2013 Scottetal Biotechnol Biofuels 6171
2013 Scottetal Biotechnol Biofuels 6171
net/publication/259001820
CITATIONS READS
14 246
3 authors, including:
Some of the authors of this publication are also working on these related projects:
FONDECYT 1160220. Improvement of Methane Bio-oxidation in Fungal Biofilters Inoculated with Methanotrophic Bacteria View project
Fondecyt 1181040. TREATMENT OF INDUSTRIAL GASEOUS EMISSIONS CAUSING ODOUR NUISANCE USING A SERIES OF BIOTRICKLING FILTERS View project
All content following this page was uploaded by German E Aroca on 30 December 2013.
Abstract
Background: Despite its semi-commercial status, ethanol production from lignocellulosics presents many
complexities not yet fully solved. Since the pretreatment stage has been recognized as a complex and
yield-determining step, it has been extensively studied. However, economic success of the production process
also requires optimization of the biochemical conversion stage. This work addresses the search of bioreactor
configurations with improved residence times for continuous enzymatic saccharification and fermentation
operations. Instead of analyzing each possible configuration through simulation, we apply graphical methods to
optimize the residence time of reactor networks composed of steady-state reactors. Although this can be easily
made for processes described by a single kinetic expression, reactions under analysis do not exhibit this feature.
Hence, the attainable region method, able to handle multiple species and its reactions, was applied for continuous
reactors. Additionally, the effects of the sugars contained in the pretreatment liquor over the enzymatic hydrolysis
and simultaneous saccharification and fermentation (SSF) were assessed.
Results: We obtained candidate attainable regions for separate enzymatic hydrolysis and fermentation (SHF) and
SSF operations, both fed with pretreated corn stover. Results show that, despite the complexity of the reaction
networks and underlying kinetics, the reactor networks that minimize the residence time can be constructed by
using plug flow reactors and continuous stirred tank reactors. Regarding the effect of soluble solids in the feed
stream to the reactor network, for SHF higher glucose concentration and yield are achieved for enzymatic
hydrolysis with washed solids. Similarly, for SSF, higher yields and bioethanol titers are obtained using this substrate.
Conclusions: In this work, we demonstrated the capabilities of the attainable region analysis as a tool to assess the
optimal reactor network with minimum residence time applied to the SHF and SSF operations for lignocellulosic
ethanol production. The methodology can be readily modified to evaluate other kinetic models of different
substrates, enzymes and microorganisms when available. From the obtained results, the most suitable reactor
configuration considering residence time and rheological aspects is a continuous stirred tank reactor followed by a
plug flow reactor (both in SSF mode) using washed solids as substrate.
© 2013 Scott et al.; licensee BioMed Central Ltd. This is an open access article distributed under the terms of the Creative
Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly cited.
Scott et al. Biotechnology for Biofuels 2013, 6:171 Page 2 of 16
http://www.biotechnologyforbiofuels.com/content/6/1/171
recovery) has been extensively used in Brazil [6]. For mod- arrangements in the hydrolysis and fermentation areas
ern bioethanol production plants, the working volume of for the production of bioethanol from lignocellulosic
bioreactors is on the order of thousands of cubic meter. As materials?
an example, a total of 20 bioreactors, with a working vol- Since the conventional use of graphical methods for
ume of 3000 m3 each, were constructed in the Shandong residence time minimization of a reactor network is no
province, China in 2003 [1]. For such large facilities, longer applicable to the system under study due to its
batch bioreactors are unattractive because of the lon- high number of reactions, we focus on more general
ger operational downtimes associated with mash adding, optimization methodologies. Optimization of reacting
broth harvesting and facility cleaning [1]. Continuous PFR systems involves solving the following reactor network
conditions are difficult to achieve in a fermentation synthesis (RNS) problem as stated by Biegler et al. [11]:
process due to its extended residence time and gas pro- “Given the reaction stoichiometry and rate laws, initial
duction, which induce mixing. In fact, residence time can feeds, a desired objective, and system constraints, what is
be as long as 48 to 72 hours to achieve an ethanol concen- the optimal reactor network structure? In particular: (i)
tration of 10 to 12% [7]. Since a cascade of continuous What is the flow pattern of this network? (ii) Where
stirred tank reactors (CSTR) also contributes reducing should mixing occur in this network? (iii) Where should
end-product inhibition, this strategy has been practiced in heating and cooling be applied in this network?” Ques-
the bioethanol industry [8]. Generally, a train of four to tion (i) addresses the mixing patterns of the reactors in
six CSTR connected in series are preferred because such the reactor network. In idealized reactors, two extremes
design presents an adequate trade-off between the glucose exist: no axial dispersion inside the reactor (PFR) and
fermentation kinetics and the capital investments for tank full axial dispersion (CSTR) [5]. Question (ii) inquires
manufacture [1]. This widely known use of a cascade of about which reactors in the network should be fed with
CSTRs as a way to minimize the residence time of the sys- fresh feed (F) and which reactors should be fed with a
tem is theoretically valid only for processes with a fixed mixture of intermediate product streams. Finally, (iii)
overall reaction stoichiometry, and that can be described refers to the heat supply or withdrawal in the network,
by a single kinetic expression. Although this may hold for e.g. to improve selectivity by increasing the rate of certain
ethanol fermentation kinetics [8], for enzymatic saccharifi- reactions over the rest of the reactions in the reaction
cation and simultaneous saccharification and fermentation network.
operations in lignocellulosic ethanol production, the reac- The problem of RNS can be addressed by an approach
tion network cannot be reduced to a single kinetic expres- based in mathematical optimization of a reactor network
sion. Hence, the classic graphical methods for residence superstructure or by graphical methods. Optimization
time optimization of continuous bioreactors are no longer based approaches start by proposing a reactor super-
applicable. structure where all the possible reactors, mixing streams
Bioethanol production from lignocellulosic substrates and heat streams are included. Then, optimal candidates
comprises a pretreatment of the feedstock to increase its are determined by searching in this superstructure. The
reactivity to further enzymatic degradation [9]. These first attempt using this strategy considered axial dispersion
biocatalysts break the structure of cellulose and hemicel- models and recycle PFRs [12] and the resulting candidate
lulose, producing sugar monomers and oligomers, which structures were found using nonlinear programming.
are subsequently fermented to bioethanol. Even at high Later, the concept of modeling the superstructure as a
solid concentration in the enzymatic hydrolysis step, glu- mixed integer nonlinear programming (MINLP) formu-
cose concentration at the beginning of the fermentation lation was introduced [13]. Although this formulation
stage will not normally exceed 145 g/L, even considering allows a more natural modeling approach, the resulting
full cellulose to glucose conversion of a pulp with 20% optimization problems are generally non-convex and,
DW solid content with 65% of cellulose. This value is therefore, it is difficult to obtain a global solution. In
rather modest compared with first generation bioethanol recent years, research in this area has been devoted to
production. Although, inhibition by ethanol or sugar overcoming difficulties associated with the non-convexity
concentrations is reduced in bioethanol production from of the optimization problems using global optimization
lignocellulosics, the enzymatic hydrolysis process has its techniques [14,15].
own inhibition effects. Glucose, cellobiose and xylose Graphical methods for RNS include the Attainable
have been reported to inhibit the reaction rates of cellu- Region (AR) analysis. This method has originated from
lolytic enzymes [10]. Considering that in conventional the work of Horn [16], who defined the AR as the set of
fermentation processes using sugar and starchy materials, all possible values of the outlet stream variables which
the inhibition problems have been minimized using can be reached by any possible (physically realizable)
adequate reactor configuration, the following question steady-state reactor system from a given feed stream
naturally arises: which are the most advantageous reactor using only the processes of reaction and mixing [17,18].
Scott et al. Biotechnology for Biofuels 2013, 6:171 Page 3 of 16
http://www.biotechnologyforbiofuels.com/content/6/1/171
Horn [16] showed that once the AR is obtained, then an would be natural to describe the AR in a four-dimensional
optimization problem with reactor output concentration concentration space; however, species concentrations are
as decision variables was essentially solved. The attain- not independent, and this allows calculating the changes
able region can be constructed for a given reaction net- in the number of moles in the enzymatic hydrolysis
work with n chemical compounds in an n- dimensional network as a function of cellulose and glucose molar
space. Its construction is supported by the application of changes (see the Dimensionality reduction techniques in
proposition and theorems [17,19-22] that describe pro- the Methods section). We choose to display results in a
perties of the AR. Despite these powerful theoretical dimensionless format using reaction conversions and
advances, there exist no sufficient conditions for the AR. yields (see Eq. (11) in the Methods section). In this two
Hence, the regions that are calculated applying the known dimensional space (cellulose conversion and glucose
necessary conditions are termed candidate attainable yield), the enzymatic hydrolysis reaction network produces
regions (ARc). For two and three dimensions, graphical the ARc shown in Figure 1 when the feed stream is com-
constructive methods can be derived from these proposi- posed of washed solids and a solid fraction of 0.2 is used.
tions and theorems, thus greatly facilitating its application. This corresponds to the minimum possible dimensionality
A detailed treatment of the methods used in this work is of the ARc, in the following sections it will be expanded by
given in the Methods section. For the readers acquainted incorporating the effect of the residence time.
with the existing theory and results of the AR, this section Figure 1 will be used to illustrate the construction of a
can be skipped. However, we recommend consulting the two dimensional ARc. Point F corresponds to feed stream
details concerning the kinetic models used for the enzym- composition, with zero glucose yield and cellulose conver-
atic hydrolysis and fermentation reaction networks. sion. To calculate the CSTR trajectory, the rate definition
In this work, we analyzed the process synthesis of the equations in Table 1 were substituted into Eq. (8), then
enzymatic hydrolysis and fermentation operations for the non-linear system of equations was solved for in-
bioethanol production, applying for the first time the creasing residence time values until full conversion was
concept of the Attainable Region to these systems. Two
scenarios are analyzed: (i) conversion of washed pre- 1 1 B B
treated material to bioethanol and (ii) production of
bioethanol from the discharge stream of the pretreat- 0.9
B)
0.9 ,B)
ment reactor (solids and reaction liquor), from this point x(F
(A,
0.8 mi
on non-separated pretreated material (nSPM). In each mix A
scenario, production of bioethanol from pretreated ma- 0.7 0.8
Glucose yield
0.4 B mix(F,B)
F FAB
atic hydrolysis system and the stream leaving this oper- 0.3
F,
ix(
m
Table 1 Rate balance equations per compound for cSHF derived ARc satisfies all the necessary conditions
and cSSF operations listed for a two dimensional AR.
Processes Operations Rate equations (r(c) vector) Since, the two dimensional ARc for enzymatic hydroly-
cSHF (1) Enzymatic saccharification rS = − r1 − r2 sis does not provide information about the residence
rB = r1 − r2 time of the reactors, and as this parameter is related to
the reactor capital cost, we constructed the ARc in a
rG = r2 + r3
three dimensional space of residence time, cellulose con-
(2) Ethanol fermentation r x ¼ r Fx
version and glucose yield. The stepwise procedure to con-
r G ¼ r FG struct the ARc in this space is depicted in Figures 2 and 3.
r P ¼ r FP The first step is shown in Figure 2. From feed point F, the
cSSF 1 and 2 rS = − r1 − r2 PFR trajectory ðFBÞ is calculated up to a residence time of
rB = r1 − r2 150 h. Then the CSTR trajectory with feed composition F
r G ¼ r 2 þ r 3 −r FG is calculated and the convex hull of both trajectories is
computed. It is clear that the PFR trajectory is extreme,
r x ¼ r Fx
while the CSTR trajectory ðFAÞ it is not since it is within
r P ¼ r FP
the convex hull (shaded gray volume). It is possible to
Reaction rates for enzymatic saccharification are from Kadam et al. [24]. For
glucose fermentation reaction rate definitions are presented in the text. connect the PFR and CSTR trajectories using PFRs with
Process cSSF combines both reaction rates through glucose as a feed points along the CSTR trajectory. These trajectories
common intermediate.
play an important role from a practical point of view as it
will be discussed later. The next step is to calculate a set
achieved. This procedure is detailed in the Methods of constant α values DSRs (Figure 3), and the extreme
section; from this point on, we will refer to it as the calcu- DSR reactor (connecting the points F and C). These reac-
lation of a CSTR trajectory with a given feed composition. tors further extend the ARc from the situation shown in
The PFR trajectory was calculated by integrating the sys- Figure 2, and the extreme DSR is completely built from a
tem of differential equations obtained by substituting the collection of extreme points (they lie in the boundary of
enzymatic hydrolysis rate equations in Table 1 into Eq. (7). the ARc and not in its interior, see definition and notation
From now on, this procedure will be identified as the in the Methods section). However, this reactor is of little
calculation of a PFR trajectory from a given point, which practical significance since along its trajectory, almost no
corresponds to its feed stream composition. Results show conversion of cellulose is obtained. This is due to a very
that the ARc is bounded (below) by a PFR from feed point
(F) up to point A. Figure 1 also shows the rate field, the
rate vector evaluated for each point in concentration
space. As it can be seen, the PFR trajectory is tangent to
the rate field at every point along its path. Between point
A and the equilibrium point B, the PFR trajectory is not
convex and hence the ARc is bounded by a by-pass reactor
with a feed stream with the composition of point A (line
mix (A, B) in Figure 1). This by-pass reactor can be either
a CSTR or a PFR fed with a stream of composition A and
operating with a residence time such that the composition
of the outlet stream is B. To build the line connecting A
and B, mix(A, B), the by-pass stream with composition A
is mixed with the outlet stream of a PFR or CSTR with
composition B according to the mixing equation, Eq. (10).
The subplot in Figure 1 gives a detailed view of this
section, indicating also that all the rate vectors along the
ARc boundary points inward or are tangent to the
boundary and no rate vector outside the ARc, points
Figure 2 Step 1 in ARc construction for enzymatic hydrolysis.
inwards to the ARc when reflected. As was proven by PFR and CSTR from feed point F, PFR with feed points over the CSTR
Glasser et al. [17], this indicates that the ARc cannot trajectory and the convex hull of these trajectories (gray shaded
be further extended by a PFR, a CSTR or mixing op- region). The ARc feed stream is washed solids at 0.2 solid fraction.
erations because all necessary conditions are met. The Letters in italics above the fed streams to each reactor correspond
to its composition, while the letters above the outlet streams denote
line connecting F and B corresponds to a bypass PFR
all the composition produced for different residence times.
or CSTR with feed composition equal to F. The
Scott et al. Biotechnology for Biofuels 2013, 6:171 Page 5 of 16
http://www.biotechnologyforbiofuels.com/content/6/1/171
Figure 4 ARc for enzymatic hydrolysis. The ARc is made of three zones: the plane FBCF, made of mixing lines connecting point B and the
; the mixing lines connecting point F and points along the PFR trajectory (in magenta); and, in the back of the figure, by PFR
extreme DSR line FC
with feed composition along the extreme DSR line FC.
Scott et al. Biotechnology for Biofuels 2013, 6:171 Page 6 of 16
http://www.biotechnologyforbiofuels.com/content/6/1/171
A B
A C
A BDC
A B
ABC
A B mix(A,B)
Figure 6 Candidate attainable region for ethanol production using S. cerevisiae. Left (A), the feed stream to the CSTR does not contain
cells and right (B) the feed stream to the CSTR contains 1 g/L of S. cerevisiae. In both cases, the feed stream to the PFR reactor contains 1 g/L of
cells and 100 g/L of glucose. Gray arrows correspond to the rate vector field, r(c), green arrows indicate the direction of the rate vector along the
reactors trajectories.
Scott et al. Biotechnology for Biofuels 2013, 6:171 Page 7 of 16
http://www.biotechnologyforbiofuels.com/content/6/1/171
connecting the feed composition to point B (line AB ) CSTR exists in this system (with the exception of points
can be used to extend the AR. Point B coincides with F and A of course).
the point on the curve of the CSTR where the rate vec- Figure 8 shows constant feeding policy DSR trajector-
tor starts pointing outside the AR. Thus, at point B the ies starting from F. As α values (see Eq. (9)) increase
ARc can be extended by a PFR with feed concentrations from 0 to 500 m3/h, the trajectories of the DSRs bend
and the CSTR followed by PFR trajectory
in B. The line AB and do not reach the point A, but they intersect the
define the boundary of the attainable region. Along this CSTR trajectory. This implies that no extreme DSR tra-
boundary lies the minimum residence time reactor configu- jectory from F exists, and hence the ARc is not expanded
rations for a given bioethanol concentration (or yield). by these reactors. When the trajectories of the constant
α DSRs from point A are included (Figure 9) these form
an extreme DSR path (red points along the AF line) and
Attainable region candidate for cSSF the PFRs with fed point along the extreme DSR trajec-
Accordingly to the analysis presented in the Methods tory (exDSR → PFR) form new extreme points. However,
section, the changes in the number of moles in the cSSF the newly included exDSR → PFR are not extreme for
reaction network can be expressed as a function of the every residence time along their trajectories, in fact as it
changes in the number of moles of cellulose, glucose can be seen in Figure 9B all the exDSR → PFR start at
and ethanol. We start the ARc construction for the cSSF the extreme DSR points and after some residence time
system by drawing the CSTR trajectory from the feed- they dive into the convex hull. At each of the final points
point (F) as well as the PFR from this point, the of these exDSR → PFR trajectories (the points where the
CSTR → PFR trajectories and the convex hull of this trajectories dive into the convex hull), a bypass reactor
region (Figure 7), the algorithmic procedure used for the connecting point A and these points exists. Although
construction of the ARc for cSSF is presented in the these exDSR → PFR are important as they constitute
Additional file 1. Up to this point, the extreme points part of the ARc boundary, they have little practical value
are F (feed point), A (the equilibrium point of complete for two reasons. Firstly, they originate along the extreme
cellulose conversion) and all the points on the PFR DSR trajectory starting on point A, this means that they
trajectory with F as feed composition. The CSTR trajec- start at a very high residence time, and they further ex-
tory lies within the convex hull, and hence no extreme tend it. Secondly, along its trajectory reactions produce
Figure 7 First step in the ARc construction for cSSF. PFR and
CSTR from feed point F to point A (full ethanol yield). Green
trajectories correspond to PFR with feed points along the CSTR. The Figure 8 Step 2 in the ARc construction for cSSF. Addition of
gray shaded region represents the convex hull of all the trajectories. constant fed policy DSRs trajectories with F as feed composition
The feed stream corresponds to washed solids at 0.2 solid fraction (no ethanol or glucose) and side-feed composition equal to F.
and all reactors are fed with F as denoted by italic letters above the These trajectories do not enlarge the ARc from the situation shown
reactor’s feeds. in Figure 7.
Scott et al. Biotechnology for Biofuels 2013, 6:171 Page 8 of 16
http://www.biotechnologyforbiofuels.com/content/6/1/171
Figure 9 Step 3 in the ARc construction for cSSF. Constant fed policy DSRs from A (feed composition) and with side-feed composition equal
to F. These DSRs enlarge the ARc from the situation shown in Figure 8. Left (A) a transparent view of the convex hull showing its interior and
right (B) the convex hull was shaded gray.
glucose but almost no bioethanol until a very high resi- connecting points A and B can be used to complete the
dence time (or cellulose conversions in Figure 9A). convex hull, this creates a plane (AFBA) made of by-
Finally, the complete ARc is shown in Figure 10. In pass reactors.
this view of the ARc, the extreme points along the PFR As residence time is of great importance from a cost
(which are also extreme points for the DSR from F) are engineering point of view, the projection of the ARc into
shown as red dots in the trajectory FB . In point B, the a residence time and bioethanol yield plane is presented
PFR trajectory is no longer extreme since a mixing line in Figure 11. As it can be seen, constant α DSRs do not
play a relevant role (particularly for large values of α since
at the same residence time, yield decreases with incre-
ments in α) as they produce small ethanol yields even at
elevated residence times. The minimal residence time
Figure 10 ARc for continuous saccharification and fermentation Figure 11 Residence time for the reactors in the ARc for cSSF.
of pretreated corn stover. PFR from feed point F is extreme up to Projection in the ethanol yield and residence time space. The
point B. The rest of the AR is composed of mixing lines, except by minimum residence time reactor network, for ethanol yields
the lines shown in magenta in Figure 9B (it is not possible to see above 0.35, is composed of a CSTR reactor with feed composition
these lines in the view shown in Figure 10). F followed by a PFR reactor.
Scott et al. Biotechnology for Biofuels 2013, 6:171 Page 9 of 16
http://www.biotechnologyforbiofuels.com/content/6/1/171
reactor configuration change as the residence time or yield pretreated material, this does not imply a higher glucose
progresses. From F to C, the minimum residence time concentration. In fact, when non-separated pretreated
configuration is a by-pass CSTR connecting point F and material is used, an important fraction of the soluble
C. This is so, because for any given ethanol yield between solids corresponds to xylose. This implies that, at equal
0 and 0.35, a horizontal line ð‘Þ extended from the yield total solid and insoluble solid fractions there is more
value in the ordinate intersects the by-pass reactor trajec- potentially obtainable glucose for washed solids. With
tory in the first place. Although intersections of ‘ and potentially obtainable glucose, we refer to the glucose
other reactors for higher residence times are possible, they that would be obtained if all the cellulose could be con-
are neither relevant nor convenient. For yields greater verted to glucose in an enzymatic hydrolysis process.
than 0.35, the minimal residence time configurations is When washed solids and non-separated pretreated ma-
represented by a PFR with feed point in C. This is a terial operations are compared in a common potentially
remarkable result as it suggests that a very simple reactor obtainable glucose basis (15% solid fraction for washed
arrangement (CSTR → PFR) can be used as the minimal solids and 20% for non-separated pretreated material),
residence time configuration. In addition, as it was dis- cellulose conversion is higher for washed solids as it is
cussed for the minimal residence time configurations for shown in Figure 12.
cSHF, the CSTR → PFR arrangement is of practical When glucose yields at 100 h, for washed solids and
value since allows taking advantage of a CSTR’s prop- nSPM, are plotted against the solid content, then nega-
erty: the reactor always operates at the outlet condi- tive slope straight lines are obtained with correlation co-
tions and not in the feed conditions. This results in efficients of 0.9998 and 0.9996 for washed solids and
and operation with a pourable liquid instead of a viscous non-separated pretreated material respectively. This
solid/liquid mixture. behavior was already observed for both SSF and enzym-
atic hydrolysis along several experimental data sets
Comparison of cSSH and cSHF operations with washed independently published by several authors and analyzed
solids and non-separated pretreated material by Kristensen et al. [25]. It is interesting to point out
For enzymatic hydrolysis, the boundary of the ARc is in- that we are using a kinetic model published in 2004, and
variably specified by a PFR reactor, despite the feed point the observation of Kristensen et al. [25] was made on
F corresponds to washed solids or non-separated pre- 2009, this means that with an appropriate simulation
treated material. Similarly, the solid fraction does not effort, this conclusion could have been drawn from in
change this situation. Although Figure 12 shows higher silico analysis several years earlier.
glucose yields for cSHF operation with non-separated The effect of the solid loading over cSSF operation
and the effect of cSSF operation with washed solids or
non-separated material is shown in Figure 13. It is very
1
0.9 0.9
0.8 0.8
0.7 0.7
Glucose yield
0.6 0.6
Ethanol yield
0.1 0.1
0 0
0 10 20 30 40 50 60 70 80 90 100 0 6 12 18 24 30 36 42 48
τ [h] Residence time τ [h]
Figure 12 ARc for cSHF at different solid loading and feed Figure 13 ARc for cSSF at different solid loading and feed
composition. Effect of solid loading on continuous enzymatic composition. Effect of solid loading on cSSF and comparison of
hydrolysis and comparison of the operation with washed solids cSSF operation with washed (solid lines) and non-separated
(solid lines) and non-separated pretreated material (dashed lines). pretreated material (dashed lines).
Scott et al. Biotechnology for Biofuels 2013, 6:171 Page 10 of 16
http://www.biotechnologyforbiofuels.com/content/6/1/171
interesting to note that, as opposed to enzymatic hydro- identify the reactor network configurations that pro-
lysis (Figure 12), at short times all the solid fractions vide lower residence times for both processes. Due to
results in the same bioethanol yield. This result opposes the high number of chemical species involved in the
to the linear decrease reported by Kristensen et al. [25] reaction network, and hence the high dimensionality
for different SSF experimental sets. The effect of oper- of the AR, it was expected that the by-pass and/or
ation with non-separated solids is far more harmful on DSR would shape the boundaries of the AR for mini-
cSSF compared to enzymatic hydrolysis. Figure 13 shows mum residence time, however these are not involved
that when non-separated pretreated material is used, in the configurations that resulted in the lowest resi-
bioethanol yield decreases by nearly 5% at 48 h residence dence time.
time. This effect can only be surpassed when the initial For SHF, the saccharification reaction must be per-
xylose fraction in the feed is taken as zero (instead of formed in a PFR to achieve the minimum residence
0.279) indicating that the model predicts a strong inhi- time; however because it is unfeasible from a tech-
bitory effect of this sugar over the enzymatic conversion nical point of view due to the rheological restrictions
of cellulose. of the system, the most adequate configuration with
Results suggest that non-separated pretreated mater- technical feasibility and with the closest residence
ial should only be used if a xylose co-fermenting time to the optimum is a CSTR followed by a PFR.
microorganism is available; otherwise, the strong in- For the fermentation operation, the minimum residence
hibitory effect exerted by xylose over the cellulolytic time is achieved in a reactor configuration of a CSTR
enzymes causes an important reduction of cellulose followed by a PFR.
conversion, and hence in the amount of bioethanol For SSF, the minimum residence time was obtained
obtained from the cellulosic fraction of the pretreated using a CSTR followed by a PFR, being the enzymatic
material. saccharification and fermentation reactions carried out
simultaneously in both reactors at isothermal conditions.
Validity of the results Regarding the effect of soluble solids in the reactor
Results presented so far suggest that a CSTR followed by network feed stream; for cSHF, higher glucose concen-
a PFR has the minimal residence time for cSSF and tration and yield are achieved for enzymatic hydrolysis
bioethanol production, and a near minimal residence with washed solids compared with non-separated pre-
time for cSHF. Furthermore, this design entails signifi- treated material. For cSSF, higher yields and bioethanol
cant benefits from a rheological point of view. However, titers were obtained when using washed solids.
our results were obtained with two among the many In this work, we demonstrated the capabilities of the
available reaction kinetics for the processes under ana- attainable region analysis as a tool to assess the optimal
lysis. Hence, we do not claim that the suggested reactor reactor network with minimum residence time applied
configuration will be the optimal case for any reaction to the SHF and SSF operations for lignocellulosic etha-
network and kinetic expressions in the cSHF and cSSF nol production. According to the kinetic models used in
systems. However, literature evidence supports that for this study, the most appropriate reactor configuration
auto-catalytic reactions and product-inhibited bio- for ethanol production from pretreated corn stover is a
reaction networks, a combination of CSTR followed by CSTR followed by a PFR, both operating in cSSF mode,
PFR or a series of CSTRs often have the minimal resi- and with washed pretreated material as substrate. The
dence time despite its particular kinetic parameter values methodology can be readily modified to evaluate other
[8,26] for a reaction network that can be expressed as a kinetic models of different substrates, enzymes and mi-
single reaction kinetic. croorganisms when available.
From a practical point of view, the PFR operation it is
not technically possible because of the gas production in Methods
the fermentation, thus a series of CSTR can be used to All the methodology described in this section is oriented
mimic this reactor. to construct the ARc for the different scenarios de-
scribed in the Background section. cSHF and cSSF ARcs
Conclusions were constructed for washed solids and nSPM. Unless
An attainable region analysis was performed over the otherwise specified, the solid fraction is equal to 0.2 total
conversion of pretreated corn stover to bioethanol, con- dried solids. For enzymatic hydrolysis simulation the
sidering two processes: SHF and SSF and washed and temperature was taken as 50°C, and for cSSF and fer-
non-washed material. Independent kinetic models were mentations temperature is 32°C. In both cSHF and cSSF
used for each operation, i.e.: enzymatic saccharification, operations, enzyme doses were established as 45 mg
fermentation, and simultaneous saccharification and protein/g cellulose (CPN commercial cellulase, Iogen
fermentation, in continuous operation. Our aim was to Corp., Ottawa, Ontario, Canada) [27].
Scott et al. Biotechnology for Biofuels 2013, 6:171 Page 11 of 16
http://www.biotechnologyforbiofuels.com/content/6/1/171
in the concentration space that are tightly connected with network. The operation of mixing applies over the outlet
the rate field. streams of reactors in the network, and over any given
combination of points already attained in the AR (but
c ¼ fc1 ; c2 ; …; cn ; τ g ð5Þ not outside it, i.e. in the complement of the AR). When
r ¼ fr 1 ; r 2 ; …; r n ; τ g ð6Þ two streams with compositions c1 and c2 are mixed, at
constant density, the compositions lie in the straight line
between c1 and c2, Eq. (10).
As it was stated in the definition of the AR given earl-
ier, mixing and reaction are the two operations that
allow reaching all the points in the attainable region c ¼ γc1 þ ð1−γ Þc2 ð10Þ
[17]. Furthermore, it was shown that only three idealized
reactors, along with mixing between their input and out- With γ a real number in the range [0,1]. This is usually
put streams, are required to construct the AR [20]. These referred to as the lever-arm rule, and can be derived
reactors are: the plug flow reactor (PFR), the continuous from mass balance equations. To clarify the mixing
stirred tank reactor (CSTR) and the differential sidestream operation concept, consider two streams 1 and 2 with
reactor (DSR). Their trajectories can be investigated by mass flows F1 and F2 respectively. Streams 1 and 2 have
analyzing the equations that define its behavior (under compositions cA1 and cA2 of component A and cB1 and cB2
constant density and isothermal operations). of component B. Under this conditions and assuming
constant density, what is the composition in A of the
dc stream produced by mixing streams 1 and 2? Clearly,
¼ r ðcÞ; cðτ ¼ 0Þ ¼ co ð7Þ
dτ the mass flow of the resultant stream is F = F1 + F2. A
mass balance for component A indicates that: FcA ¼ F 1
Eq. (7) defines the PFR reactor trajectory in the con- cA1 þ F 2 cA2 , then if γ = F1/F, we have: cA ¼ γcA1 þ ð1−γ ÞcA2 ,
centration space as a function of its residence time (τ). as in Eq. (10). Clearly, any point along a mixing line is
From Eq. (7) it is evident that the concentrations attainable, and the duty of the mixing operations is to fill
mapped out by integrating the PFR equations produce in concave regions in space. This mixing definition is
a trajectory that is tangent to the rate vector at every intimately connected to the concepts of convex sets
point along the reactor’s path. On the other hand, a and convex hulls. Let us consider a subset S of the
CSTR is defined by Eq. (8). Whereas PFR trajectories space of n-tuples (S ⊂ Rn), we will say that S is convex
are calculated by integration, the trajectory associated if for every pair of points in S, the line connecting
with a CSTR is found by solving a system of nonlin- them is completely contained in S. The set shown in
ear equations for a given value of residence time. For Figure 14 is convex, and the convex hull is the inter-
a particular value of τ, the CSTR has the property section of all the sets in Rn that contain S. In two di-
such that the vector defined by the difference between the mensions it can be envisioned as the tightest rubber
outlet and feed concentrations ðc−co Þ is collinear with the band that bound the set (as in Figure 14), and in
rate vector. higher dimensions as a convex polytope enclosed by a
c−co ¼ r ðcÞτ ð8Þ finite number of hyper planes.
Finally, extreme points are defined as points in Rn that
For two dimensional systems, the AR is constructed lie in a vertex of the convex hull. They can neither lie in
using only CSTRs and PFRs. However, in three or more the interior of the convex hull, nor in the interior of one
dimensions differential sidestream reactors (DSR) play a of the hyper planes (lines) that bound the convex hull.
role in shaping the AR boundary, DSRs are defined by In Figure 14 points A and B are not extreme points
Eq. (9). since they lie in the interior of the convex hull. Point C
is not extreme either because it is along one of the lines
dc
¼ r ðcÞ þ αðco −cÞcðτ ¼ 0Þ ¼ co ð9Þ between two vertices.
dτ
Now that the necessary terminology has been intro-
Physically, a DSR corresponds to a PFR with a side duced, we are in position to present some necessary
feed stream distributed all along its length. It is inter- conditions that characterize the attainable region [17],
esting to note that, if α is equal to zero, then we have this list is not exhaustive and more properties can be
a PFR and if α is equal to 1/τ and the reactor oper- founded elsewhere [20]: (i) the AR must contain the
ates in stationary state, then the reactor behaves as a feed point, (ii) the AR must be convex, (iii) all reac-
CSTR. tion rate vectors in the boundary of the AR (δAR)
The particular combination of reactor types and their must be tangent, point inward or be equal to 0 and
arrangement is called a reactor structure or reactor (iv) no negative of a rate vector in the complement
Scott et al. Biotechnology for Biofuels 2013, 6:171 Page 13 of 16
http://www.biotechnologyforbiofuels.com/content/6/1/171
xS ¼ 1−S=S o
G−Go
xG ¼
f SG S o ð11Þ
P
xP ¼
f SP S o þ f GP Go
the dependent components change of moles can be cal- and mp are not zero, then the ethanol production rate
culated as: Δndep ¼ −A−1dep Aind Δnind . Clearly, Adep has to
(rp) will be larger than the glucose production rate,
be square and non-singular. which is clearly impossible. Secondly, if mp and ms are
For the enzymatic hydrolysis reaction network, the equal to zero, no important differences in the model
atomic balance is given by Eq. (12) with compounds i = {S: predictions are observed under the conditions used in
Cellulose (C6H10O5), G: Glucose (C6H12O6), B: Cellobiose this study. In fact, if 100 g/L of glucose are taken as the
(C12H22O11), W: Water (H2O)} and atoms j = {C,H,O} initial concentration in a PFR, the only effect is a 2%
2 3 increase in the residence time required for total glu-
2 3 ΔnS cose consumption and a 0.88% decrease in ethanol yield
6 6 12 0 6
ΔnG 7 at 32°C.
AH ΔnH ¼ 4 10 12 22 2 56 7
4 ΔnB 5 ¼ 0 ð12Þ
Another important benefit of taking the values of mp
5 6 11 1
ΔnW and ms as zero is that the ARc for glucose fermentation
can be constructed in only two dimensions (ethanol
However, AH clearly it is not a full rank matrix. In fact, yield and residence time). To understand why this is
rank(AH) = 2; that is, a row in AH can be written as a linear possible, note that we can calculate the reaction rates of
combination of the remaining two rows (the third row can glucose, ethanol and biomass as functions of the ethanol
be expressed as the first row times zero plus the second production rate:
row times 0.5). Hence, partitioning between independent
(cellulose and glucose) and dependent components (cello-
1 1
biose and water) and taking only the independent rows of rx; rG ; rp ¼ ; ; 1 rP ð15Þ
Y P Y PY x
AH, we have:
−1 This implies that glucose and biomass concentrations
dep ¼ − Adep
ΔnH H
ind Δnind
AH H
ð13Þ can be expressed as a function of ethanol concentration:
ΔnB ðP−P 0 Þ
dep ¼
ΔnH X ¼ X0 þ
ΔnW YP
−1
12 0 6 6 ΔnS P−P0
¼− G ¼ G0 −
12 ΔnG
Y PY x
22 2 10
1 −ΔnS −ΔnG
¼ ð14Þ Finally, our ability to calculate X and S as a function of
2 ΔnS −ΔnG
P allow us to also calculate the reaction rates as a
This demonstrates that the change of the number of function of P exclusively. In other words, for each value
moles of water and cellobiose during the reaction course of P in the {P, τ} plane we can calculate a reaction vector
can be calculated as a function of the changes of glucose {rp, 1} that uniquely determines the trajectories of the
and cellulose. This also means that the AR of the en- CSTR and PFR reactors from a given feed point.
zymatic hydrolysis reaction has to be constructed in a Finally, to construct the ARc for cSSF only three
two-dimensional space of glucose and cellulose con- dimensions in the concentration space are required.
centration or cellulose conversion and glucose yield Although a more rigorous analysis can be performed
(and not in a four-dimensional one). Since we are inter- using the dimensionality reduction technique used by
ested in the residence time of the different reactor con- Omtveit et al. [31], the same results can be obtained by
figurations, we add this variable as the third dimension applying the following reasoning. If the ARc for cSHF
of the AR. Hence, the AR of enzymatic hydrolysis must can be built in the bi-dimensional space of {xS, xG} and
be build in the 3-dimentional space {xS, xG, τ}. the ARc for glucose fermentation can be reduced to only
In the original model of ethanol fermentation, the one dimension of ethanol yield, then as the two reaction
parameters ms and mp in Eq. (4), have values that are networks are linked by a component present in both
close to zero so in this study these values were taken as networks (glucose) then 3 dimensions are needed to
zero. Two reasons explain this simplification. Firstly, build the ARc for cSSF: {xS, xG, xP}. This result implies
under SSF conditions glucose concentrations reach a that every reaction rate in the cSSF network can be
very low value during the reaction course. This is caused calculated from conversions and yields {xS, xG, xP}.
by the greater glucose demand by the biomass compared
with the rate of glucose production from cellulose. Attainable region construction
Clearly, in these conditions bioethanol rate is not For glucose fermentation and enzymatic saccharification
controlled by the glucose to ethanol rate, but by the (without considering the reactors residence time), ARc
cellulose to glucose rate. However, if the parameters ms can be constructed in two dimensions. In this space, it is
Scott et al. Biotechnology for Biofuels 2013, 6:171 Page 15 of 16
http://www.biotechnologyforbiofuels.com/content/6/1/171
possible to build the ARc using the following steps maximum user defined value of residence time is
[18,19]: achieved. Calculate the convex hull formed by these
trajectories.
(i) Calculate the PFR trajectory starting from the feed Create a set of constant feed rate (α) values such
point. This trajectory can be calculated by solving that α = [0, α1, α2, …, αlarge]. Calculate the DSR
Eq. (7) up to a pre-established residence time. trajectories (Eq. (9)) for each α value from each
(ii) If the PFR trajectory is not convex, find the convex available extreme point (such as feed point and
hull of the PFR by drawing mixing lines to fill the equilibrium points). Then calculate the convex hull
non-convex parts. of these trajectories, eliminate the interior points
(iii) Next, check along the boundary of the convex hull and store the extreme points. These extreme points
to see if any reaction vector points outward. If the lie on the extreme DSR as defined by Feinberg [21].
reaction vector point outward over certain regions, If necessary, refine the set of α values to produce
then find the CSTRs that extend the region the more points in the extreme DSR trajectory. A
most. If no reaction vector point outward, check stopping criterion suitable for automation of the
if there are vectors in the complement of the ARc algorithm is given elsewhere [32], however we
that can be extrapolated back into the ARc. If this refined the set of α values manually.
is the situation, extend the region using appropriate From each extreme point on the DSR extreme
reactors. trajectory, generate PFRs with feed points along
(iv) Find the new, enlarged convex hull. If a CSTR lies these points. Calculate the convex hull of the
in the boundary, the reaction vector at this point enlarged region created by these trajectories.
must point out of the ARc, and a PFR with feed
point on the CSTR will extend the region. We verified our ability to apply the above described
(v) Repeat steps (iii) and (iv), alternating between PFRs methodology by reproducing the results of Example 1:
and CSTRs until no reaction vectors point out over 3D Van de Vusse type kinetics in Seodigeng et al. [32].
the ARc, and the necessary conditions are met.
Software and computational tools
As was stated by Glasser and Hildebrandt [17], this MATLAB® was used to perform all calculations in this
constructive procedure implies that for a two dimen- work. To solve systems of ordinary differential equations
sional system, the boundary of the attainable region (ODE), such as the ODEs that define the PFR and DSR
“must be achieved by a sequential process and must con- trajectories, we used the MATLAB built-in ODE45 algo-
sist of alternate straight lines and plug-flow trajectories”. rithm based on explicit Runge–Kutta formula. Systems
For cSSF and cSHF (considering the residence time), of algebraic equations, defining CSTR trajectories, were
the ARc must be built in a three dimensional space. For solved using fmincon solver and its built-in interior
cSSF, we choose cellulose conversion, glucose and etha- point method [33]. For convex hull calculation, the
nol yields as these dimensions since they provide useful MATLAB convhull solver was used. This tool is based
insights regarding: the liquefaction process, as this on the Qhull algorithm developed by Barber et al. [34].
process depends on cellulose conversion; the yield and
productivity of the product of interest, related to ethanol Additional file
conversion and the glucose yield since glucose is the
compound that links the enzymatic hydrolysis and Additional file 1: Reaction invariants and algorithmic construction
fermentation processes. of the attainable region.
The construction of a three dimensional ARc is far
more difficult than the previously described process for Abbreviations
two dimensions. Regardless of these difficulties, powerful ARc: Candidate attainable region; cSSF: Continuous simultaneous
saccharification and fermentation; cSHF: Continuous separated hydrolysis and
theoretical results were derived in a series of papers fermentation; DW: Dry weight; nSPM: Non-separated pretreated material;
[20-22]. These theoretical results were recently used to RNS: Reactor network synthesis.
formulate an automated algorithm for ARc construction
[32] and we follow this algorithm to analyze the cSSF Competing interests
The authors declare no financial or non-financial competing interests.
and cSHF reaction networks and build the candidate
attainable regions. The algorithm can be summarized in Authors’ contributions
the following steps: FS, GA, and RC conceived of and participated in the design of the study. FS
performed the calculations and wrote a draft of this work. RC and GA
provided oversight during the calculations and development of the first draft
Calculate the PFR and CSTR trajectories from of the manuscript. FS, GA, and RC edited several versions of the manuscript.
the feed point. Stop the calculations when the All authors read and approved the final manuscript.
Scott et al. Biotechnology for Biofuels 2013, 6:171 Page 16 of 16
http://www.biotechnologyforbiofuels.com/content/6/1/171