Computer-aided molecular design of solvents for accelerated reaction kinetics Heiko Struebing, Zara Ganase, Panagiotis G. Karamertzanis, Eirini Siougkrou, Peter Haycock, Patrick Piccione, Alan Armstrong, Amparo Galindo, Claire S. Adjiman Contents Design problem... 1 Experimental methodology... 2 Computational methodology... 3 Progress of the QM-CAMD algorithm for the Menschutkin reaction... 4 Gas and liquid phase optimised conformations... 9 Design problem Full details of the mathematical problem posed can be found on the online collaborative site on Cyber-Infrastructure for Mixed Integer NonLinear Problems (MINLP), run by Carnegie Mellon University and IBM at the following address: http://www.minlp.org/library/problem/index.php?i=137&lib=minlp. The 38 functional groups used in formulating the design problem are shown Supplementary Table 1. Supplementary Table 1 Functional groups used to build solvent molecules. CH 3 CH 2 CH C CH=C ach ac acch 3 acch 2 acch OH acoh CH 3 C=O CH 2 C=O CH=O CH 3 COO CH 2 COO CH 3 -O CH 2 -O CH-O CH 2 CN COOH CH 2 Cl CHCl CHCl 2 accl CH 2 NO 2 CHNO 2 CHNO 2 I acf CH 3 CN C 7 H 5 N C 7 H 8 O CHCl 3 C 6 H 5 NO 2 CH 3 NO 2 C 4 H 8 O "ac" denotes a carbon in an aromatic ring. NATURE CHEMISTRY www.nature.com/naturechemistry 1
Experimental methodology Supplementary Table 2 Initial molar concentrations of reactants in experiments and parameter estimation. Experiment Number [1] w / mol dm -3 [1] NMR / mol dm -3 [1] EST / mol dm -3 [2] w / mol dm -3 [2] NMR / mol dm -3 [2] EST / mol dm -3 1 0.10 0.11 0.11 0.17 0.19 0.20 2 0.18 0.18 0.18 0.35 0.35 0.37 3 0.27 0.26 0.26 0.48 0.47 0.51 4 0.36 0.35 0.35 0.48 0.46 0.52 Phenacyl bromide is denoted by [1], and pyridine by [2]. Experiments in nitromethane-d3, where the subscript w denotes the value obtained based on initial weighing; the subscript NMR denotes the value calibrated to the concentration of the internal standard; the subscript EST denotes the value obtained during the parameter estimation procedure. Supplementary Figure 1 - Measured (symbols) and estimated (lines) concentration values of phenacyl bromide, where the colours represent a different reaction; blue=1, red=2, 3=green and 4=purple. Supplementary Figure 2 - Measured (symbols) and estimated (lines) concentration values of pyridine, where the colours represent a different reaction; blue=1, red=2, 3=green and 4=purple. NATURE CHEMISTRY www.nature.com/naturechemistry 2
Supplementary Figure 3 - Measured (symbols) and estimated (lines) concentration values of the product, where the colours represent a different reaction; blue=1, red=2, 3=green and 4=purple. Computational methodology Solvent Predicted description, e.g., solvent bulk 2 x CH 3, 4 x CH 2 GC properties, e.g., SMD techniques dielectric constant surface tension + DFT Activation free energy of solvation CTST k QM Supplementary Figure 4 - Overall methodology for rate constant calculation in a given solvent, such as hexane. GC techniques refer to group contribution techniques, SMD + DFT to the use of density functional theory and the SMD continuum solvation model to compute free energies of solvation of the reactants and transition state, CTST to conventional transition state theory and k QM to the calculated rate constant. Modelling choices made in computing k QM are discussed in more details before presenting the results of the calculations with the chosen approach. Transmission coefficient. Although numerous approximations to the transmission coefficient exist, the simplest transmission coefficient in common use (Garrett & Truhlar 1979, Sumathi & Green 2002, Barreto et al. 2003, Ashcraft et al. 2008, Henriksen & Hansen 2008) is the Wigner tunnelling correction factor. This transmission coefficient can be implemented when a small degree of tunnelling occurs or when the system is at a high temperature (Henriksen & Hansen 2008). Tunnelling is not expected to be a significant factor in this reaction and the Wigner coefficient is therefore used. Calculation of vibrational frequencies. The vibrational frequencies required in the calculation of the rate constant are computed in the gas phase. Whether vibrational frequencies calculated in the liquid phase should be used has been the subject of some debate in the literature 1,2. Given that the SMD model has been parameterised using solutes for which the difference between gas phase and liquid phase vibrational energies is small, liquid phase vibrational frequencies may in principle lead to improved accuracy. However, gas phase vibrational frequencies have been found to yield reliable results in many cases 2 and our findings are consistent with this. 1 J. Ho, A. Klamt, M.L. Coote, J Phys. Chem. A 2010, 114, 13442-13444. 2 R. F. Ribeiro, A. V. Marenich, C. J. Cramer, D. G. Truhlar, J. Phys. Chem. B 2011, 115, 14556 14562 NATURE CHEMISTRY www.nature.com/naturechemistry 3
Choice of functional and basis set. There are extensive benchmarking studies available for choosing computational models to obtain accurate barrier heights for gas phase reactions 3,4. There are, however, no such studies for liquid phase reactions. In choosing a computational model, we considered a number of options and found that a good qualitative assessment of solvent effects can be found with many different options. It also quickly became apparent that if the functional and basis set chosen differ significantly from those used to parameterise SMD, the quantitative agreement is poor. Of the models tested, M05-2X/6-31G(d) resulted in the closest match to experimental data, followed by B3LYP/6-31+G(d) and B3LYB/6-31+G(d,p) (see Supplementary Figure 5). In the paper, we report the use of B3LYP/6-31+G(d) in the QM-CAMD approach. We have also used M05-2X/6-31G(d) within the QM-CAMD methodology, and have found the same final solvent design, indicating the robustness of the approach. Supplementary Figure 5 - Comparison of rate constants calculated with 3 different computational models and experimental rate constants. Progress of the QM-CAMD algorithm for the Menschutkin reaction Step numbers make reference to Figure 1 in the main article. QM calculations are performed using B3LYP/6-31+G(d). Iteration 1 Step 2: Obtain reaction rate constants Toluene Chlorobenzene Ethyl acetate Tetrahydrofuran Acetone Acetonitrile Solvent Rate constant [dm 3 mol -1 s -1 ] 1.61E-06 3.94E-05 4.88E-05 8.70E-05 2.86E-04 4.70E-04 3 Y. Zhao, N. González-García, and D. G. Truhlar, J. Phys. Chem. A 2005, 109, 2012-2018; Y. Zhao, N. González-García, and D. G. Truhlar,, J. Phys. Chem. A 2006, 110, 4942(E) 4 J. Zheng, Y. Zhao, and D. G. Truhlar, Journal of Chemical Theory and Computation 5, 808-821 (2009). NATURE CHEMISTRY www.nature.com/naturechemistry 4
Step 3: Build surrogate model log k = -22.43-100.72 A + 8.55 B + 7.61 S + 2.31 δ + 11.99 δ H 2 /100 Step 4: CAMD solvent design Best solvent found: Iodonitromethane Groups Number of groups CH2NO2 1 I 1 k CAMD (dm 3 mol -1 s -1 ) 6.17E+11 Step 5: Test for convergence New solvent generated - Continue Iteration 2 Step 2: Obtain reaction rate constant for new solvent Solvent Toluene Chlorobenzene Ethyl acetate Tetrahydrofuran Acetone Acetonitrile Iodonitromethane Rate constant [dm 3 mol -1 s -1 ] 1.61E-06 3.94E-05 4.88E-05 8.70E-05 2.86E-04 4.70E-04 2.47E-04 Step 3: Build surrogate model log k = -1.32-13.40 A - 7.61 B + 3.36 S - 4.13 δ - 1.20 δ H 2 /100 Step 4: CAMD solvent design Best solvent found: cf table; several isomers possible Groups Number of groups CH2NO2 1 CH3 3 CH2 2 C=C 1 k CAMD (dm 3 mol -1 s -1 ) 2.22E-02 NATURE CHEMISTRY www.nature.com/naturechemistry 5
Step 5: Test for convergence New solvent generated - Continue Iteration 3 Step 2: Obtain reaction rate constant for new solvent Solvent Toluene Chlorobenzene Ethyl acetate Tetrahydrofuran Acetone Acetonitrile Iodonitromethane 1xCH2NO2,3xCH3,2xCH2,1xC=C Rate constant [dm 3 mol -1 s -1 ] 1.61E-06 3.94E-05 4.88E-05 8.70E-05 2.86E-04 4.70E-04 2.47E-04 8.46E-05 Step 3: Build surrogate model log k = -4.50-0.84 A - 1.11 B - 0.11 S - 1.46 δ + 1.21 δ H 2 /100 Step 4: CAMD solvent design Best solvent found: Nitromethanol Groups Number of groups CH2NO2 1 OH 1 k CAMD (dm 3 mol -1 s -1 ) 8.25E-03 Step 5: Test for convergence New solvent generated - Continue Iteration 4 Step 2: Obtain reaction rate constant for new solvent Toluene Chlorobenzene Solvent Rate constant [dm 3 mol -1 s -1 ] 1.61E-06 3.94E-05 NATURE CHEMISTRY www.nature.com/naturechemistry 6
Ethyl acetate Tetrahydrofuran Acetone Acetonitrile Iodonitromethane 1xCH2NO2,3xCH3,2xCH2,1xC=C Nitromethanol 4.88E-05 8.70E-05 2.86E-04 4.70E-04 2.47E-04 8.46E-05 5.76E-04 Step 3: Build surrogate model log k = - 4.18-4.71 A - 1.79 B - 0.35 S - 1.75 δ + 1.41 δ H 2 /100 Step 4: CAMD solvent design Best solvent found: Nitromethane Groups Number of groups CH3NO2 1 kcamd 6.23E-02 Step 5: Test for convergence New solvent generated - Continue Iteration 5 Step 2: Obtain reaction rate constant for new solvent Solvent Toluene Chlorobenzene Ethyl acetate Tetrahydrofuran Acetone Acetonitrile Iodonitromethane 1xCH2NO2,3xCH3,2xCH2,1xC=C Nitromethanol Nitromethane Rate constant [dm 3 mol -1 s -1 ] 1.61E-06 3.94E-05 4.88E-05 8.70E-05 2.86E-04 4.70E-04 2.47E-04 8.46E-05 5.76E-04 6.09E-04 NATURE CHEMISTRY www.nature.com/naturechemistry 7
Step 3: Build surrogate model log k = -4.25-4.12 A - 1.59 B - 0.16 S - 1.66 δ + 1.24 δ H 2 /100 Step 4: CAMD solvent design Best solvent found: Nitromethane Groups Number of groups CH3NO2 1 kcamd 6.71E-04 Step 5: Test for convergence No new solvent generated. Step 6: Test best solvent found experimentally Best solvent: Nitromethane NATURE CHEMISTRY www.nature.com/naturechemistry 8
Gas and liquid phase optimised conformations All conformations are obtained via minimisations at the B3LYP/6-31+G(d) level of theory. 1. Optimised gas phase conformations a. Transition state conformation Supplementary Figure 6 - Cartesian coordinates for the transition state conformation for the specified Menschutkin reaction in the gas phase. NATURE CHEMISTRY www.nature.com/naturechemistry 9
b. Phenacyl bromide conformation Supplementary Figure 7 - Cartesian coordinates for the conformation of phenacyl bromide for the specified Menschutkin reaction in the gas phase. c. Pyridine conformation Supplementary Figure 8 - Cartesian coordinates for the conformation of pyridine for the specified Menschutkin reaction in the gas phase. NATURE CHEMISTRY www.nature.com/naturechemistry 10
2. Optimised liquid phase conformations a. Toluene i. Transition state conformation Supplementary Figure 9 - Cartesian coordinates for the transition state conformation for the specified Menschutkin reaction in the solvent toluene. NATURE CHEMISTRY www.nature.com/naturechemistry 11
ii. Phenacyl bromide conformation Supplementary Figure 10 - Cartesian coordinates for the conformation of phenacyl bromide for the specified Menschutkin reaction in the solvent toluene. iii. Pyridine conformation Supplementary Figure 11 - Cartesian coordinates for the conformation of pyridine for the specified Menschutkin reaction in the solvent toluene. NATURE CHEMISTRY www.nature.com/naturechemistry 12
b. Chloroform i. Transition state conformation Supplementary Figure 12 - Cartesian coordinates for the transition state conformation for the specified Menschutkin reaction in the solvent chloroform. NATURE CHEMISTRY www.nature.com/naturechemistry 13
ii. Phenacyl bromide conformation Supplementary Figure 13 - Cartesian coordinates for the conformation of phenacyl bromide for the specified Menschutkin reaction in the solvent chloroform. iii. Pyridine conformation Supplementary Figure 14 - Cartesian coordinates for the conformation of pyridine for the specified Menschutkin reaction in the solvent chloroform. NATURE CHEMISTRY www.nature.com/naturechemistry 14
c. Chlorobenzene i. Transition state conformation Supplementary Figure 15 - Cartesian coordinates for the transition state conformation for the specified Menschutkin reaction in the solvent chlorobenzene. NATURE CHEMISTRY www.nature.com/naturechemistry 15
ii. Phenacyl bromide conformation Supplementary Figure 16 - Cartesian coordinates for the conformation of phenacyl bromide for the specified Menschutkin reaction in the solvent chlorobenzene. iii. Pyridine conformation Supplementary Figure 17 - Cartesian coordinates for the conformation of pyridine for the specified Menschutkin reaction in the solvent chlorobenzene. NATURE CHEMISTRY www.nature.com/naturechemistry 16
d. Ethyl acetate i. Transition state conformation Supplementary Figure 18 - Cartesian coordinates for the transition state conformation for the specified Menschutkin reaction in the solvent ethyl acetate. NATURE CHEMISTRY www.nature.com/naturechemistry 17
ii. Phenacyl-bromide conformation Supplementary Figure 19 - Cartesian coordinates for the conformation of phenacyl bromide for the specified Menschutkin reaction in the solvent ethyl acetate. iii. Pyridine conformation Supplementary Figure 20 - Cartesian coordinates for the conformation of pyridine for the specified Menschutkin reaction in the solvent ethyl acetate. NATURE CHEMISTRY www.nature.com/naturechemistry 18
e. Tetrahydrofuran i. Transition state conformation Supplementary Figure 21 - Cartesian coordinates for the transition state conformation for the specified Menschutkin reaction in the solvent tetrahydrofuran. NATURE CHEMISTRY www.nature.com/naturechemistry 19
ii. Phenacyl bromide conformation Supplementary Figure 22 - Cartesian coordinates for the conformation of phenacyl bromide for the specified Menschutkin reaction in the solvent tetrahydrofuran. iii. Pyridine conformation Supplementary Figure 23 - Cartesian coordinates for the conformation of pyridine for the specified Menschutkin reaction in the solvent tetrahydrofuran. NATURE CHEMISTRY www.nature.com/naturechemistry 20
f. Acetone i. Transition state conformation Supplementary Figure 24 - Cartesian coordinates for the transition state conformation for the specified Menschutkin reaction in the solvent acetone. NATURE CHEMISTRY www.nature.com/naturechemistry 21
ii. Phenacyl bromide conformation Supplementary Figure 25 - Cartesian coordinates for the conformation of phenacyl bromide for the specified Menschutkin reaction in the solvent acetone. iii. Pyridine conformation Supplementary Figure 26 - Cartesian coordinates for the conformation of pyridine for the specified Menschutkin reaction in the solvent acetone. NATURE CHEMISTRY www.nature.com/naturechemistry 22
g. Acetonitrile i. Transition state conformation Supplementary Figure 27 - Cartesian coordinates for the transition state conformation for the specified Menschutkin reaction in the solvent acetonitrile. NATURE CHEMISTRY www.nature.com/naturechemistry 23
ii. Phenacyl bromide conformation Supplementary Figure 28 - Cartesian coordinates for the conformation of phenacyl bromide for the specified Menschutkin reaction in the solvent acetonitrile. iii. Pyridine conformation Supplementary Figure 29 - Cartesian coordinates for the conformation of pyridine for the specified Menschutkin reaction in the solvent acetonitrile. NATURE CHEMISTRY www.nature.com/naturechemistry 24
h. Nitromethane i. Transition state conformation Supplementary Figure 30 - Cartesian coordinates for the transition state conformation for the specified Menschutkin reaction in the solvent nitromethane. NATURE CHEMISTRY www.nature.com/naturechemistry 25
ii. Phenacyl-bromide conformation Supplementary Figure 31 - Cartesian coordinates for the conformation of phenacyl bromide for the specified Menschutkin reaction in the solvent nitromethane. iii. Pyridine conformation Supplementary Figure 32 - Cartesian coordinates for the conformation of pyridine for the specified Menschutkin reaction in the solvent nitromethane. NATURE CHEMISTRY www.nature.com/naturechemistry 26