Energy Fuels 2010, 24, 4756–4765 Published on Web 03/24/2010
: DOI:10.1021/ef901346f
Adsorption on Activated Carbons of Five Selected Volatile Organic Compounds Present in Biogas: Comparison of Granular and Fiber Cloth Materials† Benoit Boulinguiez*,‡,§ and Pierre Le Cloirec‡,§ ‡
Ecole Nationale Sup erieurede Chimie de Rennes, Centre National de la Recherche Scientifique (CNRS), UMR 6226, Avenue du G en eral Leclerc, CS50837, 35708 Rennes CEDEX 7, France, and §Universit e europ eenne de Bretagne, 5 Boulevard La€ ennec, 35000 Rennes, France Received November 12, 2009. Revised Manuscript Received March 15, 2010
The adsorption of a selected panel of five volatile organic compounds, common in biogas, on four different activated carbons was evaluated in laboratory-batch experiments. The experiments were performed with a synthetic biogas. The four tested commercial adsorbents consisted of two granular and two fiber cloth activated carbons. The adsorptions were satisfactorily fitted by the Langmuir-Freundlich model and related to the inherent porosity characteristics of the adsorbents. The most porous adsorbents showed the best compromise in adsorption capacities for the five compounds, but the adsorption kinetics confirmed the importance of fiber cloth morphology in activated carbon materials. The early step of thermal regeneration was investigated and suggested an optimal temperature of regeneration higher than 200 °C for any carbon, despite being partially hampered by the formation of non-volatile compounds.
covery considerations, methane, which is the main component of biogas (about 60%), is an environmental threat, that contributes ultimately to the greenhouse effect by a factor of about 20 compared to carbon dioxide. Methanogenesis depends upon various parameters, such as temperature, humidity, process, and organic source, which generates biogases as very complex matrixes, including a large number of volatile organic compounds (VOCs). Screening and quantification studies of the VOCs in biogases from different origins have established a pattern of the most recurrent chemical families or compounds.9-12 Hence, the VOC composition of any biogas consists of a group of sulfur, aromatic, aliphatic, siloxane, and halogenated compounds. Among the trace components of any biogas, hydrogen sulfide is notable for its relatively high concentration, more than 100 ppm(v), regardless of the organic source or process considered.9,10,13 It is thus often removed by a specific process, such as scrubbing or dedicated adsorption on doped activated carbon (AC).14,15 While there are various technologies available to purify biogas of carbon dioxide (pressure/temperature swing adsorption, scrubbing, selective adsorption, and membrane separation), efficient and inexpensive technologies are still under development for the removal of VOCs.12,16,17 Although the trace VOCs account for less than 1% of
Introduction The predication of a global waste and energy crisis has led to a number of different options being investigated to anticipate and regulate this phenomenon. Among these, the use of biogas is of growing importance because of its considerable environmental benefits.1 Many projects have been established in different regions of the globe because biogas plants can use various sources of organic material.2 For example, biogas production from municipal solid waste has been a driving technology in the U.S.A. and the U.K., while in Europe, a significant number of projects are based on agricultural residues, food industry waste, and wastewater sludge.1,3 In northern Europe, the number of co-digestion and global treatment plants is rising. In the latter, the organic constituents from landfill, agricultural, food industry, and wastewater treatment processes are mixed in an “all in one” unit to generate biogas.4-7 Technical, social, economic, and environmental studies of biogas production are all encouraging the use of such technology to generate energy locally.8 In fact, energy production has remained an important parameter in sustaining the development of biogas production, especially with the rise in energy prices. However, despite energy re† This paper has been designated for the Bioenergy and Green Engineering special section. *To whom correspondence should be addressed. E-mail: benoit.
[email protected]. (1) De Baere, L. Water Sci. Technol. 2000, 41, 283–290. (2) Li, R.; Chen, S.; Li, X.; Lar, J. S.; He, Y.; Zhu, B. Energy Fuels 2009, 23, 2225–2228. (3) Rulkens, W. Energy Fuels 2008, 22, 9–15. (4) la Cour Jansen, J.; Gruvberger, C.; Hanner, N.; Aspegren, H.; Sv€ ard, A. Water Sci. Technol. 2004, 49, 163–169. (5) Neves, L.; Oliveira, R.; Alves, M. Waste Manage. 2006, 26, 176–181. (6) Sosnowski, P.; Wieczorek, A.; Ledakowicz, S. Adv. Environ. Res. 2003, 7, 609–616. (7) Amon, T.; Amon, B.; Kryvoruchko, V.; Machmuller, A.; HopfnerSixt, K.; Bodiroza, V.; Hrbek, R.; Friedel, J.; Potsch, E.; Wagentristl, H.; Schreiner, M.; Zollitsch, W. Bioresour. Technol. 2007, 98, 3204–3212. (8) Murphy, J.; McKeogh, E. Energy 2006, 31, 294–310.
r 2010 American Chemical Society
(9) Jaffrin, A.; Bentounes, N.; Joan, A. M.; Makhlouf, S. Biosyst. Eng. 2003, 86, 113–123. (10) Rasi, S.; Veijanen, A.; Rintala, J. Energy 2007, 32, 1375–1380. (11) Schweigkofler, M.; Niessner, R. Environ. Sci. Technol. 1999, 33, 3680–3685. (12) Shin, H.; Park, J.; Park, K.; Song, H. Environ. Pollut. 2002, 119, 227–236. (13) Spiegel, R. J.; Preston, J. L. J. Power Sources 2000, 86, 283–288. (14) Le Leuch, L.; Subrenat, A.; Le Cloirec, P. Langmuir 2003, 19, 10869–10877. (15) Truong, L.; Abatzoglou, N. Biomass Bioenergy 2005, 29, 142–151. (16) Herdin, G.; Gruber, F.; K€ uffmeier, R.; Brandt, A. Solutions for siloxane problems in gas engines utilizing landfill and sewage gas. Spring Technical Conference, San Antonio, TX, 2000. (17) Baudu, M.; Le Cloirec, P.; Martin, G. Water Sci. Technol. 1991, 23, 1659–1666.
4756
pubs.acs.org/EF
Energy Fuels 2010, 24, 4756–4765
: DOI:10.1021/ef901346f
Boulinguiez and Le Cloirec
the total gaseous emissions, they exert a disproportionate environmental burden because of their physical and chemical properties. The use of biogas as a high-quality fuel requires pre-upgrading by the removal of undesirable VOCs before the separation process of the methane/carbon dioxide mixture.10,18 Indeed, these VOCs are responsible for degradating equipment by the corrosion or abrasion of mechanical parts. A valuable method to control the VOCs in biogas is the adsorption-desorption process. Among adsorbent materials, AC is commonly used in gas purification because it is relatively inexpensive; the price per kilogram for granular AC is about 2 euros. The recent development of a new form of AC, activated carbon fiber cloth (ACFC), draws attention because of the process opportunity of the Joule effect heating to achieve the desorption and thus regenerate the adsorbent.19-24 However, investment costs of adsorption processes based on ACFC are significant in comparison to technologies using granular activated carbon (GAC), because the price per kilogram of ACFC is 100 times higher than that of GAC. The objective of this study is to determine for a selected panel of five VOCs, typically found in biogas, the adsorption performances of four different ACs to remove these undesirable VOCs, to achieve biogas upgrading, and to evaluate the feasibility of the regeneration of these adsorbents. Thus, the adsorption capacity and kinetics of toluene, dichloromethane, isopropanol, ethyl mercaptan, and D4 siloxane (octamethylcyclotetrasiloxane) on two commercial GACs and two commercial ACFCs were investigated. Batch adsorptions were carried out in a synthetic biogas matrix of methane and carbon dioxide to generate results close to field conditions. The Langmuir and Langmuir-Freundlich adsorption models are compared to fit the adsorption data. The porosity of the ACs is discussed to explain the observed adsorption characteristics of the VOCs. The early step of desorption investigated for the ACs is presented to assess the feasibility of their regeneration.
performed according to a batch procedure in 2 L glass vessels. Fixed known weighed amounts of adsorbent were suspended in identical batch reactors, to which specific weighed amounts of the liquid VOC were added and then volatilized. Prior to the addition of the adsorbent and the adsorbate, reactors were purged and filled with a mixture (55:45, v/v) of methane and carbon dioxide to reproduce a common biogas matrix. The synthetic biogas was set up from gas cylinders of methane and carbon dioxide (Air Liquid). Mass gas flow controllers, calibrated and certified by the manufacturer (EL-Flow F-201, Bronkhorst), ensured the accurate mixing of methane and carbon dioxide in the desired ratio to obtain a stream of 55% methane and 45% carbon dioxide by volume under normal conditions of pressure and temperature. The 4 min purge time of each vessel with the synthetic biogas was 5 times higher than of the residential time at the 150 L h-1 stream flow to ensure adsorption conditions, with regard to the biogas matrix effect, as reproducible as possible among the vessels and throughout the experiments. The relative humidity in the synthetic biogas was below the 10% limit because it was produced from gas cylinders, which are almost water-free, because the considered adsorption process would be set up after the dehydration process. All vessels were placed in a thermostatted bath regulated at 25 ( 0.2 °C for 20 h. This period corresponds to the equilibrium time determined from prior kinetic adsorption experiments reported later in this paper. Equilibrium was considered achieved with a deviance smaller than the measurement error in the concentration measured between three consequent withdrawals from the gas phase. Aliquots of the gas phase were manually injected into a gas chromatograph (FOCUS GC, Thermo Scientific) equipped with a capillary column (RTX-1, 15 m 0.32 mm 0.3 μm, Restek) and a flame ionization detector to quantify the VOC concentration by external calibration. The mass balance between the initial and the equilibrium concentration was used to assess the amount of VOC adsorbed on the AC. The same batch procedure in the 2 L vessel was performed to characterize the adsorption kinetics of the VOCs on each AC Siloxane D4 excluded; however, because its low volatility (133 Pa at 20 °C) was manageable in long-term experiments, it was too low to carry out reliable kinetic experiments. Identical initial conditions were applied for each VOC to provide a reliable basis to determine and compare the kinetics of adsorption. Thus, for a specific amount of AC suspended in the reactor, a specific amount of pure liquid VOC was added, volatilized, and continuously stirred at 25 ( 0.2 °C. Gas samples were withdrawn at regular intervals with a 250 μL gastight syringe and injected into the gas chromatograph. Table 1 presents the general characteristics and dimensional information about the molecules of the studied VOCs. Surface Area and Porosity Measurements. Nitrogen adsorption isotherms on the ACs were determined at 77 K using an Autosorb-1-MP (Quantachrome Instruments). The samples were outgassed at 350 °C under a 10-5 Torr vacuum (1.33 10-7 Pa) overnight.25 The specific surface area (SBET) was calculated using the multipoint Brunauer-Emmett-Teller (BET) method in the relative pressure (p/p0) range of 0.01-0.1.26,27 The total pore volume (V0) was calculated from the amount of nitrogen adsorbed at the relative pressure of 0.995. The micropore volume for a pore width of less than 2 nm (Vmicro), the average pore width (Dave), and the mode pore width (Dw) were calculated using the density functional theory approach, with the assumption of an AC surface heterogeneously composed by slit-shaped pores.
Materials and Methods Adsorbents. The adsorption of the VOCs was tested on four commercial ACs of various origins. The Picabiol, referred to as B1, and N60 are granular adsorbents (Pica, Saint Maurice, France), whereas FM30K (CCI, U.K.) and THC515 (Dacarb, Asnieres, France) are ACFCs. Prior to all experiments, the supplied materials were washed between 3 and 6 times, with regard to their respective content of ashes using a B€ uchner filtration apparatus with ultra-pure water to remove dust and then soaked in ultra-pure water overnight to remove any fine particles and pollutants from the production process. Adsorbents were then dried at 150 °C overnight. VOC Adsorption. The technical position adopted for this study in a global biogas treatment process was to consider a biogas stream after an initial desulfurization process to remove hydrogen sulfide and after a drying step prior to compression. The adsorption isotherms of the VOC with each AC were (18) Schweigkofler, M.; Niessner, R. J. Hazard. Mater. 2001, 83, 183– 196. (19) Subrenat, A.; Baleo, J. N.; Le Cloirec, P.; Blanc, P. E. Carbon 2001, 39, 707–716. (20) Le Cloirec, P.; Brasquet, C.; Subrenat, E. Energy Fuels 1997, 11, 331–336. (21) Grande, C. A.; Rodrigues, A. E. Int. J. Greenhouse Gas Control 2008, 2, 194–202. (22) Luo, L.; Ramirez, D.; Rood, M. J.; Grevillot, G.; Hay, K. J.; Thurston, D. L. Carbon 2006, 44, 2715–2723. (23) Moon, S.; Shim, J. J. Colloid Interface Sci. 2006, 298, 523–528. (24) Sullivan, P. D.; Rood, M. J.; Grevillot, G.; Wander, J. D.; Hay, K. J. Environ. Sci. Technol. 2004, 38, 4865–4877.
(25) Rouquerol, J.; Avnir, D.; Fairbridge, C. W. Pure Appl. Chem. 1994, 66, 1739–1758. (26) Brunauer, S.; Emmet, P.; Teller, E. J. Am. Chem. Soc. 1938, 60, 309–319. (27) Condon, J. Surface Area and Porosity Determinations by Physisorption: Measurements and Theory, 1st ed.; Elsevier: Amsterdam, The Netherlands, 2006.
4757
Energy Fuels 2010, 24, 4756–4765
: DOI:10.1021/ef901346f
Boulinguiez and Le Cloirec
Table 1. Molecular Information and Dimensional Characteristic for the Studied VOCs36,37
Figure 1. Adsorption isotherms at 25 of dichloromethane (left) and siloxane D4 (right) on the carbon materials. The scatter points represent the experimental adsorption data, and respective Langmuir-Freundlich model curves are symbolized by dashed lines. The calculation limits of the Langmuir-Freundlich model for siloxane D4 are symbolized by dotted lines.
models to these experimental points generates model parameters that enable the adsorption capacity of the carbons to be predicted over a relevant concentration range for biogas upgrading. The siloxane adsorption isotherm consists of two regions that are concave to the gas concentration axis and separated by a region that is convex. The concave region that occurs at low gas concentrations is associated with the formation of a single layer of adsorbate molecules over the surface. The convex portion represents the buildup of additional layers, while the other concave region is the result of capillary or intergranular condensation of the adsorbate.28 Indeed, after the experiments at high concentration with siloxane D4, a thin film of liquid phase was observed in the interstices of the ACs, especially with the GACs. Typical concentrations of siloxane in biogas are reported to be about 50 mg m-3; thus, condensation is unlikely to occur in the adsorption of siloxanes from biogas.11,18,29-31 On the other hand, this possibility should not be minimized during the regeneration process because siloxane will be desorbed and the concentration might rise to dozens of millimoles per cubic meter, where
Thermal Analysis. Thermal analysis was performed using a thermogravimetric analyzer (DTA-50, Schimatzu). The desorption of about 20 mg of AC was carried out in a 50 mL min-1 flow rate of nitrogen. The sample was heated from room temperature to 400 °C with a 5 °C min-1 heating rate. The ACs were saturated at 50% of their maximum adsorption capacity with the studied VOC according to the same procedure and with the same equipment described in the adsorption section.
Results and Discussion Adsorption Isotherms. The successful selection of a proper adsorbent for specific applications depends upon a good description of the equilibrium phenomenon between the gas and solid phases. Adsorption equilibrium is established when the amount of adsorbate being adsorbed is equal to the amount being desorbed from the adsorbent. At this point, the equilibrium concentrations in both phases are constant. Plotting the solid-phase concentration against the gas-phase concentration graphically depicts the equilibrium adsorption isotherm, as shown in Figure 1. Except for siloxane D4, which presented an adsorption isotherm of type IV, all VOCs were favorably adsorbed on the carbons with an isotherm curve shape of type I, such as dichloromethane.26 The experimental adsorption conditions were chosen so that the concentration values attained at equilibrium included the typical range of VOC concentrations measured in real biogases produced from either landfill sites or wastewater plants or digesters. Thereby, the application of adsorption
(28) Fournel, L.; Mocho, P.; Fanlo, J. L.; Le Cloirec, P. Environ. Technol. 2005, 1277. (29) Accettola, F.; Guebitz, G.; Schoeftner, R. Clean Technol. Environ. Policy 2008, 10, 211–218. (30) Dewil, R.; Appels, L.; Baeyens, J. Energy Convers. Manage. 2006, 47, 1711–1722. (31) McBean, E. A. Can. J. Civ. Eng. 2008, 35, 431–436.
4758
Energy Fuels 2010, 24, 4756–4765
: DOI:10.1021/ef901346f
Boulinguiez and Le Cloirec
such a phenomenon could occur. One can take advantage of VOC condensation when they are desorbed from ACs.24 Because the first concave part could be fitted as a type I adsorption isotherm, the adsorption models used in this study for siloxane D4 were limited to the first part of the curves, which is the range worth investigating for biogas applications. For the other compounds, the regression was performed on the whole range of the experimental data. Different models, such as Langmuir, Freundlich, or Langmuir-Freundlich, may be used to fit experimental adsorption data. Although it remains a widely used model, the Freundlich model was not considered in this study because it does not give any limit to the adsorption capacity, so that the amount adsorbed goes to infinity as the concentration increases.32 Instead, the experimental data were fitted by the Langmuir and Langmuir-Freundlich models, defined by eqs 1 and 2, respectively qm bCe ð1Þ qe ¼ 1 þ bCe qe ¼
qm bCe 1=p 1 þ bCe 1=p
Table 2. Parameter Estimates of the Langmuir Model of the VOC Adsorption Performed at 25 °Ca Langmuir VOC toluene
sample
B1 NC60 FM30K THC515 isopropanol B1 NC60 FM30K THC515 dichloromethane B1 NC60 FM30K THC515 siloxane D4 B1 NC60 FM30K THC515 ethyl mercaptan B1 NC60 FM30K THC515
ð2Þ
b qm (mmol g-1) (m3 mmol-1) 5.17 (6%) 3.58 (6%) 3.39 (5%) 4.01 (4%) 2.46 (9%) 2.24 (3%) 2.15 (5%) 2.82 (3%) 1.80 (4%) 1.78 (6%) 1.32 (5%) 1.85 (5%) 1.55 (9%) NA 0.67 (5%) 0.45 (51%) 2.09 (10%) 2.07 (9%) 2.05 (4%) 2.03 (3%)
χ2
1.19 (18%) 13.6 6.54 (19%) 8.4 6.75 (18%) 6.4 9.81 (20%) 5.2 0.17 (30%) 34.5 0.13 (10%) 6.7 0.19 (17%) 17.3 0.14 (6%) 2.6 0.02 (8%) 1.8 0.03 (11%) 7.3 0.06 (13%) 10.75 0.04 (13%) 6.10 11.29 (31%) 11.4 NA NA 27.31 (24%) 3.26 7.42 (52%) 234.7 0.18 (34%) 45.7 0.03 (35%) 17.2 3.60 (58%) 49.8 0.04 (14%) 2.03
fit R2 0.92 0.93 0.92 0.94 0.87 0.98 0.97 0.99 0.99 0.99 0.99 0.98 0.86 NA 0.93 0.32 0.73 0.95 0.72 0.97
a The numbers in parentheses are the relative standard error of the estimate.
where qe is the solid-phase equilibrium concentration (mmol g-1), qm is the adsorption maximum capacity (mmol g-1), b is the Langmuir constant (m3 mmol-1), Ce is the gas-phase equilibrium concentration (mmol m-3), and 1/p is the Freundlich isotherm exponent. With the LangmuirFreundlich equation, the unit of b depends upon the value of p; thereby, it becomes (m3/p mmol-1/p). The determination of the model parameters was achieved by a nonlinear regression procedure, in preference to the usual linear transformations.33 The model parameter values for the Langmuir and Langmuir-Freundlich models are reported in Tables 2 and 3, respectively. The convergence criterion chosen to perform the nonlinear regression was the residual square sum weighted by the experimental standard deviation of the experimental data. The coefficient of variation for the experimental adsorption data sets ranges between 5 and 8% throughout the concentration range investigated. The Langmuir-Freundlich model is derived from the Langmuir isotherm by assuming each adsorbate occupies p sites or considering the adsorption on non-uniform surfaces, in other words, the heterogeneous energy of the adsorption sites.34 For both models, the qm parameter represents the maximum adsorption capacity of the AC. Such a value only occurs at a high concentration of adsorbate when the pores in the AC are filled up with adsorbate molecules. In these conditions, qm is the major parameter driving the equation and the adsorption reaches the plateau of saturation. As the concentration of the selected adsorbate falls to low values, typically around a millimole per cubic meter, the influence of the model parameter b increases compared to the parameter qm. The comparison of ACs in such a concentration range needs to consider both of these model parameter values, i.e., b and qm. To compare the two adsorption models, R2 and a modified 2 χ are used. R2 indicates the overall goodness of fit over the
whole range of the independent variable. It is of major interest to note that this indicator does not take into account the degree of freedom of a model. A comparison of models to different degrees of freedom exclusively based on this indicator is likely to lead to a biased conclusion because the more parameters a model has, the more adaptive it is to any set of data, exhibiting good fitting at the expense of the physical meaning of the parameters of the model. As a consequence, the Langmuir-Freundlich model is, by definition, more flexible than the Langmuir model to fit an identical data set, leading to greater R2. For this reason, an indicator independent of the degree of freedom of a model is required to compare sensibly the goodness of fit among the tested models. A modified χ2 was used; its expression includes the degree of freedom of the considered model d. The better the fit, the closer to zero the χ2 value, defined as follows: ! 2 n X ^ 1 ðq q Þ e e χ2 ¼ ð3Þ n - d i ¼1 sqe i
where qe is the experimental value, q^e is the value predicted by the model, n is the number of experimental points for the isotherm, d is the number of parameters of the model, and sqe is the experimental standard deviation of the measured qe. For the Langmuir model, the fitting results are acceptable for most cases; only 3 of 19 R2 values are below 0.85. The interpretation of these parameter values sensibly represents the adsorption of the studied VOCs on the ACs. Relevant trends among the ACs or among the VOCs can be drawn, such as the better adsorption of toluene compared to dichloromethane on any AC or the better adsorption capacities at VOC saturation of the B1 sample compared to the others. Nevertheless, except for dichloromethane, the fittings of the experimental adsorption data by the Langmuir model are more approximate than those by the LangmuirFreundlich model. The Langmuir-Freundlich model is a three-parameter model, whereas the Langmuir model is a two-parameter
(32) Boulinguiez, B.; Le Cloirec, P. Energy Fuels 2009, 23, 912–919. (33) Boulinguiez, B.; Le Cloirec, P.; Wolbert, D. Langmuir 2008, 24, 6420–6424. (34) Bansal, R. C. Activated Carbon Adsorption; CRC Press: Boca Raton, FL, 2005.
4759
Energy Fuels 2010, 24, 4756–4765
: DOI:10.1021/ef901346f
Boulinguiez and Le Cloirec
Table 3. Parameter Estimates of the Langmuir-Freundlich Model of the VOC Adsorption Performed at 25 °Ca Langmuir-Freundlich VOC toluene
isopropanol
dichloromethane
siloxane D4
ethyl mercaptan
a
sample
qm (mmol g-1)
b (m3/p mmol-1/p)
1/p
χ2
fit R2
B1 NC60 FM30K THC515 B1 NC60 FM30K THC515 B1 NC60 FM30K THC515 B1 NC60 FM30K THC515 B1 NC60 FM30K THC515
7.38 (7%) 4.37 (4%) 3.93 (6%) 4.81 (3%) 20.74 (51%) 2.78 (4%) 2.85 (2%) 3.06 (4%) 1.96 (13%) 2.54 (8%) 1.80 (13%) 3.16 (12%) 2.69 (15%) NA 0.85 (10%) 0.74 (1%) 8.97 (19%) 4.78 (23%) 3.00 (51%) 2.53 (5%)
0.48 (13%) 1.71 (15%) 2.20 (28%) 2.99 (19%) 0.03 (52%) 0.14 (3%) 0.19 (2%) 0.15 (5%) 0.02 (10%) 0.04 (5%) 0.06 (8%) 0.04 (6%) 0.95 (33%) NA 1.76 (36%) 21.96 (46%) 0.04 (24%) 0.04 (21%) 0.83 (90%) 0.10 (8%)
0.54 (6%) 0.59 (7%) 0.65 (12%) 0.57 (10%) 0.42 (4%) 0.73 (4%) 0.64 (2%) 0.88 (5%) 0.94 (8%) 0.76 (4%) 0.75 (8%) 0.71 (4%) 0.45 (10%) NA 0.31 (19%) 0.60 (28%) 0.54 (9%) 0.40 (5%) 0.20 (78%) 0.62 (7%)
1.42 0.94 2.91 0.77 0.46 0.52 0.23 1.66 2.06 0.77 2.69 0.45 0.78 NA 0.62 0.08 0.23 0.11 0.16 2.85
0.993 0.993 0.983 0.993 0.998 0.999 0.999 0.995 0.998 0.998 0.999 0.998 0.993 NA 0.998 0.987 0.998 0.999 0.950 0.998
The numbers in parentheses are the relative standard error of the estimate.
Table 4. Specific Surface Area and Porosity Characteristics of the ACs Determined by Nitrogen Adsorption at 77 Ka samples
SBET (m2 g-1)
V0 (cm3 g-1)
Vmicro (cm3 g-1)
Dave (nm)
Dw (nm)
Picabiol B1 NC60 FM30K THC515
1930 1670 1200 1540
1.36 0.70 0.51 0.79
0.70 0.42 0.42 0.57
3.0 2.2 1.8 2.0
0.5 0.9 0.6 0.7
a
SBET, specific surface area; V0, total pore volume; Vmicro, micropore volume; Dave, average pore width; and Dw, mode pore width.
model. Hence, as expected, the R2 values are optimal with the former. For instance, the lowest value of R2 with the Langmuir-Freundlich model is 0.950, while it is 0.32 with the Langmuir model. Because the χ2 values for the Langmuir-Freundlich model are enhanced by a factor greater than 1 for 17 of the 19 isotherms, this suggests that the Langmuir-Freundlich model provides statistically a more adequate equation to fit the adsorption results than the Langmuir model, independent of the VOC or the AC used. However, in the two cases of dichloromethane-B1 and ethyl mercaptan-THC515, for which the χ2 values are of the same order of magnitude in both models, the R2 values show major deviations, whereas there is no significant difference between the model adequacies to fit these sets of data. These examples highlight the bias introduced by the R2 values when they are used to evaluate the fitting of experimental data sets by models with unequal degrees of freedom. The additional parameter 1/p provided by the threeparameter model takes into account adsorption phenomena not considered by the Langmuir model, which assumes ideal monolayer adsorption on energetically similar adsorption sites. For 18 of 19 isotherms, the values of 1/p differ significantly from 1. The adsorption of the VOC on AC is unlikely to occur on uniform energy adsorption sites in the conditions of this study. However, this parameter has no effect on the precision of the values of the adsorption parameters because the model shows no dependence upon the relative standard errors of the parameter estimates. The physical interpretation of this third parameter relies on the non-ideal adsorption conditions of a single compound. The biogas matrix competitively challenges the adsorption of a VOC on the ACs, although the adsorption
of methane and carbon dioxide is known to be limited to a low amount under ambient temperature and pressure conditions.23 Besides, the functional group of the VOC might chemically interact with the functional groups present at the AC surface. The adsorption performance of any AC is a combination of physisorption, which relies on the porosity features of the AC and the dimensions of the adsorbate molecules to build up the van der Waals interactions, plus the chemical interactions between the AC surface functional groups and the adsorbate molecule functional groups.34,35 The porosity characteristics of the ACs are collected in Table 4. These results were determined from nitrogen adsorption performed at 77 K, as shown in Figure 2. Nitrogen adsorption on the B1 sample shows a specific hysteresis loop at high relative pressure, which is characteristic of a material with a significant amount of mesoporosity.27 The adsorption curves of nitrogen on the other samples, however, do not show such a pattern in the adsorption-desorption loop. The cumulative quantity of adsorbed molecules of nitrogen on the B1 sample leads to the determination of the highest pore volume, 1.36 cm3 g-1, which is double the measured value of the second most porous sample, THC515. In addition, the specific surface area of 1930 m2 g-1 developed by the B1 sample is 15% greater than the second largest one measured for the other granular NC60 sample. These differences between the ACs are related to the various origins of the material and the activation process used to produce them. (35) Yang, R. T. Adsorbents: Fundamentals and Applications; WileyInterscience: Hoboken, NJ, 2003; p 424.
4760
Energy Fuels 2010, 24, 4756–4765
: DOI:10.1021/ef901346f
Boulinguiez and Le Cloirec
Figure 2. Nitrogen adsorption at 77 K on the adsorbents to determine their porosity characteristics. Figure 3. Comparison of the relative adsorption capacities at biogas concentrations and concentrations of saturation between the samples B1 and THC515 for the five studied VOCs. (/) For dichloromethane, the values are converted to absolute values because they are negative as a result of the better adsorption on the THC515 sample.
The adsorption results obtained for the studied VOC on the ACs are of the order of magnitude of those reported in the literature.18,28,38-40 For the four largest molecules, i.e., toluene, siloxane D4, ethyl mercaptan, and isopropanol, the B1 sample has the highest maximum adsorption capacities, which are double those of the second highest capacities measured with another AC. This is due to the physisorption that rules the adsorption in these conditions; the large cumulative pore volume and specific surface area shown by the B1 sample enable more molecules to be adsorbed than by an AC with a small cumulative pore volume and specific surface area, such as the FM30K sample. The uniqueness in terms of porosity of the B1 sample compared to the other ACs explains the observed discrepancies between the maximum adsorption capacities for these four VOCs. A large porous volume allows for additional layers of adsorbate to build up on the carbon and, thus, directly increases the capacity at concentrations close to the saturation of the adsorbent.34,35 For toluene and siloxane, the sample THC515 with the second largest cumulative pore volume of 0.79 cm3 g-1 shows the second highest maximum adsorption capacities, 4.81 and 0.74 mmol g-1, respectively. With the latter, the maximum adsorption capacity with FM30K has an uncertainty of 10%, thus remaining of the same order of magnitude as the THC515 results. On the other hand, for ethyl mercaptan and isopropanol, molecules of smaller dimension with heteroatomic functional groups, the trends and amplitudes in the maximum adsorption capacities between THC515, NC60, and FM30K are less pronounced or even inverted, although THC515 undeniably has the best porosity characteristics among these three ACs. The effects of the interactions between the functional groups on the AC surface and the molecule can explain these results. These considerations make even more sense for the adsorption of dichloromethane on the ACs. The NC60 sample shows a maximum adsorption capacity of 2.54 mmol g-1, 20% higher than that of the B1
sample, although none of the pore characteristics of the former is greater than those of the B1 sample. A comparison of these maximum adsorption capacities demonstrates the duality of the adsorption process between physisorption driven by the porosity of the AC and the size of the molecule and chemisorption involving chemical interactions at the boundary between the AC and the adsorbate. Thus, the adsorption of dichloromethane on ACs is probably not limited by physisorption but rather includes chemical interactions that differ in magnitude between adsorbents from different origins and having undergone different activation processes.28 In biogases, VOC concentrations are of the order of magnitude of a few millimoles per cubic meter.11,18,29 Thus, the valuable information from the adsorption isotherms on the ACs is located at the beginning of the curve, where the equation is dependent upon both parameters b and qm and not only the maximum adsorption capacity parameter qm. Because all of the assessed values of the b parameter for the VOCs with the B1 sample are the lowest ones, the high maximum adsorption capacities demonstrated by this AC are hampered at low concentration because of the adverse effect of the weight of b in the equation. Consequently, the amount of adsorbed VOC between the B1 sample and a sample for which the values of qm are lower but the values of b are greater, such as THC515, decreases between the adsorption at saturation and the adsorption at low concentration. This effect is illustrated by Figure 3, by comparing the relative deviances of adsorption capacities between the B1 sample and the THC515 sample at saturation and at biogas concentration, i.e., 3 mmol m-3 for toluene and 1 mmol m-3 for other VOCs. The relative deviance is calculated as follows: (qB1 - qTHC515)/qB1. Isopropanol is the only compound that shows better relative adsorption at low concentration than at saturation, with a deviance of 11% in favor of the B1 sample. The particularly high value of maximum adsorption capacity for the B1 sample in this case could not be compensated by the determined values of b, although they present a consequent discrepancy. It must noted that, for isopropanol, the values
(36) Kurita, Y.; Kondo, M. Bull. Chem. Soc. Jpn. 1954, 27, 160–163. (37) Lide, D. R. Handbook Chemistry and Physics, 85th ed.; CRC Press: Boca Raton, FL, 2004; p 2656. (38) Kim, K.; Kang, C.; You, Y.; Chung, M.; Woo, M.; Jeong, W.; Park, N.; Ahn, H. Catal. Today 2006, 111, 223–228. (39) Finocchio, E.; Montanari, T.; Garuti, G.; Pistarino, C.; Federici, F.; Cugino, M.; Busca, G. Energy Fuels 2009, 23, 4156–4159. (40) Wu, J.; Str€ omqvist, M. E.; Claesson, O.; F€angmark, I. E.; Hammarstr€ om, L. Carbon 2002, 40, 2587–2596.
4761
Energy Fuels 2010, 24, 4756–4765
: DOI:10.1021/ef901346f
Boulinguiez and Le Cloirec
of the model parameters resulting from the regression do not make sense physically; the model predicts an adsorption capacity at saturation of about 20 mmol g-1, which would be 10 times higher than any other adsorbent. Because the regression provides a mathematical solution that fits the experimental data but is not consistent in the physical sense, the relative standard error of the estimate rises to 50%, statistically indicating the inexactness of this pair of parameter values. Nonetheless, the graphical observation confirms the higher adsorption capacity of the B1 sample compared to the other ACs. For toluene, ethyl mercaptan, and siloxane D4, the deviances of adsorption between the compared ACs decrease by 31, 14, and 44%, respectively, between the capacities at high and low concentrations. For dichloromethane, the same tendency is observed to the extent that THC515 is already the best adsorbent at high concentrations; thus, adsorbed dichloromethane at low concentration is shifted even further by 163%. This evolution of adsorption at low concentration can be explained by the difference between the porosity characteristics of these two ACs. For the B1 sample, 52% of the porosity is due to pores with a width of less than 2 nm, defined as the threshold between meso- and micropores.27 As a result, this AC shows an average pore size of 3.0 nm higher than any other material. The pore structure of this sample may be pictured as having a substantial quantity of small pores branching off from larger pores that are only filled up and used for adsorption at VOC concentrations higher than those encountered in biogases. On the other hand, the two ACFCs display the smallest average pore widths, highlighting the microporous character of these samples. Their porosity is due to a significant number of narrow pores, which is more effective than a large volume, allowing for the build-up of additional layers that is unlikely to occur at low concentration, where VOC molecules cover the surface of the pores before filling up the volume of the pores. For the biogas application, the B1 sample remains undeniably the best adsorbent as it adsorbs greater quantities of four of five VOCs than any other AC. Adsorption kinetic investigations are of great interest because a continuous adsorption process is required to carry out the upgrading of biogas. Kinetics of Adsorption. The initial adsorption coefficient γ, the external film mass-transfer coefficient ksA, and the internal diffusion coefficient of Weber KW are used to quantify the kinetics adsorption behavior of an adsorbent.17,41,42 The γ coefficient is determined by the initial adsorption kinetic, which could be described by the Adams-BohartThomas relation. The derivative equation is as follows: dq ¼ KABT Cðqm - qÞ - K 0 ABT q ð4Þ dt
Table 5. Kinetic Estimates of the Coefficients for the Adsorption of the VOC at 25 °Ca VOC toluene
isopropanol
dichloromethane
ethyl mercaptan
sample
γ (109) (m3 g-1 s-1)
ksA (106) (m3 s-1)
KW (mmol m-3 s-0.5)
B1 NC60 FM30K THC515 B1 NC60 FM30K THC515 B1 NC60 FM30K THC515 B1 NC60 FM30K THC515
3.52 (0.97) 3.39 (0.96) 4.61 (0.96) 5.87 (0.99) 3.50 (0.96) 3.69 (0.96) 4.62 (0.96) 6.77 (0.97) 2.29 (0.99) 2.05 (0.99) 3.07 (0.99) 6.93 (0.97) 6.74 (0.98) 6.04 (0.98) 7.70 (0.98) 8.64 (0.98)
0.59 (0.98) 0.54 (0.96) 0.78 (0.97) 2.30 (0.96) 1.12 (0.96) 1.07 (0.97) 1.47 (0.96) 3.05 (0.96) 1.12(0.96) 1.07 (0.98) 1.81 (0.96) 2.66 (0.96) 1.50 (0.96) 0.87 (0.98) 1.99 (0.96) 2.08 (0.98)
2.08 (0.97) 2.12 (0.96) 2.61 (0.96) 3.06 (0.97) 2.40 (0.97) 2.45 (0.96) 3.06 (0.96) 6.32 (0.96) 2.51 (0.96) 2.70 (0.96) 3.03 (0.97) 6.60 (0.96) 4.60 (0.99) 2.83 (0.96) 4.96 (0.98) 5.26 (0.97)
a
The linear coefficients of determination are reported in parentheses for each estimate.
follows:43 V
dC ¼ - ks AðC - Cs Þ dt
ð6Þ
At the initial stage of the adsorption reaction, when t f 0, Cs f 0 and C f C0. Equation 6 can be reformulated as C ks A t ð7Þ - ln ¼ C0 V When the adsorption kinetics are essentially due to internal diffusion, the gas-phase concentration variation is proportional to the square root of time (Fick’s law). This is considered valid when the adsorbed quantity is less than 20% of the adsorption maximum capacity. Under this threshold, adsorption sites in the micropore volume are available in large amounts; thus, kinetics is only rate-controlled by diffusion through meso- and macropores. The internal diffusion coefficient of Weber KW is obtained from the following relation: C ¼ C0 - KW t1=2
ð8Þ
Kinetic coefficients were determined by regression with linear coefficients of determination higher than 0.95 to ensure the validity of the estimates. The adsorption kinetic results are presented in Table 5, where the coefficients of determination of each regression are reported in parentheses next to the estimates of the coefficients. Figure 4 shows the experimental adsorption data as a function of time for toluene on the FM30K sample and the fitting of both linear driving force and Weber models. Because these models are only valid at the early stage of the adsorption curve, the values of the parameters of the models are assessed below the limit of 3600 s. Figure 5 shows the fitting of the Adams-Bohart-Thomas model to the experimental data points during the first hour of adsorption of toluene on the FM30K sample. For each kinetic parameter, the ACFCs show higher values than any GAC. Between the two ACFCs, THC515 obtains even better results than FM30K. Such materials are constituted of fibers less than 10 μm in diameter that develop a high external surface area, larger than the GACs. Thereby, the contact surface between the fluid and the solid is greater when using a fiber cloth material as an adsorbent, explaining the higher film diffusions (ksA) obtained with these materials.24
At the initial stage of the adsorption reaction, when t f 0, q f 0 and C f C0. Equation 4 can be reformulated as V dC γ ¼ - KABT qm ¼ ð5Þ C0 m dt t f 0 The external film mass-transfer coefficient ksA is determined by the linear driving force for the mass-transfer relation as (41) Sphan, H.; Schlunder, E. S. Chem. Eng. Sci. 1975, 30, 529–537. (42) McKay, G.; Bino, M. J. Water Res. 1988, 22, 279–286. (43) Kadirvelu, K.; Faur-Brasquet, C.; Le Cloirec, P. Langmuir 2000, 16, 8404–8409.
4762
Energy Fuels 2010, 24, 4756–4765
: DOI:10.1021/ef901346f
Boulinguiez and Le Cloirec
Figure 4. Adsorption curve of toluene on the FM30K sample fitted by the linear driving force and Weber models.
Figure 5. Adsorption curve of toluene on the FM30K sample for the first hour fitted by the Adams-Bohart-Thomas model during the first 3600 s.
The microporous volume of THC515 is the second highest among the four materials, and the average pore width is the second smallest after FM30K. In a first approximation, if micropores are considered spherical to estimate their number in the material, the quantity of micropores would be 6 times greater in THC515 than in any other tested material, followed by FM30K. Adsorbates reach adsorption sites through micropores without the additional diffusion resistance of mesopores, which is usually the rate-controlling step in the case of porous adsorbents. These pores are referred to as feeder or transport pores.34,35 Thus, the highest values obtained for the KW coefficient are explained by the large number of micropores present in the fiber materials. In fact, mesoporosity determines the adsorption kinetics of the VOCs on the granular materials and explains the results observed with the B1 sample. The initial adsorption coefficient includes the effect of the film diffusions and the surface diffusion at the early stage of adsorption. Because THC515 exhibits the best behavior toward both, the values for γ are the highest. The ACFCs, particularly the THC515 sample, show better kinetics of adsorption for all selected VOCs than the GACs tested because of the morphology of the material. Thermal Regeneration. The first derivatives of the thermogravimetric data, referred to as DTG curves, are used to discuss the regeneration of the ACs. This method has already been reported to study the possible regeneration of carbon.44-46 The DTG curves of the samples loaded with the VOCs are presented in Figure 6. The VOC concentrations applied in the adsorption phase prior to the thermogravimetric experiments were higher than those commonly found in biogas but were required because of the sensitivity of the equipment. Thus, the DTG curves are related to the adsorption capacities close to the maximum for each material. This hypothesis confirms the quantity of the VOCs that are desorbed from the ACs. The sample THC515 desorbs more dichloromethane than the B1 sample, which
desorbs more toluene, isopropanol, and siloxane D4 than any other AC. The DTG curves for isopropanol and dichloromethane demonstrate, for all of the ACs, a main peak centered at 100 °C. The shift to a higher temperature of this main desorption peak for more microporous ACs indicates the increase in the energy of adsorption of the VOC in smaller pores. Molecules, when adsorbed on the carbon surface, are held by adsorption forces that are enhanced in narrow pores.38,45 A mesoporous carbon, such as B1, begins to desorb any VOC at a lower temperature than a sample in which porosity is mostly due to microporosity, such as the ACFCs. The desorption peak for the ACFCs presents a shoulder close to 50 °C, which can be assigned to the desorption of water. The desorption of toluene from the B1 sample confirms its higher adsorption capacity. The toluene DTG curves spread from 50 to 250 °C for the B1 sample and to more than 300 °C for the other ACs, confirming the effect of the structural porosity of the samples. The DTG curve of B1 shows that most of the desorbed toluene occurs at the early stage of the desorption during unloading of larger pores. This confirms the assumption made earlier that the remarkable capacity of the B1 sample is due to the buildup of multi-layers of adsorbate molecules in larger pores that require less energy to desorb than molecules adsorbed on the ACFCs, for which the ends of the toluene DTG curves trail along at temperatures higher than 300 °C. The DTG curves for ethyl mercaptan present two peaks, varying in intensity, which underlines the heterogeneity among our ACs tested. The first peaks of the curves represent weakly bound products, such as water, although for NC60 and THC515, the first peak contains a shoulder close to 50 °C. Following our previous assumption about the desorption of water with the other VOCs, the first whole peak does not represent the desorption of water only. Its shape seems to be an overlapping of different peaks, which can be interpreted as the desorption of more than one product, similar to methyl mercaptan.47,48 Byproducts from the oxidation of
(44) Seredych, M.; Bandosz, T. J. Energy Fuels 2008, 22, 850–859. (45) Bagreev, A.; Bashkova, S.; Bandosz, T. J. Langmuir 2002, 18, 8553–8559. (46) Bashkova, S.; Bagreev, A.; Bandosz, T. J. Langmuir 2003, 19, 6115–6121.
(47) Bashkova, S.; Bagreev, A.; Bandosz, T. J. Environ. Sci. Technol. 2002, 36, 2777–2782. (48) Bashkova, S.; Bagreev, A.; Bandosz, T. J. Ind. Eng. Chem. Res. 2002, 41, 4346–4352.
4763
Energy Fuels 2010, 24, 4756–4765
: DOI:10.1021/ef901346f
Boulinguiez and Le Cloirec
Figure 6. DTG curves after adsorption of the VOCs on the AC samples.
ethyl mercaptan are probably desorbed at this stage, because sulfur species are known to be reactive products during their thermal desorption from AC.45,46 FM30K does not strictly follow the assumption about byproduct formation because the first peak on the FM30K curve seems to represent the desorption of water molecules. Moreover, the desorption of ethyl mercaptan in this case does not follow the assumptions drawn from the porosity characteristics of the adsorbents. Indeed, the material with a high microporosity (FM30K) desorbs ethyl mercaptan at lower temperatures than the material with the highest mesoporosity (B1). The polarity of the molecule probably interacts differently with the functional surface groups in these two fundamentally different ACs, because the Boehm titration of the ACs presented in another study has shown major differences among these ACs.49 The DTG curves of samples B1 and FM30K show a third desorption peak at 350 °C, which is not observed on the other VOC curves; hence, the release of carbon dioxide from the adsorbent cannot be assumed in this case. Some byproduct with a higher boiling point or stronger bonding interaction with the carbon matrix are likely to be generated to a greater extent on these materials. It is probable that the
functional group discrepancies among the materials or their inherent composition, noticeably the traces of metal, act as a catalyst for the oxidation reaction.50 The desorption of siloxane from the carbons follows a more complex pattern, with several peaks appearing on the curves. The desorption of water is visible at temperatures below 100 °C, while temperatures higher than 350 °C are required to desorb siloxane from the material. The peak of the desorption ranges from 200 °C for the B1 sample to 350 °C for the THC515 sample. This corresponds to the temperature range of desorption reported by other research groups.30,39 Considering the shape of the desorption curves, the conversion of siloxane D4 into byproduct (linear siloxanes and silica) through the opening of the ring probably occurs on the surface of the ACs.18,39,51 Finocchio and coworkers39 have shown the growth of amorphous silica on AC used as an adsorbent to upgrade biogas containing siloxane. Because this amorphous species does not desorb from the graphite structure of ACs, a phenomenon of exhaustion of the adsorbent is measurable. The mass balances of siloxane (50) Bashkova, S.; Bagreev, A.; Bandosz, T. J. Catal. Today 2005, 99, 323–328. (51) Ortega, D. R.; Subrenat, A. Environ. Technol. 2009, 30, 1073– 1083.
(49) Boulinguiez, B.; Le Cloirec, P. Carbon 2010, 48, 1558–1569.
4764
Energy Fuels 2010, 24, 4756–4765
: DOI:10.1021/ef901346f
Boulinguiez and Le Cloirec
desorbed from the carbons deviate from the values predicted by the isotherm model. The siloxane desorbed from the sample is about 63% of that adsorbed in the experimental conditions. The regeneration is hampered by the formation of byproduct from siloxane D4; thus, the regeneration of the AC is only partial.
tions, the adsorption capacities of the B1 sample are less outstanding in comparison to the other adsorbents. Porosity discrepancies among the adsorbents can explain the observed adsorption for these four compounds. The adsorption of dichloromethane shows the need to consider interaction other than simple physisorption to interpret the different results between the carbons. Moreover, the adsorption kinetics confirm the value of ACFC, which morphology enhances the kinetics of adsorption by increasing the exchange surface area between the fluid and the adsorbent phase. The desorption of the adsorbates from the carbons is achievable, although siloxane and ethyl mercaptan may severely hamper regeneration processes at temperatures below 200 °C. An exhaustion effect of the adsorbent by the accumulation of non-volatile byproduct is predicted. The same observations are expected using the electrothermal regeneration process with fiber cloth materials, which is a reason to consider this form despite their lower adsorption capacities compared to the granular materials.
Conclusion The Langmuir-Freundlich model provides the best fit of the experimental adsorption data. The third parameter enables the model to be adequately adjusted to the non-ideal conditions of adsorption by the biogas matrix and allows our adsorption discussion to be based on the estimates of the model parameters. The effective adsorption of toluene, siloxane D4, isopropanol, and ethyl mercaptan, by the B1 sample, a GAC derived from wood, makes this the best adsorbent in terms of equilibrium adsorption capacities at both biogas and saturation concentrations. However, at biogas concentra-
4765