Anda di halaman 1dari 10

Bioresource Technology 97 (2006) 47–56

Conversion of municipal solid waste to carboxylic acids


using a mixed culture of mesophilic microorganisms
Cateryna Aiello-Mazzarri a, Frank K. Agbogbo b, Mark T. Holtzapple b,*

a
Universidad del Zulia, Facultad de Ingenierı́a, Coordinación de Ciclo Básico, Departamento de Quı́mica, Maracaibo, Venezuela
b
Texas A&M University, Department of Chemical Engineering, College Station, TX 77840-3122, USA

Received 22 October 2004; received in revised form 10 February 2005; accepted 11 February 2005
Available online 12 April 2005

Abstract

Waste biomass was anaerobically converted to carboxylate salts by using a mixed culture of acid-forming microorganisms. Muni-
cipal solid waste (MSW) was the energy source (carbohydrates) and sewage sludge (SS) was the nutrient source (minerals, metals,
and vitamins). Four fermentors were arranged in series and solids and liquids were transferred countercurrently in opposite direc-
tions, which allows both high conversions and high product concentrations. Fresh biomass was added to Fermentor 1 (highest car-
boxylic acid concentration) and fresh media was added to Fermentor 4 (most digested biomass). All fermentations were performed
at 40 C. Calcium carbonate was added to the fermentors to neutralize the acids to their corresponding carboxylate salts. Iodoform
was used to inhibit methane production and urea was added as a nitrogen source. Product concentrations were up to 25 g/L, with
productivities up to 1.4 g total acid/(L liquid d). Mass balances with closure between 93% and 105% were obtained for all systems.
Continuum particle distribution modeling (CPDM) was applied to correlate batch fermentation data to countercurrent fermentation
data and predict product concentration over a wide range of solids loading rates and residence times. CPDM for lime-treated MSW/
SS fermentation system predicted the experimental total acid concentration and conversion within 4% and 16% respectively.
 2005 Published by Elsevier Ltd.

Keywords: MixAlco process; Municipal solid waste; Sewage sludge; Fermentation; Carboxylic acids; Mixed acids

1. Introduction amount of carbon dioxide that biomass combustion


generates (Bungay, 1981; Sterzinger, 1995).
Many energy experts are concerned about the deple- Another environmental problem is the generation
tion of petroleum, the main source of liquid fuels and and accumulation of the following wastes: Municipal
many chemicals. Oil prices are volatile and supplies solid waste (MSW)—in the United States (US), ap-
are unstable. Oil combustion releases carbon dioxide, proximately 220 million tons of MSW were generated
the most important greenhouse gas implicated in global in 1998, four million tons more than in 1997 (EPA,
warming (Sterzinger, 1995; Klass, 1998). In contrast, 1998). Municipal solid waste contains about 65% biode-
biomass is a source of liquid fuels that does not result gradable components, such as paper, food scraps, and
in a net increase of carbon dioxide in the atmosphere yard waste (EPA, 1998). For the purpose of this
(Holtzapple et al., 1997; Hileman, 1999) because bio- research, we will refer to MSW as the biodegradable
mass growth removes from the atmosphere the same organic fraction, which excludes glass, metal, plastic,
ceramic, rock, or soil also found in MSW. Sewage
*
Corresponding author. Tel.: +1 979 845 9708; fax: +1 979 845 6446.
sludge (SS)—SS is the residual solids from conventional
E-mail addresses: caiello@luz.ve (C. Aiello-Mazzarri), m-holtzapple@ aerobic or anaerobic sewage treatment. The Environ-
tamu.edu (M.T. Holtzapple). mental Protection Agency (EPA, 1999) estimated that

0960-8524/$ - see front matter  2005 Published by Elsevier Ltd.


doi:10.1016/j.biortech.2005.02.020
48 C. Aiello-Mazzarri et al. / Bioresource Technology 97 (2006) 47–56

6.9 million tons of dry SS were generated in 1998 in the carboxylate salts by adding fresh liquid to the most di-
US of which about 60% were land applied, composted, gested biomass. Both high conversions and high product
or used as landfill cover. The remaining 40% was dis- concentrations are possible by using countercurrent
carded with no attempt to recover nutrients. operation.
MSW is an excellent energy source, but lacks nutri- Long residence times are associated with countercur-
ents. SS has nutrients, but lacks energy-yielding carbohy- rent fermentations. Obtaining a wide range of operating
drates. In this research, MSW and SS were combined, so conditions experimentally would be tremendously time
they complement each other, making the process an consuming. Mathematical modeling of the counter-
attractive alternative for managing two different streams current fermentations was performed using continuum
that are produced in every community. particle distribution modeling (CPDM) developed by
Holtzapple et al. (1997) have developed a process to Loescher (1996). CPDM can predict the operating con-
obtain fuels and chemicals from biomass, the MixAlco ditions for the countercurrent fermentations from batch
process (Fig. 1). In this process, biomass is first pre- fermentation data.
treated with lime, and then a mixed culture of acid-form-
ing anaerobic microorganisms produces carboxylate salts.
These salts are subsequently concentrated and thermally 2. Methods
converted to mixed ketones and finally hydrogenated
to mixed alcohols. The advantages of this process 2.1. Fermentor
include the use of low-value substrates without the
use of a sterile environment or the need for enzyme addi- Each fermentor consisted of a 1-L plastic centrifuge
tion. bottle (Beckman # 355676, Fisher, PA), which was
A countercurrent fermentation (Fig. 2) allows the capped with a rubber stopper with a glass tube inserted
least reactive biomass to contact the lowest carboxylic through the stopper (Ross, 1998). The glass tube was
acid concentration, which in batch fermentations could capped with a rubber septum for gas sampling and vol-
not be digested because of accumulating carboxylic acid. ume measurement. A large hole was drilled in the centri-
As the solids are transferred from one fermentor to the fuge bottle cap, so that when the cap was replaced it
next upstream fermentor (i.e., from F1 to F2, F2 to F3, could fit over the glass tube, and still be tightened to
and F3 to F4), the biomass becomes less reactive and the hold the stopper in the bottle. Inside the bottle there
carboxylate salt concentration drops. This flow arrange- were two stainless steel (0.25-in. welded 304) tubes with
ment reduces the inhibitory effect from the accumulating welded ends that mixed the components of the fermenta-
tion. The fermentors were placed on a Wheaton Modu-
lar Cell Production Roller Apparatus (Model III,
Waste Pretreatment Fermentation Concentration Fisher, PA) located in an incubator and rotated at
Biomass 1 rpm.

Thermal
Conversion
2.2. Substrates

2.2.1. Municipal solid waste (MSW)


Mixed
Alcohols
Hydrogenation To ensure uniform substrate composition through all
the MSW experiments, 136 kg of MSW were prepared.
Fig. 1. MixAlco process. To simulate the organic fraction of landfill waste (Holtz-

Product Fresh
Liquid Liquid

F1 F2 F3 F4
Fresh Product
Biomass Solid

Fig. 2. Four-stage countercurrent system fermentation (F1: Fermentor 1, F2: Fermentor 2, F3: Fermentor 3, and F4: Fermentor 4).
C. Aiello-Mazzarri et al. / Bioresource Technology 97 (2006) 47–56 49

apple et al., 1992). The components of MSW were col- material from Bee Creek Park (College Station, TX),
lected (Bryan, College Station, TX), sun-dried, ground and compost material from domestic and commercial
and passed through a 10-mm screen in a hammer mill piles. The compost samples were taken from the middle
(Forest Science Research Laboratory, Texas A&M Uni- of the pile to ensure a large quantity of anaerobic micro-
versity). Once combined, the MSW was ground and organisms. To minimize exposure to oxygen, the swamp
passed through a 6-mm screen to ensure a uniform mix- material and compost were collected into bottles filled
ture. Fats and oils were not added to the MSW to pre- with deoxygenated distilled water.
vent spoiling during storage.
2.6. Methanogens inhibitor
2.2.2. Sewage sludge
Aerobically treated sewage sludge was obtained from Iodoform (CHI3) was used as a methanogen inhibitor
Bryan Wastewater Treatment Plant Number 3 (Bryan, in all fermentations. An iodoform solution (20 g CHI3/L
TX). The sewage had undergone activated sludge treat- ethanol) was added individually to each reactor contin-
ment in which the incoming sewage stream was sent to uously throughout the fermentations. Iodoform is light
an aeration basin for approximately 15 d. After leav- and air sensitive, so the solution was kept in amber-
ing the aeration basin, the sludge was digested aero- colored glass bottles and special care was taken to
bically for approximately 40 d. Upon removal from replace the cap immediately after use.
the digestor, the sludge was coagulated, dried for 10 d,
and ground in a hammer mill fitted with a 3-mm screen. 2.7. Batch experiments

2.3. Lime treatment Batch experiments were performed to determine


experimental reproducibility, and to collect the neces-
The components of lignocellulose are complex and sary data for continuum particle distribution modeling
include solubles, cellulose, hemicellulose, lignin, and (CPDM). These data were used to develop a model rate
ash. As biomass digests, the remaining fraction is less equation for the countercurrent fermentation. The pro-
reactive (Gizjen et al., 1988), which is logical considering cedure involved operating five different fermentors with
that biomass is heterogeneous; some components are different initial substrate concentrations (20, 40, 70,
highly digestible whereas others are almost inert. To 100 g dry substrate/L liquid). An additional fermentor
enhance digestibility, lime treatment was performed at with high substrate concentration (100 g dry substrate/
100 C for 1 h with loadings of 0.1 g Ca(OH)2/g dry bio- L liquid) and high carboxylic acid concentration
mass and 10 mL of distilled water/g dry biomass. After (approximately 20 g carboxylic acids/L liquid) was eval-
treatment, CO2 was bubbled through the biomass slurry uated. Samples were taken by removing the fermentor
to neutralize the lime and the slurry was dried at 105 C. from the incubator, collecting the gas, removing the
Detailed procedures of the lime treatment are available stopper, purging with nitrogen, measuring pH, centri-
in Aiello-Mazzarri (2002). fuging for 10 min, collecting a liquid sample under nitro-
gen purge, resealing the fermentor, and placing it back
2.4. Media and nutrients in the roller apparatus. For the CPDM experiments,
the resulting carboxylic acid concentrations were con-
The fermentation media consisted of deoxygenated verted into acetic acid equivalents to be used in the
distilled water, 0.28 g sodium sulfide/L distilled water, modeling equations developed by Loescher (1996).
and 0.28 g cysteine hydrochloride/L distilled water. The batch carboxylic acid concentrations can be con-
Dry nutrients were added to the fermentations. The verted to acetic acid equivalents (a)
dry nutrient mixture used in all the experiments corre-
sponds to the modified Caldwell and Bryant medium a ðmol=LÞ ¼ acetic ðmol=LÞ
(Caldwell and Bryant, 1966). þ 1:75  propionic ðmol=LÞ

2.5. Inocula þ 2:5  butyric ðmol=LÞ


þ 3:25  valeric ðmol=LÞ
Rumen fluid was used as the main inoculum. Rumen
contents were removed from a fistulated steer (Univer- þ 4:0  caprioc ðmol=LÞ
sity Nutrition and Field Laboratory, Texas A&M Uni- þ 4:75  heptanoic ðmol=LÞ ð1Þ
versity), filtered through four layers of cheesecloth into
1-L propylene centrifuge bottles, transported to the lab- On mass basis, the acetic acid equivalent can be ex-
oratory, and inoculated (5% v/v) into each fermentor pressed as
within 1 h after collection to ensure microorganism via-
bility. Other sources of microorganisms included swamp Ae ðg=LÞ ¼ 60:05 ðg=molÞ  a ðmol=LÞ ð2Þ
50 C. Aiello-Mazzarri et al. / Bioresource Technology 97 (2006) 47–56

2.8. Countercurrent fermentations The broth was mixed with equal parts of an internal
standard (4-methyl-n-valeric acid) and 3-M H3PO4.
In countercurrent operation, liquid and solids flow in The analysis was performed using an Agilent 6890 series
opposite directions in four-fermentor trains. In the labo- gas chromatograph (Agilent Technologies, Palo Alto,
ratory scale, the fermentors operate in a semi-continuous CA) equipped with a flame ionization detector (FID)
manner, whereas on an industrial scale, the fermentation and a 7683 series injector. A 30-m fused-silica capillary
would be continuous. Countercurrent fermentations were column (J&W Scientific, Model # 123-3232 CX, Agilent
initiated as batch cultures under anaerobic conditions by Technologies, CA) was used. The column head pressure
adding the substrates, calcium carbonate, urea, nutrients, was maintained between 90 and 103 kPa (13–15 psig).
and inocula to deoxygenated water media in each fer- At every sample injection, the gas chromatograph tem-
mentor. The experiments were conducted as batch fer- perature program allowed the temperature to rise from
mentations until the culture was established (7–10 d). 50 C to 200 C at a 20 C/min rate. The temperature
Countercurrent operation was initiated with the transfer was subsequently held at 200 C for 10 min. Helium
of liquid and solids occurring every 1, 2, or 3 d. Counter- was used as carrier gas, and the total run time per sam-
current fermentations were conducted at varying liquid ple was 17 min. The fermentation broth consists of a
residence times (LRT) and volatile solid loading rate mixture of carboxylate salts and carboxylic acids. This
(VSLR). The operating parameters for each fermentation analytical procedure converts all salts to their corre-
with lime-treated MSW/SS are shown in Table 1. sponding acids, allowing product concentrations to be
reported as g carboxylic acid/L.
2.9. Reaction conditions Gases produced during fermentation were accumu-
lated within the reactor. Every sampling day, the volume
The fermentations were performed under anaerobic of gas produced since the last transfer session was mea-
conditions at 40 C. Every 3 d, after each liquid/solid sured. The volume was measured by displacing water in
transfer, 2.0 g calcium carbonate was added to each fer- an inverted glass graduated cylinder apparatus that was
mentor to neutralize carboxylic acids. To maintain filled with 30% CaCl2 solution. The CaCl2 minimized
anaerobic conditions, nitrogen from a high-pressure microbial growth in the water tank, and reduced water
liquid-nitrogen cylinder (Praxair, Bryan, TX) was flushed evaporation. The CaCl2 solution had an acidic pH
whenever the fermentors were open to the atmosphere. (5.6), which prevented CO2 adsorption. To check meth-
The solid and liquid transfer procedures are detailed in anogen inhibition, every 2 or 3 d the gas from the
Aiello-Mazzarri (2002). fermentation was analyzed for CH4 by gas chromatog-
raphy. A 5-mL sample was taken through the reactor
2.10. Analytical methods septum and analyzed using an Agilent 6890 series gas
chromatograph. The chromatograph was equipped with
The fermentor broth was analyzed by gas chromatog- a thermal conductivity detector (TCD). A 4.6-m stain-
raphy to measure the concentration of carboxylic acids. less steel packed column with 2.1-mm ID (60/80 Carbo-
xen 1000, Supelco 1-2390 U, Agilent Technologies,
Table 1
CA) was used. Samples were injected manually. The
Operating parameters for lime-treated MSW/SS countercurrent
fermentations inlet temperature was fixed at 230 C, and the detector
temperature was set at 250 C. The oven temperature
Fermentation trains A B C D E
was maintained at 225 C for 5 min. Helium was used
LRT (d) 20.7 19.0 22.7 20.7 17.0
as carrier gas. The total elution time for a sample was
VSLR (g VS/(L liquid in 3.0 4.0 8.1 3.1 6.4
all fermentors d)) 5 min.
VS feed at each 11.6 19.3 24.1 11.6 15.4 Volatile solids in the initial substrates and solid fer-
transfer (g VS) mentation residues were determined by first drying the
Solid feed at each 14.4 24.0 30.0 14.4 19.2 material at 105 C and then ashing the material at
transfer (g dry)
550 C for at least 3 h. Volatile solid determination
Liquid feed to F4 at 0.15 0.20 0.10 0.15 0.10
each transfer (L) in the liquid fermentation broth followed the same
Frequency of transfer Every 3 d heating procedure above; except prior to drying, the
Iodoform addition rate 5.33 8 8 5.33 8 liquid was mixed with lime to ensure that the carbox-
(mg iodoform added/L ylic acids would not volatize and alter the measure-
liquid fed to F4)
ment.
Nutrients addition rate 1.33 1.00 2.00 1.33 2.00
(g dry nutrients added/L
liquid fed to F4) 2.11. Mass balance
Urea addition rate 1.0 0.75 1.5 1.0 1.5
(g urea added/L For all the countercurrent fermentation experiments,
liquid feed to F4)
a complete mass balance was obtained on the entire
C. Aiello-Mazzarri et al. / Bioresource Technology 97 (2006) 47–56 51

train during a steady-state period. Closure is defined as VS digested


conversion ðxÞ ¼ ð9Þ
follows: VS fed
mass out total carboxylic acids produced
closure ¼ ð3Þ yield ðyÞ ¼ ð10Þ
mass in þ water of hydrolysis VS fed

undigested VS þ dissolved VS þ carboxylic acids produced þ biotic CO2 þ CH4


closure ¼ ð4Þ
mass in þ water of hydrolysis

total acid selectivity ðsÞ


To calculate the water of hydrolysis, it was assumed that total carboxylic acids produced
¼ ð11Þ
the biomass could be represented as cellulose, which has VS digested
a monomer weight of 162 g/mol. When cellulose is
hydrolyzed, it gains one molecule of water per mono- total acid productivity ðpÞ
mer; therefore, the water of hydrolysis is calculated as total carboxylic acids produced
¼ ð12Þ
18 L liquid in all reactors  time
water of hydrolysis ¼ VS digested  ð5Þ
162
2.13. Statistical analyses
2.12. Operational parameters
The statistical analyses were performed using Excel
The liquid residence time determines how long the software. Analysis of variance for fermentations A and
liquid remains in the system, which affects the final prod- D were performed using the student t-test at 5% level
uct concentration. Long liquid residence times allow of significance. The mean and standard deviations of
high product concentrations whereas shorter liquid the total acid concentrations for fermentations A and
residence times allow lower product concentrations D were determined from the steady-state operational
(Holtzapple et al., 1999). Liquid residence time is calcu- data from days 126 to 257. Correlations between volatile
lated as solid loading rate (VSLR) and acid productivity (p),
TLV selectivity (s), yield (y), and conversion (x) were ob-
liquid residence time ðLRTÞ ¼ ð6Þ tained by fitting experimental results to linear equations
Q
using Excel software.
where, Q = flowrate of liquid out of the fermentor set
(L/d), TLV = total liquid volume.
X
TLV ¼ ðK i  w þ F i Þ ð7Þ 3. Results and discussion
i

where, K i ¼ average wet mass of solid cake in Fermentor 3.1. Batch reproducibility
i ðgÞ; w = average liquid fraction of solid cake in
Fermentor i (L liquid/g wet cake); F i ¼ average For the following experiments, simulated MSW and
volume of free liquid in Fermentor i ðLÞ. SS were combined in an 80:20 ratio. To determine the
The volatile solids loading rate is calculated as reproducibility of MSW/SS batch experiments, two sets
of three fermentors were operated simultaneously under
VS fed=d
volatile solids loading rate ðVSLRÞ ¼ ð8Þ identical conditions. Untreated MSW/SS was used as
TLV the substrate for one set, whereas lime-treated MSW/
At a low VSLR, the solid residence time increases, SS was used for the other. The fermentors contained
allowing for more complete digestion. the substrate at a concentration of 100 g dry substrate/
Biomass is composed of volatile solids (VS) and ash, L liquid. So that the microorganisms would already be
and except for the lignin most VS are reactive. The adapted to the substrate, the inoculum used in these
digestion process converts part of the VS into gas and experiments was inocula from countercurrent experi-
liquid products, with some solids remaining undigested. ments with the same substrate.
In the liquid products, VS consist of carboxylic acids, The initial pH for all fermentations was approximately
extracellular proteins and energy-storage polysaccha- 6.5. During the fermentations, the pH varied from 5.4 to
rides (Ross, 1998). The following terms are used 6.5. The off-gas was analyzed for methane every other
throughout this paper. day, but no methane was found. The combination of
52 C. Aiello-Mazzarri et al. / Bioresource Technology 97 (2006) 47–56

18 treated MSW/20% SS) and the same inocula mixture.


16 Both fermentations were started on the same day and
were operated identically to determine the reproducibil-
14
ity of the countercurrent fermentations.
Total carboxylic acid
concentration (g/L)

12 In Fig. 4, the total carboxylic acid concentration as a


10 function of time during fermentations A and D are pre-
8
sented. Both fermentations have the same behavior
through the operation time (258 d). A t-test was con-
6
ducted over the steady-state operation (days 126–257),
4
Treated MSW/SS and the results showed that there were no significant
2
Untreated MSW/SS differences between these data set at 5% level of sig-
nificance. The results show excellent reproducibility for
0
0 2 4 6 8 10 12 14 16 18 20 countercurrent fermentations.
Time (days) Fermentation C had the highest total carboxylic acid
Fig. 3. Average carboxylic acid for untreated MSW/SS and treated
concentration obtained for lime-treated MSW/SS
MSW/SS batch fermentations at 100 g substrate/L liquid, error bars ±1 (26.0 g carboxylic acids/L liquid) which surpassed the
standard deviation. economic goal of 22 g carboxylic acid/L liquid (Holtzap-
ple et al., 1999). This fermentation was operated at pH
5.7, LRT = 22.7 d, and VSLR = 8.1 g VS/(L liquid d)
low pH and iodoform addition effectively inhibited meth- for 403 d, and also had the highest acid productivity
anogenesis. Average total carboxylic acids concentra- (1.36 g total acids/(L liquid d)). The conversion was
tions for the untreated and lime-treated MSW/SS are 0.302 g VS digested/g VS fed, the yield was 0.18 g car-
shown in Fig. 3 along with one-standard-deviation errors boxylic acid/g VS digested, and the selectivity was
bars. The variation was less as the beginning and 0.577 g VS digested/g VS fed. Fermentation D (pH
increases as the fermentation progresses. The largest 5.9, LRT = 20.7 d, and VSLR = 3.1 g VS/(L liquid d))
standard deviation was 0.46 g/L untreated MSW/SS had the highest conversion (0.449 g VS digested/g VS
and 0.62 g/L for treated MSW/SS. This experiment dem- fed). The total carboxylic acid productivity was 1.12 g
onstrated that batch experiments using lime-treated or total acids/(L liquid d), the yield was 0.264 g carboxylic
untreated simulated MSW/SS as substrate system were acid/g VS digested, and the total carboxylic acid produc-
very reproducible. tivity was 0.577 g total acids/(L liquid d).
The correlations between volatile solid loading rate
3.2. Countercurrent fermentations (VSLR) and acid productivity (p), selectivity (s), yield
(y), and conversion (x) are shown in Figs. 5–8. The re-
The results from the lime-treated countercurrent fer- sults show that at higher volatile solids loading rates,
mentations are shown in Table 2. Fermentations A and higher total acid productivities are obtained. Conver-
D were initiated using the same substrate (80% lime- sion, selectivity, and yield decrease with increasing

Table 2
Results for lime-treated MSW/SS countercurrent fermentations
Fermentation trains A B C D E
Average pH in all fermentors 5.96 ± 0.15 5.79 ± 0.02 5.68 ± 0.06 5.93 ± 0.20 5.85 ± 0.11
Total carboxylic acid concentration (g/L) 16.90 ± 0.77 16.29 ± 1.52 25.99 ± 1.46 17.07 ± 0.99 18.62 ± 0.97
Acetic acid (wt%) 39.57 ± 4.72 40.55 ± 4.30 40.40 ± 3.10 45.06 ± 4.48 43.89 ± 4.40
Propionic acid (wt%) 15.70 ± 3.57 14.18 ± 2.10 14.17 ± 1.40 17.81 ± 2.39 12.24 ± 2.21
Butyric acid (wt%) 23.31 ± 5.00 21.33 ± 3.71 22.15 ± 2.99 18.69 ± 3.63 22.08 ± 3.31
Valeric acid (wt%) 8.04 ± 1.41 7.77 ± 1.44 6.97 ± 0.96 7.35 ± 1.49 7.25 ± 1.01
Caproic acid (wt%) 9.47 ± 2.49 12.33 ± 2.96 11.09 ± 2.54 6.98 ± 2.22 9.35 ± 2.82
Heptanoic acid (wt%) 1.61 ± 0.86 2.64 ± 1.27 1.41 ± 0.74 1.70 ± 1.12 1.45 ± 0.83
Conversion (g VS digested/g VS fed) 0.432 0.366 0.302 0.449 0.392
Yield (g total acids/g VS fed) 0.276 0.221 0.175 0.264 0.237
Selectivity (g total acids/g VS digested) 0.639 0.603 0.577 0.589 0.558
Total carboxylic acid productivity (g total acids/(L liquid d)) 0.832 0.886 1.360 0.829 1.125
Biotic CO2 productivity (g CO2/(L liquid d)) 0.390 0.251 0.372 0.290 ND
Methane productivity (g CH4/(L liquid d)) 0.002 0.001 0.007 0.002 ND
Mass balance closure (g VS digested/g VS in) 0.96 101.9 100.3 0.93 ND
Note: All errors are ±1 standard deviation.
ND = not determined.
C. Aiello-Mazzarri et al. / Bioresource Technology 97 (2006) 47–56 53

8 mg iodoform/L 0.7

(g VS digested/g VS fed)
25
fed to F4
0.6
0.5

Conversion
20
Total carboxylic acid
concentration (g/L)

0.4
15 0.3
Fermentation A 0.2 Experimental data
Fermentation D Best line fit
10 Iodoform/Bromoform 0.1
Steady-state operation
8 mg/L fed to F4 0
Switch to 4 mg iodoform/L 0 2 4 6 8 10
5 fed to F4 VSLR (g VS/(L liquid in all fermentors·d))

0 Fig. 8. Correlation of conversion with volatile solid loading rate.


10 30 50 70 90 110 130 150 170 190 210 230 250 270
Time (days)

Fig. 4. Total acid concentration for lime-treated MSW/SS fermenta- VSLR. These correlations are described by the following
tion A and D.
equations:
p ¼ 0:0965VSLR þ 0:5317 ð13Þ
Experimental data
2
Best Line Fit s ¼ 0:0090VSLR þ 0:6274 ð14Þ
(g total acids/g/(L liquid·d))
Total acid productivity

1.5
x ¼ 0:0263VSLR þ 0:5037 ð15Þ
1
y ¼ 0:0178VSLR þ 0:3086 ð16Þ
0.5 The substrate system, lime-treated MSW/SS, is a hetero-
geneous lignocellulosic material in which some of the
0
0 2 4 6 8 10 components are more digestible by the microorganisms
VSLR (g VS/(L liquid in all fermentors·d)) (Ross, 1998). MSW is the source of carbon and energy,
Fig. 5. Correlation of total acid productivity with volatile solid
and SS is the source of micronutrients. At high VSLR,
loading rate. more substrate is available for the microorganisms,
which means that more nutrients and easily digested
substrate are present. The mixed culture of microorgan-
isms digested the primarily accessible fraction of the
1
substrate; however, less carboxylic acids were produced
(g total acids/g VS digested)

0.9
0.8 per unit of digested biomass. Part of the digested sub-
0.7 strate is non-acid soluble volatile solids (dissolved VS).
Selectivity

0.6
When substrate and nutrients are abundantly available,
0.5
0.4
the microorganisms may produce energy-storage com-
0.3 Experimental data
pounds that can be used in periods of starvation as sug-
0.2 Best line fit gested by Domke (1999).
0.1 However, at low VSLR, acid productivity was lower,
0
0 2 4 6 8 10 but conversion, selectivity, and yield were higher. At
VSLR (g VS/(L liquid in all fermentors·d)) lower VSLR, less substrate was added, so fewer nutri-
ents and less easily digestible fractions of the substrate
Fig. 6. Correlation of selectivity with volatile solid loading rate.
were available for the microorganisms. As a conse-
quence, the microorganisms not only would consume
the more-available fractions, but also the less-available
fractions of the substrate. In natural environments, such
0.4
as rumen, substrate concentrations are low, and there-
(g total acids/g VS fed)

0.3 fore not enough for the microorganisms to reach maxi-


mum growth rates (Russell and Baldwin, 1979). At these
Yield

0.2 conditions, per unit of digested biomass, the micro-


Experimental data
organisms may produce more carboxylic acid and less
0.1
Best line fit energy-storage components.
0 The CPDM method developed by Loescher (1996)
0 2 4 6 8 10 has been used to predict the product acid concentration
VSLR (g VS/(L liquid in all fermentors·d))
and conversions for countercurrent fermentation sys-
Fig. 7. Correlation of yield with volatile solid loading rate. tems (Ross, 1998; Domke, 1999; Thanakoses, 2002;
54 C. Aiello-Mazzarri et al. / Bioresource Technology 97 (2006) 47–56

Chan, 2002; Aiello-Mazzarri, 2002). CPDM utilizes From the CPDM Mathematica program, the values
batch data to predict the results of the countercurrent of the predicted total carboxylic acid concentration
fermentations. and conversion for the countercurrent fermentation at
The acetic acid equivalents (Ae) from each of the five various LRT and VSLR were obtained. CPDM predic-
batch experiment was fit to the equation tions for conversions and product concentrations for the
lime-treated MSW/SS countercurrent fermentations
bt
Ae ¼ a þ ð17Þ (substrate = 200 g VS/L liquid) are presented in the
1 þ ct ‘‘map’’ in Fig. 9. The experimental carboxylic acid con-
where t is the time (d) of fermentation, and a, b, and c centrations and conversions from the lime-treated
are constants fit by least squares analysis. The rate r MSW/SS countercurrent fermentations are compared
was obtained from Eq. (18). with the predicted values from the CPDM ‘‘map’’
shown in Fig. 9. As shown in Table 3, the average abso-
dðAe Þ b lute error between the experimental and predicted
r¼ ¼ ð18Þ
dt ð1 þ ctÞ2 total carboxylic acid concentration and conversion
The specific rate, ^r (g Ae produced/(g VS d)) was deter- for lime-treated fermentations was 4.4% and 15%,
mined from Eq. (18) by dividing it by the initial amount respectively.
of substrate concentration, S0 (g VS/L) in each of the The error between the predicted conversion, xpred and
five fermentors the experimental conversion, xexp as a function of selec-
tivity is shown in Fig. 10. Linear regression of the data
r
^r ¼ ð19Þ gave the following correlation:
S0
The predicted rate, ^rpred , was obtained from Eq. (20); xpred  xexp
Error ¼ ¼ 2:663s  1:418 ð23Þ
where the rate of acid production depends on volatile xexp
solids conversion (x) and product concentration (Ae)
f
Future improvements to the CPDM model should
eð1  xÞ incorporate a varying selectivity, r, to better describe
^rpred ¼ h
ð20Þ
1 þ g½/Ae the countercurrent fermentation process.
Higher substrate concentrations would be allowed if
Least square analysis was used to determine the empir-
the process were done on a large scale (Holtzapple
ical parameter constants e, f, g, and h for ^rpred (Eq.
et al., 1999). A higher VS concentration should result in
(20)) from the specific rate ^r (Eq. (19)). Ae was converted
higher total carboxylic acid concentrations. CPDM
back to carboxylic acid concentration by / (the ratio of
was used to predict the conversion and product concen-
total grams of actual acids to grams of Ae). The (1  x)
trations for a VS concentration of 300 g VS/L liquid; the
term in the numerator of Eq. (20), is the conversion pen-
CPDM ‘‘map’’ is shown in Fig. 11. As observed in the
alty function by South and Lynd (1994).
CPDM ‘‘map,’’ total acid concentrations as high as
The conversion, x was calculated using
40 g/L can be reached at LRT of 35 d and VSLR of
Ae  TLV 12 g/(L d). Also, conversions as high as 82% can be
xðtÞ ¼ ð21Þ
M0  r achieved at LRT of 10 d and VSLR of 2 g/(L d). Both,
where, r = selectivity (g Ae produced/g VS digested); high conversions (>70%) and high product concentra-
M0 = the initial substrate mass (g VS). tions (>30 g/L) can be achieved at LRT of 35 d and
The selectivity r, for Eq. (21) was calculated from the VSLR between 2 and 3 g/(L d).
selectivity s (g total acids produced/g VS digested) deter-
mined in the countercurrent experiment.
s ¼ /r ð22Þ 35
Total carboxylic acid
concentration (g/L)

The selectivity (s) used in the CPDM Mathematica pro- 30

gram was 0.60 g total acids/g VS digested, the average 25 LRT (days)

value of selectivity for the lime-treated MSW/SS coun- 20 35


30
tercurrent fermentations with terrestrial inocula. Selec- 15 25

tivity slightly decreases with increasing VSLR (Fig. 6) 10


12 8 6 5
15
VSLR 4 3
and was described by Eq. (14). Eq. (20) was used in 5 (g/(L·d)) 2 10
a Mathematica program (Aiello-Mazzarri, 2002) to 0
predict acetic acid equivalent concentration (Ae) and 0 0.2 0.4 0.6 0.8 1

conversion (x) for the countercurrent fermentation at Conversion (g VS digested/g VS fed)


various VSLR and LRT. Ae was converted back to Fig. 9. CPDM ‘‘map’’ for lime-treated MSW/SS countercurrent
carboxylic acid concentration by multiplying by /. fermentation (200 g VS/L liquid).
C. Aiello-Mazzarri et al. / Bioresource Technology 97 (2006) 47–56 55

Table 3
Comparison of experimental and predicted carboxylic acid concentration and substrate conversion for lime-treated MSW/SS countercurrent
fermentations A–F
Fermentation trains A B C D E F Averagea (%)
LRT (d) 20.7 19.0 22.7 20.7 17 19.7
VSLR (g VS/L d) 3.0 4.0 8.1 3.1 6.4 5.7
Experimental carboxylic acid concentration (g/L) 16.90 16.29 25.99 17.07 18.62 20.66
Predicted (CPDM) carboxylic acid concentration (g/L) 16.12 17.54 23.53 17.10 18.39 19.97
Errorb (%) 4.6 7.7 9.5 0.2 1.2 3.3 4.4
Experimental conversion 0.43 0.37 0.30 0.45 0.39 0.33
Predicted (CPDM) conversion 0.54 0.49 0.29 0.56 0.36 0.38
Errorb 20.3 24.3 3.3 24.3 7.7 15.1 15.8
a P
Average absolute error ¼ ðjðpredicted  experimentalÞ=experimentaljÞ=6.
b
Error = ((predicted  experimental)/experimental) · 100.

30 60
25
50 Mesophilic
Error (%)

acid concentration (g/L)


20
-------- Thermophilic

Total carboxylic
15 40
10
30 LRT (days)
5 35
30
0 20 25
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 12
VSLR 8 15
Selectivity 10 (g/(L·d)) 6 6 5 4
4 3 3 10
(g total acid/g VS digested) 2
0
Fig. 10. Correlation of error between experimental and predicted 0 0.2 0.4 0.6 0.8 1
conversion and selectivity for lime-treated MSW/SS fermentations Conversion (g VS digested/g VS fed)
with terrestrial inocula.
Fig. 12. CPDM ‘‘maps’’ for lime-treated MSW/SS four-stage coun-
tercurrent fermentation under mesophilic and thermophilic conditions
(80/20 MSW to SS ratio, and 300 g VS/L liquid for both conditions.
Thermophilic data from Chan, 2002).
45
40
Total carboxylic acid
concentration (g/L)

35
30 LRT (days)
25 35 4. Conclusions
30
20
25
15
VSLR
12 8 6 5
15 The batch and countercurrent fermentations using
10 (g/(L·d) 4
3 10 lime-treated simulated MSW/SS as substrate were very
5 2
reproducible. MSW/SS countercurrent fermentations
0
0 0.2 0.4 0.6 0.8 1 have been well characterized, and countercurrent fer-
Conversion (g VS digested/g VS fed) mentations can be operated for extended periods of time
under steady-state conditions. Addition of 4–8 mg iodo-
Fig. 11. CPDM ‘‘map’’ for lime-treated MSW/SS countercurrent
fermentation for substrate concentration 300 g VS/L liquid. form/(L liquid added to F4) inhibited methane genera-
tion. The highest total carboxylic acid concentration
obtained for lime-treated MSW/SS with terrestrial in-
ocula was 26.0 g carboxylic acids/L liquid, which sur-
The CPDM ‘‘maps’’ for lime-treated MSW/SS coun- passed the economic goal of 22 g carboxylic acid/L
tercurrent fermentations under mesophilic and thermo- liquid (Holtzapple et al., 1999). The fermentation was
philic conditions (Chan, 2002) at 300 g substrate/L operated at pH 5.7, LRT = 22.7 d and VSLR = 8.1 g
are compared in Fig. 12. For the same LRT and VSLR VS/(L liquid d) for 403 d. Industrial-scale fermentors
lower than 8 g VS/(L d), using mesophilic conditions with higher substrate concentrations would achieve
(40 C) higher acid concentrations and conversions can higher product acid concentrations and conversions.
be achieved than with thermophilic conditions (55 C). The CPDM model predicts product concentrations
At high VSLR using thermophilic conditions, acid con- and conversions within 4% and 16%, respectively. The
centrations as high as 60 g acids/L can be obtained, but CPDM model was a useful tool in predicting product
conversions may be lower than 40%. concentration and conversion. The predicted values
56 C. Aiello-Mazzarri et al. / Bioresource Technology 97 (2006) 47–56

can be used to estimate the costs and process economics Holtzapple, M.T., Lundeen, J.E., Sturgiss, R., Lewis, J.E., Dale, B.E.,
of industrial-scale fermentors. 1992. Pretreatment of lignocellulosic municipal solid waste by
ammonia fiber explosion (AFEX). Appl. Biochem. Biotechnol. 34,
5–21.
Holtzapple, M.T., Ross, M.K., Chang, N.S., Chang, V.S., Aldelson,
References S.K., Brazel, C., 1997. Biomass Conversion to Mixed Alcohol
Fuels Using the MixAlco Process. In: Saha, B.C., Woodward, J.
Aiello-Mazzarri, C., 2002. Conversion of municipal solid waste to (Eds.), ACS symposium series 666. ACS, Washington, DC, pp.
carboxylic acids by anaerobic countercurrent fermentation. Ph.D. 130–142.
Dissertation. Texas A&M University, College Station, Texas. Holtzapple, M.T., Davison, R.R., Ross, K., Aldrett-Lee, S., Nagwani,
Bungay, H.R., 1981. Energy, the Biomass Options. John Wiley & M., Lee, C.M., Lee, C., Adelson, S., Karr, W., Gaskin, D., Shiraga,
Sons, New York. H., Chang, N.S., Chang, S., Loescher, M., 1999. Biomass conver-
Caldwell, D.R., Bryant, M.P., 1966. Medium without rumen fluid for sion to mixed alcohol fuels using the MixAlco process. Appl.
non-selective enumeration and isolation of rumen bacteria. Appl. Biochem. Biotechnol. 77–79, 609–631.
Microbiol. 14, 794–801. Klass, D.L., 1998. Biomass for Renewable Energy, Fuels, and
Chan, W., 2002. Thermophilic anaerobic fermentation of waste Chemicals. California Academic Press, San Diego, pp. 651–653.
biomass for producing acetic acid. Ph.D. Dissertation. Texas Loescher, M.E., 1996. Volatile fatty acids fermentation of biomass and
A&M University, College Station, Texas. kinetic modeling using the CPDM method. Ph.D. Dissertation.
Domke, S.B., 1999. Fermentation of industrial biosludge, paper fines, Texas A&M University, College Station, Texas.
bagasse, and chicken manure to carboxylate salts. Ph.D. Disser- Ross, M.K., 1998. Production of acetic acid from waste biomass.
tation. Texas A&M University, College Station, Texas. Ph.D. Dissertation. Texas A&M University, College Station,
EPA, 1998. Municipal solid waste generation, recycling and disposal in Texas.
the United States: facts and figures for 1998. US EPA 530-F-00- Russell, J.B., Baldwin, R.L., 1979. Comparison of substrate affinities
024, Washington, DC. Environmental Protection Agency. Avail- among several rumen bacteria: a possible determinant of rumen
able from: <http://www.epa.gov/garbage/pubs/mswfinal.pdf>. bacterial competition. Appl. Environ. Microbiol. 37, 531–536.
EPA, 1999. Biosolids generation, use and disposal in the United States, South, C.R., Lynd, L.R., 1994. Analysis of conversion of particulate
1977. US EPA 530-R-99-009, Washington, DC. Environmental biomass to ethanol in continuous solids retaining and cascade
Protection Agency. Available from: <http://www.epa.gov/oigearth/ bioreactors. Appl. Biochem. Biotechnol. 45–46, 467–481.
ereading_room/BIOSOLIDS_FINAL_REPORT.pdf>. Sterzinger, G., 1995. Making biomass energy a contender. Technol.
Gizjen, H.J., Zwart, K.B., Teunissen, M.J., Voegels, G.D., 1988. Rev. 98, 34–40.
Anaerobic digestion of cellulose fraction of domestic residue by Thanakoses, P., 2002. Conversion of bagasse and corn stover to mixed
means of rumen microorganisms. Biotechnol. Bioeng. 32, 749–755. carboxylic acids using a mixed culture of mesophilic microorgan-
Hileman, B., 1999. Case grows for climate change. Chem. Eng. News isms. Ph.D. Dissertation. Texas A&M University, College Station,
77, 16–23. Texas.