R E S E A R C H

Program Obtains Complex Pi-Electron Data I IT scheme combines quantum chemistry and large computer to get electron distribution in unsymmetrical conjugated systems

READY. NT's Dr. Howard N. Schmeising examines punch tape that's ready to use to compute geometry and excited state reactivity criteria of molecules. The tape is fed to a UNIVAC 1105; data are punched on a Frieden Flexowriter

142ND ACS NATIONAL M E E T I N G

Physical Chemistry

Substitution of methyl groups on ben­zene changes electron distribution on the conjugated ring system by induc­tion, making classical or Hiïckel mo­lecular orbital calculations hazardous. But the resulting electronic distribu­tion can be calculated for any member of the 13-member series by computer-aided quantum chemistry, say Dr. R. L. Flurry, Jr., and Dr. P. G. Lykos of Illinois Institute of Technology, Chicago. In addition to calculating energies and basicities for the ground state, the IIT group calculates pre­dicted values for excited states.

The IIT scientists developed a com­prehensive, flexible program for a UNIVAC 1105 computer, which is able to handle the gross volume of cal­

culations needed to obtain pi-electron data useful to chemists, Dr. Flurry told the ACS National Meeting, while Dr. Lykos described the IIT work be­fore the International Symposium on Molecular Structure and Spectroscopy, held in Japan. The international meet­ing and the ACS sessions were held the same week.

Only a modern, high-speed com­puter such as the UNIVAC 1105 or IBM 7090 can handle quantum chemi­cal calculations at this level of sophisti­cation, Dr. Flurry notes. Other calcu­lations and methods that yield similar results are restricted to simplified models and small molecules having a high degree of symmetry.

The program worked out at IIT has handled a coupled pi-electron system with 29 electrons—the tetraphenyl-pyryl radical. And only computer-aided quantum chemistry can include electron repulsion interaction in mole­cules that have intricate geometry, and

can optimize orbital description for the actual states of interest—ground, triplet, or doublet states. These two abilities of the method (which was developed and is being continually refined at IIT) distinguish it from "classical" quantum chemistry, Dr. Flurry says. Programs are also being developed by other workers in the U.S., Europe, and the Far East; among workers involved are G. G. Hall and A. T. Amos, now at the University of Nottingham, Nottingham, England, and Dr. L. C. Snyder of Bell Tele­phone Laboratories.

Objective. The objective is to take a chemically oriented and highly graphical input—that is, chemical com­position and basic geometrical struc­ture—and calculate electron densities at each atom in the conjugated struc­ture, and other chemically desirable parameters. Classical quantum chem­istry theory (Hùckel molecular orbital method) gives similar results on sim­ple, highly symmetrical molecules; but these results differ significantly from the IIT results on molecules with low symmetry, Dr. Flurry says. The pro­gram for the UNIVAC 1105 ignores symmetry, thus retains flexibility for calculations on substituted, unsym­metrical molecules. The burden of making the transition from the lan­guage of the chemist to that of the computer is placed on the computer.

The program handles neutral mole­cules or ions of substituted or unsub-stituted pi-electron systems in their ground or excited states. Complex molecules are studied in much the same way as benzene—a simple, sym­metrical model of pi-electron mole­cules—has been extensively studied. In conjugated molecules, pi electrons are relatively free to move over the entire molecule, and detailed struc­ture of the core doesn't have much chemical importance. Important, though, and useful to chemists, are electron distributions in ground and excited states. These and other results play a critical role in understanding

42 C & E N S E P T . 24, 1962

reactivity, nuclear magnetic resonance and electron spin resonance spectra, and basicity.

The first program was prepared for the UNIVAC 1105, and uses the Hartree-Fock method as worked out for an open-shell matrix by Dr. C. C. J. Roothaan. Input consists of a mini­mum number of bond lengths and angles. The program computes all unique intermolecular distances that are needed for calculating integrals. The program may also produce maxi­mum overlap molecular orbitals or Hiickel orbitals as starting points for molecular orbitals, and has some inte­grals over atomic orbitals already stored in its memory. Both of these sets of information can be supplied with initial input.

Speed and size of the computer play a large part in solving the eigen­value problem. After obtaining a solution in the program, a chemist can get output in the form of charge den­sities, spin densities, orbital energies, or frontier electron densities. IIT's Roy Blomquist is developing an auxil­iary program that prints pertinent data

on geometric diagrams of molecules. How It's Done. After input defines

the problem, the program builds up basic information in the form of start­ing wave functions, complete descrip­tion of molecular geometry, and or­ganization of subsequent calculations. The next step includes consideration of kinetic energy of the pi electrons, interactions of pi electrons with the core (molecular framework of sigma electrons and nuclei), and the cou-lombic mutual repulsion of electrons.

Product wave functions for sigma electrons alone and pi electrons alone are combined with a partial antisym-metrizer to form a total wave function. An energy operator is written to assess the kinetic energy, nuclear attraction energy, mutual repulsion of the pi electrons, and the repulsions between each of the sigma electrons and each of the pi electrons.

Calculations give the pi-electron contribution to the total energy, in­cluding sigma-pi coulomb and ex­change interaction. The total energy of the pi electrons depends upon fac­tors that include integrals over atomic

orbitals. Before this, theoretical cal­culations have used atomic orbitals suitable for free atoms. To circum­vent the problem of distortion as atoms approach each other closely, integrals over atomic orbitals are evaluated empirically. Estimates of parameters are continually improved as calcula­tions are compared with chemical and physical data on a variety of con­jugated systems that have been well characterized. Calibration with ob­served behavior of molecules in such areas as NMR or ESR permits transi­tion from simple structures to more complex ones made up of similar build­ing blocks.

The program finds a solution to the eigenvalue problem by repeated ap­proximation until it satisfies a con­vergence criterion. For a large mole­cule, there may be a dozen iterations that take a total of 88 minutes on the UNIVAC 1105. An IBM 1620 would take 20 to 25 times as long, while an IBM 7090 would take one tenth as long, Dr. Flurry says.

The final wave function solution ob­tained may be used to compute charge and spin densities at the various unique positions in the molecule. Programing is flexible because of a matrix interpretive program that was written by Dr. H. N. Schmeising of IIT explicitly for this research pro­gram. Revisions dictated by experi­ence and advances in theory can be inserted. The calibrated semiempiri-cal theory can be used not only to assess ground state properties of com­plicated systems, but also as a starting point for studying excited states of pi-electron systems.

Useful Results. Theoretical esti­mates of the various reactivity criteria that chemists use are direct products of the program. These numbers al­low insight as to points of nucleophilic or electrophilic attack. They also in­dicate positions of attack by radicals. The program gives bond lengths and bond angles according to a minimum energy criterion. And the informa­tion helps in analysis of magnetic resonance spectra, especially ESR.

In organic chemistry, Dr. Flurry explains, he calculates localization energies for each distinct benzene compound having one to six methyl groups, then he predicts basicities in the ground and first excited triplet states. For the ground state, his pre­dictions agree well with the basicity constants in hydrofluoric acid as worked out by others.

DIAGRAM. Dr. Peter G. Lykos (left) and Roy Blomquist check a diagram (of a protonated imidazole), which was produced by the computer readout printer. Mr. Blomquist worked out the method for instructing the computer to print the diagram

S E P T . 2 4, 1962 C & E N 43

Mechanism Explains Surfactant Effects Agents help glycerol tristearate crystals stabilize emulsions by decreasing crystal interaction

Spin densities of methylbenzene negative ions also give good agreement with ESR spectra. Such agreement gives confidence in predictions for ex­cited state basicities calculated with the same parameters. The computer-aided quantum chemistry overcomes the obstacles of extensive calculations imposed by low symmetry and explicit inclusion of electron repulsion.

Other Fields. R. L. Miller, also at IIT, is using computer-aided quantum chemistry on molecules of biological interest—the purines and substituted purines. He is looking for data on reactivity and on geometry at lowest energy states. Harold Weber of IIT's Armour Research Foundation is tackling the problem of understanding the nature of the parameters that en­ter into the theory. And Dr. Schmeis-ing, assistant director of the IIT com­putation center, is studying molecular geometry, especially in its relationship to triplet states.

One step in continual improvement of computer programs and theoretical frameworks is a study of atomic distor­tion. Electron orbitals about atoms appear to be distorted by close ap­proach to other atoms—at least one reason for empiricism in parts of the calculations. To understand distortion better, the IIT group is studying the extent to which atoms are distorted when molecules form. A basic tool here is the change in dipole moment and quadrupole moment as distortion occurs.

This form of computer-aided quan­tum chemistry, while initially expen­sive, has shown its possible use, Dr. Flurry says. And it is starting to move out of the academic world into indus­trial research. Dr. O. W. Adams of Abbott Laboratories is developing a quantum chemistry program similar to IFTs.

Use of the theory and programs of computer-aided quantum chemistry might accelerate after a planned test of its predictive value. The program's nature ties it closely to experiment. It represents an attempt to find a suitable middle ground between the limita­tions that molecular size imposes on an exact solution of the Shroedinger equa­tion, and the empirical regularities found for many homologous series of complex molecules. Dr. Lykos, Dr. Flurry, and associates are confident enough to plan the prediction of trip­let states of certain molecules, and then the unmeasured ESR behavior of these states.

142ND ACS NATIONAL M E E T I N G

Colloid a n d Surface Chemis t ry

A mechanism to explain how the ad­dition of surfactants affects water-in-oil emulsions stabilized by solid par­ticles has been proposed by Unilever research workers. Key to the mech­anism is decreased interaction energy of the solid particles.

Emulsions of water in paraffin oil can be stabilized by glycerol tristearate crystals if small amounts of surfactants are added, Dr. M. van den Tempel told the Symposium on Liquid Disper­sion. Surfactant doesn't affect the con­tact angle of the water and oil against the crystals. It stabilizes the emulsion by decreasing interaction of the crys­tals. This permits the crystals to reach the surface of the water droplets be­fore any great amount of coalescence takes place.

Solid particles can be used to sta­bilize emulsions if two conditions are met:

• Contact angle between the water and oil and the particle surface is such that the particle is adsorbed at the liquid-liquid interface.

• Particles are in a state of incipient flocculation. Basis for these conditions is the theory that a dense layer of solid particles at the liquid-liquid in­terface is needed to stabilize emul­sions.

Research workers have studied how contact angles affect emulsion stability in various systems—for example, water and benzene, stabilized by small crys­tals of barium sulfate. The contact angle in this system, they find, can be varied by adding surfactant. And the angle affects the stability of the emul­sion formed.

But other systems act differently. Unilever chemists find, for instance, that the contact angle in the system of water, paraffin oil, and triglyceride crystals isn't affected by surfactants. Still, emulsion stability is markedly increased.

Lower Viscosities. To determine the effect of surfactants in this sys­

tem, Dr. van den Tempel and co­worker Dr. E. H. Lucassen-Reynders carried out a substantial investigation of water-paraffin oil emulsions. First, they measured emulsion stability. They find that it is very high if the system contains both 1% of glycerol tristearate crystals and about 0 .1% of surfactant (either oil- or water-solu­ble). In contrast, stability was 100 to 1000 times lower in emulsions con­taining only the tristearate or only the surfactant.

The Unilever chemists also deter­mined contact angle, Here, studies show that the angle isn't affected by adding Aerosol OT, cetyl alcohol, gly­cerol monooleate, or sodium dodecyl sulfate, in concentrations up to the critical micelle concentrations of the surfactants. Measured in the water phase, the contact angle is 110°.

Steady state viscosity measurements (at very low shear rates) were used to study the flocculation of the crystals in oil. Experiments were carried out on suspensions containing 1% of tri­stearate crystals in paraffin oil; sur­factants investigated were glycerol monooleate, Aerosol OT, and cetyl al­cohol in concentrations of 2.5, 10, and 10 micromoles, respectively, per cc. of oil. Results show that at shear rates less than about 0.1 per second, viscosi­ties of the surfactant-containing dis­persions are markedly lower than those of the dispersions without sur­factant.

Chain-Like Network. Triglyceride crystals dispersed in oil are always highly flocculated, Dr. van den Tempel points out. This stems from the fact that attraction forces exist at all dis­tances between the particles. In quiescent, dilute dispersions, there isn't any sedimentation. Reason is that the triglyceride crystals flocculate to form a three-dimensional network which is strong enough to withstand the pull of gravity. Unilever scien­tists have measured the strength of this network in more concentrated sus­pensions, and estimate that the average energy of the bond between consecu­tive particles is about 40 kT units. So, apparently, adding surfactant to this system reduces the interaction

44 C & E N S E P T . 24, 1962