Professional Documents
Culture Documents
Dalton Transactions
Dalton Transactions
Dalton Transactions
Transactions
View Article Online
PAPER View Journal | View Issue
A series of Li4SiO4 was synthesized using LiNO3 and six different silicon precursors. The precipitated-
silica-derived Li4SiO4 presented the highest CO2 capacity in a 10 h sorption test, and ZSM-5-derived
Li4SiO4 demonstrated the most rapid CO2 sorption. The CO2 sorption kinetics predominantly followed
the nucleation mode and could be accurately described by the Avrami-Erofeev model. The Avrami-
Erofeev model provided an in-depth analysis of correlation between sorption performance and material
properties. Both the nucleation speed and nucleation dimensionality affected the overall sorption kinetics.
The kinetics and pore-size distribution suggest that the sorption kinetics was dependent on the quantity
of ∼4 nm-pores which favors nucleation dimensionality. For the cyclic tests, the precipitated-silica-
derived sample presented the poorest performance with the capacity decreasing from 31.33 wt% at the 1st
cycle to only 11.52 wt% at the 30th cycle. However, the sample made from fumed silica displayed an
opposite trend with the capacity increasing from 19.90 wt% at the 1st cycle to 34.23 wt% at the 30th cycle.
The radically distinct behaviour of samples during cycles was on account of the alternation of sorption
Received 23rd April 2018, kinetics. The decrease in ∼4 nm-pores after cycles was responsible for the decrease of nucleation dimen-
Accepted 10th June 2018
sionality for the precipitated-silica-derived sample. The rearrangement during cycles could enrich the
DOI: 10.1039/c8dt01617h pores of ∼4 nm for the fumed silica-derived sample, which improved the nucleation growth, thus enhan-
rsc.li/dalton cing the kinetics with cycles.
9038 | Dalton Trans., 2018, 47, 9038–9050 This journal is © The Royal Society of Chemistry 2018
View Article Online
One such sorbent which has been widely used is a CaO- The consequent CO2 capacity was increased from 2.0 mmol g−1
based material due to its high theoretical sorption capacity (8.8 wt%) to 4.3 mmol g−1 (18.9 wt%). Venegas et al.43 employed
(78 wt% or 17.7 mmol g−1) and low cost. However, it has been a precipitation method to achieve a small particle size of
well known to suffer from rapid decays over multiple cycles of Li4SiO4, being around 3 μm as opposed to 36 μm via the solid-
sorption and desorption due to sintering-related state reaction, and the smaller particles lowered the reaction
deactivation.13–16 In order to minimize the sintering effect and activation energy by half. Subha et al.44 found that Li4SiO4 pre-
improve the cycling stability, sacrifice of the sorption capacity pared via sol–gel routes showed improvements over the solid-
was accepted by incorporating high-melting-point inert state samples in terms of both CO2 capacity (30 wt% or
Published on 11 June 2018. Downloaded by Dalian University of Technology on 8/6/2018 8:55:53 AM.
materials such as Ca9Al6O18,17 Ca2Al2SiO7,18 CaZrO3,19 6.8 mmol g−1 for sol–gel and 17 wt% or 3.8 mmol g−1 for the
Ca2SiO4,19 MgO 20 and Al2O3 21,22 into CaO as a stabilizer (or solid state) and sorption kinetics (22.5 mg g−1 min−1 for sol–gel
‘spacer’). To avoid the complexity and potential cost increment and 13.2 mg g−1 min−1 for the solid state). It was believed that
by making such ‘hybrid’ type sorbents, it is motivated to doping hetero elements may create lattice vacancies which
develop more efficient alternative sorbents that can be used accelerate O2− hopping, thus favouring CO2 sorption. Gauer
over repeated cycles. Recently, alkali metal oxide CO2 sorbents et al.45 incorporated Al and Fe into Li4SiO4 and promoted the
such as Li4SiO4, Li2ZrO3 and Na2ZrO3 have attracted increasing sorption rate from 0.01 to 0.12 and 0.15 wt% min−1. Seggiani
attention for their good CO2 chemisorption capacities and et al.46 developed K2CO3-doped Li4SiO4 and achieved a CO2
excellent cycling stability.23,24 They could react with CO2 to capacity of 27 wt% (or 6.1 mmol g−1). Chen and co-workers47
form carbonate as demonstrated in eqn (1)–(3)23,25,26 synthesized Ca-doped Li4SiO4 which reached a CO2 capacity of
35.1 wt% (or 8.0 mmol g−1). The Si precursor also plays a role in
Li4 SiO4ðsÞ þ CO2ðgÞ Ð Li2 CO3ðsÞ þ Li2 SiO3ðsÞ ; affecting the performance of Li4SiO4 sorbents. Shan et al.48
ð1Þ
ΔH298 ¼ 142 kJ mol1 compared the performance of Li4SiO4 derived from diatomite
and pure SiO2. The sorption capacity of diatomite-derived
Li2 ZrO3ðsÞ þ CO2ðgÞ Ð Li2 CO3ðsÞ þ ZrO2ðsÞ ; Li4SiO4 reached 28.62 wt% (or 6.5 mmol g−1) which was
ð2Þ
ΔH298 ¼ 160 kJ mol1 higher than that for pure SiO2-derived Li4SiO4 (22.08 wt% or
5.0 mmol g−1). Kim et al.40 prepared Li4SiO4 by fumed silica
Na2 ZrO3ðsÞ þ CO2ðgÞ Ð Na2 CO3ðsÞ þ ZrO2ðsÞ ; with three different Li precursors: LiOH, LiNO3 and Li2CO3. The
ð3Þ sorbent made from LiOH was identified as the best in their
ΔH298 ¼ 149 kJ mol1
study, showing a CO2 capacity of 29.8 wt% (or 6.8 mmol g−1)
The theoretical capacities of Li-based sorbents (36.7 wt% or and a sorption rate of 56.1 mg g−1 min−1. Pan et al.49 developed
8.3 mmol g−1 for Li4SiO4 and 28.7 wt% or 6.5 mmol g−1 for Li4SiO4 from very porous silicon precursor SBA-15 and reported
Li2ZrO3) are higher than that of the corresponding Na-based the highest CO2 capacity (36.3 wt% or 8.3 mmol g−1) in the
one (23.8 wt% or 5.4 mmol g−1 for Na2ZrO3). This was likely open literature so far. Another important concern of Li4SiO4 is
the reason why Li species were reasonably more attractive for the decay of capacity during the cyclic sorption–desorption.
researchers. From an industrial perspective, Li4SiO4 is advan- Almost all the reported cyclic capacities of Li4SiO4 declined
tageous over Li2ZrO3 because silicon (Si) is the second most during cycles with different levels, and the mechanistic reason
abundant element on the earth’s surface and silicon oxide for this decline is still yet to be disclosed.
(silica) as a precursor is more widely available compared with However, Li-based sorbents have been criticized for their
zirconium oxide (zirconia). In addition to the materials above, slow CO2 sorption kinetics which limits their development and
a number of newly developed Li-based ceramics were also application.50–53 Since kinetics is the limiting factor of the
investigated for CO2 capture, such as Li2CuO2,27–29 Li2TiO3,30 application of Li-based sorbents, theoretical investigation of
Li5AlO4,31,32 Li5FeO4,33–36 Li2O–Bi2O3 37 and LiNiO2.38 Current sorption kinetics is of great importance for interpreting the
research on these materials focuses on the improvement of reaction mechanism and improving material design. The
sorption capacity. Even though some of them showed higher double exponential model, being a popular data fitting tool,
CO2 capacity than Li4SiO4, the material cost still remains as an has been frequently used to fit the kinetic sorption profile for
issue. Moreover, further investigations on the cycling stability alkali metal ceramics.39,43,44,48,54 The shrinking core model, as
of these newly developed materials are still needed. a common model for solid–fluid interaction, has also been
In view of the potential of Li4SiO4 as a CO2 sorbent, a used in studying the CO2 sorption kinetics using alkali metal
number of studies have been conducted to optimize the per- based sorbents.55,56 Zhang et al.57 applied the shrinking core
formance of cyclic sorption. Li4SiO4 is commonly prepared by model and the double exponential model as well as the
solid state reactions between Li and Si precursors.39,40 It has Avrami-Erofeev nucleation model to interpret the CO2 sorption
been found that a smaller particle size of the precursor results kinetics in Li4SiO4 and found that only the Avrami-Erofeev
in a more complete reaction during synthesis and thus higher model is able to explain the reaction mechanism and give
purity of Li4SiO4, which naturally leads to greater CO2 capture reasonable fitting.
capacity.41 Romero-Ibarra et al.42 deployed post-treatment of As discussed above, Li4SiO4 sorbents developed from
Li4SiO4 by mechanical ball-milling and succeeded in reducing diverse precursors resulted in different sorption performances.
the crystal size and increasing the surface area of the material. This work aims to model and interpret the sorption kinetics of
This journal is © The Royal Society of Chemistry 2018 Dalton Trans., 2018, 47, 9038–9050 | 9039
View Article Online
Li4SiO4 samples made from six different Si precursors and of 30 °C min−1 to remove the pre-sorbed substances such
develop the correlation of sorption kinetics with the textual as H2O and CO2. Then the chamber temperature was con-
property of the material. Since cyclic sorption–desorption gen- trolled to 650 °C under a stream of pure CO2 at 60 ml min−1
erally alters the performance of CO2 capture, to the best of our for 10 h.
knowledge, there has not been any theoretical study to explain Cyclic sorption/desorption tests were also carried out to
the mechanism of this performance change during cycles. evaluate the long term stability of CO2 capture performance.
This work applied the Avrami-Erofeev model, which has been After pre-calcination in a 60 ml min−1 flow of pure N2 at
proved as a reasonable model of CO2 sorption in Li4SiO4, to 650 °C, the sample was kept at 650 °C under a stream at 60 ml
Published on 11 June 2018. Downloaded by Dalian University of Technology on 8/6/2018 8:55:53 AM.
gain insight into the kinetic change caused by cycles. min−1 periodically switched between pure CO2 and pure N2 30
times. The duration of purging the CO2 stream in a cycle was
60 min and for the N2 purging it only lasted for 10 min.
2. Experimental and methodology The CO2 capacity of each cycle (Qi) is defined in eqn (4)
2.1 Material synthesis mmax;i mmin;i
Qi ¼ 100% ð4Þ
m0
The Li4SiO4-based sorbents were prepared by using the solid-
state reaction of LiNO3 with pure SiO2 with a molar ratio of where mmax,i and mmin,i are the maximum mass and initial
nLi : nSi = 4 : 1. Six types of pure SiO2, namely SBA-15 (JCNANO, mass at the ith cycle and m0 is the minimum mass of the
Nanjing, China), MCM-41 (JCNANO, Nanjing, China), ZSM-5 sample during the entire test. The total capacity (Qtotal ) over
(JCNANO, Nanjing, China), fumed silica (S5130; Sigma- the entire 30 cycles was also calculated in eqn (5)
Aldrich, USA), precipitated silica (Sipernat-306; Evonik, Essen,
X
30
Germany) and waste-derived silica, were selected as the Si pre- Qtotal ¼ Qi ð5Þ
cursors. The waste-derived SiO2 was synthesized through the i¼1
hydrolysis of SiCl4 (99 wt%; LDK Solar Ltd, Xinyu, China) at
Since the only difference among these Li4SiO4 samples is
150 °C and was collected and dried at 105 °C for 2 h. Initially,
the Si precursors, the Si precursor’s names are used to rep-
0.0666 mol LiNO3 (Fengshun Fine Chemicals Ltd, Shanghai,
resent the samples investigated in this study. For instance,
China) was dissolved in 40 ml of absolute ethanol (99.8%,
LS-SBA-15 means the Li4SiO4 made from the SBA-15 precursor.
Beijing Chemical works) and then sonicated for 20 min until it
completely dissolved. Then, 0.0166 mol pure SiO2 was added
to the solution and the mixture was stirred for 24 h at 300 rpm
at 60 °C until it evaporated completely. After drying in a 3. Results and discussion
vacuum oven at 60 °C for 8–12 h, the mixed powders were cal-
3.1 Characterization results
cined at 750 °C for 2 h with a ramping rate of 10 °C min−1.
After the calcination process, the products were ground to Fig. 1 shows the XRD patterns of the samples developed from
powder again in an agate mortar for later use. the six different Si precursors. The presence of Li4SiO4 was
confirmed in all the samples. No impurities were observed
2.2 Material characterization
Powder X-ray diffraction (XRD) was carried out on a Dmax/2500
X-ray diffractometer (Rigaku, Japan) with a 2θ range of 10–90°
(Cu Kα radiation, λ = 0.1542 nm). Morphological properties of
samples were detected using scanning electron microscopy
(SEM, Zeiss MERLIN VP Compact). All the sorbents for SEM
were sputter-coated with a layer of platinum (∼5 nm). N2 physi-
sorption was conducted on a Quantachrome autosorb iQ-C
instrument. Prior to analysis, the sorbent was degassed at
300 °C until the pressure increment rate in the sample
chamber was less than 20 μmHg min−1. The Brunauer–
Emmett–Teller (BET) surface area was estimated based on the
N2 adsorption isotherms with a P/P0 range of 0.05–0.3. The
total pore volume and pore size distribution were calculated
according to the Barrett–Joyner–Halenda (BJH) method.
9040 | Dalton Trans., 2018, 47, 9038–9050 This journal is © The Royal Society of Chemistry 2018
View Article Online
This journal is © The Royal Society of Chemistry 2018 Dalton Trans., 2018, 47, 9038–9050 | 9041
View Article Online
Table 1 Surface area, pore size and pore volume of the as-prepared also present on the external surface. Therefore as shown in
samples Fig. 4, Li2CO3 and Li2SiO3 were more likely to be mixed
together than to be separated.
Average Pore
BET surface External pore size volume The CO2 sorption kinetics of the freshly calcined samples
Samples area (m2 g−1) area (m2 g−1) (nm) (cm3 g−1) in 10 h at 650 °C was plotted as CO2 uptake versus time
(Fig. 5a) and the CO2 capture rate (Fig. 5b) which is defined as
LS-SBA-15 1.09 0.45 3.94 0.009
LS-MCM-41 0.99 0.99 4.18 0.008 the first order derivative of CO2 uptake with respect to time.
LS-ZSM-5 1.46 1.46 3.94 0.012 The sample derived from precipitated silica presented the
Published on 11 June 2018. Downloaded by Dalian University of Technology on 8/6/2018 8:55:53 AM.
dN ¼ V Ṅdτ ð6Þ
Fig. 4 CO2 uptake profile (a) and CO2 capture rates (b) of as-prepared where dN is the number of nuclei in volume V generated
Li4SiO4 samples during a single cycle test (in pure CO2 at 650 °C for during time dτ and Ṅ is the nuclei generation rate.
10 h). The second assumption is that growth rate of the nuclei
size is independent of the transformation. So the radius of
nuclei is only a function of time.
were formed with Li2SiO3 as the internal layer and Li2CO3 as r ¼ ṙðt τÞ ð7Þ
the external layer.40,41,47,60 However, the surface of the fully
chemisorbed sorbent presented the Si element in the EDX where ṙ is the radius growth rate, τ is the time when nuclei are
mapping in this study (Fig. S1†) indicating that Li2SiO3 was generated and t is the time of growth (so t > τ).
9042 | Dalton Trans., 2018, 47, 9038–9050 This journal is © The Royal Society of Chemistry 2018
View Article Online
The ideally converted volume of the solid is then expressed to the number of available germ nuclei remaining in the
as: material
Upon integration and by defining α = Vconvert/V, eqn (9) By assuming k = k″N0 and omitting lower order terms pro-
yields: vided that t is large, the final expression of eqn (15) is
ideal
Vconvert π lnð1 αÞ ¼ kt 3 ð16Þ
lnð1 αÞ ¼ ¼ ṙ 3 Ṅt4 ð10Þ
V 3
Eqn (11) for homogeneous nucleation and eqn (16) for
Therefore, eqn (10) can be simplified as: heterogeneous nucleation presented a similar format for
3-dimensional nucleation but heterogeneous nucleation
lnð1 αÞ ¼ kt 4 ð11Þ showed a lower order for time terms.
where πṙ3Ṅ/3 is lumped as k. Eqn (11) described the relation of The above deductions were based on the assumption that
conversion and time for homogeneous nucleation. But some- nuclei undergo three dimensional growth, which leads to the
times nucleation is not absolute homogeneous. For hetero- power exponent of t as 4 and 3 for homogeneous and hetero-
geneous nucleation, the nucleation growth starts from a geneous nucleation, respectively. For the situation where the
number of germ nuclei in the material. At the commencement nuclei evolve in other numbers of dimensionalities (n), a
of the nucleation, there are a number of N0 germ nuclei. With general format of the nuclei growth model for homogeneous
the progression of nucleation, the number of germ nuclei and heterogeneous nucleation should be:
available for nucleation declines due to two possible phenom- lnð1 αÞ ¼ kt nþ1 ð17Þ
ena: the first is that nucleation already occurred from some
germ nuclei and this part of nuclei is no longer available; the lnð1 αÞ ¼ kt n ð18Þ
second effect comes from the heterogeneity of the reaction as
Eqn (17) and (18) imply that the reaction kinetics depends
some germ nuclei might be incorporated or ingested by the
on the constant k and nucleation dimensionality n + 1 (or n).
actively growing growth nuclei. Hence, the rate of dis-
k indicates the speed of nucleation, and the nucleation dimen-
appearance of germ nuclei is given by
sionality number indicates the available dimensions for nuclei
dN ¼ dN′ þ dN″ ð12Þ to grow. Taking logarithm for both sides, eqn (17) and (18)
provide,
The reduction of germ nuclei due to the transformation
from germ nuclei to growing nuclei is proportional ln½lnð1 αÞ ¼ ln k þ ðn þ 1Þln t ð19Þ
This journal is © The Royal Society of Chemistry 2018 Dalton Trans., 2018, 47, 9038–9050 | 9043
View Article Online
ln½lnð1 αÞ ¼ ln k þ n ln t ð20Þ the product shell became even thicker, some germ nuclei may
be fully surrounded by the converted volume, so they had no
Plotting ln[−ln(1 − α)] vs. ln(t ) yields a straight line with available dimensionality to grow. Consequently, the kinetics
slope = n + 1 or n. The magnitude of n implies the reaction would eventually shift to a much slower stage (ln(t ) > 3) as
mechanism. When n > 1, the model is known as ‘nucleation shown in Fig. 6. Since the kinetics in the second stage seem to
and nuclei growth model’ which means that the reaction rate be different from the stage prior to the inflection point,
is determined by the rate of nuclei growth.57 For the situation forcing it to fit the Avrami-Erofeev model to get a slope value
when n < 1, the reaction generally proceeds by diffusion does not have any physical sense. It is more appropriate to
Published on 11 June 2018. Downloaded by Dalian University of Technology on 8/6/2018 8:55:53 AM.
control.63 examine what the reaction mechanism of this stage is. The
Reaction (1) can be regarded as a transformation from first possible cause of the reduced reaction kinetics is that
Li4SiO4 to Li2CO3 and Li2SiO3, so Li2CO3 and Li2SiO3 are the nucleation can only proceed towards the centre of a particle
nuclei. Fig. 6 displays the plot of ln[−ln(1 − α)] to ln(t ) in the after the continuous product shell is formed. Hence, the
first cycle for the six as-prepared samples. Two distinct stages nucleation rate is limited by the geometry of the particle. This
were observed for all the samples as fast and slow periods, type of process could be described by the reaction-controlled-
respectively. By comparing with the data of pore size distri- shrinking-core-model (or known as the contracting model);
bution in Fig. 2, the slopes of the initial fast stages in Fig. 6 the second possibility is that the product shell inhibits the
showed a consistent order with the order of ∼4 nm pores contact of reactants; therefore CO2 has to diffuse though the
(nLS-ZSM5 > nLS-Precipitated > nLS-MCM41 > nLS-Fumed ≈ nLS-Waste > product shell for further reaction. This type of process can be
nLS-SBA15). This strongly implied that ∼4 nm pores have a posi- generally described by the diffusion-controlled-shrinking-core-
tive effect on the sorption kinetics. A possible reason might be model.
that pores in this size allow rapid penetration of CO2 gases For the reaction controlled model (contracting model), if
into the interior of sorbent particles, and therefore some germ the nucleation proceeds towards the centre at a constant
nuclei grow from the core with all the dimensionalities avail- speed, the conversion with time is expressed as61
able to propagate. In contrast, without these pores, germ
nuclei were only generated from the particle surface. The avail- ð1 αÞ1=3 þ ð1 α0 Þ1=3 ¼ kðt t0 Þ ð21Þ
able direction for surface germ nuclei to grow is only towards
where k is a constant and α0 is the conversion at time t0. So if
the particle centre. The nucleation dimensionality was then
the plot of −(1 − α)1/3 against t follows a straight line, the
significantly reduced. The existence of ∼4 nm pores tended to
second stage after the inflection point corresponds to a reac-
proceed with the nucleation more homogeneously, leading to
tion controlled model.
a higher value of the kinetic slope (growth dimensionality).
However, in the second scenario where the reaction is
With the evolution of the product shell, the nucleus was
limited by CO2 diffusion across Li2SiO3 and the Li2CO3 shell,
more likely to grow into the converted volume, but growth in
the conversion is governed by eqn (22)
this manner did not contribute to the total CO2 capture. When
ð1 αÞ2=3 1 α ð1 α0 Þ2=3 1 α0
þ ¼ kðt t0 Þ ð22Þ
2 3 2 3
A straight line of [(1/2)(1 − α)2/3 − (1/3)(1 − α)] against t
should be expected if the reaction in the second stage is
limited by diffusion.
Fig. 7 shows that the diffusion controlled model gave a
better fitting to the experimental data than the reaction con-
trolled model, indicating that the slow kinetics in the second
stage is more likely to be diffusion limited.
9044 | Dalton Trans., 2018, 47, 9038–9050 This journal is © The Royal Society of Chemistry 2018
View Article Online
reach the same value of the previous cycle, but also the desorp-
tion process failed to come back to the original base line. This
incomplete sorption and desorption resulted in poor reversi-
bility of Reaction (1). This kind of negative phenomenon was Fig. 8 Sorption and desorption performance in 30 cycles. Conditions:
sorption in pure CO2 for 60 min and desorption in pure N2 for 10 min at
especially evident for samples made from precipitated silica, 650 °C.
MCM-41 and waste silica. To inspect the effect of the desorp-
tion extent on the sorption capacity of the following cycle, the
30-cycled LS-Precipitated was further tested with five
additional sorption–desorption cycles. But the desorption time
in each additional cycle was extended to 30 min to achieve
more complete desorption. Remarkable improvement of sorp-
tion capacity was observed after more complete desorption
(Fig. S2†), but the capacity was still much lower than the
theoretical value. This indicated that incomplete desorption
did not solely result in the poor cyclic capacity. The uptake
slopes at the end of each cycle were still steep for these three
samples, suggesting that 60 min were insufficient to reach
their CO2 uptake ceiling. Surprisingly, the sorbent derived
from fumed silica captured more and more CO2 cycle by cycle Fig. 9 CO2 capture capacities for the 30 cycles. (a) Capacity of individ-
until the end of Cycle No. 30. At the last cycle, the CO2 uptake ual cycle; (b) total capacity over the 30 cycles.
was 34.23 wt%, corresponding to 93.27% conversion of
Li4SiO4. It seemed that the slope of CO2 uptake was nearly flat
at the end of each cycle, indicating that 60 min sorption was
sufficient for a full carbonation. the 1st cycle. The sample derived from precipitated silica,
The cyclic capacity in the 30 cycles is summarized in Fig. 9. which was identified as the best sorbent based on the single
The order of total uptake over the 30 cycles (Fig. 9b) was: LS- 10 h test (Fig. 5a), exhibited the most disappointing perform-
fumed > LS-ZSM-5 > LS-SBA-15 > LS-MCM-41 > LS-waste > LS- ance for the long term cyclic test. Its individual capacity for
precipitated, which is identical to the order of the individual Cycle No. 30 was 11.52 wt%, which was only 1/3 of that at
capacity at the 30th cycle (Fig. 9a). This means that the capacity Cycle No. 3 (the cycle it peaked in terms of individual cycle
at the last cycle was more important than the initial capacity at uptake). Its total CO2 captured all over 30 cycles was as low as
This journal is © The Royal Society of Chemistry 2018 Dalton Trans., 2018, 47, 9038–9050 | 9045
View Article Online
1 7 17
5 42 29
10 61 27
15 72 25
Fig. 10 XRD patterns of LS-fumed and LS-precipitated samples after 20 71 23
30 cycles. 30 71 20
9046 | Dalton Trans., 2018, 47, 9038–9050 This journal is © The Royal Society of Chemistry 2018
View Article Online
Table 3 Surface area, pore size and pore volume of LS-fumed and LS- tics was mainly dependent on the ∼4 nm pores, the enrich-
precipitated samples before and after 30 cycles ment of such pores was responsible for this improved cyclic
kinetics.
BET surface Average pore Pore volume
Samples areaa (m2 g−1) size (nm) (cm3 g−1) With higher cycle numbers (>10), the evident distinction
between LS-Precipitated and LS-Fumed was that
LS-fumed before 30 cycles 1.05 3.94 0.007
LS-Precipitated started from a higher conversion value but
LS-fumed after 30 cycles 3.65 3.94 0.034
LS-precipitated before 1.44 3.95 0.009 grew with a lower number of dimensionality afterwards. The
30 cycles higher starting conversion value of LS-Precipitated came from
Published on 11 June 2018. Downloaded by Dalian University of Technology on 8/6/2018 8:55:53 AM.
This journal is © The Royal Society of Chemistry 2018 Dalton Trans., 2018, 47, 9038–9050 | 9047
View Article Online
will swell after being converted from Li4SiO4 to Li2SiO3 and 10 h sorption test in a pure CO2 stream at 650 °C. It was found
Li2CO3, because the molar volume of Li2SiO3 and Li2CO3 that the sorption kinetics had a positive correlation with the
(35.542 cm3 mol−1 and 35.019 cm3 mol−1) is higher than the quantity of ∼4 nm pores. As such, LS-ZSM-5 which had the
molar volume of Li4SiO4 (50.146 cm3 mol−1). If the pores are most number of ∼4 nm pores exhibited the most rapid sorp-
too small, the volumetric expansion would close the narrow tion kinetics. The cyclic performance of these sorbents was
pore as depicted in Fig. 13, thus causing the blockage of CO2 tested for 30 cycles of sorption/desorption. After 30 cycles, all
transfer. In this small pore scenario, nucleation can only com- the materials presented a smaller particle size due to a self-
mence from the external part of the sorbent. The experimental rearrangement effect with cycles. Only the sorbent made from
Published on 11 June 2018. Downloaded by Dalian University of Technology on 8/6/2018 8:55:53 AM.
observation and theoretical analysis in this study indicated fumed silica presented outstanding cycling stability with its
that ∼4 nm was highly likely the optimum size for Li4SiO4 capacity growing from 19.90 wt% for Cycle No. 1 to 34.23 wt%
sorbents. Pores of this size were highly possible from the for Cycle No. 30. Even though the cycles reduced the particle
minor cracks of particles, which allowed CO2 diffuse rapidly size, all the other sorbents showed deteriorated sorption
into the interior of particle without being blocked when the capacity with cycles, especially for the sorbent made from pre-
product was generated. In such a manner, the nucleation cipitated silica. The diverse cyclic behaviour of different sor-
could commence from anywhere of the particle. Hence, the bents was attributed to the sorption kinetics after cycles. The
nuclei had many dimensionalities to grow. However, the Avrami-Erofeev model was used to interpret the rates of nuclea-
nucleation only started from the particle surface if the pores tion for the cycled LS-Fumed sorbent. The kinetics was
are too small. Once the product shell was formed, the inner boosted due to more homogeneous nucleation which might be
part did not have a chance to be converted by the nucleation related to the enrichment of ∼4 nm pores after 30 cycles. The
mechanism, instead, the inner part could only be converted cycled LS-Precipitated sample only showed an improved start-
via the diffusion limited mode. The favourable sorption kine- ing point due to the remaining Li2CO3 and Li2SiO3 from the
tics by ∼4 nm pores was also evident in a number of other incomplete regeneration in the previous cycle. But the
Li4SiO4 studies.65–67 decrease of ∼4 nm pores in the cycled LS-Precipitated sample
The Li4SiO4 conversion at the inflection point from the reduced the dimensionality of nucleation and thus slowed
nucleation stage to the diffusion-controlled stage denoted the down the overall sorption kinetics.
extent of sorption via nucleation growth. As shown in Table 2,
the conversion ratio at the inflection point increased with
cycle numbers for fumed silica derived sorbents. In the first Abbreviations
cycle, the sorbent conversion was only α ≈ 7% in the nuclea-
tion stage. After the inflection point, all the remaining Li4SiO4 D Diffusion coefficient in the product shell, m2 s−1
had to be converted in the diffusion limited mode. With k Constant
increasing cycle numbers, the conversion after nucleation k′ Rate of germ nuclei disappearance
stage increased to α ≈ 42% at the 5th cycle, 61% at the 10th m0 Minimum mass of the sample during the entire test,
cycle, 72% at the 15th cycle, 71% at the 20th cycle and 71% at g
the 30th cycle, respectively. Although the slow diffusion would mmax,i Maximum mass at the ith cycle, g
limit further CO2 capture, the diffusion limited stage only mmin,i Minimum mass at the ith cycle, g
accounted for a minor portion of the total CO2 sorption. This N Number of nuclei
implied that nucleation growth played a progressively impor- Ṅ Generation rate of nucleus, s−1
tant role in a cyclic sorption test, especially with higher cycle N′ Reduction of germ nuclei due to transformation
numbers. For the precipitated silica derived sample, the N″ Reduction of germ nuclei due to ingestion
nucleation stage converted 17% of the Li4SiO4 at the first cycle, N Number of dimensions
which is still rather a low conversion. But the diffusion con- Qi CO2 capture capacity at the ith cycle, wt%
trolled stage demonstrated a decent rate of CO2 capture in the Qtotal Total CO2 capture capacity, wt%
first cycle, so this material performed great CO2 capacity R Radius of nuclei, m
initially. With increasing cycle numbers, the conversion by ṙ Growth rate of nuclei radius, m s−1
nucleation growth kept declining from the 5th cycle to the 30th t Time for sorption, s
cycle for the LS-Precipitated sample. Likewise, the sorption V Volume, m3
rate in the diffusion controlled stage also deteriorated with Videal
convert Ideally converted volume, m3
cycles, thus leading to a declining trend of sorption capacity. Vconvert Real converted volume, m3
α Sorbent carbonation conversion
τ Time when nuclei are generated, s
4. Conclusions
Six different Si precursors were used to synthesize Li4SiO4 sor- Conflicts of interest
bents. The sorbent derived from precipitated silica demon-
strated the highest CO2 capacity (0.35 g-CO2 per g-sorbent) in a There are no conflicts to declare.
9048 | Dalton Trans., 2018, 47, 9038–9050 This journal is © The Royal Society of Chemistry 2018
View Article Online
This journal is © The Royal Society of Chemistry 2018 Dalton Trans., 2018, 47, 9038–9050 | 9049
View Article Online
47 X. Chen, Z. Xiong, Y. Qin, B. Gong, C. Tian, Y. Zhao, D. Lardizábal Gutiérrez and V. Collins-Martínez,
J. Zhang and C. Zheng, Int. J. Hydrogen Energy, 2016, 41, Int. J. Hydrogen Energy, 2014, 39, 16656–16666.
13077–13085. 57 Q. Zhang, D. Han, Y. Liu, Q. Ye and Z. Zhu, AIChE J., 2013,
48 S. Shan, Q. Jia, L. Jiang, Q. Li, Y. Wang and J. Peng, Chin. 59, 901–911.
Sci. Bull., 2012, 57, 2475–2479. 58 D. M. Ruthven, Principles of adsorption and adsorption pro-
49 Z. Xu, C. Shan, B. Xie, Y. Liu and B. Pan, Appl. Catal., B, cesses, John Wiley & Sons, 1984.
2017, 200, 439–447. 59 D. D. Duong, Adsorption analysis: equilibria and kinetics,
50 K. B. Yi and D. Ø. Eriksen, Sep. Sci. Technol., 2006, 41, 283– Imperial College Press, 1998.
Published on 11 June 2018. Downloaded by Dalian University of Technology on 8/6/2018 8:55:53 AM.
9050 | Dalton Trans., 2018, 47, 9038–9050 This journal is © The Royal Society of Chemistry 2018