Information Textbooks Media Resources

Computational Chemistry in the Undergraduate Chemistry

Curriculum: Development of a Comprehensive Course
Formula
Zbigniew L. Gasyna and Stuart A. Rice*
Department of Chemistry, The University of Chicago, 5735 South Ellis Avenue, Chicago, IL 60637;
*sarice@rainbow.uchicago.edu

The intellectual basis for including computational

chemistry in the undergraduate chemistry curriculum can be
summarized as follows. Progress in the past decade in the
capability for carrying out realistic computations of chemical
phenomena and processes has been breathtaking, and that
rapid progress is continuing. It is a truism that the availability
of modern computer hardware and software has transformed
the manner in which many chemical problems are analyzed,
making possible the testing of ideas and the design of processes
that previously could only be dreamed of. Drawing on these
developments, there are already many parts of the undergraduate chemistry curriculum into which computation has
been integrated, typically as a tool that eases the burden of
didactic calculation, or for molecular modeling, or to illustrate
the solutions of, for example, the Schrdinger equation for
model systems. Yet, despite being a common component of
the undergraduate chemistry curriculum and notwithstanding
its various implementations, there is still considerable disagreement as to what is to be achieved by the study of computational chemistry.
An excellent survey of the relevant pedagogical issues,
and a review of the approaches to teaching computational
chemistry in place at several institutions, has been published
by DeKock et al. (1). Moreover, particular undergraduate
programs in computational chemistry have frequently been
described in this Journal (see, for example, articles dealing with
electronic structure calculations, refs 24). For that reason
we will focus attention in this report on the character of the
recently introduced juniorsenior level Computational
Chemistry course at The University of Chicago, using specific
examples to illustrate the approach taken.
We believe a study of computational chemistry should
generate an understanding of the interplay between basic
theory and computational methodology, that it should illuminate the circumstances under which computation is the
preferred tool for solving problems and illustrate the accuracy
with which those problems can be solved. Above all, we
believe that a study of computational chemistry should enhance
critical thinking about scientific goals which computation can
assist in achieving. A course of this type requires both lecture
presentation and extensive practice in the computational
laboratory. Moreover, the diversity of subject matter to be
illustrated and the complexity of the underlying theories
imply that the students progress needs to be closely monitored
and that adequate methods for evaluation of the students
work must be devised.
Presented at the 30th Great Lakes Regional Meeting of the
American Chemical Society, Loyola University of Chicago, Chicago,
IL, May 2830, 1997.

There are, of course, many ways of achieving these goals.

Our implementation focuses on problem solving through
applications and does not involve any level of code writing
by the student. The lecture material is addressed to review of
the relevant theoretical basis for various chemical properties,
processes, and phenomena and to the teaching of methods of
calculation that can be used to carry out numerical evaluation
of quantities of physical and chemical interest. The course
material has been organized to provide the maximum possible
correlation between the laboratory exercises and the lecture
material. Since the lecture material in computational chemistry encompasses theoretical topics from several fields of
chemistry, completion of a course in physical chemistry is a
prerequisite.
Course Structure

The Computational Laboratory

The computational laboratory hardware consists of a
cluster of 12 Silicon Graphics O2 workstations running IRIX
6.3 (Silicon Graphics, Inc., version of UNIX), an HP LaserJet
5M black-and-white printer, and an HP DeskJet 1600CM
color printer. The computers are sufficiently powerful (180 MHz
R5000SC, 64 Mb memory, 4.3 Gb disk, 17 monitors,
system CAM) to carry out extensive computations efficiently,
and universal enough to have application software available
from various vendors. All of the machines are connected via
a 100BaseT, 100Mbps Ethernet local area network and share a
server that stores a common database of names and passwords
as well as some of the software.
The design and operation of a laboratory to which students
have unlimited access plays an important role in their learning.
Instruction relevant to particular applications programs is
given in specified laboratory sessions. Other than those
scheduled periods, the laboratory is open to students 12 hours
each weekday and at other times by special arrangement. The
laboratory is supervised by a full-time professional chemist
who is qualified to deal with both the computer technology
and the tutoring of individual students in Computational
Chemistry. In addition, teaching assistants recruited from a
pool of experienced and knowledgeable graduate students are
directly involved in the instructional laboratory periods.

Manuals
The students are provided with a list of textbooks on the
theory and practice of numerical analysis and various subjects
in theoretical and computational chemistry (see Appendix);
only the numerical analysis textbook is required. Notes covering all of the lecture material are distributed to the students
following the pertinent lecture.

JChemEd.chem.wisc.edu Vol. 76 No. 7 July 1999 Journal of Chemical Education

1023

Information Textbooks Media Resources

To assist a student in carrying out the laboratory exercises

we have prepared a manual and written instructions pertinent
to each of the course sections. In addition, the laboratory
manual is intended to serve three purposes. First, it gives
summaries of the most important elements of the theory
related to the computational exercises. Second, it reviews
some fundamental aspects of the IRIX operating system and
introduces the reader to the particular program under the
IRIX environment. Third, the manual contains a series of
exercises designed to demonstrate how to utilize the program
and to illustrate the importance of the proper approach to
solving a particular problem.

The Lecture Material

The scope and applicability of computational methods
in chemistry are too large to be covered in a one-quarter
course, so some compromise between depth and breadth of
coverage must be reached. As currently constructed, the
course in computational chemistry is organized around
sample scientific problems (see below). The computational
approach to the solution of all such problems requires precise
formulation of the mathematical description of the system,
the introduction of concepts which permit the selection of
appropriate approximations, the transformation of the mathematical representation to a form that is most convenient for
numerical solution of the equations, and execution of the
computations. The course does not cover the writing of
computer code. Instead, emphasis has been placed on the
use of available applications programs, some examples of
which are Mathematica, the suite of packages associated with
MSI/Biosym, and GAMESS. These general applications
programs are regarded as tools for solving scientific problems,
and attention is focused on the assumptions used in developing
the code (e.g., cutoffs in potential functions) and the expected
accuracy and systematic limitations of the algorithms used.
The lecture and laboratory contents of our Computational Chemistry course include four major topics that are, in
our opinion, essential ingredients of the subject. The topics we
have chosen to emphasize are Numerical Methods of Analysis,
Molecular Mechanics for Conformational Analysis, Molecular
Dynamics and Monte Carlo Simulation, and Electronic
Structure Calculations. In each of these major areas of computational chemistry, specific application software is used. The
application software chosen reflects the expertise available in
the Department of Chemistry. In use, each application package
is supplemented with utility programs running under the
IRIX environment.

In the computational laboratory, instruction in numerical

analysis is based on Mathematica (5). This program has found
application in a variety of topics in the undergraduate curriculum, many of them reported in this Journal (69).
Mathematica allows for handling of formulas, numbers, text,
and graphics. In its window-based environment, it provides
an interactive interface in which the equations can be solved
analytically or numerically, and various functions can be
introduced from a pull-down menu. Descriptive text may be
placed inside the document and two- and three-dimensional
plots can be embedded sequentially. The program has an
online reference and help system to assist students. The
exercises covering numerical methods can be based on other
similar programs, of which Mathcad (10) is the most closely
related.
The students are given homework assignments that focus
attention on particular numerical methods. We start with
evaluating expressions and simple symbolic computation, as,
for example, symbolic differentiation of functions, and then
introduce various methods of data manipulation, such as
interpolation, and fitting using the linear and polynomial least
squares methods. We show how to use the program to plot
functions and data. More advanced topics are described below.
Iteration
Iterations using Mathematica can be performed in many
ways. As an example, we employ an iteration method to
produce the Fibonacci numbers. Other examples of iteration
procedures are then introduced along with more complex
applications of the program.
Symbolic, Numerical, and Monte Carlo Integration
Monte Carlo integration is a method based upon the
generation of random numbers and statistical sampling. In
order to evaluate an integral
b

f x dx

I=

(1)

of some particular function f (x) in the interval [a,b] it is

approximated by
b

b a
N

f x dx

i=1

xi

It is pointed out that the error of the integration is of the

order of 1/N 1/2, reflecting the fact that the variance in f (x),
which is a measure of the extent to which f (x) deviates from
its average value over the region of integration, is

Course Material

Numerical Methods of Analysis

The lecture material dealing with numerical analysis is
intended to sample several topics that are ubiquitous components of applications software. The goal of the presentation
is to make the student aware of the mathematical basis for
an algorithm and of the trade-offs between accuracy and
convenience of use of an algorithm, and to provide a basis for
understanding how algorithms for solving particular problems
can be selected. The topics covered are computer arithmetic,
interpolation and extrapolation, numerical differentiation,
numerical integration, and the solution of ordinary differential
equations.
1024

(2)

I2 1 f2 = 1 1
N
N N

2
fi

i=1

1
N

fi

i=1

(3)

As an example, we show how to implement the Monte

Carlo procedure in Mathematica to evaluate the integral and
standard deviation for various large values of N. The example
is integration of
1

I =4
0

dx
1 + x2

(4)

This integral is equal to and is also calculated, for com-

Journal of Chemical Education Vol. 76 No. 7 July 1999 JChemEd.chem.wisc.edu

Information Textbooks Media Resources

parison, using the internal Mathematica integrating function.

Students are asked to tabulate the computed values of and
their standard deviations for different values of N. They should
then be able to notice the improvement in the accuracy of the
calculated value with increasing N and to examine the tradeoff of accuracy for computational time. Other mathematical
functions are then integrated using the Monte Carlo method,
and techniques for reducing the variance and improving the
efficiency of the calculations are described and discussed.
Complex Numbers and Fourier Transforms
Mathematica allows the execution of operations on
complex numbers. Fourier transformation of data, which
inherently involves operations on complex numbers, is one
of the meaningful applications, because Fourier transform
methods have become important in chemical applications
since the advent of FTIR and FT-NMR. Mathematica supports
calculation of both the forward and the inverse Fourier
transform of discrete data. The purpose of this exercise is to
calculate the Fourier transform of a spectrum and to carry
out the inverse transform to reproduce the spectrum. The
original data can be modified in many ways, including the
addition of random noise. The noisy spectral data are then
transformed by applying a filter function that passes only the
low-frequency information, allowing the student to remove
the noise. The result of the inversion calculation is then
compared with the starting spectrum.
Matrices: Hckel Molecular Orbital Calculations
Mathematica provides an excellent vehicle with which to
illustrate the application of Hckel theory in the calculation
of molecular orbital energies and the coefficients in the wave
function representation, because the program handles matrices
and operations on matrices with great ease. A good model
compound for such calculations is benzene. Hckel theory
calculations of the -electronic structure of the benzene
molecule deal with a general case of degenerate energy levels,
which requires special attention. When there is degeneracy,
the standard Mathematica procedure for solving the secular
matrix only guarantees that the eigenvectors are linearly
independent. To use the Hckel eigenvectors as the wave
function coefficients, they all must be mutually orthonormal.
The orthonormalization of the eigenvector matrix for benzene
is accomplished by application of a suitable procedure from
the Mathematica online linear algebra package.
Solving Equations
Solving equations with Mathematica can be carried out in
various ways. There are simple internal procedures which solve
polynomial equations. The limitations in the application of these
procedures to certain equations is demonstrated with suitable
examples. For solving simultaneous linear equations, the
matrix method is implemented.
Differential Equations in Chemical Kinetics
There are many procedures that can be used to solve
differential equations numerically. We have selected the
fourth-order RungeKutta method for demonstration purposes.
The examples demonstrate the usefulness of this technique
in studies of chemical kinetics. A rate equation for a simple
first-order reaction A B, for which dx/dt = {kx, is solved
and compared with the exact solution. The method is then
applied to coupled differential equations that describe various

chemical processes; for example, calculation of the rate of

product formation for the consecutive chemical reactions
A B, B C is investigated. The time-dependent consumption of the substrate A and the evolution of the intermediate
B and the product C are computed. Several more complex
chemical processes are also studied using this technique,
among them oscillatory chemical reactions.
Three-Dimensional PlotsVisualization of Atomic
Orbitals
The graphical display of atomic and molecular orbitals
is a focus of attention throughout the chemistry curriculum.
With Mathematica the wave functions of simple atomic orbitals (s, p, d) can be computed and displayed. An example
of such a wave function representation is a mathematical description of the 2p orbital
2pz ze { kr

(x 2

y2

(5)

z 2)1/2,

with r =
+ +
and k a constant. The wave function
isosurface (i.e., the surface on which | 2pz| has a constant
value) can be plotted employing graphics functions from the
Mathematica add-on package. Introduction of different
shading or color schemes for the positive and negative regions
of the wave function provides an additional challenge in
generating this graphical representation.

Molecular MechanicsMultidimensional Minimization

The lecture material dealing with molecular mechanics
examines the various contributions to the potential energy of
an isolated molecule and the definition of suitable potential
energy functions. The use of interpolation procedures and tables
in the calculation of components of the potential energy
function, the treatment of solvent effects on molecular
conformation, and multidimensional minimization are then
discussed. Primitive versions of Monte Carlo and molecular
dynamics augmentations of conformation searches are addressed.
Attention is called to the differences between the energies and
conformations of molecules predicted by the several variants
of molecular mechanics applications software.
In the computational laboratory, molecular mechanics
and molecular dynamics exercises are carried out using the
Insight II (Molecular Simulations, Inc.) suite of programs.
Some examples of analyses carried out by the students are
cited below.
Cyclohexane and 1,4-Cyclohexadiene Conformational
Analyses
Using the Builder or Biopolymer module of Insight II,
students assemble cyclohexane and 1,4-cyclohexadiene
molecules and analyze their conformations. Using the Discover
module, the students optimize the structure of the molecule
under different force fields (Amber, cvff, cff91) and explore
various minimization algorithms, such as steepest descent and
conjugate gradient. The effects of including cross terms and
of use of the Morse potential are investigated.
Cyclooctane Conformational Analysis
See the discussion under this subheading in the following section.

Monte Carlo and Molecular Dynamics Simulations

Monte Carlo methods are used for many purposes. We
have focused attention on the calculation of the equilibrium

JChemEd.chem.wisc.edu Vol. 76 No. 7 July 1999 Journal of Chemical Education

1025

Information Textbooks Media Resources

properties of a many-body system of interacting particles.

One example of a Monte Carlo procedure, namely Monte
Carlo integration, is discussed in the numerical analysis part
of the lectures; it is developed with Mathematica. In this part
of the course we show that similar mathematical problems
can be solved using simple Fortran programs compiled under
IRIX. The general lecture material dealing with Monte Carlo
simulations discusses all the major procedures associated with
calculating the thermodynamic properties of many-molecule
assemblies and includes the discussion of various sampling
methods and of simulations in several different ensembles.
The analysis of molecular dynamics methods includes the
discussion of algorithms for the integration of the equations
of motion, simulations in various ensembles, and formulations
of the equations of motion that permit the definition of
thermodynamic boundary conditions.
The computational laboratory exercises carried out by
the students are listed below.
Monte Carlo Simulations
Typical exercises deal with the simulation of a simple
molecular liquid in the constant-NVT ensemble; the example
calculations are based on Fortran programs. The differences
between the structures (i.e., the pair correlation functions)
of fluids that interact with hard-sphere and LennardJones
potentials are examined.
Molecular Dynamics Simulations
A molecular dynamics simulation of the structure of a
simple liquid (e.g., H2O) is carried out using the Discover
program of Insight II.
Cyclooctane Conformational Analysis
Starting from one of the conformations of the
cyclooctane molecule the student runs a conformational
search using the unrestrained Molecular Dynamics procedure
of Insight II. This procedure generates many cyclooctane
conformers. One of the parameters that affect the molecular
dynamics simulations is temperature. We suggest using
temperatures of 300, 600 and 900 K in these simulations.
The analysis of the Molecular Dynamics runs requires construction of graphs of the trajectoriesthat is, system energy vs
frame (or time) and cluster graphs. The cluster graphs are
indispensable for the identification of new conformers formed
in the simulation process. The completed assignment requires
screen printouts containing graphs and molecular structures,
identification of different conformers obtained in the simulations, and explanation of the effect of temperature based
on the analysis of the runs output.
Folding a DNA Double Helix
Because some students enrolled in the course are concentrating in one or another Biological Chemistry/Molecular
Biology program, we have introduced exercises that pertain
to biological systems. A typical example is the exercise in
which a student attempts to restore the double helix structure
of a partially denatured oligonucleotide double strand using
the Molecular Dynamics procedure of Insight II. Denaturing
of DNA structures is associated with the breakage of hydrogen
bonds connecting the complementary nucleotide bases and
partial to complete unfolding of the strands. The hydrogen
bonds of interest and their typical lengths are provided for
the complementary bases CG and TA. This Molecular
1026

Dynamics analysis requires that certain constraints be imposed

on the structure during the computations in order to subject
the system to folding. The superposition of the calculated
structure on the double helix of the reference native structure
gives the means to evaluate the outcome of calculations.

Electronic Structure Calculations

The lecture material deals with the electronic structure
of atoms and molecules from the point of view of approximations to the solution of the Schrdinger equation. The
calculation of the electronic structures of atoms is used as the
vehicle for introducing the independent electron model, the
self-consistent field scheme, the HartreeFock method, and
the importance of correlation energy. Quantum mechanical
perturbation theory and the variational principle are discussed.
The BornOppenheimer approximation is introduced, along
with a qualitative discussion of the circumstances under which
it breaks down. The role of symmetry in restricting the class
of solutions to the Schrdinger equation is illustrated for a
simple molecule. The student is also introduced to the
principal ideas involved in several semiempirical theories of
molecular electronic structure.
The computational laboratory exercises are based on use
of the GAMESS (General Atomic and Molecular Electronic
Structure System) program (11) (http://www.msg.ameslab.gov/
GAMESS/GAMESS.html ). GAMESS is a very flexible application program that can be used to calculate the equilibrium
geometries, vibrational frequencies, electric dipole moments,
and properties of the excited electronic states of molecules.
It can also be used to calculate reaction coordinates and
transition states and classical trajectories on a molecular
potential energy surface. GAMESS has built-in a wide range of
methods for the calculation of the electronic structure of a
molecule. Other similar ab initio programs can be adopted in
these calculations, one obvious choice being Gaussian (12).
The GAMESS input and output formats are not particularly user-friendly, and the program has only a text interface. A companion program Molden (http://www.caos.kun.nl/
~schaft/molden/molden.html) is available under UNIX for
viewing the results of the calculations. This freeware program
reads the GAMESS output files and displays data in twoand three-dimensional formats. Molden is particularly easy
to use and it has proved to be an indispensable graphical
interface to GAMESS.
One of the difficulties in formulating the input data for
the ab initio calculation of the electronic structure of a molecule
is associated with the necessity of providing the initial coordinates for the molecule, most commonly in the form of a
Z-matrix. In some cases, the students can take advantage of
their previous knowledge of Insight II to generate the Zmatrix in the MOPAC format, which is one of the input data
formats recognized by GAMESS.
Several of the exercises that have been implemented in
the course are described below.
The Structure of Water
We begin the hands-on exercises for GAMESS with a study
of the structure of the water molecule. The student calculates
the geometry, electric dipole moment, and normal mode
frequencies of water in the ground electronic state using three
different basis sets, and compares the results to see how each
quantity converges (or fails to converge) to the experimental

Journal of Chemical Education Vol. 76 No. 7 July 1999 JChemEd.chem.wisc.edu

Information Textbooks Media Resources

value as the quality of the representation improves. The calculations start with the STO-3G basis set and are repeated
using a larger basis set, such as the 3-21, 6-31G and 6-31
G* basis sets. The question to be answered is if the larger
basis set or the inclusion of the set of diffuse functions gives
a significant improvement when the results are compared to
experimental values. Looking at the calculated relative frequencies and the X-Y-Z components of the normal modes,
students should be able to identify the symmetric and antisymmetric stretching and the bending normal modes. We ask the
students to comment on which calculated quantities are most
reliably inferred from electronic structure calculations, and also
which are most accurate, and how strongly they depend on
the choice of basis set.
GAMESS allows students to calculate the properties of
excited electronic states of molecules as well as those of the
ground electronic state. In this case, we use an MCSCF
(CASSCF) approach to get the energy of an excited electronic
state of water. Students are required to define the configuration
interaction spacethat is, to use their own judgment in
describing the ground state and the excited electronic states
in terms of the number of singly occupied - and -spin
orbitals, assigning the number of frozen core orbitals and
allocating the number of occupied and virtual orbitals.
HCN Isomers and the Dissociation Energy
of the HCN Bond
In this exercise the students employ GAMESS to estimate
the relative energies of HCN and HNC and to estimate the
dissociation energy for the HCN bond. The restricted
HartreeFock level of description, based on an STO-3G basis
set, is used for simplicity. The linearity of HCN allows the
use of Cartesian coordinates rather than the Z-matrix in the
input file for the geometry optimization. Calculations are
repeated for the case of C 1 symmetry and the results are
compared. The input file is rearranged so that it represents
linear HNC and the energies and bond lengths and angles are
compared for the optimized geometries of HCN and HNC.
The calculation of the dissociation energy of the HCN
bond provides the opportunity to explore the large disparity
between the results obtained for the character of the dissociation limit from the restricted HartreeFock and the unrestricted HartreeFock levels of calculation. In this case, the
former extrapolates to a dissociation limit which is H+ + CN {,
whereas the latter extrapolates to the correct dissociation limit
H? and CN?.

correction term. A comparison of the optimized molecular

geometries is used to see if there are any significant changes
due to the correlation energy.
The Barrier for Inversion in NH3
The ammonia molecule has a C3v pyramidal structure,
which is known to undergo inversion through an umbrellalike motion. In this exercise the student estimates the barrier
for the isomerization using GAMESS. First, the energy is
calculated for the pyramidal configuration, then the energy
assuming a planar D3h configuration; the difference between
these energies is taken as an estimate of the barrier for inversion. Using the known literature value of the barrier
energy (2009 cm {1) students are asked to suggest possible
reasons for the discrepancy between their result and experiment.
In addition to purely computational aspects, this exercise has
features that force the student to deal with the molecular
symmetry, with encoding the symmetry using the symmetryunique atoms, and with a series of technical problems pertinent
to the program.
Computational Efficiency (Optimization of the Structure
of Benzene)
In this exercise, the benzene structure is optimized using
different symmetries and different coordinate representations,
but the focus is not on the final results of the calculations,
which should be identical, but on the relative computational
efficiency of the calculations. The system geometry is optimized
under its full D 6h symmetry, and then under C 2v and C1
symmetries. In addition to noting any changes in the numerical results, the total CPU time for each run is recorded.
Subsequently, the student checks whether it is more efficient to
perform optimization in Cartesian coordinates or in internal
coordinates for the benzene ring system.

The Electronic Structure of AgCO+

Transition metals and other heavy elements represent a
challenge to electronic structure theory because they contain
large numbers of electrons. For this exercise we suggest that
the student use the effective core potential (ECP) method to
approximate the effect of core electrons on valence electrons.
The basis set proposed by Stevens, Basch, Krauss, Jasien, and
Cundari (SKB) is used. The ECP method treats valence
orbitals in a 6-31G* basis and fits core orbitals using an
electrostatic potential represented as a sum of four empirically
fitted Gaussians. It is also often the case with heavy elements
that the correlation energy is a very significant contribution
to the total energy. The student is asked to carry out two
calculations on of the structure of AgCO+, one with the
restricted HartreeFock method, and one with the MP2

Molecular Mechanics
1997: Determine, by the method of Molecular Mechanics,
the equilibrium conformation of perfluoroeicosane,
F(CF2)20F, at T = 0 K. Using the same potential energy
surface, estimate the number of defects in a chain at T =
300 K.
1998: Calculate the relative energies and conformational
geometries of all of the conformers of cyclodecane and
cyclodecyne.

Student Assessment
The course grade is based on the results of completed
laboratory assignments and a computational project. Possible
topics for the computational project are distributed early in
the quarter, and a date for completion (near the end of the
quarter) is assigned. The student is asked to choose one of the
computational projects, carry it out, and prepare a report
describing the basic theory used, the results, and the implications for other systems that might be studied. The following
topics were proposed to the students in Spring 1997 and
Spring 1998.

Molecular Dynamics
1997: Carry out a Molecular Dynamics simulation of the
structure of water in the solvation shells surrounding a
Ne atom and a Na+ ion, and compare these structures.
Alternative choices for the pair of atom and ion solutes
whose solvation shell structures are to be compared are
Ar and K+ and Kr and Rb +.

JChemEd.chem.wisc.edu Vol. 76 No. 7 July 1999 Journal of Chemical Education

1027

Information Textbooks Media Resources

1998: Calculate the differences between the equilibrium
constants for the reaction
n-butane (all trans) n-butane (gauche)
when the n-butane molecule is in its own vapor at 165 K
and p = 0.01 torr, and when it is dissolved in dense fluid
Xe at 165 K and p = 1 atm.
Electronic Structure of Molecules
1997: Compute the barrier to rotational motion in the
ground electronic state of ethylene from calculations of
the electronic energy as a function of the angle between
the CH2 groups.
1998: Calculate the geometries of acetylene in its ground
and first excited singlet electronic states.

Have We Achieved the Goals of the Course?

Consider first the view from the students world. Student
evaluations of the content of the Computational Chemistry
course, of the division between lecture and laboratory material,
and of the amount they have learned have all been strongly
favorable, notwithstanding the pace at which the course
proceeds. These evaluations also identify two issues that are
pertinent to our understanding of undergraduate education.
First, a reasonable fraction of the student evaluations
made the comment that the instructor expected them to have
remembered and to be able to use what they learned in other
courses. Aside from the amusing character of this remark, it
in part reflects the extent to which the instructor pushed the
students to learn rapidly, and in part suggests that our undergraduate curricula do not sufficiently emphasize integration
of subject matter and evolution of understanding through
increasing depth of study.
Second, the breadth of the material covered in the course
cannot be found in any one textbook. Because the course
focused on developing understanding of the character of the
underlying numerical methods of computation, a single text,
dealing with numerical analysis, was chosen. It was assumed
that the combination of distributed lecture notes, other books
the student retained from previous courses, and the reference
texts held at the library, would adequately supplement the
chosen text. However, it appears that this textbook strategy
is not optimal; the student evaluations report that the text
was rarely used. In principle, the availability of a textbook
with requisite breadth would correct this difficulty. On the
other hand, it is plausible to interpret lack of appreciation
of the usefulness of the assigned text as an example of the
insufficiency of the current curriculum with respect to intellectual integration of the various disciplines that contribute
to chemistry. We believe that one of the advantages of a computational chemistry course is that it forces confrontation
with the need for that integration.
Consider now the view from the instructors world. In
the first section we posited several goals for a course in
computational chemistry. Specifically, we suggested that this
course should generate an understanding of the interplay
between basic theory and computational methodology, that it
should illuminate the circumstances under which computation
is the preferred tool for solving problems and illustrate the
accuracy with which those problems can be solved. We also
argued that a study of computational chemistry should enhance
critical thinking about scientific goals which computation can
assist in achieving. How well does the implementation of
1028

Computational Chemistry we have taught for two years

meet these goals, and what criteria can be used to make
that assessment?
One feature of contemporary education is the extent to
which technology has interfered with the employment of
primary tools and principles in didactic exercises. It is the case
that modern laboratory instrumentation enables a student to
carry out sophisticated experiments without having to think
about the principles that define the method of measurement.
Similarly, and more relevant for present purposes, calculations
using applications programs enable a student to visualize
complex molecular structures, for example, a protein, without
knowing anything about the underlying theory used and
without clearly thinking through the uncertainties generated
by the method of calculation. The emphasis we have placed
on the methods used for computation, their accuracy and
their limitations, are intended to redress the balance between
understanding of fundamental principles and obtaining results.
It has been our experience that a very large fraction (usually
more than half ) of the students who have taken the Computational Chemistry course at The University of Chicago have
worked with scientific applications software prior to enrollment
in the course. This experience comes by way of undergraduate
research opportunities, most often associated with biomedical
problems. Before enrollment in Computational Chemistry,
by their own testimony, the students believed everything they
saw as output on the monitor screen. At the end of the course,
again by their own testimony, these students have become
much more sophisticated and critical in their interpretation of
the results of computations. Judging by the quality of their term
papers, they have also grasped the most important elements of
the science underlying the problems they addressed, and they
have developed an understanding of the nature of approximation in calculations and of the trade-off between computational
efficiency and accuracy. At least in these senses we believe our
Computational Chemistry course has been successful.
We close with one example that leads us to the conclusion
just cited, namely, the response of a student who undertook, as
a quarter project, the calculation of the equilibrium conformation of perfluoroeicosane. Prior to enrollment in Computational Chemistry this student had three years experience in
modeling peptide and protein molecules using commercially
available applications software. He was one of those characterized above as believing whatever appeared on the monitor
screen. When he discovered that the potential parameters in
the Molecular Mechanics application program he was using
predicted that perfluoroeicosane could not exist he was
stunned. He spent many hours trying to revise the potential
parameters and the calculation so as to obtain a result he could
believe was realistic. His report contained the following emphatic statement: What Dr. Rice has been telling us all quarter
is true! These programs contain hidden assumptions that can
lead to calculated results that are nonsense. It is necessary to
know why a calculation is done in a specific way.
Acknowledgments
We thank the National Science Foundations Division
of Undergraduate Education for financial support through
the Instrumentation and Laboratory Improvement program
(DUE-9751224). We thank Dima Chekmarev, Mark Kobrak,
Jeanne Siemion, R. Michael Townsend, and Meishan Zhao

Journal of Chemical Education Vol. 76 No. 7 July 1999 JChemEd.chem.wisc.edu

Information Textbooks Media Resources

for their invaluable contribution to the Computational

Chemistry project at the University of Chicago.

D. J.; Baker, J.; Stewart, J. P.; Head-Gordon, M.; Gonzalez, C.;

Pople, J. A. Gaussian 94; Gaussian, Inc.: Pittsburgh PA, 1995.

Appendix

Literature Cited
1. DeKock, R. L.; Madura, J. D.; Rioux, F.; Casanova, J. Reviews in
Computational Chemistry, Vol 4; VCH: Weinheim, 1993; pp
149228.
2. Duke, B. J.; OLeary, B. J. Chem. Educ. 1995, 72, 501504.
3. Williams, D. L.; Minarik, P. R.; Nibler, J. W. J. Chem. Educ. 1996,
73, 608613.
4. Lehman, J. J; Goldstein, E. J. Chem. Educ. 1996, 73, 10961098.
5. MATHEMATICA; Wolfram Research, PO Box 6059,
Champaign, IL 61826.
6. Healy, E. F. J. Chem. Educ. 1995, 72, A120A121.
7. Ramachandran, B.; Kong, P. C. J. Chem. Educ. 1995, 72, 406408.
8. David, C. W. J. Chem. Educ. 1995, 72, 995997.
9. Lang, P. L. Towns, M. H. J. Chem. Educ. 1998, 75, 506509.
10. Mathcad; MathSoft, One Kendall Square, Cambridge, MA
02139.
11. Schmidt, M. W.; Baldridge, K. K.; Boatz, J. A.; Elbert, S. T.; Gordon, M. S.; Jensen, J. H.; Koseki, S.; Matsunaga, N.; Nguyen,
K. A.; Su, S. J.; Windus, T. L.; Dupuis, M.; Montgomery, J. A. J.
Comput. Chem. 1993, 14, 13471363.
12. Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Gill, P. M. W.;
Johnson, B. G.; Robb, M. A.; Cheeseman, J. R.; Keith, T.;
Petersson, G. A.; Montgomery, J. A.; Raghavachari, K.; Al-Laham,
M. A.; Zakrzewski, V. G.; Ortiz, J. V.; Foresman, J. B.; Cioslowski,
J.; Stefanov, B. B.; Nanayakkara, A.; Challacombe, M.; Peng, C. Y.;
Ayala, P. Y.; Chen, W.; Wong, M. W.; Andres, J. L.; Replogle, E. S.;
Gomperts, R.; Martin, R. L.; Fox, D. J.; Binkley, J. S.; Defrees,

Required Textbook

Burden, R. L.; Faires, J. D. Numerical Analysis, 6th ed.;

Brooks/Cole: Pacific Grove, CA, 1997; ISBN 0-534-95532-0.
Recommended Books

Wolfram, S. The Mathematica Book, 3rd ed.; Wolfram Media/

Cambridge University Press, 1996; ISBN 0-9650532-0-2.
Koonin, S. E.; Meredith, D. C. Computational Physics
(FORTRAN ed.); Addison-Wesley: Redwood City, CA,
1989; ISBN 0-201-12779-2.
Haile, J. M. Molecular Dynamics Simulations: Elementary
Methods; Wiley: New York, 1992; ISBN 0-471-81966-2.
Kalos, M. H.; Whitlock, P. A. Monte Carlo Methods; Wiley:
New York, 1986; ISBN 0-471-8~839-2.
Frenkel, D.; Smit, B. Understanding Molecular Simulation:
From Algorithms to Applications; Academic: San Diego, 1996;
ISBN 0-12-267370-0.
Hasanein, A. A.; Evans, M. W. Computational Methods in
Quantum Chemistry; World Scientific: River Edge, NJ, 1996;
ISBN 981-02-2611-X.
Szabo, A.; Osthend, N. S. Modern Quantum Chemistry:
Introduction to Advanced Electronic Structure Theory; Free
Press: New York; Collier Macmillan: London, 1982; ISBN
0-486-69186-1.

JChemEd.chem.wisc.edu Vol. 76 No. 7 July 1999 Journal of Chemical Education

1029