Lecture 14 Beyond Hartree-Fock A better way to do it ©2013, Jordan, Schmidt & Kable Lecture 14 Copyright Notice Some images used in these lectures are taken, with permission, from “Physical Chemistry”, T. Engel and P. Reid, (Pearson, Sydney, 2006); denoted “ER” throughout the lectures and other sources as indicated, in accordance with the Australian copyright regulations. ©2013, Jordan, Schmidt & Kable Lecture 14 Learning outcomes 14.1 Can compare common basis sets and arrange in order of increasing quality 14.2 Discuss Variational theorem in context of single determinant and multideterminant wavefunctions. 14.3 Explain why Hartree-Fock wavefunction cannot describe bond cleavage 14.4 Explain how to improve the Hartree-Fock wavefunction using the N-electron basis of configurations 14.5 Recognize terms such as CI, SCF, HF 14.6 Interpret output of quantum chemistry software ©2013, Jordan, Schmidt & Kable Lecture 14 Hamiltonian operator is like a matrix Vectors are changed (rotated and stretched/shrunk) by matrices. If we represent the wavefunction of a vector is some basis function space, then the Hamiltonian operator acts like a matrix. k y = Ĥ = + Ey k j i + = E j i Where the vector is a solution to the Schrödinger equation, it will be only stretched or shrunk upon operation with the Hamiltonian. In this case it is called an eigenvector. ©2013, Jordan, Schmidt & Kable Lecture 14 The Slater determinant Generally, for a closed-shell molecule: y1 (1) ( ) Y 1,2....N = 1 N y1 ( 2) y1 (1) y N /2 (1) y1 ( 2) y N /2 ( 2) y1 ( N ) y1 ( N ) y N /2 ( N ) Exchanging two electrons is like exchanging rows, which changes the sign of the determinant. If the third electron went into y1, then two rows would be the same, and the determinant is zero. ©2013, Jordan, Schmidt & Kable Lecture 14 The Hartree-Fock Hamiltonian The Hamiltonian is the sum of kinetic and potential energies: 2 Z e Ĥ = å Ñi2 + å å -å å J i=1 2me i=1 j=i 4pe0rij i=1 J =1 4pe0 riJ N 2 N N e2 N M j j = å a jk c k k The basis coefficients ajk are optimized to minimize the energy of the oneelectron orbitals. But, since the Hamiltonian depends on the wavefunctions of the other electrons, this must be iterated until “self-consistency” is achieved. Hartree-Fock (HF) is a Self-Consistent Field (SCF) Method. ©2013, Jordan, Schmidt & Kable Lecture 14 The Hartree-Fock Approximation The electronic wavefunction is written as a single Slater determinant. y1 (1) ( ) Y 1,2....N = y1 ( 2) 1 N y1 (1) y N /2 (1) y1 ( 2) y N /2 ( 2) y1 ( N ) y1 ( N ) y N /2 ( N ) The spin-orbitals are products of spatial orbitals and spin functions: y j (i ) = j j (i ) a (i ) y j (i ) = j j (i ) b (i ) The spatial orbitals are constructed from basis functions, c. j j = å a jk c k k ©2013, Jordan, Schmidt & Kable Lecture 14 The Hartree-Fock Hamiltonian The Hamiltonian is the sum of kinetic and potential energies: 2 Z e Ĥ = å Ñi2 + å å -å å J i=1 2me i=1 j=i 4pe0rij i=1 J =1 4pe0 riJ N 2 N N e2 N M j j = å a jk c k k The basis coefficients ajk are optimized to minimize the energy of the oneelectron orbitals. But, since the Hamiltonian depends on the wavefunctions of the other electrons, this must be iterated until “self-consistency” is achieved. Hartree-Fock (HF) is a Self-Consistent Field (SCF) Method. ©2013, Jordan, Schmidt & Kable Lecture 14 The Hartree-Fock Method HF procedure obtains optimized orbitals according to the variational principle. The total wavefunction is a single Slater determinant. {ck } Ĥ ©2013, Jordan, Schmidt & Kable j j = å a jk c k k {e j } E Lecture 14 Example – N2 using minimal STO-3G basis The STO-3G basis set is the smallest that one would ever use. It has just the core and valence orbitals represented, and is pretty much only good for qualitative calculations. Core and valence orbitals are ATOMIC BASIS SET ---------------sums of three gaussians. THE CONTRACTED PRIMITIVE FUNCTIONS HAVE BEEN UNNORMALIZED THE CONTRACTED BASIS FUNCTIONS ARE NOW NORMALIZED TO UNITY SHELL TYPE PRIM EXPONENT 2p orbitals are made by multiplication with x, y or z to create node. CONTRACTION COEFFICIENTS N 3 3 3 S S S 1 2 3 99.106169 18.052312 4.885660 4 4 4 L L L 4 5 6 3.780456 0.878497 0.285714 3.454881 ( 3.341410 ( 1.041372 ( 0.154329) 0.535328) 0.444635) -0.193164 ( -0.099967) 0.258372 ( 0.399513) 0.194997 ( 0.700115) TOTAL NUMBER OF SHELLS TOTAL NUMBER OF BASIS FUNCTIONS NUMBER OF ELECTRONS CHARGE OF MOLECULE STATE MULTIPLICITY NUMBER OF OCCUPIED ORBITALS (ALPHA) NUMBER OF OCCUPIED ORBITALS (BETA ) TOTAL NUMBER OF ATOMS ©2013, Jordan, Schmidt & Kable = = = = = = = = 1.171553 ( 0.736704 ( 0.116706 ( 0.155916) 0.607684) 0.391957) 4 10 14 0 1 7 7 2 Lecture 14 Example – N2 using minimal STO-3G basis The coefficients of these 3-Gaussian basis functions are optimized. ITER EX DEM TOTAL ENERGY E CHANGE DENSITY CHANGE ORB. GRAD 1 0 0 -107.370987008 -107.370987008 0.438830221 0.000000000 ---------------START SECOND ORDER SCF--------------2 1 0 -107.494559234 -0.123572227 0.118946520 0.063938252 3 2 0 -107.500374882 -0.005815648 0.033282163 0.013749873 4 3 0 -107.500654278 -0.000279396 0.000295497 0.000288385 5 4 0 -107.500654311 -0.000000032 0.000001456 0.000014352 6 5 0 -107.500654311 0.000000000 0.000000145 0.000000320 INTEGRALS 1050 SKIPPED 117 1050 1050 1050 1041 1015 117 117 117 122 128 -----------EIGENVECTORS ------------ 1 2 3 4 5 6 7 8 9 10 N N N N N N N N N N 1 1 1 1 1 2 2 2 2 2 S S X Y Z S S X Y Z 1 -15.5063 AG 0.703182 0.012857 0.000000 0.000000 0.001709 0.703182 0.012857 0.000000 0.000000 -0.001709 2 -15.5050 B1U 0.702819 0.025712 0.000000 0.000000 0.009237 -0.702819 -0.025712 0.000000 0.000000 0.009237 ©2013, Jordan, Schmidt & Kable 3 -1.4085 AG -0.173700 0.500000 0.000000 0.000000 0.230275 -0.173700 0.500000 0.000000 0.000000 -0.230275 4 -0.7275 B1U -0.172555 0.746615 0.000000 0.000000 -0.252769 0.172555 -0.746615 0.000000 0.000000 -0.252769 5 -0.5486 B3U 0.000000 0.000000 0.629644 0.000000 0.000000 0.000000 0.000000 0.629644 0.000000 0.000000 1 2 3 4 5 6 7 8 9 10 N N N N N N N N N N 1 1 1 1 1 2 2 2 2 2 S S X Y Z S S X Y Z 6 -0.5486 B2U 0.000000 0.000000 0.000000 0.629644 0.000000 0.000000 0.000000 0.000000 0.629644 0.000000 7 -0.5303 AG 0.069560 -0.399585 0.000000 0.000000 0.604238 0.069560 -0.399585 0.000000 0.000000 -0.604238 8 0.2653 B2G 0.000000 0.000000 0.822656 0.000000 0.000000 0.000000 0.000000 -0.822656 0.000000 0.000000 9 0.2653 B3G 0.000000 0.000000 0.000000 0.822656 0.000000 0.000000 0.000000 0.000000 -0.822656 0.000000 10 1.0408 B1U -0.124828 1.094636 0.000000 0.000000 1.162969 0.124828 -1.094636 0.000000 0.000000 1.162969 Lecture 14 Example – N2 using minimal STO-3G basis …and here are your orbitals. ©2013, Jordan, Schmidt & Kable Lecture 14 Example – N2 using minimal STO-3G basis Here is the MO diagram Energy (Hartree) 2 1 0 -1 -2 -15 ©2013, Jordan, Schmidt & Kable Lecture 14 Gaussian basis sets + + = STO-3G A single basis function constructed from 3 gaussians represents each core and valence orbital. 6-31G Core orbitals of heavy atoms represented with 6 gaussians. Valence orbitals represented by 3 gaussians. One extra set of extra (smaller) valence functions made with one gaussian. 6-311G As above, but an extra set of more diffuse gaussians. 6-311+G Another set of very diffuse functions for heavy atoms. 6-311++G Add some diffuse functions for the hydrogens as well. Polarization functions may also be added to atoms – e.g. d orbitals on carbon and p orbitals on hydrogen. 6-31G(d) is popular, for instance. ©2013, Jordan, Schmidt & Kable Lecture 14 Variational Theorem: Energies of H atom Common basis sets show that the larger the basis, the lower the energy of the H atom. The correct energy is approached with the largest basis set. Energy (hartree) -0.46 -0.48 Correct energy is -0.5 Eh -0.50 STO-3G 6-31G 6-311G Basis Set ©2013, Jordan, Schmidt & Kable 6-311++G Error is only 40cm-1 at 6-311++G Lecture 14 Hartree-Fock Potential energy curve of H2 We use the 6-311++G(d,p) basis set. -0.90 -0.92 -0.94 -0.96 2 H atoms = -1.0 Eh Energy (hartree) -0.98 -1.00 -1.02 -1.04 -1.06 -1.08 -1.10 -1.12 -1.14 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 H-H distance (Å) Somehow, Hartree-Fock does not describe dissociation properly…. Why? ©2013, Jordan, Schmidt & Kable Lecture 14 Hartree-Fock wavefunction for H2 HF wavefunction is antisymmetrized product of 2 1s sigma spin-orbitals. ( ) Y HF 1,2 = y1s (1) y1s ( 2) - y1s (1) y1s (1)) ( 2 1 () () = j1s 1 j1s 2 Symmetric space function j1s = ©2013, Jordan, Schmidt & Kable () ( ) () ( ) éa 1 b 2 - b 1 a 2 ù ë û 2 Antisymmetric spin function j1sa + j1sb ) ( 2 1 Lecture 14 Hartree-Fock wavefunction for H2 HF spatial wavefunction is simple product of 2 1s sigma orbitals. ( 2 j1s (1) j1s ( 2) = = 1 ( j1sa (1) + j1sb (1) () () ) 2 (j1sa (2) + j1sb (2)) 1 () () () () ( )) () 1 j1sa 1 j1sa 2 + j1sa 1 j1sb 2 + j1sb 1 j1sa 2 + j1sb 1 j1sb 2 2 As the molecule dissociates, these terms describe both electrons being on the one atom, which is H- plus a proton! This adds a coulombic attraction term which prevents proper dissociation. What we have not accounted for is electron CORRELATION ©2013, Jordan, Schmidt & Kable Lecture 14 Hartree-Fock wavefunction for H2 Look at the wavefunction as a 2d plot along bond axis Ionic terms 1.0 (1,2) 0.8 0.6 0.4 0.2 4 2 0.0 -2 0 z 1 ©2013, Jordan, Schmidt & Kable 2 z -2 2 0 -4 -4 4 Lecture 14 The set of configurations are an N-electron basis If a single determinant wavefunction is not enough, then add more determinants! The various configurations form an orthonormal basis in the N-electron space. k y = Ĥ = + Ey k j i + = E j i Here i, j, and k are determinants which represent different configurations. Calculations which allow the interaction of configurations is called….. Configuration Interaction (CI). ©2013, Jordan, Schmidt & Kable Lecture 14 Configuration Interaction for H2 Adding in (with a negative coefficent) some of the doubly excited configuration can remove ionic terms. Ionic terms 1.0 1.0 0.8 0.6 0.8 0.2 0.6 (1,2) (1,2) 0.4 0.4 0.0 -0.2 -0.4 -0.6 0.2 -0.8 1 2 ©2013, Jordan, Schmidt & Kable z -4 1 4 j1s (1) j1s ( 2) -2 0 2 -2 0 z -2 z -2 0 -4 0 -4 1s* 1s 2 z 0.0 4 2 -1.0 2 4 2 -4 4 j1s * (1) j1s * ( 2) 1s* 1s Lecture 14 Configuration Interaction for H2 We do a complete active space self-consistent field (CAS-SCF) theory calculation, to allow both configurations while simultaneously optimizing the orbitals, with the same basis as before. At 0.9Å ITER 1 2 3 4 5 TOTAL ENERGY -1.125049420 -1.141745458 -1.141768656 -1.141768724 -1.141768726 DEL(E) LAG.ASYMM. SQCDF MICIT -1.125049420 0.160266 1.377E-02 1 -0.016696038 0.004000 1.652E-04 1 -0.000023198 0.000164 7.577E-07 1 -0.000000069 0.000020 1.318E-08 1 -0.000000001 0.000003 2.313E-10 1 DAMP 0.0000 0.0000 0.0000 0.0000 0.0000 -------------------LAGRANGIAN CONVERGED -------------------FINAL MCSCF ENERGY IS -1.1417687256 AFTER 5 ITERATIONS j1s (1) j1s ( 2) 1s* 1s j1s * (1) j1s * ( 2) -MCCI- BASED ON OPTIMIZED ORBITALS ---------------------------------- 1s* 1s CI EIGENVECTORS WILL BE LABELED IN GROUP=D2H PRINTING ALL NON-ZERO CI COEFFICIENTS STATE 1 ENERGY= -1.1417687256 S= 0.00 SZ= 0.00 SPACE SYM=AG ALPH|BETA| COEFFICIENT ----|----|-----------10 | 10 | 0.9900962 01 | 01 | -0.1403909 HF energy was -1.11832. CASSCF energy is -1.14177 Eh ©2013, Jordan, Schmidt & Kable Variational Principle!! Lecture 14 Configuration Interaction for H2 -0.90 -0.92 -0.94 -0.96 Energy (hartree) -0.98 -1.00 -1.02 -1.04 ALPH|BETA| COEFFICIENT ----|----|-----------10 | 10 | 0.8803840 01 | 01 | -0.4742616 -1.06 -1.08 -1.10 -1.12 ALPH|BETA| COEFFICIENT ----|----|-----------10 | 10 | 0.9947259 01 | 01 | -0.1025694 -1.14 -1.16 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 H-H distance (Å) CI wavefunction decribes dissociation (chemistry!) because it accounts for CORRELATION. Wavefunction at equilibrium is well described by a single configuration (determinant). ©2013, Jordan, Schmidt & Kable Lecture 14 Level of correlation (# of configurations) The Quantum Chemistry Landscape The answer pointless Sensible compromise Hartree-Fock limit 1st year chemistry Quality of basis set ©2013, Jordan, Schmidt & Kable Lecture 14