kaolinite (Edgar Plastic Kaolin, Axner Pot-
tery), delaminated kaolinite (Kaopaque 10, IMERYS), alu-
minium phosphate (Fisher Scientific), aluminium oxide (J.T.
Baker), talc (Nytol 100HR, Axner Pottery), Y-type zeolite
(HS-320, Hydrogen Y, Wako Chemicals), bentonite (Asbury
Carbons), vermiculite, quartz, muscovite mica, silicon carbide
(-325 mesh, Electronic Space Products International), graphite
(grade TC306, Asbury Carbons), and aluminium sulfate (Fisher
Scientific) were used as received. Layered silicates were H+
exchanged by soaking in 1 M hydrochloric acid for 12 h, filtering
and dried at 80 ◦C overnight. Chemically delaminated kaolinite
was prepared by intercalating with urea and deintercalating by
washing with water.44 The super acid was prepared by stirring
aluminium oxide (J.T. Baker) in 2.5 M H2SO4 followed by
calcination at 600 ◦C.45
analyzed by silanizing47 the product. Dehydration products were
extracted with 60 ◦C chloroform and analyzed by GC-MS.
Thin layer chromatography
Thin layer chromatography was used to assess the composition
of the hydrolysis product. Solutions were spotted onto cellulose
plates and developed with a 20 : 7 : 10 mixture of n-butanol,
acetic acid, and water. The plates were stained with a 3% urea
and 1 M phosphoric acid in n-butanol saturated water solution.
Discrete element modeling
Discrete element models of the milling process were generated
using EDEM (DEM Solutions Ltd.).
Degree of polymerization
The degree of polmerization of the insoluble cellulose residue
was determined using viscometry according to the method
outlined in ASTM D 4243.
Results
Mechanical processing
Three milling modes were investigated for the mechanocatalytic
depolymerization of cellulose—shaking, rolling, and stirring.
Fig. 1 illustrates solubilization achievable in a SPEX shaker
mill. No appreciable solubilization was realized on samples of
microcrystalline cellulose milled without a catalyst.
Various amounts of cellulose and catalyst were ground using a
rolling mill (custom), mixer mill (SPEX Certiprep, Metuchen,
NJ), or attrition mill (Union Process Inc., Akron OH). Initial
catalyst assessment was performed using a mixer mill. Typically,
2 grams of a 1 : 1 mixture of catalyst and cellulose were ground in
a 65 mL vial (1.5¢¢ ID ¥ 2.25¢¢ deep) made of 440C steel, utilizing
three 0.5¢¢ balls made of the same material as the milling vial.
Attrition milling experiments were performed by Union Process,
Inc. in a 1-SD attrition mill run at 350 rpm with a 1.5 gallon
tank, 40 lbs of 0.25¢¢ chrome steel (SAE 52100) balls as the
milling media, and 1200 g of a 1 : 1 mixture of cellulose and
catalyst. Rolling mill experiments were performed in a custom
rolling mill constructed of 316 stainless steel with a diameter of
1.37¢¢ and a length of 4.93¢¢. The mill was charged with 25 0.5¢¢
balls made of 440C steel and 2 g of a 1 : 1 mixture of cellulose
and catalyst.
Fig. 1 Solubilization of cellulose as a function of milling time in a
SPEX 8000D mixer mill. Catalysts were milled with microcrystalline
cellulose in a 1 : 1 ratio. The most rapid solubilization was measured
using 1 : 1 aluminosilicates such as kaolinite.
Gravimetric analysis
The extent of hydrolysis was monitored gravimetrically. Con-
version of cellulose to water-soluble oligosaccharides was de-
termined by stirring 0.1 g of the reaction mixture in 30 mL
of water. Any oligosaccharide with a degree of polymerization
<5 will be solvated.46 The production of water-soluble products
was measured by filtration through a 47 mm diameter Whatman
The catalyts’ chemical and physical properties effect on
conversion efficiency was studied by choosing materials with
specific structural and chemical properties. Fig. 2 summarizes
the solubilization results for cellulose mechanocatalytically
treated for two hours in a shaker mill.
A shaker mill was chosen to assess catalyst efficacy since
cellulose hydrolysis is observed after as little as 6 min of milling.
This mode is a high-energy process with the possible realization
of localized high pressures. After 3 h of milling, up to 84% of the
cellulose can be converted to water-soluble fractions allowing
rapid assessment of catalysts parameters.
R
Nucleporeꢀ track etched polycarbonate membrane filter with a
pore size of 0.220 mm. The residue was dried in a 60 ◦C oven for
12 h and then weighed.
Gas chromatography with mass sensitive detection
GC-MS analysis was performed on an Agilent 6850 GC with
an Agilent 19091-433E HP-5MS column (5% phenyl methyl
siloxane, 30 m ¥ 250 mm ¥0.25 mm nom.) coupled with a
5975C VL mass selective detector. Saccharide composition was
The layered silicate mineral kaolinite was determined to
be a good mechanocatalyst and the composition of the sol-
ubilized fraction produced was analyzed utilizing thin layer
This journal is
The Royal Society of Chemistry 2010
Green Chem., 2010, 12, 468–474 | 469
©

View Article Online
Fig. 2 Mechanocatalytic activity for a sample of compounds inves-
tigated. Catalysts were milled with microcrystalline cellulose in a 1 : 1
ratio for 2 h. Layered compounds with a surface acidity, H0, <-3.0,
as determined by dicinnamalacetone gave the best yields. The catalytic
efficiency of kaolinite was improved through physical delamination and
proton exchange (*).
chromatography and gas chromatography with mass sensitive
detection. Both methods confirmed that depolymerization oc-
curs rapidly with no oligosaccharides larger than n = 2 detected
even after 30 min of treatment. The three major water-soluble
components detected were levoglucosan, fructose, and glucose.
The degree of polymerization of the insoluble residue was
measured and found to decrease linearly with time.
The variation in product composition was studied as a
function of milling mode and time. A study of the energy input
through milling, and its effect on products, was investigated
using a variable speed rolling-mill. Models were developed using
discrete element methods (EDEMTM, DEM Solutions Ltd.) to
estimate the compressive forces achieved during milling. These
models indicate that, in a 10 s period, a shaker mill can produce
9 impacts with forces between 400 and 3000 N; an attrition mill
can produce 9 impacts between 400 and 2000 N; a rolling mill
at 30 rpm generates 4 impacts between 60 and 110 N; and at
100 rpm, 10 impacts between 60 and 130 N.
High-energy processing in a shaker mill resulted in the
production of levoglucosan, fructose and glucose with a ratio
of 9 : 1 : 4.3 after 30 min of treatment and a ratio of 4.6 : 1 : 4.1
after two hours of treatment. The product distribution was
similar for samples prepared in an attrition mill. Low-speed
processing in a rolling mill (30 rpm) resulted in no measurable
catalytic activity; increasing the rotation velocity to 100 rpm
resulted in 13.2 0.8% solubilization after 96 h of treatment.
The product consisted of levoglucosan, fructose, and glucose
in a 1 : 1 : 5.8 ratio. With continued high energy milling the
levoglucosan fraction decreased and other dehydration products
were observed - levoglucosenone and 5-hydroxymethyl furfural
(HMF), as well as the retro-adol condensation product furfural
(Fig. 3). This result is encouraging since HMF is of interest for
use as a fuel or chemical feedstock.1,48,49
Fig. 3 Cellulose can be mechanocatalytically converted to glucose and,
with prolonged treatment, into a variety of products. The formaldehyde
produced was not observed with the analytical techniques utilized.
and bentonite (Al2Si4O11·H2O) showed good catalytic abil-
ity. Low-acidity solids such as talc (H2Mg3(SiO3)4), vermi-
culite ((MgFe,Al)3(Al,Si)4O10(OH)2·4H2O), quartz (SiO2), mica
(KF)2(Al2O3)3(SiO2)6(H2O), silicon carbide (SiC), graphite (C),
and aluminium sulfate (Al2(SO4)3) were less effective. The
hardness of the catalyst did not play a role in the efficiency
of the depolymerization. Both kaolinite and talc are soft, but
kaolinite is a much more efficient catalyst. Silicon carbide and
aluminium oxide are both very hard, but silicon carbide showed
little or no catalytic ability. The use of harder catalysts resulted in
undesirable wear on the container and milling media. The most
effective catalyst is the layered mineral kaolinite. Kaolinite is
an aluminosilicate consisting of aluminium-containing (as AlO6
units) layers covalently bound to silicon-containing (as SiO4
units) layers as in a 1 : 1 ratio. These layers are held together
by hydrogen bonds from protons on open Al–O–Al sites to
open Si–O–Si sites. The structurally similar bentonite has each
aluminium-containing layer covalently bound above and below
by a silicon-containing layer in a 2 : 1 configuration; this prevents
the active sites from interacting with the cellulose (Fig. 4). The
role of aluminium in the active sites was confirmed by comparing
the catalytic ability of quartz and aluminium phosphate. These
compounds are isostructural; substituting the SiO4 units in
quartz with AlO4 and PO4 units, as in aluminium phosphate,
results in an increase in active sites and the observed increase in
catalytic ability.
Layered compounds are effective mechanocatalysts because
the layers are typically held together by weak forces such as
hydrogen bonding and van der Waals forces. These bonds can be
easily broken via mechanical processing (grinding or rubbing).50
The result is a material with a high specific surface area (SSA)
that is only dependent upon the number of layers in each particle.
Discussion
Milling alone (without a catalyst present) is not sufficient
to hydrolyze the glycosidic bond in cellulose. Acidic solids
such kaolinite (Al2Si2O7·2H2O), alumina super acid, alu-
minium phosphate (AlPO4), alumina (Al2O3), Y-type zeolite,
3
SSA
=
(1)
hardsphere
rdsolid
470 | Green Chem., 2010, 12, 468–474
This journal is
©
The Royal Society of Chemistry 2010

View Article Online
expression is meaningless. If percent composition is used, the
resulting expression does not accurately reflect the consequences
of increasing the milling load without increasing the vessel size
(which results in a decrease in reaction rate). We have found that
reaction rates can be examined by expressing the concentration
of reactants and products in terms of mass/free volume.52 Here
the free volume is the volume of the milling container not
occupied by balls, reactants, or products. This value is calculated
by converting the masses of the reactants, products, and milling
media to volumes based on the materials’ densities. This volume
is subtracted from the container volume to give a free volume.
This approach indirectly incorporates the motions available to
the milling media. The milling media, for the same mode of
milling, in systems with large free volumes will have a greater
mean free path than in systems with small free volumes.
We determined the reaction order by generating kinetic plots.
Although attrition is quite rapid in a SPEX mill, finely ground
cellulose (Avicel) and catalysts were used to minimize the effect
of initial particle size. The reaction cannot be zeroth order
since reaction rate would be independent of concentration. The
concentration of the reactants directly affects the free volume
and, subsequently, the motion of the milling media. Higher
concentrations result in less motion. In the most severe case
the concentration would be so high that no media motion
is allowed. A zeroth order model would predict yield in this
case - an unphysical prediction. Fig. 5 compares a first and
second order plot in this system. The differentiation between
first and second order behavior is a little more subtle. Both
kinetic plots can be fit to lines representing initial and final
kinetics. Although the first order plot gives slightly better linear
correlation coefficients, a second order model more accurately
describes the data. For example, using H+ exchanged, physically
delaminated kaolin the linear correlation coefficient for the
initial kinetics is -0.9986 for first order kinetics and 0.9974 for
second order kinetics. In the final kinetic region this coefficient
is -0.9969 for first order kinetics and 0.9952 for second order
kinetics. The important feature is the data point near the
intersection of the two regression lines (inset Fig. 5). It does not
fall on the first order curve that would be generated by the sum
of the two linear fits. It does fall on the curve generated by the
two linear second order fits. This behavior was observed for all
catalysts listed in Table 1 which summarizes the rates observed
in several systems. The uncertainties reported are the standard
deviations in the slope of the second-order plot calculated
using linear regression. The uncertainty in the time change
was determined by propagating these uncertainties through the
equation describing the intersection of these lines. Second order
behavior has also been observed in the hydrolysis of the glucose
dimer cellobiose in super critical water.53
Fig. 4 The structures for the 2 : 1 smectite bentonite (left) and 1 : 1
smectite kaolinite (right). The octahedra are AlO6 units and the
tetrahedra are SiO4 units. The sheets in bentonite are charged due to an
Al : Si imbalance. The sheets are held together by ionic forces between
the sheets and a cation (dark sphere). The sheets in kaolinite are held
together by hydrogen bonds from protons on the Al layer to the oxygens
on the Si layer.
2
SSA
=
(2)
layered
(n + nf −1)tdsolid
Eqn (1) and 2 give the specific surface areas (SSA) in area/unit
mass for hard spheres and layered materials where n = the
number of layers, dsolid = the density of the solid, r = the
radius of the particle, t = the interlayer spacing, and f = the
ratio of the layer thickness to the interlayer spacing. Eqn (1)
is derived from the surface area of a solid sphere (4pr2) and
the mass of the sphere as determined by its density (mass =
4/3 pr3dsolid). Eqn (2) is derived from the surface area of a
fragment using its length and width (2lw) and the mass of
n layers using the material’s density and interlayer spacing
(mass = lw(n+nf-1)tdsolid). This derivation assumes both sides
of the sheet are catalytically active. For kaolinite, only one side
of the sheet (the AlO6 side) is catalytically active. This results
in a SSA half of that predicted by eqn (2). Using the dsolid
=
-3
˚
2.599 gcm , t = 2.892 A, and f = 1.561 for kaolinite, we find
that reducing it to a collection of single aluminosilicate sheets
(through delamination) would result in an active specific surface
area of 852 m2 g-1. This would be equivalent to 1.35 nm spherical
particles. Structures consisting of 5 layers would have a specific
surface area equivalent to 10.2 nm particles (113 m2 g-1). It has
been suggested that mechanochemical processes are dependent
on the area of surface sites available.15 This is in line with our
observation that layered materials are the ideal structure for
mechanocatalysts. Delamination of kaolinite to single sheets has
been observed during grinding.51 In fact, it is not unreasonable
to expect delamination to single layers; this has been observed
in talc.15
It was confirmed that the process was catalytic by performing
turnover studies using kaolinite and cellulose in a shaker
mill. Two hours of milling time resulted in loss of catalytic
efficiency over 5 turnovers. One hour of milling resulted in no
loss in catalytic efficacy over 8 turnovers. Although extended
milling can induce significant defects in the crystal structure
of the catalysts, the active sites on these catalysts are surface
protons and should be unaffected by the defect structure
of the solid. Prolonged milling may, instead, result in the
formation of insoluble polymerization products. In particular,
It is important to note that anhydrous kaolinite showed a
decrease in catalytic ability. This is due to the fact that water is
required to break the cellulose glycosidic bond.
Kinetics and depolymerization mechanism
Concentrations in chemical reaction are typically expressed
in terms of moles/l. In a solventless, solid-solid reaction this
This journal is
The Royal Society of Chemistry 2010
Green Chem., 2010, 12, 468–474 | 471
©

View Article Online
Table 1 Second-order rates observed in the catalytic depolymerization of cellulose in a SPEX shaker mill. The column labeled “time change”
indicates when the rate changes
Initial Rate (second order) L g-1 h-1 Second Rate (second order) L g-1 h-1 Time Change (hours)
Kaolinite, physically delaminated, H+ exchanged 0.0522 0.0022
0.1460 0.0058
0.0836 0.0054
0.0902 0.0035
0.0832 0.0027
0.0825 0.0056
0.00639 0.00018
0.00230 0.00016
0.00415 0.00026
1.21 0.14
1.08 0.29
1.13 0.15
1.16 0.11
0.75 0.17
1.86 0.18
1.383 0.049
3.567 0.029
Kaolinite, H+ exchanged
Kaolinite, chemically delaminated
Kaolinite, physically delaminated
Kaolinite
0.0464 0.0017
0.0433 0.0013
0.0330 0.0013
0.03240 0.00085
0.0268 0.0015
0.01147 0.00050
0.00943 0.00014
Bentonite
Aluminium oxide
Talc
Fig. 5 A first (left) and second (right) order plot of the hydrolysis of cellulose in a SPEX shaker mill. Both plots show strong linear relations, but the
data point near the intersection of the two fit lines (insets) indicates that a second-order process is the more accurate description. The catalyst used
was H+ exchanged, physically delaminated kaolinite.
furfural polymerizes when heated in the presence of an acid.
These insoluble by-products would interfere with the interaction
between the catalyst and the reactant. Limiting the milling time
limits the production of these by-products.
In order to understand the mechanism of cellulose depoly-
merization, the degree of polymerization (DP) of the insoluble
residue was measured. This can be compared to the change in DP
observed in acid and enzyme hydrolysis. The three approaches
to depolymerization - acid, enzymatic, and mechanocatalytic
proceed by quite different kinetics and mechanisms. By examin-
ing the change in degree of polymerization of the residue with
respect to the fraction of oligomers with a DP<5 (or degree of
solubilization) the role of these factors can be reduced and the
Fig. 6 Change in the degree of polymerization for the insoluble residue
approaches can be compared.
of cellulose hydrolysis as a function of degree of solubilization. The
dotted line represents a model where hydrolysis only occurs at the ends
of solubilization. Values for acid hydrolysis were simulated
of the polymer chain.
Fig. 6 shows the change in DP as a function of degree
using a model that all bonds have an equal probability of
cleavage.54 Change in DP from enzyme hydrolysis was taken
from the literature.55 It can be seen that mechanocatalytic
depolymerization does not follow a mechanism like acid or
enzyme hydrolysis. Mechanocatalytic hydrolysis does not ran-
domly cleave cellulose chains like acid hydrolysis. Initially,
depolymerization more closely matches the enzymatic process.
The accessibility of surface sites gives rise to the evolution of
the degree of polymerization in enzyme hydrolysis.56 Similarly,
mechanocatalysis is dominated by two processes - attrition and
hydrolysis. During the initial milling time, cellulose particles are
being broken down physically and chemically. There are three
main chemical reactions occurring. The reactions are: hydrolysis
catalyzed by the catalyst’s surface protons, dehydration by the
catalyst, and retro-aldol condensation due to the localized high
pressures. The surfaces of these particles are accessible to the
catalyst. At a certain point, attrition slows and only the end units
of the cellulose chain are accessible. This results in a change
in DP that coincides with a model where only the ends of
a polymer chain are allowed to react (dashed line in Fig. 6).
For the layered catalysts, bentonite, talc, and kaolinites the
rate changes (Table 1) when solubilization is between 30 and
40%. This corresponds to the region in Fig. 6 where the DP of
the residue matches an end-only hydrolysis model and further
corroborates the second-order model.
472 | Green Chem., 2010, 12, 468–474
This journal is
The Royal Society of Chemistry 2010
©

View Article Online
Industrial applicability and economics
In order for mechanocatalysis to be an effective industrial tool,
it must be effective for real world materials. We tested the con-
version efficiency for a wide range of relevant cellulose sources.
Fig. 7 illustrates the efficiency observed in the depolymerization
of cellulose after two hours of milling in a mixer mill. Initial
particle size was kept to less than 2 cm. In all cases, the cellulose
source and catalyst were reduced to fine powders in 5 to 10 min
due to the vigorous nature of the attrition process. Agricultural
wastes from corn (corn stover), wood (aspen), and fruit (Prunus
stone) production were examined; all showed improved water
solubility after mechanocatalytic treatment. Commercial and
residential waste such as paper and mixed waste from clearing
land can also be efficiently treated. The grasses A. gerardii
(Big Bluestem), S. scoparium (Little Bluestem) and P. virgatum
(Switchgrass) are crops that are of interest for use as a biomass
source. It should be noted that 90% of a corn kernel’s mass can
be converted to soluble matter in a single pass.
Fig. 8 The energy consumed to produce 1 gram of glucose from 1 gram
of cellulose treated in a shaker mill, 1 kg of cellulose treated in an attritor,
and modeled consumptions for 100 kg in a large attritor. The dashed
line is the energy released by burning the ethanol produced from 1 gram
of glucose.
one gram of glucose. It can be seen that the SPEX mill is an
energy intensive process. Switching to an attritor allowed the
process to be scaled-up by 1000 fold; the result was nearly
a 46-fold decrease in the energy consumption as expressed in
kJ/gram glucose produced. It was found that conversion in a
small attritor required a 4-hour initiation time before the rate
became appreciable. The kinetic data from this batch was used,
in conjunction with the behavior observed in the shaker mill, to
develop a predictive model for a 100 kg batch. The gray trace in
Fig. 8 shows the projected energy consumption for an attritor
with a 150 hp motor and fast reaction kinetics. It is important to
note that the four-hour induction period must be eliminated for
this approach to produce glucose at an energy cost lower then
the energy released by burning the ethanol prepared from the
glucose. This process is energy positive for 0.9 h of milling with
a predicted conversion efficiency of 20.2%.
Fig. 7 Unlike current approaches to cellulose hydrolysis, mechanocat-
alytic hydrolysis is relatively insensitive to feedstock.
Our survey experiments have utilized a SPEX mixer mill,
which allows us to rapidly assess the viability of catalytic
materials and develop kinetic models for cellulose conversion.
We have found that the reaction goes as a second-order
process in cellulose. However, this technology is not scalable
without significant redesign.57 Rolling-mode and stirring-mode
are among the scalable approaches. Utilizing our DEM model,
it was determined that rolling mills do not develop the high
pressures encountered in a shaking mill. Processing in a rolling
mill produced a product composed primarily of glucose. This
suggests that the forces that occur during the milling process
are directly related to the composition of the soluble fraction
produced. Low forces result in no observable solubilization;
increasing the rotational velocity of the roller mill results in
compressive forces and a measurable yield of sugars. The
most energetic process, shaking-mode, results in an increase in
the levoglucosan fraction. This implies that there is a critical
energetic region favorable for the production of fructose and
glucose.
Conclusions
The natural layered structure of clays is ideally suited for use
as a catalyst in milling processes. We have shown that layered
structures are particularly suited to mechanocatalytic processes.
This opens up a new area for catalyst synthesis. A new set of
catalysts based on this paradigm could be synthesized to take
advantage of the unique conditions in a ball mill. Using this
idea, it may be possible to depolymerize other biopolymers such
as chitin or protein as well as synthetic polymers such as nylon,
polyethylene terepthalate, polycarbonate, and polylactic acid.
The observation of the glucose dehydration products levoglu-
cosan, levoglucosenone, and 5-hydroxymethyfurfural as well as
the retro-aldol condensation product furfural suggest that it will
be possible to design mechanocatalysts for the direct conversion
of cellulose into these compounds. In fact, many of the synthetic
pathways utilized to produce derivatives from these compounds
should be directly accessible through solventless milling.
Attritors are scalable, can be run in a batch or continuous
mode, and can produce compressive forces similar to those
achieved in a shaker mill (0.4 to 3 GPa, as predicted by out DEM
models). We performed a limited number of kilogram-scale tests
using a small Union Process attritor. Fig. 8 illustrates the energy
costs associated with the two milling technologies. The dashed
line is the energy obtainable from the ethanol produced from
Mechanocatalytic processing of materials has significant
advantages over current methods. The best catalyst so far,
kaolinite, costs around $80/ton and can be reused. Any catalyst
waste produced is innocuous and there are no toxic solvents
needed. Additionally, no heating or high-pressure equipment is
This journal is
The Royal Society of Chemistry 2010
Green Chem., 2010, 12, 468–474 | 473
©

View Article Online
required, simplifying plant design. Mechanocatalytic conversion
of cellulose is insensitive to lignin and hemicellulose content
allowing any cellulosic biomass source to be utilized. This is an
improvement over methods that utilize edible biomass (such as
corn) for ethanol production.
22 E. Barraud, S. Begin-Colin, G. Le Caer, O. Barres and F. Villieras,
J. Alloys Compd., 2008, 456, 224–233.
23 M. K. Beyer and H. Clausen-Schaumann, Chem. Rev. (Washington,
DC, U. S.), 2005, 105, 2921–2948.
24 V. V. Boldyrev and K. Tkacova, J. Mater. Synth. Process., 2000, 8,
121–132.
25 P. G. Fox, J. Mater. Sci., 1975, 10, 340–369.
26 S. Granick and S. C. Bae, Nature, 2006, 440, 160–161.
27 S. Kipp, V. Sepelak and K. Dieter Becker, Chem. Unserer Zeit, 2005,
39, 384–392.
Dedication
This manuscript is dedicated to the memory of Glenn T. Blair
(1945–2009).
28 B. Kubias, M. J. G. Fait and R. Schloegl, Handb. Heterog. Catal.
(2nd Ed.), 2008, 1, 571–583.
29 V. I. Levitas, Phys. Rev. B: Condens. Matter Mater. Phys., 2004, 70,
184118.
30 B. Rodr´ıguez, A. Bruckmann, T. Rantanen and C. Bolm, Adv. Synth.
Acknowledgements
Catal., 2007, 349, 2213–2233.
31 B. M. Rosen and V. Percec, Nature, 2007, 446, 381–382.
32 L. Takacs, Hyperfine Interact., 1997, 111, 245–250.
33 Z. V. Todres, in Organic Mechanochemistry and Its Practical Applica-
tions, CRC Taylor and Francis, Boca Raton, FL, 2006, pp. 61-102.
34 F. K. Urakaev and V. V. Boldyrev, Powder Technol., 2000, 107, 93–
107.
We thank Agricol Corporation (Madison, WI) for supplying
samples of Big Bluestem, Little Bluestem, and Switch grass;
IMERYS (Roswell, GA) for supplying samples of physically
delaminated kaolinite; and Asbury Carbons (Asbury, NJ) for
supplying samples of bentonite and graphite.
35 F. K. Urakaev and V. V. Boldyrev, Powder Technol., 2000, 107, 197–
206.
36 K. Wieczorek-Ciurowa and K. Gamrat, Mater. Sci., 2007, 25, 219–
232.
References
37 H. Yang and P. G. McCormick, Trans. Mater. Res. Soc. Jpn., 1994,
1 G. W. Huber, S. Iborra and A. Corma, Chem. Rev., 2006, 106, 4044–
4098.
14A, 617–620.
38 I. J. Lin and S. Nadiv, Mater. Sci. Eng., 1979, 39, 193–209.
39 G. Kaupp, CrystEngComm, 2009, 11, 388–403.
40 L. S. Esteban and J. E. Carrasco, Powder Technol., 2006, 166, 139–
151.
2 S. J. B. Duff and W. D. Murray, Bioresour. Technol., 1996, 55, 1–33.
3 A. Demirbas and G. Arin, Energy Sources, 2002, 24, 471–482.
4 M. B. Sticklen, Nat. Rev. Genet., 2008, 9, 433–443.
5 T. S. Bayer, D. M. Widmaier, K. Temme, E. A. Mirsky, D. V. Santi
and C. A. Voigt, J. Am. Chem. Soc., 2009, 131, 6508–6515.
6 Y. Zhao, W.-J. Lu, H.-T. Wang and D. Li, Environ. Sci. Technol.,
2009, 43, 1565–1570.
7 R. Rinaldi, R. Palkovits and F. Schuth, Angew. Chem., Int. Ed., 2008,
47, 8047–8050.
8 T. Iwaya, S. Tokuno, M. Sasaki, M. Goto and K. Shibata, J. Mater.
Sci., 2008, 43, 2452–2456.
41 D. Schell and C. Harwood, Appl. Biochem. Biotechnol., 1994, 45–46,
159–168.
42 V. S. P. Bitra, A. R. Womac, N. Chevanan, P. I. Miu, C. Igathinathane,
S. Sokhansanj and D. R. Smith, Powder Technol., 2009, 193, 32–45.
43 C. Suryanarayana, Prog. Mater. Sci., 2001, 46, 1–184.
44 K. Tsunematsu and H. Tateyama, J. Am. Ceram. Soc., 1999, 82,
1589–1591.
45 K. Arata, Appl. Catal., A, 1996, 146, 3–32.
46 J. B. Taylor, Trans. Faraday Soc., 1957, 53, 1198–1203.
47 M. L. Sanz, J. Sanz and I. Mart´ınez-Castro, J. Chromatogr., A, 2004,
1059, 143–148.
48 J. N. Chheda, Y. Roman-Leshkov and J. A. Dumesic, Green Chem.,
2007, 9, 342–350.
49 Y. Roman-Leshkov, J. N. Chheda and J. A. Dumesic, Science, 2006,
312, 1933–1937.
50 J. Kano and F. Saito, Powder Technol., 1998, 98, 166–170.
51 S. Yariv and I. Lapides, J. Mater. Synth. Process., 2000, 8, 223–
233.
9 J. Shabtai, X. Xiao and W. Zmierczak, Energy Fuels, 1997, 11, 76–87.
10 R. W. Thring and J. Breau, Fuel, 1996, 75, 795–800.
11 T. Ohta, Appl. Energy, 2000, 67, 181–193.
12 A. Karpinska, C. Kajdas and K. i. Hiratsuka, Tribologia, 2006, 37,
21–37.
13 C. Kajdas and A. Karpinska, Tribologia, 2005, 36, 125–141.
14 Y. Momose, Toraiborojisuto, 2001, 46, 391–397.
15 M. Hasegawa, M. Kimata and I. Takahashi, Adv. Powder Technol.,
2007, 18, 541–554.
16 S. Ikeda, T. Takata, M. Komoda, M. Hara, J. N. Kondo, K. Domen,
A. Tanaka, H. Hosono and H. Kawazoe, Phys. Chem. Chem. Phys.,
1999, 1, 4485–4491.
52 S. M. Hick, C. Griebel and R. G. Blair, Inorg. Chem., 2009, 48,
2333–2338.
17 S. Ikeda, T. Takata, T. Kondo, G. Hitoki, M. Hara, J. N. Kondo, K.
Domen, H. Hosono, H. Kawazoe and A. Tanaka, Chem. Commun.,
1998, 2185–2186.
53 M. Sasaki, M. Furukawa, K. Minami, T. Adschiri and K. Arai, Ind.
Eng. Chem. Res., 2002, 41, 6642–6649.
54 R. Simha, J. Appl. Phys., 1941, 12, 569–578.
55 Y. H. P. Zhang and L. R. Lynd, Biomacromolecules, 2005, 6, 1510–
1515.
18 S. Murakami, M. Tabata and J. Sohma, J. Appl. Polym. Sci., 1984,
29, 291–298.
19 C. Kajdas, K. Hiratsuka and A. Borkowska, Tribologia, 2004, 35,
11–25.
20 T. Shirasaki, Shikizai Kyokaishi, 1972, 45, 737–743.
21 S. Murakami, M. Tabata, J. Sohma and M. Hatano, J. Appl. Polym.
Sci., 1984, 29, 3445–3455.
56 Y. H. P. Zhang and L. R. Lynd, Biotechnol. Bioeng., 2004, 88, 797–
824.
57 G. Janke, A. Frendel and G. Schmidt-Naake, Chem. Eng. Technol.,
1999, 22, 997–1000.
474 | Green Chem., 2010, 12, 468–474
This journal is
The Royal Society of Chemistry 2010
©