Solving multi-reference problems with a single-reference coupled- cluster method Yingbin Ge Department of Chemistry Central Washington University 65 th.

Slides:



Advertisements
Similar presentations
Introduction to Møller-Plesset Perturbation Theory
Advertisements

“Rotational Energy Transfer in o - / p -H 2 + HD” Renat A. Sultanov and Dennis Guster BCRL, St. Cloud State University St. Cloud, MN June 20, 2007 OSU.
Chapter 14 Chemical Kinetics.
KINETICS.
HIGH PERFORMANCE ELECTRONIC STRUCTURE THEORY Mark S. Gordon, Klaus Ruedenberg Ames Laboratory Iowa State University BBG.
COUPLED-CLUSTER CALCULATIONS OF GROUND AND EXCITED STATES OF NUCLEI Marta Włoch, a Jeffrey R. Gour, a and Piotr Piecuch a,b a Department of Chemistry,Michigan.
B3LYP study on the lowest energy Pt clusters and their reactivity towards small alkanes T. Cameron Shore, Drake Mith, and Yingbin Ge* Department of Chemistry,
B3LYP study of the dehydrogenation of propane catalyzed by Pt clusters: Size and charge effects T. Cameron Shore, Drake Mith, Staci McNall, and Yingbin.
Hydrates of Cytosine: The Order of Binding Water Molecules Géza Fogarasi, Péter G. Szalay Institute of Chemistry, Eötvös Loránd University, Budapest, H-1518,
Chemistry with Computers Yingbin Ge Iowa State University Central Washington University October 13, 2007.
Coupled Cluster Calculations using Density Matrix Renormalization Group "like" idea Osamu Hino 1, Tomoko Kinoshita 2 and Rodney J. Bartlett 1 Quantum Theory.
Potensial Energy Surface Pertemuan V. Definition Femtosecond spectroscopy experiments show that molecules vibrate in many different directions until an.
A non-iterative perturbative triples correction for the spin-flipping equation- of-motion coupled-cluster methods with single and double substitutions.
A Fast Algorithm for Generalized Van Vleck Perturbation Theory Wanyi Jiang, Yuriy G. Khait, Alexander V. Gaenko, and Mark R. Hoffmann Chemistry Department,
ChE 551 Lecture 19 Transition State Theory Revisited 1.
1 Entropy Explained: The Origin of Some Simple Trends Lori A. Watson a, Odile Eisenstein b a Department of Chemistry, Indiana University, Bloomington,
Ab Initio Calculations of the Ground Electronic States of the C 3 Ar and C 3 Ne Complexes Yi-Ren Chen, Yi-Jen Wang, and Yen-Chu Hsu Institute of Atomic.
1 Li Xiao and Lichang Wang Department of Chemistry & Biochemistry Southern Illinois University Carbondale The Structure Effect of Pt Clusters on the Vibrational.
Chapter 14: Rates of Reaction Chemistry 1062: Principles of Chemistry II Andy Aspaas, Instructor.
Vibrational Relaxation of CH 2 ClI in Cold Argon Amber Jain Sibert Group 1.
Multi-reference Configuration-Interaction Calculations of the Low-Lying Electronic States of Iron and Vanadium Monohydride, FeH and VH Zhong Wang, Trevor.
Chemical Kinetics Unit 11. Chemical Kinetics Chemical equations do not give us information on how fast a reaction goes from reactants to products. KINETICS:
Frank Lee Emmert III, Stephanie Thompson, and Lyudmila V. Slipchenko Purdue University, West Lafayette, IN
1 The Structure and Ring Puckering Barrier of Cyclobutane: A Theoretical Study Sotiris S. Xantheas, Thomas A. Blake Environmental Molecular Sciences Laboratory.
ChE 551 Lecture 23 Quantum Methods For Activation Barriers 1.
Chemical Kinetics The study of rates of chemical reactions, the factors that affect the rates, and the sequence of steps by which a reaction occurs. The.
AP CHEMISTRY CHAPTER 12 KINETICS. 2 Chemical Kinetics Thermodynamics tells us if a reaction can occur Kinetics tells us how quickly the reaction occurs.
The Advanced Light Source (ALS) at Lawerence Berkeley National Laboratory Berkeley, California Tunable VUV radiation from 8 – 30 eV Brian W. Ticknor 1,
Veracity through variety (of methods): Simulating dipeptides with little volume Tanja van Mourik.
Silver Nyambo Department of Chemistry, Marquette University, Wisconsin Reactive pathways in the chlorobenzene-ammonia dimer cation radical: New insights.
Theoretical Study of the Flame Synthesis of Titanium Dioxide Nanoparticles Kui Ting Lam, Doug DePrekel, Kevin Ngo, Phu Vo, and Yingbin Ge Department of.
Volker Lutter, Laborastrophysik, Universität Kassel 69 th ISMS Champaign-Urbana, Illinois HIGH RESOLUTION INFRARED SPECTROSCOPY AND SEMI-EXPERIMENTAL STRUCTURES.
1 Chapter 10 Copyright © The McGraw-Hill Companies, Inc. Permission required for reproduction or display. Chemical Bonding II: Molecular Geometry and Hybridization.
Chemical Bonding II: Molecular Geometry and Hybridization of Atomic Orbitals Chapter 10 Copyright © The McGraw-Hill Companies, Inc.  Permission required.
A Walkthrough For Quantum Chemistry Newbies Part 1: Basis Sets, Related Functions, and Usage by Peker Milas.
Some Problems (Precautions) of ab initio Quantum Chemistry Jim Jr-Min Lin Most papers only show the “good” sides of the calculation and hide the problematic.
SILYL FLUORIDE: LAMB-DIP SPECTRA and EQUILIBRIUM STRUCTURE Cristina PUZZARINI and Gabriele CAZZOLI Dipartimento di Chimica “G. Ciamician”, Università di.
Double Excitations and Conical Intersections in Time-Dependent Density Functional Theory Chaehyuk Ko, Benjamin G. Levine, Richard M. Martin, and Todd J.
Kinetics Chemistry—Introduction
Int. Symp. Molecular Spectroscopy Ohio State Univ., 2005 The Ground State Four Dimensional Morphed Potentials of HBr and HI Dimers Collaborator: J. W.
Antal Zoltan-PhD candidate 6304-Computational Chemistry March 2010.
Calculating Potential Energy Curves With Quantum Monte Carlo Andrew D Powell, Richard Dawes Department of Chemistry, Missouri University of Science and.
Computational Studies of the Electronic Spectra of Transition-Metal-Containing Molecules James T. Muckerman, Zhong Wang, Trevor J. Sears Chemistry Department,
COMPOSITE METHODS. Based on additivity (independence) of –Basis set improvements –Level of theory improvements High level = H [e.g., CCSD(T)] Low level.
Effective C 2v Symmetry in the Dimethyl Ether–Acetylene Dimer Sean A. Peebles, Josh J. Newby, Michal M. Serafin, and Rebecca A. Peebles Department of Chemistry,
1 HOONO ISOMERIZATION TO HONO 2 INVOLVING CONICAL INTERSECTIONS T. J. DHILIP KUMAR, and JOHN R. BARKER Department of Atmospheric, Oceanic and Space Sciences,
1 MODELING MATTER AT NANOSCALES 6. The theory of molecular orbitals for the description of nanosystems (part II) Perturbational methods for dealing.
Multiply Charged Ions Quantum Chemical Computations Trento, May 2002 Lecture 2.
2008 International Symposium on Molecular Spectroscopy Anion Photoelectron Spectra of CHX 2 - and CX 2 - Properties of the Corresponding Neutrals Scott.
Spectroscopic and Theoretical Determination of Accurate CH/  Interaction Energies in Benzene-Hydrocarbon Clusters Asuka Fujii, Hiromasa Hayashi, Jae Woo.
Dissociation of H 2 Do HF calculations for different values of the H-H internuclear distance (this distance is fixed since we are in the Born- Oppenheimer.
Orbital Hybridisation & VSEPR Learning Goals Students will be able to predict the hybridization in a variety of compounds using Lewis Structures & energy.
Chapter 9 Alkynes. 9.1 Sources of Alkynes Industrial preparation of acetylene is by dehydrogenation of ethylene CH 3 CH 3 800°C 1150°C cost of energy.
FAR-IR ACTION SPECTROSCOPY OF AMINOPHENOL AND ETHYLVANILLIN: EXPERIMENT AND THEORY Vasyl Yatsyna, Daniël Bakker*, Raimund Feifel, Vitali Zhaunerchyk, Anouk.
The Rotational Spectrum of the Water–Hydroperoxy Radical (H 2 O–HO 2 ) Complex Kohsuke Suma, Yoshihiro Sumiyoshi, and Yasuki Endo Department of Basic Science,
Chemical Kinetics The rate of a reaction is the positive quantity that expresses how the concentration of a reactant or product changes with time. The.
AP CHEMISTRY Chapter 14 Chemical Kinetics Chemical Kinetics Study of how rapidly a reaction will occur. In addition to speed of reaction, kinetics.
1 Molecular Geometry and Hybridization of Atomic Orbitals.
Rotational spectra of C2D4-H2S, C2D4-D2S, C2D4-HDS and 13CH2CH2-H2S complexes: Molecular symmetry group analysis Mausumi Goswami and E. Arunan Inorganic.
Substitution Structures of Large Molecules and Medium Range Correlations in Quantum Chemistry Calculations Luca Evangelisti Dipartmento di Chimica “Giacomo.
Structure and Bonding in Organic Chemistry
Statistical Mechanics and Multi-Scale Simulation Methods ChBE
Chemical Bonding II: Molecular Geometry and Hybridization of Atomic Orbitals Chapter 9 Copyright © The McGraw-Hill Companies, Inc.  Permission required.
Electronic Structure Theory
Chapter 1B Carbon Compounds and Chemical Bonds
Chapter 13 Chemical Kinetics.
Potential Energy Main Concept:
Wafaa Fawzy Murray State University (MSU)
Presentation transcript:

Solving multi-reference problems with a single-reference coupled- cluster method Yingbin Ge Department of Chemistry Central Washington University 65 th Northwest/22 nd Rocky Mountain Regional Meeting of the American Chemical Society, June 20-23,

Contents Why solve multi-reference problems with a single- reference method? Background of the left-eigenstate completely renormalized coupled-cluster method (CCL). 1-3 Using the CCL method on o Bi-radicals (Eg. H 3 C–H  CH 3 + H). o Tri-radicals (Eg. H 2 C–H  3 CH 2 + H). o Tetra-radicals (Eg. H 2 Si=SiH 2  3 SiH SiH 2 ). o Unimolecular decomposition of ethylene (C 2 H 4 ) P. Piecuch and M. Wloch, J. Chem. Phys. 123, (2005). 2. P. Piecuch, M. Wloch, and J. R. Gour, Chem. Phys. Lett. 418, 467 (2006). 3. M. Wloch, P. Piecuch, and J. R. Gour, J. Phys. Chem. A 111, (2007).

Why solve multi-reference problems with a single-reference method? To study the mechanism of chemical vapor deposition such as silicon carbide CVD. >100 gas-phase reactions at 1000 – 2000 K; many of which involve bi/tri/tetra-radicals. 1 Need an accurate, size-extensive, and inexpensive black-box method to study CVD. 1.Y. Ge, M. S. Gordon, F. Battaglia, and R. O. Fox, JPCA, 111, 1462 & 1475 (2007). 2.Y. Ge, M. S. Gordon, F. Battaglia, and R. O. Fox, JPCA, 114, 2384 (2010). 3

Background of CCL “Left eigenstate completely-renormalized coupled-cluster singles, doubles, and non-iterative triples”, a.k.a. CR-CCSD(T) L or CR-CC(2,3), 1-3 abbreviated as CCL. An accurate, size-extensive, and relatively inexpensive black-box method. RHF and ROHF based CCL are implemented in GAMESS ($contrl cctyp=cr-ccl $end). 1. P. Piecuch and M. Wloch, J. Chem. Phys. 123, (2005). 2. P. Piecuch, M. Wloch, and J. R. Gour, Chem. Phys. Lett. 418, 467 (2006). 3. M. Wloch, P. Piecuch, and J. R. Gour, J. Phys. Chem. A 111, (2007). 4

5 Breaking bonds of closed-shell species in silicon carbide CVD 1 A–B  A + B A = H, Cl, CH 3, SiH 3 B = H, Cl, CH 3, SiH 3 1. Y. Ge, M. S. Gordon, and P. Piecuch, J. Chem. Phys. 127, (2007).

6 H 3 C–SiH 3  CH 3 + SiH 3 CCL, CCSD(T), and FCI energies MINI basis set

7 H 3 C–SiH 3  CH 3 + SiH 3 Errors (in mE h ) relative to FCI/MINI

Quantitative Assessment of the quality of the CCL method?  E CCL (i) = E CCL (i) – E FCI (i) at the i-th structure. NPE: nonparallelity error. NPE CCL = max[  E CCL (1),  E CCL (2), …,  E CCL (N)] – min[  E CCL (1),  E CCL (2), …,  E CCL (N)] REE: reaction energy error. REE CCL = |  E CCL (3R e ) –  E CCL (R e )| 8

999 H 3 C–SiH 3  CH 3 + SiH 3 Errors (in mE h ) relative to FCI/MINI

10 H 3 C–SiH 3  CH 3 + SiH 3 Errors (in mE h ) relative to FCI/MINI NPE REE

11 Nonparallelity error (in mE h ) a. Excluding the H 3 C−Cl data

12 Reaction energy error (in mE h ) a. Excluding the H 3 C−Cl data

Breaking bonds of open-shell species in silicon carbide CVD 1 A–B  A + B A = 3 CH 2, 1 SiH 2 B = H, Cl, CH 3, SiH 3 1. Y. Ge, M.S. Gordon, P. Piecuch, M. Wloch, and J.R. Gour,J. Phys. Chem. 112, (2008). 13

Methods to be compared ROHF-based CCL. UHF-based CCSD(T). Multi-reference MP2 (MRMP2). Basis Sets:  MINI  6-31G  6-31G(d)  cc-pVDZ  cc-pVTZ 14

Benchmark methods Full configuration interaction (FCI) gives exact energy within a given basis set. Full second-order configuration interaction (FSOCI). Internally-contracted multi-reference configuration interaction (MRCI). Davidson quadruple correction for FSOCI(Q) and MRCI(Q). FCI >> FSOCI(Q) >> MRCI(Q) in cost. 15

H 2 C–H  3 CH 2 + H Errors (in mE h ) relative to FCI/6-31G(d) 16

Nonparallelity error (in mE h ) of FSOCI(Q) and MRCI(Q) MRCI(Q) will be used as benchmark when larger basis sets are used. 17

ROCCL vs. UCCSD(T) ROCCL vs. MRMP2 UCCSD(T): often used to treat near- equilibrium open-shell species with little multi- reference character. MRMP2: can be used to treat bond-breaking reactions, bi-radical and tri-radical systems with significant multi-reference character. Eg. H 2 C–H  3 CH 2 + H 18

H 2 C–H  3 CH 2 + H Errors (in mE h ) relative to FCI/6-31G(d) 19

H 2 C–H  3 CH 2 + H Errors (in mE h ) relative to MRCI(Q)/cc-pVTZ 20

H 2 C–H  3 CH 2 + H Nonparallelity error (in mE h ) 21

H 2 C–H  3 CH 2 + H Reaction energy error (in mE h ) 22

Average (over several basis sets) nonparallelity error (in mE h ) 23

Average (over several basis sets) reaction energy error (in mE h ) 24

Average NPE and REE (in mE h ) for bond-breaking reactions of open-shell species with tri-radical character 25

How about tetra-radicals? Neither is good; CCL is better. 26 MINI Basis Set

Quality of the CCL method For bi-radicals, NPE & REE: R-CCL << R-CCSD(T). For tri-radicals, NPE: RO-CCL ≈ MRMP2 < U-CCSD(T). REE: RO-CCL ≈ U-CCSD(T) < MRMP2. CCL is ideal for calculations on PES study such as the unimolecular decomposition pathways of C 2 H 4. 27

Why C 2 H 4  C 2 H 2 + H 2 ? Chemical vapor deposition of diamond, graphite, carbon nanotubes, and silicon carbide. Lack of accurate potential energy surface (PES) of the decomposition of C 2 H 4. C 2 H 4  C 2 H 2 + H 2 : is there a direct path? 28

Computational Methods Geometry optimization calculations at the CCL/cc-pVTZ level. Hessian calculations at CCL/cc-pVTZ level. Harmonic-oscillator/rigid-rotor approximation. Single point energies obtained at the CCL/cc- pV5Z level. 29

30 cc-pVDZcc-pVTZ Unsigned Mean Error Bond length0.017Å0.002Å Bond angle0.7 o 0.3 o Unsigned Max Error Bond length0.025Å0.006Å Bond angle1.5 o 0.4 o Quality of the CCL geometry 1 1. Experimental data of H 2, C 2 H 2, C 2 H 3, and C 2 H 4 molecular geometries are used.

31 Quality of the CCL Energy 1-2 CCL error compared to exp. 1-2 (kJ/mol) cc-pVTZcc-pVQZcc-pV5Z C 2 H 4  C 2 H 2 + H C 2 H 4  H 2 C=C: + H C 2 H 4  C 2 H 3 + H C 2 H 4  C 2 H 2 + 2H -212 Unsigned Mean 433 Unsigned Max Mebel, A. M.; Morokuma, K.; Lin, M. C. JCP 1995, 103, Chang, N. Y.; Shen, M. Y.; Yu, C. H. JCP, 1997, 106, 3237.

32 Quality of CCL vibrational frequency 1 1. Experimental data of H 2, C 2 H 2, C 2 H 3, C 2 H 4, and H 2 C=C: vib. frequencies are used.

C 2 H 4  H 2 C=C: + H 2  C 2 H 2 + H 2 33 C 2 H 4  H 3 C−CH:  C 2 H 2 + H 2

C 2 H 4  H 2 C=CH + H  C 2 H 2 + H 2 34 C 2 H 4  C 2 H 2 + H 2 (2 imaginary frequencies) C 2 H 4  H 2 C=CH + H  C 2 H 2 + H + H

CCL and CASSCF(12,12) imaginary vibrational frequencies for C 2 H 4  C 2 H 2 + H G**6-311G**cc-pVDZ aug-cc- pVDZ cc-pVTZ CCL 1824i 887i 1623i 919i 1643i 861i 1347i 977i 1438i 947i CASSCF 2526i 385i 2331i 581i 2378i 555i 2083i 729i 2074i 663i

CCL/cc-pVTZ imaginary vibrational frequencies for C 2 H 4  C 2 H 2 + H i947i

37 C 2 H 4  H 2 C=C: + H 2  C 2 H 2 + H 2

38 C 2 H 4  H 2 C=CH + H  C 2 H 2 + H + H

Conclusions (I) For bi-radicals, CCL is much more accurate than R- CCSD(T). For tri-radicals, CCL is slightly better than U-CCSD(T) and MRMP2. CCL/cc-pVTZ predicts accurate molecule geometry (0.002Å, 0.3 o ). Both CCL/cc-pVQZ and CCL/cc-pV5Z predicts reaction energy within 4 kJ/mol error. 39

Conclusions (II) Accurate CCL/cc-pV5Z C 2 H 4 PES and Gibbs energy surfaces at CVD temperatures are obtained. There’s no direct path from ethylene to acetylene. The dominant reaction path below 1800 K is C 2 H 4  H 2 C=C: + H 2  C 2 H 2 + H 2. The dominant reaction path above 1800 K is C 2 H 4  H 2 C=CH + H  C 2 H 2 + H + H. 40

Acknowledgements T. Cameron Shore (Central Washington University) Mark S. Gordon (Iowa State University) Piotr Piecuch & Jeff R. Gour (Michigan State University) Marta Wloch (Michigan Tech. University) This work is financially partly supported by the Central Washington University, and partly by the U. S. Department of Energy, Grant No. DE-FC07-05ID14661 and Grant No. DE-FG02-01ER

Questions and Comments? 42

43

44

45

Experimentally determined reaction energies for [C 2 H 4  C 2 H 2 + H 2 ], [C 2 H 4  H 2 C=C + H 2 ], [C 2 H 4  C 2 H 3 + H], and [C 2 H 4  C 2 H 2 + H + H] are 1.72 eV, 3.63 ‒ 3.76eV (3.63 is used in the table), 4.75 eV, and 6.20 eV, respectively. Chang, N. Y.; Shen, M. Y.; Yu, C. H. Journal of Chemical Physics 1997, 106, Shimanouchi, T. Tables of Molecular Vibrational Frequencies, Consolidated Vol. I, Natl. Stand. Ref. Data Ser.; Natl. Bur. Stand. (US),

47 CCL cc-pVTZcc-pVQZcc-pV5Zcc-pVQZ* C 2 H 4  C 2 H 2 + H C 2 H 4  H 2 C=C + H C 2 H 4  C 2 H 3 + H C 2 H 4  C 2 H 2 + 2H Abs. Mean 4332 Abs. Max. 8544

48 CCL/cc-pVDZ & errorCCL/cc-pVTZ & errorExp. C2H2C2H2 R(C-C) R(C-H) C2H3C2H3 R(C-C) =CHR(C-H) =CH 2 cisR(C-H) =CH 2 transR(C-H) C=CHθ(CCH) C=CH 2 cisθ(CCH) C2H4C2H4 R(C-C) R(C-H) θ(CCH) H2H2 R(H-H) Abs. Mean Error R θ Abs. Max Error R θ

49 C2H4C2H4 Symm.Exp 22,24 CCLExp/theory 1AgAg AgAg AgAg AuAu B 1u B 1u B 2g B 2u B 2u B 3g B 3g B 3u H 2 C=CSymm.Exp 25 CCLExp/theory 1A1A A1A A1A B1B B2B B2B H2H2 Symm.Exp 19 CCLExp/theory 1ΣgΣg C2H2C2H2 Symm.Exp 22 CCLExp/theory 1ΣgΣg ΣgΣg ΣuΣu ΠgΠg ΠuΠu C2H3C2H3 Symm.Exp 23 CCLExp/theory 1A' A' A' A' A' A' A' A" A"

How much data? 8 species: H 2 A–B  H 2 A + B A = C or Si; B = H, Cl, CH 3, or SiH 3 5 basis sets: MINI or MIX, 6-31G, 6-31G(d), cc-pVDZ, cc-pVTZ. 6 methods: FCI, FSOCI(Q), MRCI(Q), CCL, UCCSD(T), MRMP2. 8*5*6=240 potential energy surfaces. 50

How to evaluate the method X?  E X (i) = E X (i) – E benchmark (i) at the i-th structure. NPE: nonparallelity error. NPE X = max[  E X (1),  E X (2) …,  E X (N)] – min[  E X (1),  E X (2) …,  E X (N)] REE: reaction energy error. REE X =  E X (3R e ) –  E X (R e ) STD: standard deviation of errors (1/3 of NPE). STD X = 51

How to get geometries and potential energy surfaces? Optimize structures with the breaking bond distance fixed at every 0.2 Å from R e to 3R e. Use FCI geometries, if possible. Or, use full-valence CASSCF geometries. CCL, UCCSD(T), and MRMP2 single-point energies are obtained to construct PES. 52

CCL/cc-pVQZ* E CCL/cc-pVQZ* = E CCL/cc-pVTZ + (E MP2/cc-pVQZ – E MP2/cc-pVTZ ) Assuming additivity of basis set effect and correlation correction. Additivity approximation is used in G1-4 theory. 53

CCL/cc-pVTZ* E CCL/cc-pVTZ* = E CCL/cc-pVDZ + (E MP2/cc-pVTZ – E MP2/cc-pVDZ ) Assuming additivity of basis set effect and correlation correction. Additivity approximation is used in G1-4 theory. Does G-n like approximation apply to bond- breaking reactions of open-shell species? 54

CCL nonparallelity errors (in mh) relative to MRCI(Q)/cc-pVTZ Smallest errors in each row are in bold font. 55

E(cc-pVTZ)-E(cc-pVDZ) MP2 overestimates basis set effect in the middle; it cancels out CCL error humps. CCL/TZ* is better than CCL/TZ* due to this fortuitous error cancellation. 56

Computational cost of CCL/cc-pVTZ* on a 2GHz machine SiCl 3 + CH 3 SiCl 3  SiCl 4 + CH 3 SiCl 2 cc-pVDZ:173 basis functions. cc-pVTZ: 344 basis functions. CCL/cc-pVTZ*: 1 day. CCL/cc-pVTZ: 36 days. 57

MRCI(Q) nonparallelity error (in mE h ) relative to FSOCI(Q) To evaluate CCL: Small basis sets: FCI as benchmark. Larger basis sets: MRCI(Q) as benchmark, lower cost than FSOCI(Q). 58

H 3 C–H  CH 3 + H Errors (in mE h ) relative to FCI/MINI 59

H 2 Si–H  1 SiH 2 + H Errors (in mE h ) relative to FCI/6-31G(d) 60

H 2 Si–Cl  1 SiH 2 + Cl Errors (in mE h ) relative to MRCI(Q)/cc-pVTZ 61

H 2 Si–Cl  1 SiH 2 + Cl Nonparallelity error (in mE h ) relative to MRCI(Q) 62

H 2 Si–Cl  1 SiH 2 + Cl Reaction energy error (in mE h ) relative to MRCI(Q) 63

64

65 H 3 C–SiH 3  CH 3 + SiH 3 Errors (in mE h ) relative to FCI/MINI NPE REE

How about tetra-radicals? 66 MINI Basis Set