You are on page 1of 7

Bioresource Technology 102 (2011) 2897–2903

Contents lists available at ScienceDirect

Bioresource Technology
journal homepage: www.elsevier.com/locate/biortech

Calculating sugar yields in high solids hydrolysis of biomass


Yongming Zhu a,⇑, Marco Malten b,1, Mads Torry-Smith a, James D. McMillan c, Jonathan J. Stickel c
a
Novozymes North America, Franklinton, NC 27587, United States
b
Novozymes China, Beijing 100085, China
c
National Renewable Energy Laboratory (NREL), Golden, CO 80401, United States

a r t i c l e i n f o a b s t r a c t

Article history: Calculation of true sugar yields in high solids enzymatic hydrolysis of biomass is challenging due to the
Received 7 July 2010 varying liquid density and liquid volume resulting from solid solubilization. Ignoring these changes in
Received in revised form 25 October 2010 yield calculations can lead to significant errors. In this paper, a mathematical method was developed
Accepted 26 October 2010
for the estimation of liquid volume change and thereafter the sugar yield. The information needed in
Available online 3 November 2010
the calculations include the compositions of the substrate, initial solids loading, initial liquid density,
and sugar concentrations before and after hydrolysis. All of these variables are measurable with conven-
Keywords:
tional laboratory procedures. This method was validated experimentally for enzymatic hydrolysis of
High solids
Cellulose enzymatic hydrolysis
dilute sulfuric acid pretreated corn stover at solid loadings up to 23% (w/w). The maximum relative error
Sugar yield of predicted glucose yield from the true value was less than 4%. Compared to other methods reported in
Biomass the literature, this method is relatively easy to use and provides good accuracy.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction sugars and should be incorporated in the yield calculation. In this


case, the yield (of glucose) would be expressed as
As hydrolytic enzymes are continuously improved to perform
C g  C g0
well at high sugar concentrations, operating enzymatic hydrolysis Yg ¼ ð2Þ
reactors at high levels of lignocellulosic solids becomes a natural
uG C is0 xG0 þ ugos C gos0
choice for reducing capital and energy costs (Hodge et al., 2009; In Eqs. (1) and (2), it is assumed that (i) the density of the
Kristensen et al., 2009a). In order to design and develop an efficient hydrolysis slurry is 1000 g/L, (ii) the volume of the liquid equals
high solids hydrolysis process, it is necessary to accurately quantify to that of the hydrolysis slurry, and (iii) the volume of the liquid
the sugar yields, i.e., the production of soluble sugars from the remains unchanged during the hydrolysis. These equations could
insoluble polysaccharides. A commonly used laboratory practice give sufficient accuracy of yield prediction when the solid concen-
is to take slurry samples and measure the sugar concentrations tration is very low, e.g., less than 2% (w/w). However, as the solid
in the liquid phase of the hydrolyzate with high performance liquid concentration increases, the yield obtained from these equations
chromatography (HPLC). The sugar yields are then calculated based may deviate significantly from the actual value since none of these
on the sugar concentrations using the following formula (taking three assumptions remains valid.
glucose yield as an example): Several studies have emphasized the importance of using prop-
er methods for yield determination in high solids hydrolysis.
C g  C g0 Hodge et al. (2009) modeled a fed-batch process for high solids cel-
Yg ¼ ð1Þ
uG C is0 xG0 lulose hydrolysis where numerical computations were employed
to calculate glucose yield. The variations of liquid density and
where all variables are defined in the Nomenclature. The denomina- insoluble solids weight (or liquid mass) were taken into account
tor in Eq. (1) stands for the potential glucose input to the reactor and expressed as functions of glucose concentrations. Roche et al.
and the numerator is the production of glucose from enzymatic (2009) developed a method for calculating hydrolysis conversions
hydrolysis. at high solids, in which the sugar concentrations in the liquid (g/L)
If the substrate contains soluble sugar oligomers, these compo- are converted to mass fractions of the hydrolysis slurry to facilitate
nents can also be hydrolyzed by enzymes to form fermentable yield calculation. In another paper by Kristensen et al. (2009b), an
experimentally based method was proposed for yield determina-
⇑ Corresponding author. Tel.: +1 919 4943402; fax: +1 919 4943422. tion. In this method, the slurry sample was diluted extensively
E-mail address: yggz@novozymes.com (Y. Zhu). (e.g., 10 times) and the diluted sugar concentrations were
1
Present address: Novozymes A/S, 2880 Bagsværd, Denmark. measured by HPLC. Due to the significant dilution, the volume

0960-8524/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.biortech.2010.10.134
2898 Y. Zhu et al. / Bioresource Technology 102 (2011) 2897–2903

Nomenclature

ai amount of water used to form soluble sugar i from its fis0 initial mass fraction of insoluble solids in slurry
polysaccharide (grams per gram of sugar i) fts0 initial mass fraction of total solids (soluble sol-
qh density of hydrolyzate liquid at 25 °C (g/L) ids + insoluble solids) in slurry
qh,c calculated density of hydrolyzate liquid at 25 °C (g/L) Xis0 initial mass fraction of insoluble solids in total solids
qh0 initial density of liquid at 25 °C (g/L) xG0 initial mass fraction of glucan in insoluble solids
qh0,c calculated initial density of liquid at 25 °C (g/L) MW molecular weight (Da)
qw density of water at 25 °C (g/L) Vh volume of hydrolyzate liquid (L)
ug molecular weight ratio of glucose to glucan monomer; Vh0 initial volume of liquid (L)
uG = 180/162 = 1.11 Vw volume of water input to hydrolysis reactor (L)
ugos ratio of glucose molecular weight to average monomer V w volume of water before ‘‘pre-hydrolysis’’ (L)
weight of glucose oligomers; ugos = 180/166.5 = 1.08 Wh0 initial mass of liquid (g)
(assuming glucose oligomers have an average DP of 4) Wis0 initial mass of insoluble solids (g)
Cg concentration of glucose (g/L) Wt total weight of hydrolysis assay (g)
Cg0 initial concentration of glucose (g/L) Ww the weight of water that is input to hydrolysis reactor
Cgos0 initial concentration of glucose oligomers (g/L) (g)
Ci concentration of sugar i (g/L) Ww,h the weight of water in the hydrolyzate (g)
Cis0 initial concentration of insoluble solids (g/L) Ww,p the weight of water that is consumed to hydrolyze poly-
Ci0 initial concentration of sugar i (g/L) saccharides (g)
Cos cumulative concentration of soluble components other Yg glucose yield (of theoretical maximum)
than sugars (g/L)
Cos0 initial cumulative concentration of soluble components
other than sugars (g/L)

displacement by the solids in the diluted sample could be ignored and xylose mixtures over a wide range of concentrations. Since glu-
and the liquid density was very close to that of pure water. The su- cose and xylose are typically the most abundant sugars found in bio-
gar mass in the samples was thus obtained by multiplying the su- mass hydrolyzate, Eq. (3) will be used in the following calculations.
gar concentrations and the volume of the diluted sample. Since the
total weight of the hydrolysis assay was known, the theoretical 2.2. Estimating the change of liquid volume after hydrolysis
maximum mass of sugars that can be produced from hydrolysis
could then be calculated. This method could work to determine su- The change of liquid volume after hydrolysis can be estimated
gar yields, but requires taking good representative samples, which by tracking the amount of water in the process. Assume one col-
becomes increasingly difficult in high solids hydrolysis especially lects the liquid phase of hydrolyzate after hydrolysis of lignocellu-
in the early stage of the reaction. losic biomass. The volume of the liquid at 25 °C is Vh (in liter) and
In this paper, we report on an alternative, improved method for the density qh (in g/L). The hydrolyzate liquid sample is analyzed
estimating sugar yields in high solids enzymatic hydrolysis of cellu- by HPLC for sugars (usually cellobiose, glucose, xylose, galactose,
lose. This method uses sugar concentrations determined by HPLC to arabinose, mannose, and possibly fructose). The cumulative con-
calculate the change of liquid volume after enzymatic hydrolysis. In centration of the other soluble components, which is not usually
case the biomass substrate contains significant amounts of soluble measured, is denoted as Cos (in g/L). The other soluble components
components other than HPLC-detectable sugars, the measured ini- may comprise soluble salts, soluble lignin, sugar oligomers, acetic
tial liquid density is incorporated to reduce the error. Experiments acid, and so forth.
using dilute acid pretreated corn stover as substrates (water washed
or unwashed) showed that this method could accurately predict the
yield values at relatively high solids concentrations.

2. Theoretical analysis

2.1. Correlation of hydrolyzate liquid density (qh) with sugar


concentrations

Plenty of data plotting the density of pure glucose and xylose


solutions against sugar concentrations was reported by Ji et al.
(2007). Correlation of the densities (in g/L) of pure glucose and
xylose solutions ranging from 0 to 240 g/L showed that the densities
increased linearly with the increase of sugar concentrations. Further
examination in our laboratory showed that a linear relationship also
exists for glucose and xylose mixtures. A general correlation of the
density and sugar concentrations can be described as:
X
qh ¼ qw þ 0:35 Ci ð3Þ

where qw = 1000 g/L (25 °C). Eq. (3) implies that the contribution of Fig. 1. Predicted density using Eq. (3) vs. experimentally measured density of
sugar concentrations to the liquid density is additive. Fig. 1 shows solutions comprising glucose and xylose. The numbers in the parenthesis represent
that Eq. (3) can provide good prediction of the densities of glucose glucose (g/L) + xylose (g/L).
Y. Zhu et al. / Bioresource Technology 102 (2011) 2897–2903 2899

Based on the above information, the weight of water in the Substituting Eqs. (11) and (12) in Eq. (10) yields:
hydrolyzate liquid (Ww,h, in grams) can be calculated as follows: P
 Vh qh0;c  ð1  ai ÞC i0
X  ¼ P ð13Þ
W w;h ¼ qh V h  C os þ Ci V h V h0 qh;c  ð1  ai ÞC i
Eq. (13) demonstrates that the ratio of liquid volumes before
or
 and after enzymatic hydrolysis can be estimated using sugar
X 
W w;h ¼ qh  C os  Ci V h ð4Þ concentrations measured by HPLC. In some cases, water is also
consumed during hydrolysis of acetyl groups from hemicellulose.
During the hydrolysis, some water was consumed when the For example, in autohydrolysis pretreatment, a significant amount
polysaccharides were cleaved by enzymes to form shorter chains of xylo-oligomers is present in the liquid phase. When the whole
and soluble sugars. For example, to produce 1 g of glucose from slurry is subject to enzymatic hydrolysis, some of the acetyl groups
glucan, 0.1 g of water was used (MW H2 O =MW glucose ¼ 18=180 ¼ in the xylo-oligomers can be hydrolyzed. However, the liquid vol-
0:1). If ai denotes the amount of water used to form soluble ume reduction due to the water consumption by the acetyl groups
sugar i, one will have acellobiose = 0.05, aglucose = agalactose = amannose = is less than 1% of the total liquid volume (data not shown), and is
0.1, and axylose = aarabinose = 0.12. thus ignored in the calculation.
If the liquid contained no soluble sugars before hydrolysis, the
weight of water (g) that was consumed to hydrolyze polysaccha- 2.3. Calculation of sugar yield
rides is
X In most enzymatic hydrolysis research on lignocellulosic bio-
W w;p ¼ ai C i V h ð5Þ mass, sugar yields are simply reported as the amount of ferment-
Based on the above analysis, the weight of water that was input able sugar produced from insoluble polysaccharide, as described
to the hydrolysis reactor is: by Eq. (1). This approach is reasonably accurate when the substrate
h X i contains no or little soluble oligomers before hydrolysis (as is the
W w ¼ W w;h þ W w;p ¼ ðqh  C os Þ  ð1  ai ÞC i V h ð6Þ case using water washed substrates). However, under certain cir-
cumstances, considerable sugar oligomers may be found in the
Because W w ¼ qw V w , Eq. (6) can be rewritten as substrate (for example, when unwashed steam exploded corn
stover is used for hydrolysis), and the oligomers can also be hydro-
Vh qw
¼ P ð7Þ lyzed by the enzymes to produce fermentable sugars. As such, the
V w ðqh  C os Þ  ð1  ai ÞC i
initial sugar oligomers should be counted as a source of ferment-
Eq. (7) applies to the situation where the liquid before enzy- able sugar production. Take the calculation of glucose yield for
matic hydrolysis contains no soluble sugars. In the case that the li- example, the true value of which can be represented as
quid before hydrolysis contains soluble sugars [as is true when the C g V h  C g0 V h0
whole slurry from pretreatment is subject to enzymatic hydrolysis Yg ¼
uG W t fts0 X is0 xG0 þ ugos C gos0 V h0
(Schell et al., 1999; Kova et al., 2009)], the relationship of hydroly-
zate liquor volume to water volume is more complex. In this case, or
let us assume that the sugars initially present in the hydrolyzate li-  
quid have been produced through an additional, hypothetical, C g VV h  C g0
h0
Yg ¼   ð14Þ
‘‘pre-hydrolysis’’ step in which no soluble sugars were present at
uG VW t fts0 X is0 xG0 þ ugos C gos0
the onset of pre-hydrolysis. If V w is used to denote the volume of h0

water input (liters) before this hypothetical ‘‘pre-hydrolysis’’ step In Eq. (14), the term Wt
can be evolved as below:
V h0
and Vh0 is used to denote the volume of hydrolyzate following
pre-hydrolysis, from Eq. (7) it follows that Wt Wt
¼ 
V h0 W h0
V h0 qw qh0
¼ P ð8Þ  
V w ðqh0  C os0 Þ  ð1  ai ÞC i0 Wt
¼ qh0
W h0
and  
Wt
Vh qw ¼ qh0
¼ P ð9Þ W t  W is0
V w ðqh  C os Þ  ð1  ai ÞC i !
1
¼ qh0
Dividing Eq. (9) by Eq. (8) gives 1  WWis0t
P
Vh ðq  C os0 Þ  ð1  ai ÞC i0
¼ h0 P ð10Þ or
V h0 ðqh  C os Þ  ð1  ai ÞC i
Wt qh0
It is noted that in Eq. (10), qh0 and qh are actual densities of the ¼ ð15Þ
V h0 1  fts0 X is0
liquid before and after hydrolysis. As mentioned earlier, since Cos is
usually not measured, the following assumptions are used to sim- Insertion of Eq. (15) to Eq. (14) gives:
plify the calculation:  
C g VV h  C g0
qh0  C os0  qh0;c ð11Þ Yg ¼   h0
qh0
uG 1fts0 Xis0 fts0 X is0 xG0 þ ugos C gos0
qh  C os  qh;c ð12Þ
or
where the subscript c means that the value of density was calcu- h   i 
Vh 1
lated using Eq. (3). Eqs. (11) and (12) apply when ‘‘the other soluble Cg V h0
 C g0 fts0 X is0
1
Yg ¼   ð16Þ
components’’ do not significantly contribute to the liquid density 1
uG qh0 xG0 þ ugos C gos0 fts0 X is0
1
and therefore can be neglected.
2900 Y. Zhu et al. / Bioresource Technology 102 (2011) 2897–2903

Table 1
Composition of corn stover pretreated by dilute sulfuric acid.a

fts0b Xis0c Insoluble solids (of dry weight)d Pretreatment liquid (g/L)e
Glucan Xylan Galactan Arabinan Lignin Ash Others Cellobiose Glucose Xylose Arabinose
Batch 1 0.328 ± 0.005 0.563 ± 0.001 52.9 ± 0.8% 2.4 ± 0.2% 0.0 ± 0.0% 0.2 ± 0.1% 31.8 ± 0.5% 8.0 ± 1.6% 4.7% 2.7 ± 0.1 15.8 ± 0.1 54.5 ± 0.5 10.7 ± 0.1
(18.5 ± 0.2) (62.5 ± 0.9) (10.1 ± 0.4)
Batch 2 0.305 ± 0.003 0.606 ± 0.002 55.7 ± 0.4% 3.7 ± 0.2% 0.9 ± 0.1% 0.7 ± 0.0% 27.2 ± 0.3% 6.5 ± 0.6% 5.2% 4.9 ± 0.5 14.1 ± 0.9 55.3 ± 1.8 10.2 ± 0.1
(19.4 ± 0.5) (71.2 ± 1.8) (10.7 ± 0.2)
a
Value are expressed as averages and standard deviations (n = 2 for fts0 and Xis0, n = 3 for compositional analysis). Pretreatment conditions: 1.1% (w/w) H2SO4, 190 °C, 1 min.
b
Initial mass fraction of total solids in slurry.
c
Initial mass fraction of insoluble solids in total solids.
d
The composition of insoluble solids are based on oven-dry weight.
e
The values in the parentheses are concentrations of total sugars (monomer + dimer + oligomers) and are expressed as of equivalent monomers.

If the sugar of interest is other than glucose, one needs to Protocols (Sluiter et al., 2008a,b). The fts0 were measured using
replace the concentrations and mass fractions in Eq. (16) with an IR-200 moisture analyzer (Denver Instrument Company, Den-
those of the corresponding sugar and adjust the u coefficients ver, CO). The sample was heated at 105 °C until less than 0.05%
accordingly. of the initial weight was lost within 1 min. The fts0 was calculated
by dividing the final dry weight by the initial weight of the sample.
3. Methods For fis0 measurement, a weighed slurry sample was washed on a
Whatman GF/D glass microfiber membrane until pH 5. The washed
3.1. Hydrolysis substrates solid cake was then dried in a 105 °C oven. The fis0 level was deter-
mined by dividing the dry weight of the washed solid cake by the
The substrates used in this work were dilute sulfuric acid pre- weight of the original slurry sample. The Xis0 was then calculated
treated corn stover generated from the NREL’s pilot demonstration from the fts0 and fis0, that is, X is0 ¼ ffts0
is0
. All fts0 and fis0 measurements
unit (Schell et al., 2003). The conditions for pretreatment were: were made in duplicates (n = 2). The densities of the liquid before
1.1% (w/w) sulfuric acid, 190 °C, 1 min. Two different batches were enzymatic hydrolysis and the hydrolyzate liquid were measured
used. The compositions of the pretreated corn stover are shown in by taking 200 lL liquid samples with a HandyStepÒ repeating pip-
Table 1. It is noted that dilute acid pretreated corn stover slurry ette (BrandTech Scientific, Essex, CT) and weighing the samples on
contains substantial amount of sugars in the liquid (Table 1). Be- an analytical balance (accuracy 0.1 mg). The density of the liquid
fore loading to hydrolysis reactors, Batch 1 material was washed was obtained from the ratio of the sample weight to sample
extensively with de-ionized water on a Whatman GF/D microfiber volume.
membrane until no detectable sugars were found in the washout To prepare HPLC samples, 1 g of slurry was extracted from each
(HPLC detection limit 0.025 g/L for glucose), while Batch 2 material hydrolysis tube with a cut-off pipette tip after the tube was vor-
was used as received. texed. The weight of the remaining content in each sampled assay
was weighed for mass balance closure. The samples were acidified
3.2. Enzymatic hydrolysis to pH < 2 with 10 lL of 40% (w/w) H2SO4 (1% (v/w) of slurry sam-
ple) to stop the hydrolysis reaction. The hydrolyzate was filtered
The enzymatic hydrolysis was carried out in 50 ml Oak Ridge through 0.45 lm syringe membranes and diluted with 5 mM
polypropylene copolymer centrifuge tubes (29  107 mm) with H2SO4 to prepare for HPLC analysis. An Agilent HPLC system
caps and silicone sealing rings (LabSource Inc., Romeoville, IL). equipped with Aminex HPX-87H column (Bio-Rad Laboratories,
The working weight of each assay was 20.0 g. All the hydrolysis Hercules, CA) running at 0.6 mL/min of 5 mM H2SO4 (pH1.8)
tubes were incubated in a rotisserie incubator (Model Combi- was used for sugar quantification by integration of signal from a
D24, Finemould Precision Industrial Company, Seoul, Korea) set refractive index detector. Xylose, mannose and galactose had the
at 50 °C and 12 rpm. All the assays were run in triplicate, and the same retention time on the column and co-eluted (Um et al.,
average values and standard deviations are reported. 2003). As minor sugars in corn stover substrate, mannose and gal-
To load the hydrolysis tubes, pre-determined amounts of sub- actose were combined with and counted as xylose in the yield cal-
strates, de-ionized water, 10% (w/w) NaOH (for neutralization of culation. Carbohydrate analysis of the substrates was performed
the acid in the unwashed substrate), sodium citrate buffer following the NREL Standard Analytical Protocols (Sluiter et al.,
(0.05 M in hydrolysis) and penicillin (antibacterial, 2.5 ppm in 2008c). An Aminex HPX-87P (Pb2+ form cation) column running
hydrolysis) were added. The content was mixed with a stainless at 0.6 mL/min of Milli-Q water was used for sugar quantification
steel spatula and settled in a 5 °C refrigerator overnight. Then cel- on an Agilent HPLC system.
lulase enzyme was added to the tubes to start the hydrolysis and
the tubes were placed in the incubator. Since the high solid slurry 3.4. Quantification of actual sugar yields
was very viscous, a 1 cm-diameter stainless steel ball was added to
each tube to improve the mixing during hydrolysis. The pH of the After sampling, the remaining slurry in each assay was acidified
assays was 5 before incubation. The enzymes used in the study with 0.19 ml (1% (v/w) of the remaining slurry) of 40% (w/w) H2SO4
were Novozymes CellicÒ CTec cellulase produced from Trichoderma to stop the hydrolysis reaction. The acidified assays were then
reesei. The hydrolysis proceeded 2 or 4 days. Additional ‘‘dummy’’ washed five times each using either 60 mL (for washed substrates)
tubes were constituted with the same contents as the hydrolysis or 40 mL (for unwashed substrates) of de-ionized water through a
tubes and were sacrificed at time zero to get the initial data points. Whatman GF/D glass microfiber membrane. The reason for using
different water volumes was that the washed substrates contained
3.3. Analysis more insoluble solids than the unwashed substrates and thus re-
quired more water to wash out the sugars. In addition, the volumes
The fts0, fis0, and Xis0 of the dilute acid pretreated corn stover of the water were pre-determined before washing so that the con-
slurry were determined following the NREL Standard Analytical centrations of glucose in the washout fell in the HPLC linear range
Y. Zhu et al. / Bioresource Technology 102 (2011) 2897–2903 2901

Table 2
Comparison of glucose yields (of theoretical maximum) obtained using different calculation methods.a
b
Substrate fts0 Xis0c Enzyme Hydrolysis Actual yield Yield Relative Yield using Relative Yield using Relative Yield using. Relative
dosage time using error (%) Eqs. (3), error (%) Eqs. (3), (13), error (%) Eq. (16), error (%)
(g/g-glucan) (hours) Eq. (1) (13), and (16) and measured
or (2) (16) measured qh0 but
qh0 ignoring
volume
change
Diluted acid 0.15 1.00 0.04 96 62.7 ± 0.3% 73.5 ± 0.7% 17.2 64.8 ± 0.7% 3.3 – – 62.2 ± 0.6% 0.9
pretreated 0.15 1.00 0.12 96 83.1 ± 0.6% 96.1 ± 0.8% 15.6 85.5 ± 0.8% 2.9 – – 81.3 ± 0.6% 2.2
corn stover 0.23 1.00 0.04 96 44.0 ± 1.2% 56.8 ± 1.8% 29.1 45.7 ± 0.7% 3.9 – – 43.6 ± 0.6% 1.1
(washed) 0.23 1.00 0.12 96 72.7 ± 0.1% 90.3 ± 0.6% 24.1 74.6 ± 0.5% 2.5 – – 69.2 ± 0.4% 4.9
Dilute acid 0.15 0.606 0.06 48 59.3 ± 0.3% 65.1 ± 0.4% 9.8 60.2 ± 0.4% 1.6 57.9 ± 0.3% 2.3 56.4 ± 0.3% 4.8
pretreated 0.15 0.606 0.18 48 82.0 ± 0.5% 89.9 ± 0.2% 9.5 83.7 ± 0.2% 2.1 80.4 ± 0.2% 1.8 77.8 ± 0.2% 5.2
corn stover 0.23 0.606 0.06 48 44.4 ± 0.2% 52.5 ± 0.1% 18.2 45.9 ± 0.4% 3.4 43.3 ± 0.1% 2.4 42.0 ± 0.1% 5.4
(unwashed) 0.23 0.606 0.18 48 64.8 ± 0.1% 76.3 ± 0.1% 17.5 67.7 ± 0.2% 4.4 64.1 ± 0.2% 1.0 61.2 ± 0.1% 5.6
a
Hydrolysis was done in triplicates (n = 3). Yield values are expressed as averages and standard deviations;
b
Initial mass fraction of total solids in slurry. fts0 in this table was reported as the ratio of dry weight of pretreated corn stover and total working weight of the assay. The
other soluble components introduced in assay preparation (sodium citrate, penicillin and NaOH) were not included.
c
Initial mass fraction of insoluble solids in total solids. Xis0 of the assays using unwashed corn stover was taken as the same as the pretreated slurry.

(0.5–15 g/L). The use of the excessive amount of water for washing initial liquid density (Eq. (16)). Table 3 summarizes the liquid den-
ensured a nearly complete removal of the soluble components sities predicted using Eq. (3) and measured experimentally. It is
from the remaining insoluble solids, as no sugars were detected seen that when washed substrate was used for hydrolysis, the li-
by the HPLC in the last droplets of the filtrates. After washing, quid densities after hydrolysis could be accurately predicted with
the volumes (mL) of the collected washout were measured and HPLC data, which suggested that only small quantity, if any, of
the sugar concentrations (in g/L) were analyzed by HPLC equipped non-sugar substances was solubilized during hydrolysis. However,
with Aminex HPX-87H column (Bio-Rad Laboratories, Hercules, the prediction of liquid density was much less accurate for un-
CA) running at 0.6 mL/min of 5 mM H2SO4 (pH1.8). The soluble washed substrate apparently because of the presence of a variety
sugars recovered from each hydrolysis assay were then quantified of non-sugar components in the liquid. The inaccuracy of density
taking into account of the losses of mass by sampling before wash- estimation could largely explain the increasing errors in the yield
ing. Due to the small sample size (1 g of slurry sample vs. 20 g of prediction of unwashed substrate hydrolysis when using Eqs. (3),
total) and vortex homogenization before sampling, the change in (13), and (16). In order to improve the accuracy of prediction, a
insoluble solids concentration of the hydrolysis slurry, if any, was modified method is proposed here. In this modification, the liquid
negligible after sample withdrawal.2 volume ratio, VV h , was still calculated based on HPLC data using
h0
Eqs. (3) and (13), but in Eq. (16), the measured initial density of
4. Results and discussion liquid was assigned for qh0. This modification further reduced the
error of the yield prediction particularly at the higher solids con-
Table 2 compares the glucose yields calculated using Eqs. (1) centration of unwashed substrate. As can be seen in Table 2, the
and (2) with those using the methods developed in this study. relative error of calculated yield of unwashed corn stover hydroly-
Two different levels of total solids (fts0 = 0.15 and 0.23) were em- sis at fts0 of 0.23 declined from 3.4–4.4% to 1.0–2.4%. Measuring the
ployed in the hydrolysis of both washed and unwashed substrates. density of initial liquid is a simple operation in the laboratory. One
It is noted that the fts0 of the washed substrate equals to the fis0 (or just needs to pipette a small volume of the liquid sample (the same
Xis0 = 1) since all the soluble components were removed during as for HPLC sample preparation) and weigh the sample on a bal-
washing. However, the fts0 of unwashed substrate were higher than ance. The density of the liquid can thus be calculated by dividing
the fis0. The fraction of insoluble solids of the total solids, Xis0, in un- the sample weight by the volume.
washed substrate (Batch 2) was 0.606. From Table 2, it can be seen As can be seen in Eq. (16), liquid volume ratio before and after
that ignoring the liquid density and volume changes in yield hydrolysis (VV h ) plays an important role in the yield calculation. For
h0

calculation (using Eqs. (1) and (2)) led to significant overestima- the hydrolysis in our experiment, the liquid volumes were esti-
tions of the yields (relative error 15.6–29.1% for washed substrate mated to increase by 2–8% (Table 3). Ignoring the volume changes,
and 9.5–18.2% for unwashed substrate), the error generally that is, assuming VV h ¼ 1, would lead to considerable underestima-
h0

increasing with solids loading. However, the relative errors tions of the yields (Table 2). Roche et al. (2009) account for the
dramatically dropped after using the equations from this study change in the mass of insoluble solids, and conversely the liquid
(Eqs. (3), (13), and (16)), decreasing to only 2–4% and 1–5% for mass, during enzymatic hydrolysis of washed substrate. However,
washed and unwashed substrates, respectively. Kristensen et al. there was an error in their equations that is important for the case
(2009b) also reported an overestimation of 23% in the yield calcu- of unwashed substrates (see Appendix). After correcting for the
lation when using an equation similar to Eq. (1) for enzymatic error, and with appropriate substitutions and rearrangements,
hydrolysis of pretreated straw (washed). the same equation depicting the liquid volume change can be
Proper estimation of liquid density is important for the predic- successfully reached (neglecting soluble components other than
tion of sugar yields, as a direct relation exists for the yield and the HPLC-detectable sugars).

2
For example, 1 g of slurry sample is taken from 20 g assay containing 23% 5. Conclusions
insoluble solids. Because of insufficient vortexing before sampling, this sample
contains 25% of insoluble solids, which is 8.7% higher than that of the assay. However,
this imperfect sample withdrawal results in only 0.11% of absolute drop, or 0.46% of An improved, mathematically-based method was developed for
relative decrease, in the insoluble solids concentration in the bulk slurry. calculating sugar yields in high solids hydrolysis of lignocellulosic
2902 Y. Zhu et al. / Bioresource Technology 102 (2011) 2897–2903

Table 3
Liquid density (g/L) and volume change in enzymatic hydrolysis of dilute acid pretreated corn stover.

Substrate fts0a Xis0b Enzyme dosage Hydrolysis Liquid density before hydrolysis Liquid density after hydrolysis Estimated liquid
(g/g-glucan) time (h) c c volume ratio before
Measured Calculated Relative Measured Calculated Relative
and after hydrolysis
with Eq. (3)c error (%) with Eq. (3)c error (%) Vh
V h0 using Eq. (13)

Dilute acid 0.15 1.00 0.04 96 995 ± 1 – – 1016 ± 1 1032 ± 1 1.5 1.042
pretreated 0.15 1.00 0.12 96 995 ± 0 – – 1026 ± 1 1032 ± 1 0.5 1.052
corn stover 0.23 1.00 0.04 96 998 ± 1 – – 1028 ± 2 1030 ± 1 0.2 1.050
(washed) 0.23 1.00 0.12 96 998 ± 0 – – 1044 ± 0 1030 ± 0 1.3 1.078
Dilute acid 0.15 0.606 0.06 48 1056 ± 1 1014 ± 0 4.0 1089 ± 2 1028 ± 0 5.6 1.023
pretreated 0.15 0.606 0.18 48 1056 ± 0 1015 ± 0 3.9 1084 ± 2 1034 ± 0 4.7 1.031
corn stover 0.23 0.606 0.06 48 1086 ± 1 1023 ± 0 5.8 1101 ± 1 1040 ± 1 5.6 1.027
(unwashed) 0.23 0.606 0.18 48 1082 ± 2 1024 ± 0 5.4 1111 ± 2 1048 ± 1 5.6 1.042
a
Initial mass fraction of total solids in slurry. fts0 in this table was reported as the ratio of dry weight of pretreated corn stover and total working weight of the assay. The
other soluble components introduced in assay preparation (sodium citrate, penicillin and NaOH) were not included.
b
Initial mass fraction of insoluble solids in total solids. Xis0 of the assays using unwashed corn stover was taken as the same as the pretreated slurry.
c
The values are expressed as averages and standard deviations (n = 3).

biomass. With this method, the sugar yields can be directly calcu- After making the correction noted above, the equation that re-
lated from the sugar concentrations that are usually determined by lates the fraction of insoluble solids to the soluble sugar concentra-
HPLC. This method was experimentally validated for relatively tions (Eq. (5) of Hodge et al. (2009) and Eq. (4) of Roche et al.
high solids hydrolysis of dilute acid pretreated corn stover (washed (2009)) can also be corrected to be:
and unwashed). At 23% of solids loading, the maximum relative er- P P
fis0 ð1  r i C i0 =qh0 Þ  r i ðC i =qh  C i0 =qh0 Þ
ror of predicted glucose yield from the true value was less than 4%. fis ¼ P ðA:3Þ
When used for other substrates, this method may need further val- 1  r i C i =qh
idations due to the variation of biomass materials. where r i is the molecular weight (MW) ratio of an polysaccharide
monomer to its soluble monomer i. Since ai is defined as amount
Acknowledgements of water used to form soluble sugar i from its polysaccharide (grams
per gram of sugar i), it is easy to see that r i ¼ 1  ai .
We thank Lindsay Jones for laboratory assistance and Prashant By putting Eq. (A.3) in terms of the mass fraction of hydrolyzate
Iyer and Hui Xu for their discussions in this work. liquor (fh = 1  fis) we arrive at the more compact expression
P
fh0 ð1  r i C i0 =qh0 Þ
fh ¼ P ðA:4Þ
Appendix 1  r i C i =qh
By making the further substitution of fh = qhVh/Wt and some
Rather than tracking the change in the volume of hydrolyzate
rearrangement, we can arrive at
liquid, Hodge et al. (2009) and Roche et al. (2009) accounted for P
the changes in the solid and liquid phases by deriving equations Vh q  ri C i0
¼ h0 P
in terms of the mass fraction of insoluble solids. However, their V h0 qh  r i C i
equations relating the change in mass fraction of a soluble species
or
to the change in its concentration in the liquid phase (Eqs. (3) and
P
(4) of Hodge et al. (2009) and Eq. (2) of Roche et al. (2009)) only Vh q  ð1  ai ÞC i0
apply to the case where no soluble sugars are present at the start ¼ h0 P ðA:5Þ
V h0 qh  ð1  ai ÞC i
of enzymatic hydrolysis (e.g., washed solids) and were mistakenly
represented to also apply to the case when soluble sugars are pres- which we recognize to be exactly Eq. (10) when the presence of sol-
ent (e.g., whole slurry). The incorrect equation was given as: uble components other than HPLC-detectable sugars are neglected.

C i  C i0 References
fi  fi0 ¼ ð1  fis Þ ðA:1Þ
qh
Hodge, D.B., Karim, M.N., Schell, D.J., McMillan, J.D., 2009. Model-based fed-batch
Here, fi denotes the mass fraction of a soluble sugar species as a for high-solids enzymatic cellulose hydrolysis. Appl. Biochem. Biotechnol. 152,
part of the total slurry, and fis denotes the mass fraction of insolu- 88–107.
ble solids as a part of the total slurry. The correct equation for the Ji, P., Feng, W., Tan, T.W., 2007. Density calculation of sugar solutions with the SAFT
model. J. Chem. Eng. Data 52, 135–140.
more general case where soluble sugars are present at the onset of Kova, K., Szaka, G., Zacchi, G., 2009. Enzymatic hydrolysis and simultaneous
enzymatic hydrolysis is: saccharification and fermentation of steam-pretreated spruce using crude
Trichoderma reesei and Trichoderma atroviride enzymes. Process. Biochem. 44,
Ci C i0 1223–1329.
fi  fi0 ¼ ð1  fis Þ  ð1  fis0 Þ ðA:2Þ Kristensen, J.B., Felby, C., Jørgensen, H., 2009a. Yield-determining factors in high-
qh qh0 solids enzymatic hydrolysis of lignocellulose. Biotechnology for Fuels 2(11).
Online <http://www.biotechnologyforbiofuels.com/content/2/1/11>.
As can be observed, the mistake was to multiply the time zero con- Kristensen, J.B., Felby, C., Jorgensen, H., 2009b. Determining yields in high solids
enzymatic hydrolysis of biomass. Appl. Biochem. Biotechnol. 156, 557–562.
centration by the later time insoluble solids and density terms. It is
Roche, C.M., Dibble, C.J., Knutsen, J.S., Stickel, J.J., Liberatore, M.W., 2009. Particle
unfortunate that this error was overlooked in more than one publi- concentration and yield stress of biomass slurries during enzymatic hydrolysis
cation. Fortunately, this discrepancy is not important for studies of at high-solids loadings. Biotech. Bioeng. 104 (2), 290–300.
washed solids such as reported by Hodge et al. (2009) and Roche Schell, D.J., Farmer, J., Newman, M., McMillan, J.D., 2003. Dilute-sulfuric acid
pretreatment of corn stover in pilot-scale reactor: investigation of yields,
et al. (2009) in which the initial soluble sugar concentrations (Ci0) kinetics, and enzymatic digestibilities of solids. Appl. Biochem. Biotechnol. 105–
are all zero. 108, 69–86.
Y. Zhu et al. / Bioresource Technology 102 (2011) 2897–2903 2903

Schell, D.A., Ruth, M.F., Tucker, M.P., 1999. Modeling the enzymatic hydrolysis of Procedure (LAP). Issue Date, 3/31/2008. National Renewable Energy
dilute-acid pretreated Douglas Fir. Appl. Biochem. Biotechnol. 77–79, 67–81. Laboratory, Golden, CO, USA.
Sluiter, A., Hames, B., Hyman, D., Payne, C., Ruiz, R., Scarlata, C., Sluiter, J., Sluiter, A., Hames, B., Ruiz, R., Scarlata, C., Sluiter, J., Templeton, D., Crocker, D.,
Templeton, D., Wolfe, J., 2008a. Determination of total solids in biomass and 2008c. Determination of structural carbohydrates and lignin in biomass –
total dissolved solids in liquid process samples – Laboratory Analytical Laboratory Analytical Procedure (LAP). Issue Date, 4/25/2008. National
Procedure (LAP). Issue date, 3/31/2008. National Renewable Energy Renewable Energy Laboratory, Golden, CO, USA.
Laboratory, Golden, CO, USA. Um, B.-H., Karim, M.N., Henk, L.L., 2003. Effect of sulfuric and phosphoric acid
Sluiter, A., Hames, B., Hyman, D., Payne, C., Wolfe, J., 2008b. Determination of pretreatments on enzymatic hydrolysis of corn stover. Appl. Biochem.
insoluble solids in pretreated biomass material – Laboratory Analytical Biotechnol. 105–108, 115–125.

You might also like