Journal of Molecular Liquids 136 (2007) 294 – 299 www.elsevier.com/locate/molliq
CO2–water supercritical mixtures: Test of a potential model against neutron diffraction data F. Lo Celso a,⁎, R. Triolo a , F. Ferrante a , A. Botti b , F. Bruni b , R. Mancinelli b , M.A. Ricci b , A.K. Soper c,d a
Dipartimento di Chimica Fisica “F. Accascina”, Università degli Studi di, Palermo, viale delle Scienze ed. 17, 90128 Palermo, Italy b Dipartimento di Fisica “E. Amaldi”, Università degli Studi “Roma Tre”, Via della Vasca Navale 84, 00146 Roma, Italy c Isis Facility, Rutherford Appleton Laboratory, Chilton, Didcot, Oxfordshire, OX11, 0QX, United Kingdom d Department of Physics and Astronomy, University College London, Gower Street, London WC1E 6BT, UK Available online 30 August 2007
Abstract A neutron diffraction experiment on supercritical mixtures of water and CO2 at two concentrations is presented. Data are analyzed within the EPSR framework and the water–water and water–CO2 radial distribution functions are compared with those calculated by a Molecular Dynamics simulation performed by using the TIPS2 and EPM-M potential models for water and CO2 respectively. It is found that the Molecular Dynamics simulation reproduces the overall shape of the site–site radial distribution functions, although missing a few subtle changes brought along when the CO2 concentration is increased. © 2007 Elsevier B.V. All rights reserved. Keywords: Neutron diffraction; Molecular Dynamics; Supercritical fluids
1. Introduction Supercritical fluids (sc-fluids) and their mixtures have recently become a subject of growing scientific and industrial interest as safe and efficient solvents in a variety of industrial processes. Two interesting fluids which are becoming very popular are water and carbon dioxide. For example at normal temperatures, water is a poor solvent of nonpolar materials; while at supercritical conditions it becomes a good solvent even for organic materials produced in many industrial processes. At high enough pressures it is miscible in all proportions with both organic compounds and oxygen. Thus, these compounds can be oxidized in a homogeneous supercritical phase. Removal of reaction products can be easily accomplished by simply venting or reducing the temperature of the reaction vessel. This constitutes the essence of many promising new technologies like the Supercritical Water Oxidation (SCWO) [1,2], which utilizes sc-water to oxidize toxic wastes (dioxin, PCBs, benzene, DDT, urea, cyanide, and many others) into relatively ⁎ Corresponding author. E-mail address:
[email protected] (F. Lo Celso). 0167-7322/$ - see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.molliq.2007.08.011
harmless products. The dissolving power of a supercritical fluid offers a safe solvent in food processing, and enables quick extraction of spirits or rapid removal of unnecessary components. As a substitute for the conventional distillation method using an azeotropic mixture, it carries out separation with an energy-saving high density concentration. In addition, use of a supercritical fluid will make it possible to manufacture specially structured products and high-functional, high-quality materials, which are almost impossible to produce with conventional manufacturing methods. Quite similar advantages are presented by sc-CO2 which is considered the solvent of the future [3–6]. Moreover aqueous mixtures in the systems H2O–CO2 and H2O–CO2–NaCl are considered a good approximation for most Earth crustal and mantle fluids. Accurate knowledge of their structural characteristics would enable accurate thermodynamic modeling of hydrothermal fluids for geologically relevant conditions. In view of their strong industrial interest, the thermodynamics of sc-H2O–CO2 and H2O–CO2–NaCl mixtures have been accurately studied [7–10], while little interest has been devoted to the study of their microscopic structure [9,11]. Moreover Ref. [9] does not report a full description of the radial distribution functions, while inferring an increased
F. Lo Celso et al. / Journal of Molecular Liquids 136 (2007) 294–299
Fig. 1. Scheme of the pressure rig used for in situ sample preparation. The He bottle is used in order to fasten sample cooling in inert atmosphere.
clustering of water molecules in the mixture. Here we present the results of a Neutron Diffraction experiment with Isotopic Substitution (NDIS), performed on two sc-H2O–CO2 mixtures at different CO2 concentration, and compare the experimental Radial Distribution Functions (RDF) with those calculated by a Molecular Dynamics (MD) simulation. 2. Methods 2.1. Experimental details, data analysis and EPSR simulation Neutron diffraction measurements have been performed at the SANDALS [12] diffractometer, installed at the ISIS facility [13] (UK). The sample container was a Ti–Zr cell designed to withstand a pressure of 3 kbar: it consists of 6 cylindrical holes (1.6 mm internal diameter) drilled in a slab (0.66 cm thickness, 4.05 cm lateral width) [14] and connected to the pressure rig shown in Fig. 1 through high pressure pipes at top and bottom. Temperature control was achieved by heaters and thermocouples in contact with the top and bottom of the sample container, giving a temperature stability better than ± 0.1 K. Samples have been prepared in situ, following the procedure described below:
• Valves 9, 7 and 2 are closed; the pressure intensifier is set at half stroke; valve 1 is opened and water from syringe 1 enters the sample container from the bottom at 1 bar. • Valves 1, 3 and 4 are closed, valve 5 is opened. • Valve 10 is opened and CO2 is adjusted at ∼ 4 bar. • Valve 7 is opened and the gas goes into the sample container from the top, pushing water into syringe 2. As soon as the liquid meniscus reaches the level of the pipe at the bottom of the sample container, gas bubbles blow into syringe 2. This procedure allows to calibrate the volume occupied by the gas. • Valve 5 is closed and CO2 pressure adjusted to (5.82 ± 0.06) or (46.84 ± 0.35) bar for the 1 mol% and 10 mol% mixtures respectively. These pressure values are used to evaluate the amount of CO2 in the mixtures from the PVT data of the pure gas [15]. • Valve 7 is closed. The temperature is set to 673 K. • Once temperature stability is reached, the pressure between valves 3 and 4 is increased to the same value read on pressure gauge 2 and valve 4 is opened. • The pressure of the sample is gradually increased up to set point (P = 1300 bar), by using the pressure intensifier. According to Ref. [16] when this set point is achieved the amount of water in the mixtures matches the desired values of 1% and 10%. During the measurements pressure fluctuated in the range (1297 ± 5) bar. • When temperature and pressure are stable at their set values, valve 4 is closed and data acquisition begins. At each CO2 concentration three solutions with different H/D content have been prepared: one fully deuterated, one fully protiated and an equimolar mixture of the two, in order to exploit H/D isotopic substitution in the NDIS experiment [17]. The total acquisition time on each sample was between 8 and 10 h, with ISIS running on average at 180 μA/h; data were recorded every 3 h, in order to check the entire apparatus stability. Experimental conditions, P = 1300 bar, T = 673 K, were chosen well above the critical pressure and temperature of the two fluids, to ensure their full miscibility in a wide concentration range [8]. Data reduction has been performed by using the GUDRUN routine available on SANDALS, which performs corrections for multiple scattering, absorption and inelasticity effects, along with subtraction of the scattering from the sample container, and data reduction to an absolute scale, following the procedure described in the ATLAS manual [18]. The availability of a transmission monitor on SANDALS has allowed to double check the sample density and concentration against the high Q limit of the total differential scattering intensity, I (Q): lim IðQÞ ¼
QYl
• Valves 1, 5, 8 and 10 are closed. The rig and sample container are vacuum pumped. Syringe 1 is filled with water; syringe 2 contains a small amount of water. The sample tank is then evacuated. The temperature of the container is set to 313 K.
295
qtRa ca ra 4k
ð1Þ
where Q is the exchanged wavevector (see Eq. (3)), t is the sample thickness, cα and aα are the fraction and total scattering cross section [19] of atoms of type α, and ρ is the sample density. The latter quantity is equal to 0.0458 atoms Å− 3 and
296
F. Lo Celso et al. / Journal of Molecular Liquids 136 (2007) 294–299
0.049 atoms Å− 3 at CO2 concentration of 0.92 mol% and 9.2 mol% respectively. For the lowest CO2 concentration we have evaluated an upper limit for the excess volume of the mixture of the order of 8 cm3, in agreement with data reported in the literature [9,10]. The output of the GUDRUN routine are the three Interference Differential Cross Sections (IDCS), F(Q), defined in Eqs. (2)–(4) and measured in barn/atom sr: XX FðQÞ ¼ wab Sab ðQÞ ð2Þ a
bza
with: Q¼
4k sinh k
ð3Þ
defining the momentum transfer vector of the scattering event, for each neutron wavelength, λ, and scattering angle, 2θ; and wab ¼ ca cb ba bb ð2 dab Þ
ð4Þ
where bα and bβ are the neutron scattering lengths [19] of two atomic sites, and the Kronecker δαβ avoids double counting of like terms in the summation. The Sαβ(Q) are called Partial Structure Factors (PSF) and are defined as Fourier transforms of the Radial Distribution Functions (RDF) of the αβ pair, gαβ(r): gab ðrÞ 1 ¼
1 2k2 q
Z 0
l
Q2 ðSab ðQÞ 1Þ
sinQr dQ Qr
ð5Þ
In the hypothesis that the microscopic structure of the deuterated sample does not sensibly differ from that of the hydrogenated one, the difference between the three IDCS depends only on the different weighting factors and the maximum contrast among the experimental data is achieved for the H/D concentration used in the present experiment. In the case of water–CO2 solutions, each IDCS is the combination of 10 distinct PSF, namely SOwOw, SOwHw, SOwC, SOwO, SHwHw, SHwC, SHwO, SOO, SCO, SCC, where the subscript w labels water sites. The real space counterparts of these functions, i. e. the corresponding RDF, which are the quantities of interest in the present study, can be extracted from the experimental data only by analyzing the IDCS functions in conjunction with a computer simulation. As a matter of fact our problem is underdetermined, as we have 3 experimental data and 10 unknowns. To circumvent this difficulty, we have used the Empirical Potential Structure Refinement code (EPSR) [20]. This is a Monte Carlo routine, which builds up a 3-dimensional model of the sample constrained by the 3 independent experimental IDCS. The routine requires as input the sample composition and thermodynamic parameters and a solid guess about the site–site interactions. In the present case the SPC/E [21] pairwise additive model has been used for water and the EPM-M [22] for CO2 and the Lorentz–Berthelot summation rules have been applied. These potential models are used as a reference to start the simulation and an empirical correction is iteratively evaluated by the routine, in order to drive the simulation box towards the best fit of the real system. When the IDCS calculated from the simulation box gives a satisfactory fit
Fig. 2. Experimental IDCS data at 9.2 mol% CO2 concentration for deuterated (circles), protiated (squares) and their equimolar mixture (triangles), compared with their EPSR fit (solid lines). Data for the protiated sample and mixture have been shifted by an arbitrary quantity for clarity.
of the experimental data, the refinement of the so-called “empirical potential” is achieved, molecular configurations can be recorded and the RDF calculated from the Monte Carlo simulation. The lateral width of the simulation box was 54.15 Å and 52.94 Å for the low and high concentration mixtures respectively and the potential cut-off was set to 20 Å. The typical quality of an EPSR fit to the NDIS data is shown, for the higher CO2 concentration (9.2 mol%), in Fig. 2, where the three IDCS for the HDO–CO2, H2O–CO2 and D2O–CO2 are reported from top to bottom. 2.2. MD simulation details An MD simulation was performed, independently from the experimental data, in order to check the quality of the TIPS2 [23] and EPM-M [22] models in describing the real supercritical mixture. We have used here the TIPS2 model for water because it is expected to describe better the water–water interactions with respect to the SPC/E model, which predicts too strong hydrogen bonding at supercritical states[24]. In the EPSR code instead, the choice of the SPC/E model is less time consuming, due to the lower number of sites, and the empirical potential recovers the model inadequacy. All simulations were performed on 2236 rigid body molecules by using the program Moldy [25]. The 10 mol% solution was formed by 2030 H2O + 206 CO2 molecules, while the 1 mol% solution contained 2215 H2O + 21 CO2 molecules; all other settings were the same for both solutions. All simulations proceeded in the isothermal–isobaric (NPT) ensemble, imposing P = 130 MPa and T = 673 K, with a time step of 0.1 fs for a total run time of 200 ps. According to Ref. [22], the OCO angle in CO2 was set to 174°. The LJ parameters for the interaction between atoms in different species were calculated as geometric mean (σ parameter) and arithmetic mean (ε parameter). Pressure control was realized by the Andersen barostat [26], allowing a uniform dilation of the simulation box. A mass parameter equal to 0.00005 amu was chosen for the barostate. The temperature was kept constant by coupling the Nosé-Hoover–Hoover thermostat to every species,
F. Lo Celso et al. / Journal of Molecular Liquids 136 (2007) 294–299
297
• the second peak of the Ow–Ow RDF is at about 5.7 Å, suggesting that the tetrahedral network of H-bonds is destroyed at these states.
Fig. 3. EPSR fits of the deuterated samples at 0.92 mol% (dashed) and 9.2 mol% (solid) CO2 concentration.
with both translational and rotational inertia parameters equal to 100 kJ/mol ps2. The LJ and the real part of the Ewald sum were truncated, with a strict cutoff, at 25 Å. In the Ewald sum the surface dipole term of De Leeuw et al. [27] was included. The radial distribution functions were accumulated every 1 ps after 50 ps of simulation, with a 25 Å pair cutoff.
On the other hand we notice that the MD simulations do not reproduce the shift of peaks and minimum approach distance visible in the EPSR RDFs as a function of the CO2 concentration. Particularly meaningful are the reduction of the minimum approach distance in the Hw–Hw RDF and the shift of the two main peaks of the Ow–Hw RDF. These findings indicate a stronger H-bond interaction between first neighboring molecules, along with a higher orientational disorder associated to the increased solute concentration. Figs. 5 and 6 illustrate the hydration of the CO2 molecules. These distribution functions, although more noisy than previous ones, show some interesting features. In Fig. 5 the RDFs involving carbon atom and water sites show that although the minimum approach distance is shorter for hydrogens compared to water oxygens, nevertheless the first peak of the Ow–C RDF comes at shorter distances than that of the Hw–C RDF. Moreover the EPSR data show a shift of the main peak of the Hw–C RDF at the highest CO2 concentration, not reproduced by the MD simulation. If these RDFs are compared to the corresponding RDF for Ow–Ow pair it is clear that although the minimum approach distances, Ow–C
3. Results and discussion The quality of the EPSR fit to the experimental IDCS (Fig. 2)) is essentially good as well as the contrast between the three functions. In the low Q region a rise of the scattering cross section of the deuterated sample is observed, due to the large isothermal compressibility of the system under supercritical conditions. In Fig. 3 the EPSR fits of the D2O–CO2 data are shown for both concentrations. The increase of intensity at low Q and the shrinking of the peak width (centered at about 2.15 Å− 1), as the CO2 concentration in the mixture increases, are a possible indication of microscopic structural changes due to the suggested [9] clustering of water molecules. In the following we will discuss the most significant RDF for the mixture in question out of the full set of 10. In particular we will compare EPSR results at the two concentrations and with the MD calculated ones. Fig. 4 reports the three RDFs for water which allow to say that the particular structural characteristics, shown in previous NDIS experiments on supercritical water [28–31] are here confirmed. The EPSR functions are systematically more intense than the MD ones, nevertheless the very same trend of the peak intensity is observed when the CO2 concentration increases. Furthermore for both EPSR and MD we notice that: • the main peaks of the Hw–Hw RDF are less resolved than for pure water at about the same conditions [29,31] at both concentrations; • the first peak of the Ow–Ow RDF broadens with increasing CO2 concentration; • the H-bond peak is more intense compared to that for pure water;
Fig. 4. Water–water RDFs: open circles and squares represent the EPSR refined RDFs at 0.92 mol% and 9.2 mol% CO2 concentration respectively; open diamond and triangle refer to the MD calculated ones.
298
F. Lo Celso et al. / Journal of Molecular Liquids 136 (2007) 294–299
Fig. 5. Hydration of the CO2 carbon atoms: same symbols as in Fig. 4.
and Ow–Ow, are very close, the first peak is at much longer distances in the case of the Ow–C pair respect to the Ow–Ow pair, hence the CO2 molecule very likely does not penetrate the solvent first neighbor shell of water. Furthermore looking at the shape of the Ow–C and Hw–C RDF, a broadening of the first peak is observed especially at the highest CO2 concentration in the EPSR fits; yet another indication of orientational disorder induced by the solute. The RDFs relative to the hydration of the CO2 oxygen, O, are reported in Fig. 6. Also in this case it appears clear that the first peak is found at larger distances with respect to the Ow–Ow distance, while the minimum approach distance of Ow–Ow pair is smaller than the Ow–O distance; thus confirming the hypothesis that the solute does not penetrate the solvent first neighbor shell.
4. Conclusions NDIS data for two sc-H2O–CO2 mixtures have been analyzed by using the EPSR routine. The site–site RDFs of water have the same overall shape as those of pure sc-water, nevertheless changes are visible with increasing CO2 concentration, in particular peaks intensity and broadness. An independent MD simulation performed by using the TIPS2 and EPM-M models for water and CO2 respectively reproduces these trends. More subtle changes in the EPSR data are the shift of particular peaks, the H-bond peak and the first peak of the Ow–Ow RDF for instance, and reduction of the minimum approach distance which are not reproduced by the MD simulation. Similar observations can be done in the case of
Fig. 6. Hydration of the CO2 oxygen atoms: same symbols as in Fig. 4.
F. Lo Celso et al. / Journal of Molecular Liquids 136 (2007) 294–299
the CO2 hydration shell: the overall shape of the RDFs is reproduced by the MD simulation, where instead the shift of the main peak with increasing CO2 concentration is missing. These observations suggest that although the TIPS2 and EPM-M models catch the basic characteristics of these sc-fluids, they still require a correction at short range and should perhaps be made softer. As far as the increased water clustering hypothesized by Ref. [9] is concerned, no safe conclusion can be reached until a thorough analysis of the water clusters and chains is performed, as the height of the H-bond peak by itself cannot be considered conclusive, since it is accompanied by changes of the water partial density. This analysis is in progress and will be published in a separate report. Acknowledgments This work has been performed within the Agreement No.01/ 9001 between CCLRC and CNR, concerning collaboration in scientific research at the spallation neutron source ISIS and with partial financial support of CNR. References [1] M. Modell, in: Z.H.M. Freeman (Ed.), Supercritical Water Oxidation. Standard Handbook of Hazardous Waste Treatment and Disposal, McGraw-Hill, New York, 1989. [2] D.D. Macdonald, L.B. Kriksunov, Electrochim. Acta 47 (5) (2001) 775. [3] J.M. DeSimone, Zihibin Guan, C.S. Elsbernd, Science 257 (1992) 945. [4] J.M. DeSimone, E.E. Maury, Y.Z. Menceloglu, J.N. McClain, T.J. Romack, J.R. Combes, Science 265 (1994) 356. [5] J.B. McClain, D.E. Betts, D.A. Canelas, E.T. Samulski, J.M. DeSimone, J.D. Londono, H.D. Cochran, G.D. Wignall, D. Chillura Martino, R. Triolo, Science 274 (1996) 2049. [6] Papers in Dec. Issue of Ind. Eng. Chem. Res., vol. 39, 2000. [7] A.E. Mather, E.U. Franck, J. Phys. Chem. 96 (1992) 6. [8] Z. Duan, N. Moeller, J.H. Weare, Geochim. Cosmochim. Acta 59 (1995) 2869.
299
[9] C.M. Destrigneville, J.P. Brodholt, B.J. Wood, Chem. Geol. 133 (1996) 53. [10] J.C. Seitz, J.G. Blencoe, Geochim. Cosmochim. Acta 63 (1999) 1559. [11] A.G. Kalinichev, R.J. Kirkpatrick, Eleventh Annual V. M. Goldsmidt Conference, vol. 3298, 2001, .pdf. [12] A.K. Soper, in: D.K. Hyer (Ed.), Proceedings of the Conference on Advanced Neutron Sources 1988, IOP Conf. Proc., vol. 97, Institute of Physics and Physical Society, London, 1989, p. 353. [13] www.isis.rl.ac.uk. [14] www.isis.rl.ac.uk/disordered. [15] http://webbook.nist.gov/chemistry. [16] Z. Duan, N. Moeller, J.H. Weare, Geochim. Cosmochim. Acta 60 (1996) 1209. [17] A.K. Soper, in: J.B. Suck, D. Raoux, P. Chieux, K.C. Riekel (Eds.), Methods in the Determination of Partial Structure Factors, World Scientific, Singapore, 1993, p. 58; A.K. Soper, J.Z. Turner, Int. J. Mod. Phys. B 7 (1993) 3049. [18] A.K. Soper, W.S. Howells, A.C. Hannon, ATLAS—Analysis of the Timeof-Flight Diffraction from Liquid and Amorphous Samples. RAL Report N. RAL-89-046, 1989. [19] V.F. Sears, Neutron News 3 (1992) 26. [20] A.K. Soper, Chem. Phys. 202 (1996) 295; A.K. Soper, J. Mol. Liq. 78 (1998) 179; A.K. Soper, Chem. Phys. 258 (2000) 121; A.K. Soper, Mol. Phys. 99 (2001) 1503. [21] H.J.C. Berendsen, J.R. Grigera, T.P. Straatsma, J. Phys. Chem. 91 (1987) 6269. [22] J. Harris, K.H. Young, J. Phys. Chem. 99 (1995) 12021; Y. Zhang, J. Yang, Y.-X. Yu, J. Phys. Chem., B 109 (2005) 13375. [23] W.L. Jorgensen, J. Chem. Phys. 77 (1982) 4156. [24] P.T. Cummings, A.A. Chialvo, J. Phys., Condens. Matter 8 (1996) 9281. [25] K. Refson, Comput. Phys. Commun. 126 (2000) 309. [26] H.C.J. Andersen, Chem. Phys. 72 (1980) 2384. [27] S.W. De Leeuw, J.W. Perram, E.R. Smith, Proc. R. Soc., A 373 (1980) 27. [28] P. Postorino, R.H. Tromp, M.A. Ricci, A.K. Soper, G.W. Neilson, Nature 366 (1993) 668. [29] A.K. Soper, F. Bruni, M.A. Ricci, J. Chem. Phys. 106 (1997) 247. [30] P. Jedlovszky, J.P. Brodholt, F. Bruni, M.A. Ricci, A.K. Soper, R. Vallauri, J. Chem. Phys. 108 (1998) 8528. [31] A. Botti, F. Bruni, M.A. Ricci, A.K. Soper, J. Chem. Phys. 109 (1998) 3180.
