Biomass and Bioenergy 140 (2020) 105704

Contents lists available at ScienceDirect

Biomass and Bioenergy

journal homepage: http://www.elsevier.com/locate/biombioe

Extraction of lipids from post-hydrolysis spent coffee grounds for biodiesel

production with hexane as solvent: Kinetic and equilibrium data
Alchris Woo Go a, *, Thi Yen Nhu Pham b, Yi-Hsu Ju a, b, c, Ramelito C. Agapay b,
Artik Elisa Angkawijaya a, Kristelle L. Quijote b
a
Graduate Institute of Applied Science and Technology, National Taiwan University of Science and Technology, Keelung Road, 10607, Taipei City, Da’an District, Taipei
City, Taiwan
b
Department of Chemical Engineering, National Taiwan University of Science and Technology, Keelung Road, 10607, Taipei City, Da’an District, Taipei City, Taiwan
c
Taiwan Building Technology Center, National Taiwan University of Science and Technology, Keelung Road, 10607, Taipei City, Da’an District, Taipei City, Taiwan

A R T I C L E I N F O A B S T R A C T

Keywords: Spent coffee grounds (SCG), an agro-industrial residue produced after brewing of coffee beans contains appre­
Spent coffee grounds ciable amounts of lipids (~16 wt%). One way of valorizing SCG is to subject it to dilute acid hydrolysis for the
Dilute acid hydrolysis production of sugar-rich hydrolysate and lipid-dense post-hydrolysis residue. The post-hydrolysis spent coffee
Equilibrium
grounds (PHSCG) contains substantial amounts of lipids (~26 wt%), making it necessary to gather lipid
Kinetics
Lipid extraction
extraction kinetics and equilibrium data for process and equipment design. The effects of solvent-to-solid ratio
(4–20 mL g 1) and temperature (30–60 � C) on the kinetics and equilibrium of lipid extraction from PHSCG using
hexane as solvent was investigated. To appropriately describe the lipid extraction system, various kinetic models
were fitted, with multi-step models best representing the extraction process (R2 � 0.99). Regardless of extraction
temperature and SSR, washing step equilibrium is reached at a very short extraction time (<10 min) and over
90% of the lipids can be extracted in at least 1 h. Extraction of lipids from PHSCG required less amount of solvent
to extract and recover the same amount of lipids present in SCG saving up to 40% of the solvent required for lipid
extraction from SCG.

1. Introduction
Recovery and use of agro-industrial residues have attracted global
attention in the pursuit of developing and/or improving sustainable
technologies without resulting in competition in the use of resources for
existing food, feed, and energy applications. A unique type of agro-
industrial residues contain significant quantities of lipids, like that of
cakes and meals of oilseed after lipid extraction, rice bran and spent
coffee grounds (SCG), among others. However, when compared to most
oilseeds and oleaginous biomass, these lipid-containing agro-industrial
residues are characterized by their relatively lower lipid content (>20
wt%) [1]. These residues require higher amount of solvent for a given
quantity of lipid, sufficient to fully immerse the residue during lipid
extraction and recovery or during in-situ (trans) esterification. To
address this, recent works on lipid-containing agro-industrial residues
like rice bran [2–4], SCG [5–7] and copra cake [8] have employed dilute
acid hydrolysis as a pretreatment step to produce sugar-rich hydroly­
sates and at the same time lipid-dense post-hydrolysis residues (PHRs),
whereby lipids are densified by 1.5–2.5 times of its original lipid con­
tent. The main mechanism behind dilute acid hydrolysis as a
pre-treatment step is to breakdown and solubilize water soluble com­
ponents like carbohydrates and proteins, while leaving behind the lipids
owing to their poor solubility in water. Solubilizing part of the solid
matrix of the biomass while leaving behind the available lipids results in
the generation of lipid-dense residues [5–7]. This pretreatment
approach is most effectively applied to wet residues like SCG, which has
previously been found to reduce energy requirements for drying, saving
at least 378.2–1029.2 kJ per kg of wet SCG processed [6,7].
In view of utilizing SCG, many of the recent works have focused on
the recovery of its lipids and its use as feedstock in biodiesel production.
Several works have reviewed and revisited the extraction of lipids from
SCG, among the notable ones are those involving the use of different
solvents employing Soxhlet extraction [9–11] and accelerated solvent
extraction [10]. Polar solvents like alcohols and esters tend to result in
higher amounts of lipids relative to the lipids that could be extracted
from hexane owing to extraction of polar lipids and non-lipid

* Corresponding author.
E-mail address: awgo@mail.ntust.edu.tw (A.W. Go).

https://doi.org/10.1016/j.biombioe.2020.105704
Received 24 February 2020; Received in revised form 22 June 2020; Accepted 23 July 2020
Available online 1 August 2020
0961-9534/© 2020 Elsevier Ltd. All rights reserved.
A.W. Go et al.
components [9]. Accelerated solvent extraction at temperatures higher
than the solvents boiling point resulted in lipid extraction efficiency, yet
would require temperatures over 100 � C and at elevated pressures [10].
Unfortunately, apart from limitations for scaling up the above lipid
extraction approaches, the information provided on lipid extraction
yields using different solvents are not directly comparable. This is
because the usual bases of comparison, which are temperature, pressure
or solvent-to-solid ratio (SSR), are not the same or fixed among the
extraction operations reported. In a separate study, lipid extraction was
carried out with different solvents and under various extraction condi­
tions, it was recommended to carryout extraction with hexane at con­
stant mixing of 50 rpm for 5 min at an SSR of 9 mL g 1 and 30 � C [12].
Use of hexane as extraction solvent not only allows extraction of most
lipids at ambient conditions but also avoids the extraction of non-lipid
components.
From another study on lipid extraction from SCG with hexane as
solvent, it was found that the extraction mechanism and kinetics could
be adequately described by a single extraction step fitted either with
modified Fick’s diffusion model or modified Peleg’s empirical model
with the equilibrium reached within an hour and maximum amount of
lipids extracted at 60 � C [13]. However, the results reported on the
required extraction time seems to be significantly different from the
previously described work which only required 5 min [12]. One diffi­
culty in reconciling literature information on lipid extraction like that
from SCG are the following: i. The differences in the extraction durations
investigated, which may not allow adequate evaluation of the appro­
priate kinetic models and minimum extraction time; ii. The differences
in the definition of extraction yield, which is either based on the lipids
obtained from the entire extraction process [12] or the lipids dissolved
into the solvent within the extraction mixture [13]; and iii. The
maximum extractable lipid by the solvent used based on Soxhlet
extraction is, in some cases [12], not reported.
For the subsequent use of extracted lipids from SCG for biodiesel
production, a recent work assessing the environmental impacts and
techno economic feasibility of biodiesel production from SCG, employ­
ing solvent extraction with hexane as solvent and the (trans)esterifica­
tion via a two-step acid-base catalyzed process indicated positive
impacts to the environment and profitability of the process with the
production capacity at 10,000 tons per year using SCG yielding 14.1 wt
% lipids during extraction [11]. To further lower down the cost of bio­
diesel production from SCG, it has been suggested to carry out the
process via integrated extraction and transesterification (in-situ process),
which was separately proven possible for SCG containing 14.1 wt% lipid
[13]. If SCG pretreated by dilute acid hydrolysis could be adopted it
would potentially improve the overall sustainability of the process.
However, there is currently no detailed information regarding the
extraction and recovery of lipids from PHSCG, which hinders the
assessment of the process and its later practical industrial application.
Taking into consideration the research gaps and lack of process data
reported in literature, the objectives of this study were to: (1) Investigate
the effect of solvent-to-solid ratio, temperature and time on the rate of
extraction, equilibrium concentration, yield and recovery of lipids from
post-hydrolysis spent coffee grounds with hexane as solvent; (2) Deter­
mine the appropriate kinetic model to describe the extraction of lipids
from PHSCG with hexane as solvent and evaluate the thermodynamic
parameters of the extraction process; and (3) Assess and evaluate the
process in terms of solvent economy and required ideal extraction
stages.
2. Materials and methods
Fresh spent coffee grounds (~25 kg) samples were collected from a
branch of global chain coffeehouses located in Taipei, Taiwan. The
chemical reagents used in the experiment include: analytical grade ethyl
acetate 99.9 vol% (Echo Chemical Co., Ltd, Taiwan), sulfuric acid 95 vol
% (Scharlau, Spain) and sodium chloride 99.9 wt% (Showa, Japan);
2
Biomass and Bioenergy 140 (2020) 105704
technical grade hexane (Echo Chemical Co., Ltd, Taiwan) with at least
95 wt% n-hexane; potassium hydroxide 85 wt% (Acros organics, USA);
FAME 37-mix and boron trifluoride methanol complex solution 13–15%
(BF3 basis) (Sigma, Aldrich, Germany). Experiments carried out are
detailed in the following subsections with the overall process flow pre­
sented in Fig. S1, following the approach described by Te et al. [14].
2.1. Collection, storage and dilute acid hydrolysis of SCG
Collected SCG having a moisture content of 61.46 � 0.98 wt% were
dried to a moisture content of less than 10 wt% to avoid bacterial or
fungal growth and stored in air-tight polypropylene bottles at ambient
conditions for later use. Drying of SCG was done by placing collected wet
SCG in stainless steel trays or pans and placed in a natural draft
convective electric oven operated at 50 � 5 � C for 3–4 days or until the
desired moisture was reached. Dried SCG samples were hydrolyzed in (4
vol%) sulfuric acid (95 wt%) solution at a constant temperature of 95 � C
and SSR of 8 mL g 1 (dry lipid-free basis) for 3 h, with 30 min interval
mixing, adopting a previously optimized procedure [7]. The resulting
hydrolysate and post-hydrolysis residue were separated by vacuum
filtration with Advantec No. 2 filter paper (5 μm pore size) as the filter
medium. The collected post-hydrolysate residue here then after referred
to as post-hydrolysis SCG (PHSCG) were then transferred to a pre-dried
& pre-weighed glass beaker and dried at 50 � 5 � C in a convective oven
to constant weight. The PHSCG yield, YPHSCG (g dry residue/g dry spent
coffee grounds), was calculated using Eq. (1).
Y PHSCG ¼
ðmDRþGB mGB Þð1 MDR Þ
(1)
mSCG ð1 MSCG Þ
where mDRþGB is the mass of dry PHSCG with the glass beaker (g), mGB is
the mass of the glass beaker (g), mSCG is the initial mass of the SCG
sample (g) subjected to hydrolysis, and M is the fractional moisture
content of the SCG sample and the residue after drying. A total of about
6 kg of dried PHSCG was collected and kept in tightly sealed propylene
bottles to avoid moisture reabsorption, which could influence the later
extraction experiments. Samples were stored at ambient conditions
(25–30 � C) until further characterization and later lipid extraction
experiments.
2.2. Determination of lipid content and lipid profile or composition from
SCG and PHSCG
Moisture content of SCG and PHSCG were determined gravimetri­
cally by freeze drying representative samples (~5 g), accurately
weighed to the nearest 0.1 mg and placed in pre-dried and pre-weighed
glass tubes. The samples were allowed to dry until constant weight, the
weight loss was taken to be the moisture content of sample analyzed.
The mean particle size was determined by the ANSI (American National
Standards Institute Method) S319.4 [15]. The crude lipid content of SCG
and PHSCG samples were determined following the AACC Method
30–25 [16] using a Soxhlet extractor with hexane as solvent for an
extraction period of 6 h. Samples of SCG and PHSCG, as well as those
after solvent extraction, were also lyophilized and subjected to scanning
electron microscope (SEM) imaging using field emission electron mi­
croscope (FE-SEM) (JSM-6500F, JOLE, Ltd. Tokyo, Japan) with the
samples coated with platinum.
Lipid profile of the extracted crude lipids were determined via gas
chromatographic analysis using Shimadzu GC-2010 Plus equipped with
a split injector, ZB-5HT Inferno column (15 m � 0.32 mm x 0.1 μm), and
flame ionization detector following the program described previously
[17], to quantify the amount of free fatty acids, monoglycerides, di­
glycerides, and triglycerides. Lipid samples (20–25 mg) were dissolved
in ethyl acetate (1 mL), and passed through 0.20-μm PTFE membrane
filters (13-mm syringe filters) before subjecting to gas chromatography
analysis. The identified peaks from the chromatograms were converted
A.W. Go et al. Biomass and Bioenergy 140 (2020) 105704

to mass percentages using calibration curves established with lipid where mLV is the mass of lipids and vial (g), mV is the mass of vial (g),
standards of mono-, di- and triolein as well as oleic acid. mAV is the mass of aliquot and vial, mFH is the mass of flask and hexane
For the determination of unsaponifiable matter content, total fatty (g), mF is the mass of flask (g), LC is the crude lipid content of PHSCG or
acid content in the lipids samples was conducted adopting the principles SCG expressed in dry basis, ms is the mass of solid sample subjected to
outlined in the AOAC official methods (Method 993.08 and Method lipid extraction (g), M is the fractional moisture content of PHSCG or
972.28) [18] with modifications as described in the work of Loyao et al. SCG, and mR is the mass of total lipids recovered (lipids from the two
[9]. The collected total fatty acids were then further reacted with aliquots and from the remaining filtrate) (g).
BF3-methanol reagent to convert the fatty acids to fatty methyl esters for
fatty acid profiling. The resulting fatty acid methyl ester composition 2.3.1. Fitting of kinetic models
from different samples was determined via gas chromatography using In order to have a better understanding of the possible mechanisms
the same method mentioned previously for lipid profile determination, involved during the extraction process, theoretical and semi-empirical
but using FAME 37-mix as reference for peak identification. kinetics models (Table 1) were fitted to the experimental data gath­
ered. Regression was carried out adopting the least-square method,
2.3. Lipid extraction kinetic and thermodynamic evaluation whereby minimizing the sum of the squares of the residuals (SSR, Eq.
(7)) while the model coefficients are iteratively modified. The regression
The lipid extraction kinetics experiments for PHSCG was carried out analysis was carried out with the aid of Microsoft Excel equipped with
at various temperatures (30, 45, and 60 � C) and SSR (4, 8, 12, 16, 20 mL Solver data analysis tool pack with the regression results assessed
g 1). For each of the extraction conditions investigated 80 mL of hexane through the resulting coefficient of determination (R2 , Eq. (8)), adjusted
(~53 g, weight to the nearest 0.1 mg) was poured into a Teflon-lines R-squared (R2adj , Eq. (9)), standard error of the mean (SEM, Eq. (10)), and
screw-capped Erlenmeyer flask (250 mL), and corresponding amounts root mean square error (RMSE, Eq. (11)).
of PHSCG for the different SSRs were weighed in screw-capped conical
X
n
polypropylene centrifuge tubes. Both the flasks and tubes were sealed SSR ¼ ðCLi cLi Þ2
C (7)
and incubated in the desired extraction temperature for 30 min prior to i¼1
mixing. Extraction was then carried out in a shake-flask incubator at the
Pn
desired extraction temperature and at a constant shaking speed of 200 ðCLi cLi Þ2
C
R2 ¼ 1 Pi¼1 (8)
rpm. The incubator is equipped with a cooler and heating system along n
i¼1 ðCLi CLi Þ2
with a P&ID controller to maintain the temperature within 0.1 � C tem­
�
perature change. At least 10 replicates of the setup were made for each 1 R2 ðn 1Þ
run to account each pre-determined extraction time. The lipids con­ R2adj ¼ 1 (9)
n k
centration in the extracting solvent (hexane) as a function of time was
Pn
determined by terminating trials at different times (2.5, 5, 10, 15, 30, i¼1 ðCLi
cLi Þ2
C
60, 120, 240, 180, and 480 min). At least duplicate runs were made for SEM ¼ pffiffiffi (10)
n
each extraction time.
Once a flask is removed at the set extraction time, the lipid-
containing hexane was separated from the solids by filtering the solu­
tion through Advantec No. 2 filter paper (5 μm pore size) with the
filtrate collected in a round bottom flask. As soon as enough filtrate was Table 1
collected, two 5-mL aliquots were pipetted out and stored in a pre-dried Theoretical and semi-empirical models for lipid extraction kinetics.
and pre-weighed 7-mL screw-capped vials. The covered vials containing
Model Name Mathematical Expression
the aliquots were immediately weighed, after which the hexane in the 0 1
Modified-Fick’s Law 4Dπ2 t
mixture was evaporated using a rotary evaporator until constant weight.
[27] B
CL ¼ C∞ @1 Ae d A ¼ C∞ ð1
2 C
Ae Bt
The collected filtrate was concentrated using a rotary evaporator until Þ

constant weight to determine the total amounts of lipid recovered after Patricelli [22] CL ¼ Cw∞ ð1 e ½ kw
c ⋅t� Þþ Cd∞ ð1 e½ kdc ⋅t�
Þ
extraction. The rotary evaporator was operated under vacuum (~2 kPa) So & Macdonald [21] kw kd1 kd2
CL ¼ Cw∞ ð1 e½ c ⋅t� Þ þ Cd1
∞ ð1 e½ c ⋅t� Þ þ Cd2
∞ ð1 e½ c ⋅t� Þ
with the distillation flask immersed in a 70 � C water bath, this was to
Modified-Peleg [28] t 1 1
ensure complete removal of hexane considering its boiling point at CL ¼
k1 þ k2 t
; R0 ¼
k1
and C∞ ¼
k2
101.3 kPa is 68.5–69.1 � C, and <-15 � C at 2 kPa. Eq. (2) to Eq. (6) were Linares [23]
CL ¼ w
Cw∞ ⋅t d
þ Cd∞ ð1 e½ kc ⋅t� Þ; T1=2
w
¼
1
and Rw0 ¼
used in calculating the response variables, concentration CL (g lipids per T1=2 þ t k2nd Cw∞

100 g hexane), percent lipid extracted EL (wt.%), lipid yields relative to k2nd Cw∞ 2 at t→0
the PHSCG mass, YLPHSCG and relative to the original SCG mass, YLSCG (g A
CL is lipid concentration (g lipids/100 g hexane) at any time t (min), C∞ is lipid
lipids g 1 dry biomass), and percent lipid recovery RL (wt.%), concentration after infinite time (g lipids/100 g hexane), D is the diffusion co­
respectively. efficient (m2/min), d is diameter of the particle (m), A is pre-exponential coef­
mLV mV ficient (dimensionless), B is volumetric mass transfer coefficient (min 1), and t is
CL ¼ � 100 (2) extraction time (min).
mAV mLV b i
C∞ is lipid concentration after infinite time (g lipids/100 g hexane), and kic is
CL
mF Þ
ðmFH kinetic coefficient(s) (min 1) for the washing (w) and diffusion stages (d).
EL ¼ 100
� 100 (3) c
k1 is Peleg’s rate constant (min ⋅ 100 g hexane/g lipids), k2 is Peleg’s capacity
LCSCG or PHSCG � ms ð1 MÞ
constant (100 g hexane/g lipids), R0 is initial extraction rate (g lipids/100 g
mR hexane ⋅ min), and C∞ is lipid concentration after infinite time (g lipids/100 g
Y LPHSCG ¼ (4) hexane).
ms ð1 MÞ d w
C∞ is lipid concentration after infinite time due to the washing stage (g lipids/
100 g hexane), Cd∞ is lipid concentration after infinite time due to diffusion stage
Y LSCG ¼ Y LPHSCG � Y PHSCG (5) w
(g lipid/100 g hexane), T1=2 is half-life time in washing stage (min), kdc is kinetic
Y LSCG or PHSCG coefficient for diffusion stage (min 1), Rw0 is initial extraction rate for washing
RL ¼ � 100 (6) stage (g lipids/100 g hexane ⋅ min), and k2nd is second-order kinetic coefficient
LCSCG or PHSCG
for washing stage (100 g hexane/g lipids ⋅ min).

3
A.W. Go et al. Biomass and Bioenergy 140 (2020) 105704

rffiffiffiffiffiffiffiffiffi
SSR Table 2
RMSE ¼ (11) Characteristics of collected biomass and lipid profile.
n
Biomass Spent Coffee Grounds Post-hydrolysis Spent Coffee
where CL is the measured lipid concentration (g lipids/100 g hexane), C cL (SCG) Groundsa (PHSCG)

is the predicted lipid concentration (g lipids per 100 g hexane), CL is the Particle size (μm) 423.35 � 8.48 349.62 � 10.22
average mean value of the measured lipid concentrations (g lipids per Composition (% w/w)
a
Moisture 61.46 � 0.98 (6.18 Not Determined (2.98 � 0.11)b
100 g hexane), n is the number of experimental data points, and k is the
� 0.24)b
number of parameters in the model including the equilibrium concen­ Lipidsc 16.85 � 0.33 (16.85 26.22 � 0.48 (19.34 � 1.05)d
trations predicted. � 0.33)d
Free fatty acide 4.99 � 0.11 4.65 � 0.11
2.3.2. Thermodynamic parameters Monoglyceridee 2.88 � 0.69 4.77 � 2.03
Diglyceridee 3.30 � 0.79 5.48 � 2.33
The activation energy (Ea , J mol 1) for the extraction process was Triglyceridee 83.57 � 1.62 83.19 � 4.45
estimated by fitting the determined kinetic coefficient (k, min 1) or Unsaponifiable 5.25 � 0.27 1.91 � 0.12
initial extraction rate (g lipids per 100 g hexane ⋅ min), to the linearized Mattere
form of the Arrhenius model (Eq. (12)). Total Fatty Acid 92.20 � 1.36e (15.54 93.18 � 1.47e (18.02 � 1.81)d
Content � 1.40)d
Ea 1 Fatty Acid Distributionf
lnk ¼ þ lnk’ (12) Palmitic acid (C16:0) 35.22 � 0.07 34.84 � 0.11
R T
Heptadecenoic acid 0.22 � 0.08 –
where, R is universal gas constant (8.31447 J mol 1 K 1), T is the ab­ (C17:1)
Linoleic acid (C18:2) 44.33 � 0.76 44.88 � 1.32
solute temperature (K), k’ is the pre-exponential factor (min 1 or g lipids Oleic acid (C18:1) 9.67 � 0.73 9.04 � 1.29
per100 g hexane ⋅ min, depending on the order of diffusion model). Stearic acid (C18:0) 7.10 � 0.06 7.25 � 0.04
At equilibrium conditions, the Gibbs free energy ΔG (J mol 1) values Arachidic acid 2.35 � 0.05 2.54 � 0.04
for lipid extraction done at various temperatures (T) were calculated (C20:0)
Behenic acid (C22:0) 0.47 � 0.04 0.58 � 0.02
using Eq. (13) and Eq. (14), while the entropy ΔS (J/mol) and enthalpy
Others 0.68 � 0.05 0.90 � 0.09
ΔH (J mol 1) were estimated by regression of Eq. (15) with the calcu­
lated ΔG s at different temperatures, where R is universal gas constant Obtained after subjecting SCG to dilute acid hydrolysis with 4 %v/v acid (95%
(8.31447 J mol 1 K 1), T is absolute temperature (K). H2SO4) at an SSR of 8 mL g 1 (based on dry-lipid-free biomass) for 3 h at 95 � C
with intermittent shaking (30 min interval).
ΔG ¼ RT lnK eq (13) a
Moisture as received (expressed in wet basis).
b
Moisture after oven drying (expressed in wet basis).
c
CLe Y Le ELe C*Le Expressed in dry basis.
K eq ffi K c ¼ ¼ ¼ or K eq ffi K d ¼ (14) d
Expressed relative to the native dry biomass and in dry basis (dry matter
CSe LC Y Le 1 ELe C*Se
yield after hydrolysis ¼ 73.75 � 0.93%).
e
Expressed relative to the extracted lipids.
ΔG ¼ ΔH TΔS (15) f
Fatty acid profile/distribution expressed as percent relative abundance (%w/
The equilibrium constant (Keq Þ, could be represented in two ways, w) based on chromatographic areas of detected fatty acid methyl esters.
either as the equilibrium constant Kc or the distribution coefficient Kd ,
with the former accounting for the concentration of lipids in the liquid of sugars (~35–40 g L 1) and proteins (~4.5 g L 1) with low furan
or solid phases relative to the total solvent mass while the latter ex­ concentrations (<0.1 g L 1) [7]. Unlike other pretreatment process,
presses the concentrations relative to the non-solute material in the adopting dilute acid hydrolysis for SCG allows the generation of sub­
respective phases. Thus, CLe is the lipid concentration in the liquid phase strate useful in biofuel production, with sugar-rich hydrolysates as po­
at equilibrium (g lipids per100 g hexane), and CSe is the lipid concen­ tential substrate for bioethanol production and the lipids-dense residues
tration in the solid phase at equilibrium (g lipids per100 g hexane), since which could be advantageous as raw material for biodiesel production.
the amount of solvent in the system does not change the equilibrium A closer analysis on the obtained lipid contents or extractable lipids,
yields (YLe ) and lipid extracted (ELe ) were used to determine Kc , while C*Le expressing these quantities relative to the native SCG, higher amounts of
is the lipid concentration in the liquid phase at equilibrium (g lipids lipid could be recovered, 19.3 g of lipid per 100 g of dry SCG, which is
per100 g hexane) and C*Se is the lipid concentration in the solid phase at about 15% more lipids than directly extracting from SCG. The additional
equilibrium (g lipids per 100 g dry lipid-free solid) are used to determine amount of lipids that have been extracted may have been a result of the
Kd . release or breakdown of structural or membrane lipids [19]. This
observation could further be supported by the fact that the amount of
3. Results and discussion total fatty acids also increased from 15.5 g to 18.0 g fatty acid per 100 g
of dry SCG. This increase in the recoverable fatty acids would also mean
Summarized in Table 2 are characteristics of and collected SCG and an increase in the amount of biodiesel (fatty acid methyl ester) that
PHSCG along with their lipid composition and fatty acid profile. A could be derived from SCG. Subjecting SCG to dilute acid hydrolysis did
notable difference observed is the amount of lipids in the collected not influence the fatty acid profile of the resulting recoverable lipids
biomasses. After subjecting the collected SCG to dilute acid hydrolysis, which mainly contains linoleic (44 wt%), palmitic (35 wt%), oleic (9 wt
the lipid content of the resulting PHSCG is about 26.2 wt%, which is %) and stearic (7 wt%) acid. From the determined fatty acid profile,
higher than the native SCG (16.8 wt%). this makes the PHSCG ~1.5 estimated biodiesel properties including density, kinematic viscosity,
times more lipid dense than native SCG. This occurred owing to the iodine value, cold filter plugging point, pour point, heating value and
densification process, where in non-lipid material or the solid fraction of cetane number could potentially meet standards set by ASTM and EN
SCG has been hydrolyzed and dissolved in the hydrolyzing medium and (Table S1).
leaving behind the lipids with the residual solids. Yielding about ~74% Apart from the increase in the recoverable lipids, it would be of in­
of the initial material, ~26% is dissolved in the resulting hydrolysate terest to know the ease of lipid extraction from PHSCG as compared to
which is similar to those observed previously [5–7]. As revealed in an the native SCG. Shown in Fig. 1 are preliminary results of lipid extrac­
optimization study previously carried out on the dilute acid hydrolysis tion kinetics for SCG and PHSCG. Generally, extraction of lipids from
of SCG as a pretreatment step, hydrolysates contain significant amount SCG and PHSCG with hexane are relatively rapid with equilibrium

4
A.W. Go et al.
reached within 2 h. At a fixed SSR or 20 mL g 1 PHSCG resulted in
higher equilibrium lipid concentration (1.92 g per100 g hexane) in the
extract as compared to SCG (1.25 g per 100 g hexane), this is owing to
the fact that PHSCG has higher lipid content compared to SCG. However,
lowering the SSR for SCG to 12 mL g 1 to achieve similar amounts of
lipid in the extraction system as to SSR 20 mL g 1 for PHSCG at a
solvent-to-oil ratio (SOR) of ~118 ml g 1, similar concentrations could
be achieved at equilibrium.
In terms of the available lipids, at least 95% of the total available
lipids are dissolved into the extracting solvent for either SCG or PHSCG.
However, considering the initial rates of extraction, the apparent slopes
for SCG-hexane system is steeper than that of PHSCG-hexane system.
This might have resulted owing to the smaller average particle diameter
of PHSCG (0.39 mm) compared to SCG (0.42 mm), with about 23% of
the PHSCG particles less than 0.25 mm as compared to SCG having only
about 4.5% less than 0.25 mm (Fig. S2). In principle, smaller particles
should improve rates of extraction considering the increase in the
overall surface area available for contact with solvent and decrease in
the diffusion path. However, particles which are too small could lead to
Fig. 1. Comparison of lipid extraction kinetics for SCG and PHSCG at different SSR and 60 � C under constant shaking speed of 200 rpm in terms of (a and c) lipid
concentration; (b and d) percent lipid extracted. (Error bars indicate the standard deviation of the experimental data obtained.).
5
Biomass and Bioenergy 140 (2020) 105704
agglomeration resulting in smaller effective surface area available. The
desired average particle size and particle size distribution vary from one
material to another, as well as the material’s interaction with the sol­
vent. A closely related work, which investigated the influence of lipid
extraction from SCG with different particle size have concluded that
particle size of less than 0.5 mm typically results in lower amounts of
lipids extracted [20]. The direct and objective comparison of SCG and
PHSCG is challenging because in actual practice, it is impractical or
impossible to have the two materials achieve the same particle size
distribution. In addition, lowering of SSR may also be adopted for
PHSCG which is expected to result in higher lipid concentration in the
extract considering that there will be higher amounts of lipids in the
system, which again makes it not possible to directly compare the SCG
and PHSCG extraction systems. To better understand the lipid extraction
process involving PHSCG, the following sections focuses on the influ­
ence of temperature and SSR on the extraction, modelling of the
extraction kinetics, thermodynamic and equilibrium data assessment of
the extraction process.
A.W. Go et al.
3.1. Effect of temperature and SSR on lipid extraction
Extraction is often carried out at elevated temperatures or near the
solvent’s boiling point to improve the dissolution of the solute, viscosity
and lubricity of the materials involved. As could be observed from Fig. 2,
increasing the temperature from 30 to 60 � C has improved the rate of
extraction. However, the final equilibrium concentrations and are only
slightly improved with the increase in temperature beyond an extraction
time of 2 h (Fig. 2a and c), with the fraction of available lipids extracted
ranging from 92 to 95% (Fig. 2b and d). As previously reported in
literature on lipid extraction from SCG of having a lipid content of 14.2
%w/w and an average particle size of 0.42 mm, the fraction of lipid
extracted at equilibrium is improve from 63% to 96% as the temperature
is increased from 20 to 60 � C [13]. One advantage of extracting lipids
from PHSCG is the fact that lipids may be extracted at lower tempera­
tures of 30 � C with over 90% of the lipids extracted.
In view of varying the SSR (Fig. 3a and c), an obvious difference is
the increase in the lipid concertation in the extract, from 1.92 to 9.52 g
per100 g hexane as the SSR is decreased from 20 to 4 mL g 1. The direct
inverse linear relationship observed between lipid concentration in the
extract and the SSR is possible as the fraction of available lipids solu­
bilized in the solvent at different SSRs are not significantly different (p
¼ 0.0713). Compared to results previously published in literature for
SCG an SSR of at least 9 mL g 1 [12] to as high as 15 mL g 1 [13] is
Fig. 2. Lipid extraction kinetic profiles for PHSCG at constant SSR of 4 mL g 1 under different extraction temperature (30, 45, and 60 � C) in terms of lipid con­
centration in the liquid phase (a and c), and percent lipid extracted or solubilized in the liquid phase (b and d). (Error bars indicate the standard deviation of the
experimental data obtained.)
6
Biomass and Bioenergy 140 (2020) 105704
required to allow maximal extraction of available lipids. Another
advantage brought about in the use of PHSCG is the lesser amount of
solvent required to solubilize and extract the available lipids, which also
results in higher concentrations of the resulting extract. Regardless of
the extraction conditions investigated for both SCG and PHSCG, at least
90% of the lipids are already dissolved in the solvent within an hour
(Fig. 3b and d).
3.2. Kinetic models and model parameters
To gain a better understanding of the possible mechanism involved
in the lipid extraction process from PHSCG at different temperature &
SSR, and at the same time, a more objective basis of comparison with
SCG, the obtained kinetic data were further fitted with various kinetic
models. Representative graphs of fitted models is presented in Fig. 4. As
could be observed, models for single-step extraction (Fig. 4c and d),
modified Fick’s Law model (R2 ¼ 0.93 to 0.95) and modified Peleg’s
model (R2 ¼ 0.97 to 0.98), tend to overestimate the amount of lipids
extracted prior to equilibrium and underestimate the lipid extracted at
equilibrium, despite having high coefficient of determination
(Table S2). For the purpose of comparison with available literature data,
the activation energies based on the single-step extraction models ranges
from 2.6 to 5.5 kJ mol 1, which is lower than that of SCG (12.7 kJ
mol 1) [13]. This indicates that the diffusion coefficients or extraction
A.W. Go et al.
Fig. 3. Lipid extraction kinetic profiles for PHSCG at constant temperature of 60 � C with different SSRs (4, 8, 12, 16, 20 mL g 1) in terms of lipid concentration (a and
c), and percent lipid extracted or solubilized in the liquid phase (b and d). (Error bars indicate the standard deviation of the experimental data obtained.).
kinetic coefficients and rates for the PHSCG-hexane systems are not
greatly influenced by temperature as compared to SCG-hexane systems.
A closer inspection of the gathered kinetic data, a short equilibrium
step could be observed between 5 and 10 min, which implies that
multiple extraction steps may be involved. This observation is also true
for SCG when extraction is carried out over an extended period of up to
8 h. Lipid extraction kinetic experiments for SCG so far carried out and
reported in literature have been done only over a short period of 15 min
[12] to an hour [13], which may not have captured the multiple equi­
librium steps during extraction. Kinetic model parameters and some
measures for the goodness of fit for the different multiple-step models
are summarized in Table 3. All regression carried out with multi-step
models resulted in coefficients of determination of at least 0.99, to
further distinguish the best models to represent the kinetic data adjusted
coefficients of determination (R2adj ), root mean square error (RMSE), and
standard error of the mean (SEM) were adopted to further validate and
qualify the models while taking into consideration physical significance
of the resulting model predictors or coefficients. For SCG and PHSCG,
2-step models (Patricelli’s model and Linares’s model) are adequate in
describing the extraction kinetics of the lipids, but So and Macdonald’s
model would best describe the lipid extraction kinetics for both systems.
The multiple-step or in most cases referred to as multi-stage extraction
mechanism does not always imply distinct observable stages along the
process but rather extraction mechanism occurring simultaneously or in
7
Biomass and Bioenergy 140 (2020) 105704
parallel [21], with the most rapid step referred to as the washing step
[21–23]. The observed washing step also implies that the lipids are
mainly found on the surface of the SCG and PHSCG particles. The
washing step or stage occurs within the first 10 min, with the equilib­
rium concentration for washing accounting for at least 80% of the total
equilibrium concentration. This is consistent with previous findings
where as much as 85% could be removed by simple washing [21].
In general, the diffusion coefficients, whether first order or second
order, increases with temperature and SSR. This is observed because the
increase in temperature lowers the viscosity of both solvent and solute
[24,25], while the increasing in SSR allows better dispersion and contact
with the solvent [21], thereby improving mass transfer and resulting in
the faster diffusion. However, in the case of PHSCG where a decrease in
diffusion coefficients were observed when the SSR was increased from 4
to 8 ml g 1 and from 16 to 20 mL g 1, this may be attributed to changes
in mixing behavior or effects of dilution. Assuming that the mixing
conditions for the different SSR were the same while considering that the
lipids are primarily found on the surface, having more solvent will dilute
the lipids found at the surface and will consequently slow down its
diffusion to the bulk fluid since a concentration gradient is the main
driving force of diffusion. With the current available models, and ac­
counting the fact that the experiments were done on a laboratory scale,
the changes in the mixing behavior and its consequent influence on the
kinetic coefficients would have to be separately investigated in more
A.W. Go et al.
Fig. 4. Comparison of different fitted models based on (a and c) single and multiple step first order diffusion mechanism and (b and d) first order, second order and
combined diffusion mechanism. (Error bars indicate the standard deviation of the experimental data obtained.).
detail.
A 3-step mechanism by So and Macdonald [21] fitted best for all
extraction investigated, it was found most adequate in describing the
extraction systems considering the process that both SCG and PHSCG
have previously undergone before extraction of the lipids. Both
lipid-containing materials have undergone size reduction and partial
hydrolysis which could facilitate the partial breakdown of cellular
components containing the lipids, exposing the lipids to either with in
the biomass matrix or on the surface of the biomass, thus, involving a
three parallel extraction mechanism occurring simultaneously, with the
first being the washing of lipids from the surface of the biomass, fol­
lowed by the diffusion of lipids from the biomass matrix, and finally the
diffusion of lipids still with in cellular components, which found in the
biomass matrix. Comparing the extraction rate constants for SCG at an
SSR of 12 mL g 1 and PHSCG at an SSR of 20 mL g 1 which has a similar
extractable amount of lipids it now more evident that the lipids found on
the surface of PHSCG were more easily extractable as supported by the
larger value of its rate constant. The activation energy for the washing of
lipids from PHSCG surface was found to be around 9.8–10.6 kJ mol 1.
The activation energy is lower compared to SCG, which was previously
found to be 12.7 kJ mol 1 [13], which is also indicative that the lipids
are more easily extracted. Comparing SCG and PHSCG lipid extraction at
8
Biomass and Bioenergy 140 (2020) 105704
the same SSR of 12 or 20 mL g 1 at 60 � C, the equilibrium concentration
of dissolved lipids of the slower diffusion steps may be indicative of the
fraction of lipids which are readily extractable from the solid matrix and
those which are still in with in the lipids bodies. It is evident that more
lipids are found readily extractable with in the biomass matrix of PHSCG
(~18–24%) as compared to SCG (~8%), while the reverse is true for
lipids still intact with in cellular components found in the solid matrix
with PHSCG having less (~2–7%) as compare to SCG (>8%). This is
observed since PHSCG have undergone further hydrolysis where by
further breaking down the cellular components containing the lipids.
Thus, subjecting lipid-containing material to hydrolysis prior to
lipid-extraction not only result in lipid-dense residues but would also
allow ease of extraction of the lipids. This is further supported with SEM
images (Fig. S3) taken after hydrolysis, with a glossy layer found on the
surface of the particles indicating more lipids are exposed in the surface,
allowing easier extraction.
3.3. Equilibrium data and thermodynamic parameters
As have been previously mentioned, the fraction of extracted lipids
does not vary significantly with the change in SSR (p ¼ 0.0713) as
compared to temperature (p < 0.0001), thus the average Gibbs free
 A.W. Go et al.
Table 3
Summary of kinetic model parameters for a multiple step extraction mechanism.
Biomass Post-Hydrolysis Spent Coffee Grounds (PHSCG) SCG

Temp ( C)
�
30 45 60 30 45 60 30 45 60 30 45 60 30 45 60 60 60
SSR (mL/g) 4 4 4 8 8 8 12 12 12 16 16 16 20 20 20 12 20
Patricelli’s Model
kwc (min
¡1
) 1.1764 1.2032 7.1376 0.9928 1.1929 1.0804 1.1848 1.3827 1.0123 1.3656 1.5064 1.1351 1.4041 1.3916 1.1538 1.5599 0.8345
Cw∞ (g lipids/100 g hexane) 7.0906 7.2805 7.1040 3.5652 3.4817 3.7233 2.2866 2.3352 2.4605 1.7411 1.7548 1.8411 1.4194 1.4700 1.5207 1.7644 1.1131
kdc (min¡1) 0.0250 0.2317 0.0342 0.0274 0.0284 0.0289 0.0263 0.0256 0.0355 0.0250 0.0250 0.0318 0.0261 0.0295 0.0274 0.0227 0.0196
Cd∞ (g lipids/100 g hexane) 2.0719 1.9509 2.3394 0.9890 1.0791 0.9082 0.6719 0.6599 0.6460 0.5292 0.5155 0.5134 0.4203 0.4030 0.3737 0.2086 0.1283
C∞ (g lipids/100 g hexane) 9.1625 9.2314 9.4435 4.5543 4.5608 4.6315 2.9585 2.9951 3.1062 2.2703 2.2703 2.3545 1.8397 1.8730 1.8944 1.9731 1.2413
SEE 0.1270 0.1494 0.1360 0.0889 0.0648 0.0501 0.0388 0.0425 0.0753 0.0421 0.0340 0.0307 0.0199 0.0250 0.0243 0.0301 0.0399
R2 0.9979 0.9971 0.9978 0.9959 0.9978 0.9987 0.9981 0.9978 0.9938 0.9963 0.9975 0.9982 0.9987 0.9981 0.9982 0.9975 0.9892
R2adj 0.9976 0.9967 0.9974 0.9952 0.9975 0.9985 0.9978 0.9974 0.9927 0.9956 0.9971 0.9979 0.9985 0.9977 0.9979 0.9971 0.9873
RMSE 0.1149 0.1351 0.1230 0.0804 0.0586 0.0453 0.0351 0.0069 0.0681 0.0381 0.0308 0.0278 0.0180 0.0227 0.0220 0.0273 0.0360
SEM 0.0619 0.0857 0.0709 0.0303 0.0161 0.0096 0.0058 0.0384 0.0218 0.0068 0.0044 0.0036 0.0015 0.0024 0.0023 0.0035 0.0061
So & Macdonald’s Model
kwc (min-1) 2.0930 2.5446 3.0504 1.9572 2.5518 2.8479 2.1232 2.5538 3.0175 2.1011 2.5610 3.0306 2.0850 2.5475 3.0376 2.0545 0.9359
Cw∞ (g lipids/100 g hexane) 6.5187 6.7383 7.0643 3.1219 3.1508 3.4377 2.1043 2.1407 2.2310 1.6380 1.6738 1.7326 1.3914 1.4738 1.4324 1.7107 1.0364
kd1
c (min-1) 0.0965 0.0858 0.0403 0.1251 0.0955 0.0695 0.1111 0.1003 0.0694 0.1335 0.1148 0.0565 0.1250 0.1127 0.0582 0.1845 0.1981
Cd1
∞ (g lipids/100 g hexane) 1.5095 1.5443 2.1568 0.9006 0.9657 0.9728 0.4610 0.5563 0.8093 0.2145 0.2178 0.5404 0.1010 0.1538 0.3745 0.0809 0.1125
kd2
c (min-1) 0.0114 0.0074 0.0036 0.0100 0.0072 0.0034 0.0110 0.0067 0.0027 0.0185 0.0161 0.0028 0.0191 0.0097 0.0027 0.0189 0.0071
9

Cd2
∞ (g lipids/100 g hexane) 1.2012 1.0923 0.3637 0.5865 0.5220 0.3405 0.4276 0.3527 0.0989 0.4247 0.3890 0.1369 0.3439 0.2929 0.1533 0.1841 0.1104
C∞ (g lipids/100 g hexane) 9.2294 9.3749 9.5848 4.6090 4.6384 4.7512 2.9930 3.0497 3.1392 2.2773 2.2807 2.4100 1.8452 1.9206 1.9603 1.9757 1.2593
SEE 0.0974 0.1056 0.1460 0.0761 0.0531 0.0434 0.0345 0.0488 0.0637 0.0387 0.0353 0.0243 0.0206 0.0406 0.0211 0.0315 0.0408
R2 0.9989 0.9987 0.9977 0.9973 0.9987 0.9992 0.9987 0.9974 0.9960 0.9972 0.9977 0.9990 0.9988 0.9955 0.9988 0.9976 0.9899
R2adj 0.9986 0.9983 0.9970 0.9965 0.9983 0.9989 0.9983 0.9966 0.9948 0.9963 0.9969 0.9987 0.9984 0.9941 0.9984 0.9968 0.9868
RMSE 0.0830 0.0901 0.1245 0.0649 0.0453 0.0370 0.0294 0.0416 0.0543 0.0330 0.0301 0.0207 0.0175 0.0056 0.0180 0.0269 0.0348
SEM 0.0323 0.0380 0.0727 0.0198 0.0096 0.0064 0.0041 0.0081 0.0138 0.0051 0.0043 0.0020 0.0014 0.0346 0.0015 0.0034 0.0057
Linares’ Model
2 2
k1 (min ⋅ 100 g hexane/g lipids) 3.98 � 10 3.93 � 10 1 � 10 5 0.1232 0.0845 1 � 10 5
0.1130 0.0673 1 � 10 4
0.1239 0.0634 0.0449 0.1222 0.1377 0.1714 0.0531 0.4155
k2 (100 g hexane/g lipids) 0.1326 0.1287 0.1408 0.2541 0.2678 0.2845 0.4140 0.4158 0.4457 0.5453 0.5567 0.5500 0.6762 0.6458 0.6235 0.5570 0.8411
Rw0 (g lipids/100 g hexane ⋅ min) 25.101 25.414 100000 8.1152 11.831 99999 8.8485 14.857 10000 8.0704 15.761 22.270 8.1809 7.2602 5.8328 18.834 2.4070
k2nd (100 g hexane/g lipids ⋅ min) 0.4415 0.4211 1981.3 0.5241 0.8485 8094.5 1.5164 2.5690 1986.1 2.4001 4.8837 6.7377 3.7410 3.0275 2.2676 5.8435 1.7030
w
C∞ (g lipids/100 g hexane) 7.5404 7.7678 7.1043 3.9348 3.7341 3.5148 2.4156 2.4048 2.2439 1.8337 1.7965 1.8180 1.4788 1.5486 1.6038 1.7953 1.1889
kdc (min¡1) 0.0191 0.0157 0.0342 0.0149 0.0212 0.0443 0.0213 0.0244 0.0621 0.0199 0.0230 0.0372 0.0219 0.0231 0.0217 0.0198 0.0031
Cd∞ (g lipids/100 g hexane) 1.6566 1.5161 2.3392 0.6600 0.8456 1.0927 0.5513 0.5902 0.8427 0.4440 0.4766 0.5276 0.3657 0.3304 0.2955 0.1805 0.0913
C∞ (g lipids/100 g hexane) 9.1971 9.2840 9.4434 4.5949 4.5797 4.6075 2.9669 2.9950 3.0866 2.2778 2.2731 2.3457 1.8445 1.8790 1.8993 1.9758 1.2802

Biomass and Bioenergy 140 (2020) 105704

SEE 0.1053 0.1254 0.1359 0.0755 0.0537 0.0623 0.0333 0.0397 0.0623 0.0387 0.0337 0.0276 0.0178 0.0225 0.0204 0.0298 0.0394
R2 0.9986 0.9980 0.9978 0.9970 0.9985 0.9981 0.9986 0.9981 0.9957 0.9968 0.9976 0.9985 0.9990 0.9984 0.9987 0.9976 0.9894
R2adj 0.9983 0.9977 0.9974 0.9966 0.9983 0.9977 0.9984 0.9978 0.9950 0.9963 0.9972 0.9983 0.9988 0.9982 0.9985 0.9972 09876
RMSE 0.0952 0.1135 0.1230 0.0683 0.0485 0.0564 0.0301 0.0359 0.0564 0.0350 0.0305 0.0250 0.0161 0.0203 0.0185 0.0269 0.0357
SEM 0.0425 0.0604 0.0709 0.0219 0.0111 0.0149 0.0043 0.0060 0.0149 0.0058 0.0044 0.0029 0.0012 0.0019 0.0016 0.0034 0.0060
A.W. Go et al.
energy based on the average equilibrium constant are well below zero,
indicating that the process is spontaneous (Fig. 5a). To better under­
stand the influence of SSR, Gibbs free energy were also determined
based on the distribution constants or coefficients (Fig. 5b), which in­
dicates that lower SSRs were more thermodynamically favored over
higher SSRs, but increasing the temperature improves the spontaneity of
the process. It can be further inferred from Fig. 4 that the process is
endothermic (ΔH > 0) and irreversible (ΔS > 0). The positive enthalpy
observed indicates that energy required for solute-solvent interaction is
greater than energy released for solvent-solvent and solute-solute
interaction [26]. Reprocessing of available literature data on the lipid
extraction of lipids from SCG with hexane at the same temperature range
[13] have elucidated that the enthalpy for extraction of lipids from
PHSCG is much lower (15.9–19.7 kJ mol 1) as compared to SCG (~105
kJ mol 1). This implies less energy is required in the extraction of lipids
from PHSCG than from SCG. As for entropy, positive change is expected
as extraction involves the mixing of two substances, which increases the
systems disorderliness [26].
Summarized in Table 4 are process information or responses ob­
tained at equilibrium. The estimated equilibrium concentrations from
Fig. 5. Gibbs’s free energy as a function of temperature and thermodynamic parameters entropy (-ΔS, slope) and enthalpy (ΔH, intercept) on the extraction of lipids
from PHSCG, based on (a) average equilibrium constants (Kc) and (b) distribution coefficient (Kd) at different solvent-to-solid ratio.
10
Biomass and Bioenergy 140 (2020) 105704
curve fitting coincides well with the experimental values and both are
well below the theoretical maximum. The theoretical maximum assumes
all the available lipids are dissolved in the extracting solvent, which
serves as a limit in the attainable concentration. Actual experimental
values are expected to be lower considering that the lipids will distribute
between the solid phase and the liquid phase at equilibrium, with some
lipids still remaining in the solid phase. In view SSR’s influence on
concentration, the increase in SSR results in lower concentration owing
to dilution of the system considering that the same degree of extraction
is achieved regardless of SSR. For the influence of temperature, equi­
librium concentrations are increase as temperature is increase indicating
that the equilibrium shifted towards the dissolution of the lipids as the
temperature is increase.
Interestingly, although most lipids were extracted from the solid and
solubilized in the solvent (>91%), the lipids actually recovered were
lesser, this is due to the entrained solution in solids and in the filter
during separation. In view of yields and recoveries, these quantities
increased as SSR was increased but is not significantly influenced by the
increase in temperature. Considering that higher SSR have lower lipid
concentration, this also means lesser amount of lipids are left behind
 A.W. Go et al.
Table 4
Equilibrium dataa of lipid-hexane-inert for SCG and PHSCG at different temperature and solvent-to-solid ratio.
T (K) SSR Theoretical Kinetic Model Based Concentration (g Lipid Lipid Lipid Lipid Lipid Overflow, X Underflow, Y underflow, N
(mL/ Maximum Concentrationb (g lipid/100 g Extracted Yield_PHSCG (g Yield_SCG (g Recovery_PHSCG Recovery_SCG (g solute/g (g solute/g (g inert/g
g) Concentration (g lipid/100 g solvent) solvent) (%) lipid/g dry lipid/g Dry (%) (%) solution)e solution)f solution)g
lipid/100 g solvent) PHSCG) SCG)

Post-Hydrolysis Spent Coffee Grounds (PHSCG)

304.15 4 9.88 9.22 (0.10)c 9.22 (0.00)d 92.78 0.12 0.09 46.15 (1.87)d 52.96 (2.14)d 0.084 0.095 0.519
d
(0.01)
318.15 4 9.88 9.37 (0.11) 9.33 (0.02) 94.41 0.12 0.09 46.21 (1.69) 53.02 (1.94) 0.085 0.094 0.512
(1.12)
333.15 4 9.88 9.58 (0.15) 9.47 (0.06) 95.72 0.12 0.09 44.10 (1.03) 50.60 (1.18) 0.087 0.093 0.488
(0.53)
304.15 8 4.94 4.60 (0.08) 4.59 (0.01) 92.32 0.17 0.12 63.01 (1.87) 72.30 (2.15) 0.044 0.055 0.435
(0.82)
318.15 8 4.94 4.64 (0.05) 4.59 (0.05) 93.73 0.16 0.12 61.26 (2.10) 70.30 (2.41) 0.044 0.052 0.395
(0.72)
333.15 8 4.94 4.75 (0.04) 4.66 (0.04) 94.89 0.16 0.12 60.61 (1.66) 69.54 (1.91) 0.044 0.051 0.383
(0.17)
304.15 12 3.28 2.99 (0.03) 2.98 (0.00) 91.10 0.18 0.13 67.13 (1.79) 77.02 (2.06) 0.029 0.041 0.363
(0.46)
318.15 12 3.28 3.05 (0.05) 3.04 (0.02) 92.74 0.18 0.13 68.33 (0.03) 78.40 (0.04) 0.030 0.038 0.352
(0.39)
333.15 12 3.28 3.14 (0.06) 3.10 (0.00) 95.03 0.18 0.13 66.89 (0.00) 76.75 (0.00) 0.030 0.035 0.313
(0.00)
11

304.15 16 2.48 2.27 (0.04) 2.26 (0.03) 91.30 0.19 0.14 74.06 (1.33) 84.97 (1.53) 0.022 0.034 0.383
(1.21)
318.15 16 2.48 2.28 (0.04) 2.30 (0.00) 92.95 0.19 0.14 72.03 (0.32) 82.65 (0.36) 0.022 0.030 0.313
(1.74)
333.15 16 2.48 2.41 (0.02) 2.35 (0.02) 95.38 0.19 0.14 70.87 (0.87) 81.32 (1.00) 0.023 0.028 0.278
(1.64)
304.15 20 2.00 1.84 (0.02) 1.81 (0.05) 90.87 0.19 0.14 73.41 (1.61) 84.24 (1.85) 0.018 0.027 0.296
(2.47)
318.15 20 2.00 1.92 (0.04) 1.85 (0.04) 93.05 0.20 0.15 75.43 (1.22) 86.55 (1.40) 0.018 0.026 0.305
(2.36)
333.15 20 2.00 1.96 (0.02) 1.91 (0.02) 94.81 0.19 0.14 72.74 (1.27) 83.47 (1.46) 0.019 0.023 0.246
(0.45)
Spent Coffee Grounds (SCG)
333.15 12 1.98 1.97 (0.03) 1.97 (0.03) 99.73 0.13 0.13 75.47 (1.25) 75.47 (1.25) 0.019 0.020 0.427
(1.85)
333.15 20 1.27 1.25 (0.04) 1.25 (0.00) 98.99 0.13 0.13 77.53 (1.20) 77.53 (1.20) 0.012 0.013 0.308
(0.06)

Biomass and Bioenergy 140 (2020) 105704

a
Average of values observed from 120 min to 480 min.
b
Based on So and MacDonald’s Model.
c
Standard error of estimate.
d
Standard deviation.
e
Calculated based on the experimental lipid concentration.
f
Solute and solvent left behind determined by material balance based on lipids recovered and experimentally determined lipid concentration.
g
Moisture and solids are assumed to compose the inert fraction and is solely found in the underflow, while mass of solution is determined by material balance of lipid and solvent left in the extraction flask.
A.W. Go et al. Biomass and Bioenergy 140 (2020) 105704

with the entrained solution. Although higher temperatures result in
more lipids extracted, resulting in higher lipid concentrations, this also
translates to more lipids left behind with the entrained solution during
separation. Thus, higher SSRs are preferred to induce better recover­
ability, while temperature even at 30 � C allows similar actual recoveries
when compared to higher temperature. Furthermore, if recoveries are
expressed relative to the original spent coffee ground processed, it can
be observed that higher recovery of the lipids can be achieved with
PHSCG compared to SCG, whether it be at the same SSR or at the same
SOR. This indicates that lipids are more easily extracted and recovered
from PHSCG.
To avoid the use of high SSR, extraction could be carried out in
multiple extraction stages. For preliminary assessment of the process the
overflow and underflow concentrations of solute and inert (Table 4)
were first determined by material balance, which were then used to
estimate the ideal equilibrium stages required to achieve complete re­
covery of the lipids assuming that the overflow and underflow concen­
tration ratio were constant and that the amount solvent to inert ratio
were also maintained at each extraction stage. From Fig. 6a, it is clear
that more extraction stages are required for lower SSRs to achieve re­
coveries of over 90%. Although an SSR of 4 mL g 1 would require up to 5
ideal crossflow stages it still results in high lipid concentration in the

Fig. 6. Comparison (a) cumulative recovery, (b) cumulative extract concentration, and (c) cumulative solvent economy of simulated multi-stage crossflow extraction
of lipids from SCG and PHSCG based on equilibrium data gathered at 60 � C for different solvent-to-solid ratio..

12
A.W. Go et al.
pooled extracts (Fig. 6b). Moreover, using a lower SSR would also result
in better solvent economy (Fig. 6c), saving ~30–40% of the required
solvents for extracting lipids from native SCG employing an SSR of 20
mL g 1 to achieve the same amount of lipids recovered. The savings in
amount of solvent used also translates to energy savings required for
later solvent recovery. Although, it may be argued that lower SSRs could
be employed for SCG, the same argument could be applied for PHSCG
while having the advantage of achieving higher product concentrations
at lower SSRs.
4. Conclusions
Lipids in SCG could be densified by subjecting the native biomass to
dilute acid hydrolysis, whereby the resulting PHSCG is 1.5 times denser
in terms of its lipid content. The lipids from PHSCG have similar fatty
acid profile with the native SCG and are found suitable as raw material
for biodiesel production. The pretreatment of SCG to generate PHSCG
having higher lipid contents could be advantageous in the production of
biodiesel through in-situ (trans)esterification. In the extraction of lipids
from PHSCG, higher temperature and lower SSR improves the rates of
lipid extraction. Regardless of temperature or SSR, ~95% of the avail­
able lipids are solubilized in hexane within 2 h. In terms of recoveries,
higher SSRs allows better actual recovery of the lipids but lower SSR
results in extract with higher concentrations. The extraction of lipids
from SCG and PHSCG involves 3 mechanisms, a rapid washing step and
two diffusion steps, and is best described by So and Macdonald’s model
(R2 � 0.995). The extraction process was found to be exergonic ( 8.3 �
ΔG � 6.1 kJ mol 1), endothermic (ΔH ¼ 16.1 kJ mol 1) and irre­
versible (ΔS ¼ 0.08 kJ mol 1 K 1), with lower SSRs found to be more
thermodynamically favored. Although the process is endothermic and
favors higher temperatures, actual process recovery is not significantly
influenced by temperature. Thus, extraction of lipids from PHSCG would
best be carried out at 30 � C and at an SSR of 4–8 mL g 1 for about 1–2 h
at each extraction stage, providing a potential solvent savings of up to
40% compared to what is required for SCG in the recovery of the same
amount of lipids, recovering at least 90% of the lipids.
Authors contribution
Alchris Woo Go – Writing Original Draft, Supervision, Investigation,
Formal Analysis, Funding Acquisition, Conceptualization, Methodology,
Visualization. Thi Yen Nhu Pham – Writing Original Draft, Methodology,
Investigation, Formal Analysis, Data Curation. Yi-Hsu Ju – Writing-
Review & Editing, Conceptualization, Resources, Project Administra­
tion. Ramelito C. Agapay – Writing-Review & Editing, Visualization. Artik
Elisa Angkawijaya – Writing-Review & Editing, Resources, Kristelle L.
Quijote – Writing-Review & Editing.
Declaration of competing interest
None.
Acknowledgements
Authors would like to thank the Ministry of Science and Technology,
Taiwan, for the research support provided through the grant MOST 108-
2218-E-011-032-MY3. A.W.G. and A.E.A. would like to thank National
Taiwan University of Science and Technology for the teaching and
research start-up support and grant provided for 2019–2021 to organize
the research group involved in this work.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.biombioe.2020.105704.
13
Biomass and Bioenergy 140 (2020) 105704
References
[1] A.W. Go, S. Sutanto, L.K. Ong, P.L. Tran-Nguyen, S. Ismadji, Y.H. Ju, Developments
in in-situ (trans) esterification for biodiesel production: a critical review, Renew.
Sustain. Energy Rev. 60 (2016) 284–305, https://doi.org/10.1016/j.
rser.2016.01.070.
[2] S. Sutanto, A.W. Go, K.H. Chen, S. Ismadji, Y.H. Ju, Maximized utilization of raw
rice bran in microbial oils production and recovery of active compounds: a proof of
concept, Waste and Biomass Valorization 8 (2017) 1067–1080, https://doi.org/
10.1007/s12649-016-9685-z.
[3] S. Sutanto, A.W. Go, S. Ismadji, Y.-H. Ju, Hydrolyzed rice bran as source of lipids
and solid acid catalyst during in situ (trans)esterification, Biofuels 11 (2020)
221–227, https://doi.org/10.1080/17597269.2017.1348190.
[4] S. Sutanto, A.W. Go, K.H. Chen, P.L.T. Nguyen, S. Ismadji, Y.H. Ju, Release of sugar
by acid hydrolysis from rice bran for single cell oil production and subsequent in-
situ transesterification for biodiesel preparation, Fuel Process. Technol. 167 (2017)
281–291, https://doi.org/10.1016/j.fuproc.2017.07.014.
[5] A.W. Go, A.T. Conag, D.E.S. Cuizon, Recovery of sugars and lipids from spent
coffee grounds: a new approach, Waste and Biomass Valorization 7 (2016)
1047–1053, https://doi.org/10.1007/s12649-016-9527-z.
[6] G.F.Y. Juarez, K.B.C. Pabilo~ na, K.B.L. Manlangit, A.W. Go, Direct dilute acid
hydrolysis of spent coffee grounds: a new approach in sugar and lipid recovery,
Waste and Biomass Valorization 9 (2018) 235–246, https://doi.org/10.1007/
s12649-016-9813-9.
[7] A.W. Go, A.T. Conag, M.M.N. Bertumen, Taguchi method to improve the
production of sugar-rich hydrolysate from non-delipidated spent coffee grounds,
and subsequent recovery of lipids and bioactive compounds, Biofuels 10 (2019)
193–205, https://doi.org/10.1080/17597269.2017.1309855.
[8] R.J.A. Chato, C.C.R. Cuevas, J.S.N. Tangpuz, L.K. Cabatingan, A.W. Go, Y.-H. Ju,
Dilute acid hydrolysis as a method of producing sugar-rich hydrolysates and lipid-
dense cake residues from copra cake, J. Environ. Chem. Eng. 6 (2018) 5693–5705,
https://doi.org/10.1016/j.jece.2018.08.072.
[9] A.S. Loyao, S.L.G. Villasica, P.L.L. DelaPe~ na, A.W. Go, Extraction of lipids from
spent coffee grounds with non-polar renewable solvents as alternative, Ind. Crop.
Prod. 119 (2018) 152–161, https://doi.org/10.1016/j.indcrop.2018.04.017.
[10] I. Efthymiopoulos, P. Hellier, N. Ladommatos, A. Russo-Profili, A. Eveleigh,
A. Aliev, A. Kay, B. Mills-Lamptey, Influence of solvent selection and extraction
temperature on yield and composition of lipids extracted from spent coffee
grounds, Ind. Crop. Prod. 119 (2018) 49–56, https://doi.org/10.1016/j.
indcrop.2018.04.008.
[11] M. Kamil, K.M. Ramadan, O.I. Awad, T.K. Ibrahim, A. Inayat, X. Ma,
Environmental impacts of biodiesel production from waste spent coffee grounds
and its implementation in a compression ignition engine, Sci. Total Environ. 675
(2019) 13–30, https://doi.org/10.1016/j.scitotenv.2019.04.156.
[12] C. Mueanmas, R. Nikhom, A. Petchkaew, J. Iewkittayakorn, K. Prasertsit,
Extraction and esterification of waste coffee grounds oil as non-edible feedstock for
biodiesel production, Renew. Energy 133 (2019) 1414–1425, https://doi.org/
10.1016/j.renene.2018.08.102.
[13] V. Najdanovic-Visak, F.Y.L. Lee, M.T. Tavares, A. Armstrong, Kinetics of extraction
and in situ transesterification of oils from spent coffee grounds, J. Environ. Chem.
Eng. 5 (2017) 2611–2616, https://doi.org/10.1016/j.jece.2017.04.041.
[14] K.G.D. Te, A.W. Go, H.J.D. Wang, R.G. Guevarra, L.K. Cabatingan, I.D.F. Taba~ nag,
A.E. Angkawijaya, Y.H. Ju, Extraction of lipids from post-hydrolysis copra cake
with hexane as solvent: kinetic and equilibrium data, Renew. Energy 158 (2020)
311–323, https://doi.org/10.1016/j.renene.2020.05.096.
[15] M. Rhodes, Particle size analysis, in: Introd. To Part, Technol., 2008, pp. 1–27,
https://doi.org/10.1002/9780470727102.ch1.
[16] Aacc Approved Methods of Analysis, AACCI method 30-25.01, crude fat in wheat,
corn, and soy flour, feeds, and mixed feeds, in: AACC Int. Approv. Methods,
eleventh ed., Cereals & Grains Association, St. Paul, MN, U.S.A., 2009 https://doi.
org/10.1094/AACCIntMethod-30-25.01.
[17] A. Go, S. Sutanto, S. Ismadji, Y. Ju, Catalyst free production of partial glycerides:
acetone as solvent, RSC Adv. 5 (2015) 30833–30840, https://doi.org/10.1039/
c5ra03249k.
[18] Aoac, 18th, Official Methods of Analysis of AOAC International, AOAC
International, 2005.
[19] P.L. Tran Nguyen, A.W. Go, L.H. Huynh, Y.H. Ju, A study on the mechanism of
subcritical water treatment to maximize extractable cellular lipids, Biomass
Bioenergy 59 (2013) 532–539, https://doi.org/10.1016/j.biombioe.2013.08.031.
[20] I. Efthymiopoulos, P. Hellier, N. Ladommatos, A. Kay, B. Mills-Lamptey, Effect of
solvent extraction parameters on the recovery of oil from spent coffee grounds for
biofuel production, Waste and Biomass Valorization 10 (2019) 253–264, https://
doi.org/10.1007/s12649-017-0061-4.
[21] G.C. So, D.G. Macdonald, Kinetics of oil extraction from canola (rapeseed), Can. J.
Chem. Eng. 64 (1986) 80–86, https://doi.org/10.1002/cjce.5450640112.
[22] A. Patricelli, A. Assogna, A. Casalaina, E. Emmi, G. Sodini, Factors affecting the
extraction of lipids from sunflower defatted seeds [Fattori che influenzano
l’estrazione dei lipidi da semi decorticati di girasole], La Riv. Ital. Delle Sostanze
Grasse. (1979) 136–142.
[23] A.R. Linares, S.L. Hase, M.L. Vergara, S.L. Resnik, Modeling yerba mate aqueous
extraction kinetics: influence of temperature, J. Food Eng. 97 (2010) 471–477,
https://doi.org/10.1016/j.jfoodeng.2009.11.003.
[24] S. Sayyar, Z.Z. Abidin, R. Yunus, A. Muhammad, Extraction of oil from jatropha
seeds-optimization and kinetics, Am. J. Appl. Sci. 6 (2009) 1390–1395, https://doi.
org/10.3844/ajassp.2009.1390.1395.
A.W. Go et al.
[25] S. Sulaiman, A.R. Abdul Aziz, M. Kheireddine Aroua, Optimization and modeling of
extraction of solid coconut waste oil, J. Food Eng. 114 (2013) 228–234, https://
doi.org/10.1016/j.jfoodeng.2012.08.025.
[26] L.A. Johnson, E.W. Lusas, Comparison of alternative solvents for oils extraction,
J. Am. Oil Chem. Soc. 60 (1983) 229–242, https://doi.org/10.1007/BF02543490.
14
Biomass and Bioenergy 140 (2020) 105704
[27] J. Crank, The Mathematics of Diffusion, 2d ed., Clarendon Press, 1975.
[28] A. Buci�c-Koji�c, M. Planini�c, S. Tomas, M. Bili�c, D. Veli�c, Study of solid-liquid
extraction kinetics of total polyphenols from grape seeds, J. Food Eng. 81 (2007)
236–242, https://doi.org/10.1016/j.jfoodeng.2006.10.027.