J. Phys. Chem. B 2008, 112, 14993–14998
14993
Effects of Solute Structure on Local Solvation and Solvent Interactions: Results from UV/Vis Spectroscopy and Molecular Dynamics Simulations John L. Gohres,†,§,⊥ Charu L. Shukla,†,§,⊥ Alexander V. Popov,‡,§,⊥ Rigoberto Hernandez,*,‡,§,⊥ Charles L. Liotta,†,‡,⊥ and Charles A. Eckert*,†,‡,⊥ School of Chemical & Biomolecular Engineering, School of Chemistry and Biochemistry, Center for Computational Molecular Science and Technology, and Specialty Separations Center, Georgia Institute of Technology, Atlanta, Georgia 30332-0100 ReceiVed: July 11, 2008; ReVised Manuscript ReceiVed: September 15, 2008
Solvation of heterocyclic amines in CO2-expanded methanol (MeOH) has been explored with UV/vis spectroscopy and molecular dynamics (MD) simulations. A synergistic study of experiments and simulations allows exploration of solute and solvent effects on solvation and the molecular interactions that affect absorption. MeOH-nitrogen hydrogen bonds hinder the n-π* transition; however, CO2 addition causes a blue shift relative to MeOH because of Lewis acid/base interactions with nitrogen. Effects of solute structure are considered, and very different absorption spectra are obtained as nitrogen positions change. MD simulations provide detailed solvent clustering behavior around the solute molecules and show that the local solvent environment and ultimately the spectra are sensitive to the solute structure. This work demonstrates the importance of atomic-level information in determining the structure-property relationships between solute structure, local salvation, and solvatochromism. Introduction Gas-expanded liquids (GXLs) are a relatively new class of tunable solvents formed by the dissolution of a gaseous species, CO2 in this study, into an organic liquid.1,2 The resulting mixture is a volume-expanded liquid phase with pressure-tunable properties such as solvent strength and viscosity. Solvent interactions and polarity are tunable all the way from the pure gas to the neat organic liquid, offering a range of interactions and polarity that depends on solvent composition and the choice of components. CO2-expanded methanol (MeOH) was chosen because it combines two components of different polarity and with different interactions. MeOH is a polar protic solvent with hydrogen-bonding capabilities, while CO2 is considered nonpolar but has a quadrupolar moment and is a Lewis acid. CO2 is known to interact with the nitrogen atom on pyridine via Lewis acid/base-type interactions.3 Solvent-dependent spectroscopy, or solvatochromism, is a well-documented technique to probe molecular interactions in solution. Previous solvatochromic studies in GXLs4,5 focused on measuring the well-known Kamlet-Taft parameters to probe the solvent’s hydrogen-bond-donating (R) ability, -accepting ability (β), and polarizability (π*). Other studies in tunable fluids focused on studying the local solvent environment around a solute molecule6-14 called the cybotactic region, defined as the region of the solvent affected by the presence of a solute.15 The composition in the cybotactic region may have a different composition (or density) than the bulk solvent because of intermolecular interactions with the solute that are different than the solvent-solvent interactions in the bulk. The composition in the cybotactic region impacts reaction rates, solubility, and spectroscopy, which is the theme of this study. All of these solvatochromic studies focused on solvent characteristics like * To whom correspondence should be addressed. † School of Chemical & Biomolecular Engineering. § Center for Computational Molecular Science and Technology. ⊥ Specialty Separations Center. ‡ School of Chemistry and Biochemistry.
Figure 1. Solutes used in this study; from the left: benzene, pyridine, pyridazine, pyrimidine, pyrazine.
interactions or local solvation; none have explored the effects of the solute on solvatochromism or local solvation patterns. This study examines the absorption spectra and solvation patterns of the five heterocyclic molecules shown in Figure 1 in CO2-expanded MeOH. These solutes are simple in structure, with slight differences in the number and placement of nitrogen atoms on the aromatic ring, but theyhave very different absorbance spectra that depend on the solute structure. Additionally, these solutes can interact with both solvent components, and the different dipole moments translate to different effects of solvent polarity. The solutes (except benzene (the benzene spectrum represents a π f π* transition)) undergo an n f π* electronic transition; solvents that interact with the lone pair electrons on the nitrogen atom hinder the transition and cause a blue shift. Several groups have examined the solvatochromic behavior of diazines in organic solvents16-19 including protic solvents. Protic solvents caused a blue shift when added to a hydrocarbon and caused the disappearance of vibrational bands that were present only in the hydrocarbon. In CO2expanded alcohols, a nonpolar species with Lewis acid capabilities is added to a protic solvent. The resulting spectra are solute sensitive and provide insight into molecular interactions and local polarity. Molecular dynamics (MD) simulations provide atomic-level detail such as local solvent structure around solute molecules. MD simulations have been used extensively to study solvent structure and solvent dynamics in many different GXLs.12-14,20-23 Previous work by this group used MD simulations to create solvent maps around the laser dye Coumarin 153 (C153) in
10.1021/jp806135s CCC: $40.75 2008 American Chemical Society Published on Web 11/01/2008
14994 J. Phys. Chem. B, Vol. 112, No. 47, 2008
Gohres et al.
MeOH and acetone GXLs.13,14 The work showed that both CO2 and the organic species form different solvation patterns around C153. Also, the number density of each species, relative to the bulk density, increased in the cybotactic region, although C153 was preferentially solvated by the organic species. The focus of this work is the determination of the solute-dependent solvent structure through the creation of detailed solvent maps around solutes of varying polarity and the use of these solvation details to rationalize solvatochromic behavior. This manuscript explores the solvatochromic behavior of each solute in several CO2 + MeOH GXLs and couples these findings to 3D solvent maps and 1D radial distribution functions (RDFs). Solvation is considered from a different perspective than similar simulation/solvatochromism studies by providing structureproperty relationships between the solute structure, local solvation patterns, and absorption spectroscopy. Experimental Techniques Materials. All solutes (benzene, pyridine, pyridazine, pyrimidine, pyrazine, and 1,3,5-triazine, analytical grade) were purchased from Sigma-Aldrich. All were used as received. Carbon dioxide was purchased from Airgas and dried over molecular sieves prior to use. Anhydrous methanol and hexane from Sigma-Aldrich were used in all UV/vis experiments. UV/Vis Absorption. Absorption spectra were obtained using a Hewlett-Packard 1050 series detector. Solvatochromic experiments were performed in a high-pressure optical cell equipped with sapphire windows and a cooling jacket externally connected to a temperature-controlled ethylene glycol bath. The temperature was measured with an Omega J-type thermocouple (with 0.1 °C precision) in contact with the liquid phase. The temperature was maintained at 40 °C throughout the experiment. CO2 was added to the cell with an Isco syringe pump, and the pressure was measured with a Druck pressure transducer with 1 psi precision calibrated by a Ruska hydraulic balance system. Samples were allowed to equilibrate for several hours before recording spectra by rapidly stirring at constant temperature with a Teflon-coated magnetic stir bar. The temperature and pressure were related to the liquid-phase mole fraction via the Patel-Teja equation of state.24 A sixth solute, 1,3,5-triazine, was used in solvatochromic experiments to probe specific interactions. Its absorption spectrum is much more sensitive to solvent effects than is that of pyridine and the diazo compounds but was not included in the computational aspects because it lacked a reliable force field. Computational Methods. CO2 and MeOH were modeled as rigid collections of atomic sites with specified fixed charges interacting through pairwise-additive, site-site Lennard Jones and Coulomb forces. Equation 1 is the force field used in the CO2-MeOH systems
uij )
∑∑ i
j>i
[ {( ) ( ) } 4εij
σij rij
12
-
σij rij
6
+
qiqj rij
]
(1)
CO2 pair interactions have been modeled using the TrAPPE potential.25 The J2 potential26 was used for MeOH pair interactions. The potentials for pyridine, pyrimidine, pyridazine, pyrazine, and benzene are OPLS-derived.27,28 Each of the pairwise potentials specifies a representation for the fixed point charges, and these are assumed to remain fixed in the heterogeneous pairwise Coulombic interactions. Molecular dynamics (MD) simulations were performed using the DL_POLY software package.29 The relative composition
of the cosolvents in the liquid phase has been determined using a semiempirical procedure described in earlier work.1,20 In that work, it was also shown that the structure and dynamics around solutes in the bulk liquid phase can be obtained in a singlephase simulation as long as the appropriate relative compositions are employed. Regardless of the relative composition, each simulated system box is populated by a total of 1001 molecules, 1 solute molecule and 1000 solvent/cosolvent molecules. Cubic periodic boundary conditions (PBCs) are used throughout, but the length of the system box is scaled to satisfy the specified fixed density for each set of conditions. In order to avoid double counting because of the PBC, pairwise terms in the potential were cut off at lengths longer than half of the box length. Coulombic interactions were handled internally by DL_POLY with the Ewald summation method using the “automatic parameter optimization” option with a tolerance of 1E-5. The equations of motion were integrated using a time step of 2 fs with the velocity verlet algorithm as implemented by DL_POLY. Equilibrium configurations were generated by randomly distributing solvent molecules around a centralized solute and equilibrating using velocity rescaling for 200 ps in the canonical (NVT) ensemble. The temperature was initially established and subsequently maintained using a Nose-Hoover thermostat with a 1 ps relaxation time constant. Following equilibration, a long equilibrium simulation was run so as to sample the structures of the NVT ensemble with the temperature maintained only by the Nose-Hoover thermostat. Trajectories were saved every 5 ps and used as initial configurations for the structural studies. Radial distribution functions (RDFs), calculated by eq 2, measure solvent structures by giving the most probable distances between atoms. In this study, RDFs between either the nitrogen on the solute or the center of mass of benzene and either the MeOH protic hydrogen or the CO2 carbon are obtained from the radial distribution function g(r):
g(r) )
〈n(r)〉 F · 4π · r2∆r
(2)
where n(r) is the average number of atoms in a spherical shell of width ∆r at distance r, and F is the bulk density. The vector space (2D or 3D) distribution functions were also determined so as to visualize the solvent maps around benzene, pyridine, pyrimidine, and pyridazine. Only these solutes were treated in this work because they are sufficient for the illustration of the effects of nitrogen displacement and addition. Furthermore, this information can be generalized to the other solutes and their corresponding RDFs. In nonspherically symmetric solutes, the retention of the three-dimensional structural information is necessary because such a structure can lead to preferred chemical contact or reaction routes. The 3D solvent distribution functions, g(x,y,z), are found by dividing the simulation box into finite elements within a Cartesian grid and recording the positions of solvent atoms relative to benzene’s center of mass or a nitrogen atom on pyridine and pyridazine. The magnitude of the distribution function in each finite element is calculated by
g(x, y, z) )
〈NCell〉/VCell NBox /VBox
(3)
where NCell is the average number of a particular atom in a finite element of volume, VCell. NBox and VBox are the total number of a particular atom in the simulation box and volume of the simulation box, respectively. The DL_POLY source code was
Local Solvation and Solvent Interactions
J. Phys. Chem. B, Vol. 112, No. 47, 2008 14995
Figure 2. Pyridine absorption spectra for neat MeOH and CO2 GXLs.
altered to allow for the calculation of the 3D distribution functions. Solvent distribution maps were then created by shading the finite elements according to the magnitude of the distribution function. The maps shown in this work are 2D sample planes that depict MeOH and CO2 solvation of the solutes from in-plane perspectives. Results and Discussion Solvatochromism. Solvent-dependent absorption, or solvatochromism, probes local polarity and solvent interactions with a chromophore.30 The term “local” is used to emphasize the short-range effect that solvent polarity and specific interactions have on absorption. It was assumed that the absorption spectra are influenced by molecules within the cybotactic region. Generally, a more polar local environment decreases the energy of transition required to electronically excite a chromophore, which is inversely proportional to the wavelength of maximum absorption (λmax). This behavior is termed positive solvatochromism, defined as increased λmax with increased solvent polarity. The chromophore must be more polar in the excited state. The azo compounds used in this study display negative solvatochromism. Decreased solvent polarity (from CO2 addition) causes a blue shift (lower wavelength) in the λmax,19,31 but the dipole moments decrease upon excitation.19 There are both polarity effects and specific interactions that cause the blue shifts. Lone pair electrons on the nitrogen atoms cause the large ground-state dipole moment; after excitation, the lone pair electrons become delocalized in the ring, and the dipole moment decreases. Protic solvents hinder electron delocalization because of hydrogen bonding with the nitrogen atoms. This increases the energy of transition required for excitation. CO2-expanded MeOH can form hydrogen bonds through MeOH and Lewis acid/base interactions through CO2. Both decrease the λmax. The six probes give very different spectra and solvatochromic behavior depending on the structural features of the probe. Representative spectra (Figure 2) are provided in the main body; however, remaining spectra (Figures S1-S4) can be found in the Supporting Information. The main observations describing the effect of the varying composition on the solvent structure are summarized here and may be verified by inspection with the corresponding figures in the Supporting Information. Benzene is the only solute that cannot form specific interactions with either solvent component; and as expected, the absorption spectra (Figure S1, Supporting Information) display positive solvatochromism due to decreased solvent polarity with added CO2. The blue shift is small (∼1-2 nm) because benzene has no net dipole moment in the ground and excited states. The other nondipolar solvents, pyrazine and 1,3,5-triazine, have
Figure 3. Pyridazine absorption spectra for neat MeOH and CO2 GXLs.
nitrogen atoms that interact with both MeOH and CO2. Pyrazine’s spectra (Figure S2, Supporting Information) blue shift slightly (∼1 nm) with added CO2 because CO2 interacts with nitrogen. Other studies reported similar observations, showing that hydrogen bonding causes a large solvatochromic shift between isooctane and MeOH.19,32 There is an additional blue shift from CO2-nitrogen interactions relative to hydrogen bonds. The magnitude of the shift is small but significant. Pyridine (Figure 2) also displays this behavior because of CO2 interactions. A subsequent study with the probe 1,3,5-triazine displayed a blue shift in CO2-expanded MeOH and CO2-expanded hexane (Figure S3, Supporting Information). This confirms the influence of CO2 interactions on solvatochromism, relative to MeOH, because hexane does not form a specific interaction with the nitrogen atoms. Pyrimidine absorption in neat MeOH has an n-π* absorption peak at ∼245 nm, with an additional vibrational band at 250 nm Figure S4, Supporting Information. CO2 addition does not cause a shift in either peak, but three absorption bands between 260 and 270 nm appear when a small amount of CO2 is added. Similar behavior was seen in the absorption spectra in isopentane/ethanol liquid mixtures, where increased isopentane led to the appearance of these peaks on the red side,18 but a significant solvatochromic shift of the n-π* transition was observed because of ethanol hydrogen bonding. The vibrational bands did not shift because they are insensitive to the local solvent polarity. Several factors could cause the appearance of the vibrational bands: reduced clustering of MeOH around the nitrogen atoms or unequal hydrogen bonding between the two atoms. In GXLs, CO2 disrupts the uniform hydrogen-bonded structure between the two nitrogen atoms, causing asymmetry of the solvent structure and the additional vibrational bands. This pattern is unique to pyrimidine, which is the only solute with a dipole moment and two nonadjacent nitrogen atoms, thus more susceptible to nonuniform MeOH-nitrogen interactions. Pyridazine is the only solute that red shifts with added CO2 pressure, as seen in Figure 3. This is in agreement with data from an ethanol-hexane liquid mixture study16 that red shifted with increased hexane but is contrary to the blue shifts that pyridine, pyrazine, and 1,3,5-triazine exhibit with added CO2. RDFs indicate that decreased CO2 interactions with the nitrogen atoms at higher CO2 concentration are responsible for the red shift. Solvent Structure. Molecular dynamics simulations provide atomic-scale structural details that aid the interpretation of spectra and diffusion results. RDFs between the solvent
14996 J. Phys. Chem. B, Vol. 112, No. 47, 2008
Figure 4. 3D distribution function from an in-plane perspective to illustrate CO2 accumulation (Left) and MeOH accumulation (Right) above and below the benzene molecular plane. The horizontal line represents the benzene molecule; darker shaded regions are areas of increased MeOH or CO2 presence relative to the bulk concentration (white areas). The scale corresponds to the magnitude of the distribution function in the discrete area, as calculated by eq 3. Tick marks are ∼0.5 Å apart.
components and the solute nitrogen(s) indicate that both CO2 and MeOH influence solvatochromic properties. Solute center of mass RDFs give less insight into the effects of neighboring atoms and are not provided. Instead, solvent maps, or 3D distribution functions, are used to show the structural effects from nearby atoms. Solvent maps for representative solutes are provided in the main body to highlight solvation trends. Several RDFs are provided in the Supporting Information as the major conclusions can be drawn with the 3D solvent maps alone. This structural analysis shows that while MeOH-nitrogen hydrogen bonds are the predominant interaction, CO2 can interact with the probe and disrupt short- and long-range solvent structure, ultimately affecting the absorption spectroscopy. The solvatochromic shift relative to pure MeOH is more dependent on the CO2-nitrogen interactions than hydrogen bonds. Benzene. UV/vis absorption suggests that the microenvironment around benzene becomes increasingly nonpolar with CO2 addition. Center of mass RDFs and 3D axial distribution functions show that the solvent forms a neatly organized cage around the solute. Benzene is the only solute where center of mass RDFs and 3D distribution functions are explicitly determined because benzene lacks nitrogen atoms and many of the solvation features can be generalized to the other solute molecules. The 3D solvent distribution functions for CO2 and MeOH around benzene are shown in Figure 4. These display solvation above and below the benzene molecule from an inplane perspective. The distinct feature of MeOH solvation is the presence of two areas with significant density found at ∼3 Å above the center of the ring. MeOH forms a circular solvent cage around benzene, with significant agglomeration directly behind the two clusters. CO2 forms an ovoid solvent cage with a higher composition than MeOH and increased clustering directly behind the two MeOH clusters at 3 Å from the solute. 3D solvent maps provide a better spatial resolution than 1D RDFs and are a basis for interpreting RDFs in Figure 5. The COM-H RDFs, that is, those corresponding to the distance between the benzene center of mass and the hydrogen in MeOH, are shown in Figure 5. The small (
