MATLAB codes for teaching quantum physics: Part 1

R. Garcia, A. Zozulya, and J. Stickney

arXiv:0704.1622v1 [physics.ed-ph] 12 Apr 2007

Department of Physics, Worcester Polytechnic Institute, Worcester, MA 01609

(Dated: February 1, 2008)
Among the ideas to be conveyed to students in an introductory quantum course, we have the
pivotal idea championed by Dirac that functions correspond to column vectors (kets) and that
differential operators correspond to matrices (ket-bras) acting on those vectors. The MATLAB
(matrix-laboratory) programming environment is especially useful in conveying these concepts to
students because it is geared towards the type of matrix manipulations useful in solving introductory
quantum physics problems. In this article, we share MATLAB codes which have been developed at
WPI, focusing on 1D problems, to be used in conjunction with Griffiths introductory text.

Two key concepts underpinning quantum physics are

the Schrodinger equation and the Born probability equation. In 1930 Dirac introduced bra-ket notation for state
vectors and operators.1 This notation emphasized and
clarified the role of inner products and linear function
spaces in these two equations and is fundamental to
our modern understanding of quantum mechanics. The
Schrodinger equation tells us how the state of a particle
evolves in time. In bra-ket notation, it reads
i~

d
|i = H|i
dt

(1)

where H is the Hamiltonian operator and |i is a ket or

column vector representing the quantum state of the particle. When a measurement of a physical quantity A is
made on a particle initially in the state , the Born equation provides a way to calculate the probability P (Ao )
that a particular result Ao is obtained from the measurement. In bra-ket notation, it reads2
P (Ao ) |hAo |i|2

(2)

where if |Ao i is the state vector corresponding to the

particular result Ao having been measured, hAo | = |Ao i
is the corresponding bra or row vector and hAo |i is thus
the inner product between |Ao i and |i. In the Dirac
formalism, the correspondence between the wavefunction
(~x) and the ket |i is set by the relation (~x) = h~x|i,
where |~xi is the state vector corresponding to the particle
being located at ~x. Thus we regard (~x) as a component
of a state vector |i, just as we usually3 regard ai = ~a
as a component of ~a along the direction . Similarly, we
think of the Hamiltonian operator as a matrix
(
)

2
Z
1
~
3
H = d ~x |~xi
(~x, t) + U (~x) h~x| (3)
2m i ~x
acting on the space of kets.
While an expert will necessarily regard Eqs.(1-3) as a
great simplification when thinking of the content of quantum physics, the novice often understandably reels under
the weight of the immense abstraction. We learn much
about student thinking from from the answers given by
our best students. For example, we find a common error

when studying 1D quantum mechanics is a student treating (x) and |i interchangeably, ignoring the fact that
the first is a scalar but the ket corresponds to a column
vector. For example, they may write incorrectly

hp||i|xi = |ihp|xi

(incorrect!)

(4)

or some similar abberation. To avoid these types of misconceptions, a number of educators and textbook authors
have stressed incorporating a numerical calculation aspect to quantum courses.4,5,6,7,8,9 The motive is simple.
Anyone who has done numerical calculations cant help
but regard a ket |i as a column vector, a bra h| as a
row vector and an operator H as a matrix because that is
how they concretely represented in the computer. Introducing a computational aspect to the course provides one
further benefit: it gives the beginning quantum student
the sense that he or she is being empowered to solve real
problems that may not have simple, analytic solutions.
With these motivations in mind, we have developed
MATLAB codes10 for solving typical 1 D problems found
in the first part of a junior level quantum course based
on Griffiths book.11 We chose MATLAB for our programming environment because the MATLAB syntax is
especially simple for the typical matrix operations used
in 1D quantum mechanics problems and because of the
ease of plotting functions. While some MATLAB numerical recipes have previously been published by others,12,13
the exercises we share here are special because they emphasize simplicity and quantum pedagogy, not numerical
efficiency. Our point has been to provide exercises which
show students how to numerically solve 1 D problems in
such a way that emphasizes the column vector aspect of
kets, the row vector aspect of bras and the matrix aspect
of operators. Exercises using more efficient MATLAB
ODE solvers or finite-element techniques are omitted because they do not serve this immediate purpose. In part
II of this article, we hope to share MATLAB codes which
can be used in conjunction with teaching topics pertaining to angular momentum and non-commuting observables.

2
I.

FUNCTIONS AS VECTORS

To start students thinking of functions as column

vector-like objects, it is very useful to introduce them
to plotting and integrating functions in the MATLAB
environment. Interestingly enough, the plot command in
MATLAB takes vectors as its basic input element. As
shown in Program 1 below, to plot a function f (x) in
MATLAB, we first generate two vectors: a vector of x
values and a vector of y values where y = f (x). The command plot(x, y, r ) then generates a plot window containing the points (xi , yi ) displayed as red points ( r ).
Having specified both x and y, to evaluate the definite
RL
integral L ydx, we need only sum all the y values and
multiply by dx.
%****************************************************************
% Program 1: Numerical Integration and Plotting using MATLAB
%****************************************************************
N=1000000;
% No. of points
L=500;
% Range of x: from -L to L
x=linspace(-L,L,N);
% Generate column vector with N
% x values ranging from -L to L
dx=x(2)-x(1);
% Distance between adjacent points
% Alternative Trial functions:
% To select one, take out the comment command % at the beginning.
%y=exp(-x.^2/16);
% Gaussian centered at x=0
%y=((2/pi)^0.5)*exp(-2*(x-1).^2); % Normed Gaussian at x=1
%y=(1-x.^2).^-1;
% Symmetric fcn which blows up at x=1
%y=(cos(pi*x)).^2;
% Cosine fcn
%y=exp(i*pi*x);
% Complex exponential
%y=sin(pi*x/L).*cos(pi*x/L);% Product of sinx times cosx
%y=sin(pi*x/L).*sin(pi*x/L);% Product of sin(nx) times sin(mx)
%A=100; y=sin(x*A)./(pi*x); % Rep. of delta fcn
A=20; y=(sin(x*A).^2)./(pi*(A*x.^2));% Another rep. of delta fcn
% Observe: numerically a function y(x) is represented by a vector!

to n = 1, etc...).
An exercise we suggest is for students to verify that
the derivative matrix is not Hermitian while the derivative matrix times the imaginary number i is. This can
be very valuable for promoting student understanding if
done in conjunction with the proof usually given for the
differential operator.
%****************************************************************
% Program 2: Calculate first and second derivative numerically
% showing how to write differential operator as a matrix
%****************************************************************
% Parameters for solving problem in the interval 0 < x < L
L = 2*pi;
% Interval Length
N = 100;
% No. of coordinate points
x = linspace(0,L,N);
% Coordinate vector
dx = x(2) - x(1);
% Coordinate step
% Two-point finite-difference representation of Derivative
D=(diag(ones((N-1),1),1)-diag(ones((N-1),1),-1))/(2*dx);
% Next modify D so that it is consistent with f(0) = f(L) = 0
D(1,1) = 0; D(1,2) = 0; D(2,1) = 0; % So that f(0) = 0
D(N,N-1) = 0; D(N-1,N) = 0; D(N,N) = 0; % So that f(L) = 0
% Three-point finite-difference representation of Laplacian
Lap = (-2*diag(ones(N,1),0) + diag(ones((N-1),1),1) ...
+ diag(ones((N-1),1),-1))/(dx^2);
% Next modify Lap so that it is consistent with f(0) = f(L) = 0
Lap(1,1) = 0; Lap(1,2) = 0; Lap(2,1) = 0;
% So that f(0) = 0
Lap(N,N-1) = 0; Lap(N-1,N) = 0; Lap(N,N) = 0;% So that f(L) = 0
% To verify that D*f corresponds to taking the derivative of f
% and Lap*f corresponds to taking a second derviative of f,
% define f = sin(x) or choose your own f
f = sin(x);
% And try the following:
Df = D*f; Lapf = Lap*f;
plot(x,f,b,x,Df,r, x,Lapf,g);
axis([0 5 -1.1 1.1]);
% Optimized axis parameters
% To display the matrix D on screen, simply type D and return ...
D % Displays the matrix D in the workspace
Lap % Displays the matrix Lap

% Plot a vector/function
plot(x,y);
% Plots vector y vs. x
%plot(x,real(y),r, x, imag(y), b); % Plots real&imag y vs. x
axis([-2 2 0 7]);
% Optimized axis parameters for sinx^2/pix^2
%axis([-2 2 -8 40]); % Optimized axis parameters for sinx/pix
% Numerical Integration
sum(y)*dx
% Simple numerical integral of y
trapz(y)*dx
% Integration using trapezoidal rule

III.

INFINITE SQUARE WELL

When solving Eq. (1), the method of separation of variables entails that as an intermediate step we look for the
separable solutions
II.

DIFFERENTIAL OPERATORS AS
MATRICES

Just as f (x) is represented by a column vector |f i in

the computer, for numerical purposes a differential opera acting on f (x) is reresented by a matrix D that acts
tor D
on |f i. As illustrated in Program 2, MATLAB provides
many useful, intuitive, well-documented commands for
10
generating matrices D that correspond to a given D.
Two examples are the commands ones and diag. The
command ones(a, b) generates an a b matrix of ones.
The command diag(A, n) generates a matrix with the elements of the vector A placed along the nth diagonal and
zeros everywhere else. The central diagonal corresponds
to n = 0, the diagonal above the center one corresponds

|E (t)i = |E (0)iexp(iEt/~)

(5)

where |E (0)i satisfies the time-independent Schrodinger

equation
H|E (0)i = E |E (0)i.

(6)

In solving Eq. (6) we are solving for the eigenvalues E

and eigenvectors |E (0)i of H. In MATLAB, the command [V, E] = eig(H) does precisely this: it generates
two matrices. The first matrix V has as its columns the
eigenvectors |E (0)i. The second matrix E is a diagonal matrix with the eigenvalues Ei corresponding to the
eigenvectors |Ei (0)i placed along the central diagonal.
We can use the command E = diag(E) to convert this
matrix into a column vector. In Program 3, we solve for

3
the eigenfunctions and eigenvalues for the infinite square
well Hamiltonian. For brevity, we omit the commands
setting the parameters L, N, x, and dx.
%****************************************************************
% Program 3: Matrix representation of differential operators,
% Solving for Eigenvectors & Eigenvalues of Infinite Square Well
%****************************************************************
% For brevity we omit the commands setting the parameters L, N,
% x and dx; We also omit the commands defining the matrix Lap.
% These would be the same as in Program 2 above.
% Total Hamiltonian where hbar=1 and m=1
hbar = 1; m = 1; H = -(1/2)*(hbar^2/m)*Lap;
% Solve for eigenvector matrix V and eigenvalue matrix E of H
[V,E] = eig(H);
% Plot lowest 3 eigenfunctions
plot(x,V(:,3),r,x,V(:,4),b,x,V(:,5),k); shg;
E % display eigenvalue matrix
diag(E) % display a vector containing the eigenvalues

Note that in the MATLAB syntax the object V (:, 3)

specifies the column vector consisting of all the elements
in column 3 of matrix V . Similarly V (2, :) is the row
vector consisting of all elements in row 2 of V ; V (3, 1) is
the element at row 3, column 1 of V ; and V (2, 1 : 3) is
a row vector consisting of elements V (2, 1), V (2, 2) and
V (2, 3).
IV.

ARBITRARY POTENTIALS

Numerical solution of Eq. (1) is not limited to any particular potential. Program 4 gives example MATLAB
codes solving the time independent Schrodinger equation for finite square well potentials, the harmonic oscillator potential and even for potentials that can only
solved numerically such as the quartic potential U = x4 .
In order to minimize the amount of RAM required, the
codes shown make use of sparse matrices, where only
the non-zero elements of the matrices are stored. The
commands for sparse matrices are very similar to those
for non-sparse matrices. For example, the command
[V, E] = eigs(H, nmodes..) provides the nmodes lowest
energy eigenvectors V of of the sparse matrix H.
Fig. 1 shows the plot obtained from Program 4 for the
potential U = 12 100 x2. Note that the 3 lowest energies
displayed in the figure are just as expected due to the
analytic formula


1
(7)
E = ~ n +
2
q
k
with n = integer and = m
= 10 rad/s.
V.

A NOTE ON UNITS IN OUR PROGRAMS

When doing numerical calculations, it is important to

minimize the effect of rounding errors by choosing units
such that the variables used in the simulation are of the

%****************************************************************
% Program 4: Find several lowest eigenmodes V(x) and
% eigenenergies E of 1D Schrodinger equation
%
-1/2*hbar^2/m(d2/dx2)V(x) + U(x)V(x) = EV(x)
% for arbitrary potentials U(x)
%****************************************************************
% Parameters for solving problem in the interval -L < x < L
% PARAMETERS:
L = 5;
% Interval Length
N = 1000;
% No of points
x = linspace(-L,L,N);
% Coordinate vector
dx = x(2) - x(1);
% Coordinate step
% POTENTIAL, choose one or make your own
U = 1/2*100*x.^(2);
% quadratic harmonic oscillator potential
%U = 1/2*x.^(4);
% quartic potential
% Finite square well of width 2w and depth given
%w = L/50;
%U = -500*(heaviside(x+w)-heaviside(x-w));
% Two finite square wells of width 2w and distance 2a apart
%w = L/50; a=3*w;
%U = -200*(heaviside(x+w-a) - heaviside(x-w-a) ...
%
+ heaviside(x+w+a) - heaviside(x-w+a));
%
%
%
e

Three-point finite-difference representation of Laplacian

using sparse matrices, where you save memory by only
storing non-zero matrix elements
= ones(N,1); Lap = spdiags([e -2*e e],[-1 0 1],N,N)/dx^2;

% Total Hamiltonian
hbar = 1; m = 1;
% constants for Hamiltonian
H = -1/2*(hbar^2/m)*Lap + spdiags(U,0,N,N);
% Find lowest nmodes eigenvectors and eigenvalues of sparse matrix
nmodes = 3; options.disp = 0;
[V,E] = eigs(H,nmodes,sa,options);
% find eigs
[E,ind] = sort(diag(E));% convert E to vector and sort low to high
V = V(:,ind);
% rearrange corresponding eigenvectors
% Generate plot of lowest energy eigenvectors V(x) and U(x)
Usc = U*max(abs(V(:)))/max(abs(U));
% rescale U for plotting
plot(x,V,x,Usc,--k);
% plot V(x) and rescaled U(x)
% Add legend showing Energy of plotted V(x)
lgnd_str = [repmat(E = ,nmodes,1),num2str(E)];
legend(lgnd_str)
% place lengend string on plot
shg

order of unity. In the programs presented here, our focus being undergraduate physics students, we wanted to
avoid unnecessarily complicating matters. To make the
equations more familiar to the students, we explicitly left
constants such as ~ in the formulas and chose units such
that ~ = 1 and m = 1. We recognize that others may
have other opinions on how to address this issue. An alternative approach used in research is to recast the equations in terms of dimensionless variables, for example by
rescaling the energy to make it dimensionless by expressing it in terms of some characteristic energy in the problem. In a more advanced course for graduate students or
in a course in numerical methods, such is an approach
which would be preferable.

VI.

TIME DEPENDENT PHENOMENA

The separable solutions |E (t)i are only a subset of

all possible solutions of Eq. (1). Fortunately, they are
complete set so that we can construct the general solution

4
is time-dependent, oscillating between the (x) that corresponds to the particle being entirely in the right well

0.25
E = 4.99969
E = 14.9984
E = 24.9959

unnormalized wavefunction

0.2

(x) = | |E0 (x)| + |E1 (x)| |

(11)

0.15

and (x) for the particle being entirely in the left well
0.1
2

(x) = | |E0 (x)| |E1 (x)| | .

0.05

(12)

By observing the period with which (x, t) oscillates

in the simulation output shown in Fig. 2 students can
verify that it is the same as the period of oscillation
2~/(E1 E2 ) expected from Eq. (10).

0
0.05
0.1
1.5

0.5

0
0.5
x [m]

%****************************************************************
% Program 5: Calculate Probability Density as a function of time
% for a particle trapped in a double-well potential
%****************************************************************

1.5

FIG. 1: Output of Program 4, which plots the energy eigenfunctions V (x) and a scaled version of the potential U (x) =
1/2 100 x2 . The corresponding energies displayed within
the figure legend, 4.99969, 14.9984 and 24.9959, are, within
rounding error, precisely those expected from Eq. (7) for the
three lowest-energy modes.

via the linear superposition

|(t)i =

X
E

aE |E (0)iexp(iEt/~)

(8)

where aE are constants and the sum is over all possible

values of E. The important difference between the separable solutions (5) and the general solution (8) is that
the probability densities derived from the general solutions are time-dependent whereas those derived from the
separable solutions are not. A very apt demonstration
of this is provided in the Program 5 which calculates the
time-dependent probability density (x, t) for a particle
trapped in a a pair of finite-square wells whose initial
state |(0)i is set equal to the the equally-weighted superposition
1
|(0)i = (|E0 i + |E1 i)
2

(9)

of the ground state |E0 i and first excited state |E1 i

of the double well system. As snapshots of the program
output show in Fig. 2, the particle is initially completely
localized in the rightmost well. However, due to E0 6= E1 ,
the probability density
1
(x, t) = [ |E0 (x)|2 + |E1 (x)|2 +
2
2|E0 (x)||E1 (x)|cos2 ((E1 E2 )t/~) ]

%
%
w
U

Potential due to two square wells of width 2w

and a distance 2a apart
= L/50; a = 3*w;
= -100*( heaviside(x+w-a) - heaviside(x-w-a) ...
+ heaviside(x+w+a) - heaviside(x-w+a));

%
%
e
H

Finite-difference representation of Laplacian and Hamiltonian,

where hbar = m = 1.
= ones(N,1); Lap = spdiags([e -2*e e],[-1 0 1],N,N)/dx^2;
= -(1/2)*Lap + spdiags(U,0,N,N);

% Find and sort lowest nmodes eigenvectors and eigenvalues of H

nmodes = 2; options.disp = 0; [V,E] = eigs(H,nmodes,sa,options);
[E,ind] = sort(diag(E));% convert E to vector and sort low to high
V = V(:,ind);
% rearrange coresponding eigenvectors
% Rescale eigenvectors so that they are always
% positive at the center of the right well
for c = 1:nmodes
V(:,c) = V(:,c)/sign(V((3*N/4),c));
end
%****************************************************************
% Compute and display normalized prob. density rho(x,T)
%****************************************************************
% Parameters for solving the problem in the interval 0 < T < TF
TF = 10;
% Length of time interval
NT = 100;
% No. of time points
T = linspace(0,TF,NT);
% Time vector
% Compute probability normalisation constant (at T=0)
psi_o = 0.5*V(:,1)+0.5*V(:,2);
% wavefunction at T=0
sq_norm = psi_o*psi_o*dx;
% square norm = |<ff|ff>|^2
Usc = U*max(abs(V(:)))/max(abs(U));

% Compute and display rho(x,T) for each time T

for t=1:NT;
% time index parameter for stepping through loop
% Compute wavefunction psi(x,T) and rho(x,T) at T=T(t)
psi = 0.5*V(:,1)*exp(-i*E(1)*T(t)) ...
+ 0.5*V(:,2)*exp(-i*E(2)*T(t));
rho = conj(psi).*psi/sq_norm;
% normalized probability density
% Plot rho(x,T) and rescaled potential energy Usc
plot(x,rho,o-k, x, Usc,.-b); axis([-L/8 L/8 -1 6]);
lgnd_str = [repmat(T = ,1,1),num2str(T)];
text(-0.12,5.5,lgnd_str, FontSize, 18);
xlabel(x [m], FontSize, 24);
ylabel(probability density [1/m],FontSize, 24);
pause(0.05);
end

(10)

% rescale U for plotting

% wait 0.05 seconds

5
6

6
T = 0s

T = 2.8189s
5
probability density [1/m]

probability density [1/m]

5
4
3
2
1
0

4
3
2
1
0

1
0.6 0.4 0.2

1
0.6 0.4 0.2

0 0.2 0.4 0.6

x [m]

6
T = 5.6378s

T = 8.4566s
5
probability density [1/m]

5
probability density [1/m]

0 0.2 0.4 0.6

x [m]

4
3
2
1
0

4
3
2
1
0

1
0.6 0.4 0.2

0 0.2 0.4 0.6

x [m]

1
0.6 0.4 0.2

0 0.2 0.4 0.6

x [m]

6
T = 11.2755s

probability density [1/m]

5
4
3
2
1
0
1
0.6 0.4 0.2

0 0.2 0.4 0.6

x [m]

FIG. 2: The probability density (x, t) output from Program

5 for a particle trapped in a pair of finite-height square well
potentials that are closely adjacent. The initial state of the
particle is chosen to be |(0)i |E0 i + |E1 i. Shown is (x, t)
plotted for times T = {0, 0.25, 0.5, 0.75, 1.0}2~/(E1 E2 ).

VII.

WAVEPACKETS AND STEP POTENTIALS

Wavepackets are another time-dependent phenomenon

encountered in undergraduate quantum mechanics for
which numerical solution techniques have been typically
advocated in the hopes of promoting intuitive acceptance
and understanding of approximations necessarily invoked
in more formal, analytic treatments. Program 6 calcu-

lates and displays the time evolution of a wavepacket for

one of two possible potentials, either U = 0 or a step
potential U = Uo (x L). The initial wavepacket is
generated as the Fast Fourier Transform of a Gaussian
momentum distribution centered on a particular value
of the wavevector ko . Because the wavepacket is composed of a distribution of different ks, different parts of
the wavepacket move with different speeds, which leads
to the wave packet spreading out in space as it moves.
While there is a distribution of velocities within the
wavepacket, two velocities in particular are useful in characterizing it. The phase velocity vw = /k = E/p =
~ko /2m is the velocity of the plane wave component
which has wavevector ko . The group velocity vg = ~ko /m
is the velocity with which the expectation value < x >
moves and is the same as the classical particle velocity associated with the momentum p = ~k. Choosing U = 0,
students can modify this program to plot < x > vs t.
They can extract the group velocity from their numerical
simulation and observe that indeed vg = 2vw for a typical wave packet. Students can also observe that, while
vg matches the particle speed from classical mechanics,
the wavepacket spreads out as time elapses.
In Program 6, we propagate the wave function forward
via the formal solution
|(t)i = exp(iHt/~)|(0)i,

(13)

where the Hamiltonian matrix H is in the exponential.

This solution is equivalent to Eq. (6), as as can be shown
by simple substitution. Moreover, MATLAB has no trouble exponentiating the matrix that numerically representing the Hamiltonian operator as long as the matrix
is small enough to fit in the available computer memory.
Even more interestingly, students can use this method
to investigate scattering of wavepackets from various potentials, including the step potential U = Uo (x L/2).
In Fig. 3, we show the results of what happens as the
wavepacket impinges on the potential barrier. The parameters characterizing the initial wavepacket have been
deliberately chosen so that the wings do not fall outside
the simulation area and initially also do not overlap the
barrier on the right. If hEi Uo , the wavepacket is
completely reflected from the barrier. If hEi Uo , a
portion of the wave is is reflected and a portion is transmitted through. If hEi Uo , almost all of the wave is
transmitted.
In Fig. 4 we compare the reflection probability R calculated numerically using Program 6 with R calculated
by averaging the single-mode11 expression


 E E U 2

o
R(E) = 


 E + E Uo 

(14)

over the distribution of energies in the initial wavepacket.

While the numerically and analytically estimated R are
found to agree for large and small hEi/Uo , there is a
noticeable discrepancy due to the shortcomings of the

% Parameters for making intial momentum space wavefunction phi(k)

ko = 2;
% Peak momentum
a = 20;
% Momentum width parameter
dk = 2*pi/L;
% Momentum step
km=N*dk;
% Momentum limit
k=linspace(0,+km,N);
% Momentum vector
% Make psi(x,0) from Gaussian kspace wavefunction phi(k) using
% fast fourier transform :
phi = exp(-a*(k-ko).^2).*exp(-i*6*k.^2); % unnormalized phi(k)
psi = ifft(phi); % multiplies phi by expikx and integrates vs. x
psi = psi/sqrt(psi*psi*dx);
% normalize the psi(x,0)
% Expectation value of energy; e.g. for the parameters
% chosen above <E> = 2.062.
avgE = phi*0.5*diag(k.^2,0)*phi*dk/(phi*phi*dk);
% CHOOSE POTENTIAL U(X): Either U = 0 OR
% U = step potential of height avgE that is located at x=L/2
%U = 0*heaviside(x-(L/2)); % free particle wave packet evolution
U = avgE*heaviside(x-(L/2));
% scattering off step potential
% Finite-difference representation of Laplacian and Hamiltonian
e = ones(N,1); Lap = spdiags([e -2*e e],[-1 0 1],N,N)/dx^2;
H = -(1/2)*Lap + spdiags(U,0,N,N);

% Time displacement operator E=exp(-iHdT/hbar)

E = expm(-i*full(H)*dT/hbar); % time desplacement operator
%***************************************************************
% Simulate rho(x,T) and plot for each T
%***************************************************************
for t = 1:NT;
% time index for loop
% calculate probability density rho(x,T)
psi = E*psi;
% calculate new psi from old psi
rho = conj(psi).*psi;
% rho(x,T)
plot(x,rho,k);
% plot rho(x,T) vs. x
axis([0 L 0 0.15]); % set x,y axis parameters for plotting
xlabel(x [m], FontSize, 24);
ylabel(probability density [1/m],FontSize, 24);
pause(0.05);

% pause between each frame displayed

end
% Calculate Reflection probability
R=0;
for a=1:N/2;
R=R+rho(a);
end
R=R*dx

numerical simulation for hEi/Uo 1. This discrepancy

can be reduced significantly by increasing the number of
points in the simulation to 1600 but only at the cost of
significantly slowing down the speed of the computation.
For our purposes, the importance comparing the analytical and numerical calculations is that it gives student
a baseline from which to form an opinion or intuition
regarding the accuracy of Eq. (14).

0.05

T = 14.4271s

0.1

0.05

0
0 10 20 30 40 50 60 70 80 90 100
x [m]

probability density [1/m]

% Parameters for computing psi(x,T) at different times 0 < T < TF

NT = 200; % No. of time steps
TF = 29; T = linspace(0,TF,NT); % Time vector
dT = T(2)-T(1); % Time step
hbar = 1;

T = 0s

0.1

0
0 10 20 30 40 50 60 70 80 90 100
x [m]

probability density [1/m]

%****************************************************************
% Program 6: Wavepacket propagation using exponential of H
%****************************************************************
% Parameters for solving the problem in the interval 0 < x < L
L = 100;
% Interval Length
N = 400;
% No of points
x = linspace(0,L,N);
% Coordinate vector
dx = x(2) - x(1);
% Coordinate step

probability density [1/m]

T = 28.8543s

0.1

0.05

0
0 10 20 30 40 50 60 70 80 90 100
x [m]

FIG. 3: Output of Program 6 showing a wavepacket encountering step potential of height hEi located at x/L = 0.5 at
different times.

VIII.

CONCLUSIONS

One benefit of incorporating numerical simulation into

the teaching of quantum mechanics, as we have mentioned, is the development of student intuition. Another
is showing students that non-ideal, real-world problems
can be solved using the concepts they learn in the classroom. However, our experimentation incorporating these
simulations in quantum physics at WPI during the past
year has shown us that the most important benefit is a
type of side-effect to doing numerical simulation: the acceptance on an intuitive level by the student that functions are like vectors and differential operators are like
matrices. While in the present paper, we have only had
sufficient space to present the most illustrative MATLAB

7


codes, our goal is to eventualy make available a more

complete set of polished codes is available for downloading either from the authors or directly from the file exchange at MATLAB Central.14


  
  

 

 





   
   



8 
 













!
"#







FIG. 4: The reflection probability R vs. hEi/Uo . The dashed

line is simply Eq. (14) where we subsitute E = hEi, the solid
line is Eq. (14) averaged over the energy distribution in the
incident wavepacket, and the points are numerical results obtained using Program 6, where the horizontal distance between points is E /Uo where E is the standard deviation of
the energy distribution in the wavepacket.

Electronic address: [email protected]

P. A. M. Dirac, The Principles of Quantum Mechanics, 1st
ed., (Oxford University Press, 1930).
Borns law is stated as a proportionality because an additional factor is necessary depending on the units of |i
C. C. Silva and R. de Andrade Martins, Polar and axial vectors versus quaternions, Am. J. Phys. 70, 958-963
(2002).
R. W. Robinett, Quantum Mechanics: Classical Results,
Modern Systems, and Visualized Examples, (Oxford University Press, 1997).
H. Gould, Computational physics and the undergraduate
curriculum, Comput. Phys. Commun. 127, 610 (2000);
J. Tobochnik and H. Gould, Teaching computational
physics to undergraduates, in Ann. Rev. Compu. Phys.
IX, edited by D. Stauffer (World Scientific, Singapore,
2001), p. 275; H. Gould, J. Tobochnik, W. Christian, An
Introduction to Computer Simulation Methods: Applications to Physical Systems, (Benjamin Cummings, Upper
Saddle River, NJ, 2006) 3rd ed..
R. Spenser, Teaching computational physics as a laboratory sequence, Am. J. Phys. 73, 151-153 (2005).
D. Styer, Common misconceptions regarding quantum

10

11

12

13

14

mechanics, Am. J. Phys. 64, 31-34 (1996).

A. Goldberg, H. M. Schey, J. L. Schwartz, ComputerGenerated Motion Pictures of One-Dimensional QuantumMechanical Transmission and Reflection Phenomena,
Am. J. Phys. 35, 177-186 (1967).
C. Singh, M. Belloni, and W. Christian, Improving students understanding of quantum mechanics, Physics Today, August 2006, p. 43.
These MATLAB commands are explained in an extensive on-line, tutorial within MATLAB and which is also
independently available on the MATHWORKS website,
http://www.mathworks.com/.
D. J. Griffiths, Introduction to Quantum Mechanics, 2nd
Ed., Prentice Hall 2003.
A. Garcia, Numerical Methods for Physics, 2nd Ed., (Prentice Hall, 1994).
G.
Lindblad,
Quantum
Mechanics
with
MATLAB,
available
on
internet,
http://mathphys.physics.kth.se/schrodinger.html.
See
the
user
file
exchange
at
http://www.mathworks.com/matlabcentral/.