GENERAL I ARTICLE

Heisenberg, Matrix Mechanics, and the

Uncertainty Principle
S Lakshmibala

Werner Heisenberg was one of the key players

in the development of quantum mechanics. Be-
sides enunciating the famous Uncertainty Princi-
ple, he was also the principal architect of Matrix
Mechanics, one of the two standard formulations
of quantum mechanics.

Genesis

Albert Einstein is said to have remarked that. intuit.ion

S Lakshmibala is a is nothing but the collect.ion of prejudices one fonus up
theoretical physicist at
to the age of eighteen. What he meant was that our
the Department of
Physics, lIT Madras. Her
so-called physical intuition is little more than a rough
current research interests feel for the way the physical world around us behaves
are in dynamical systems on everyday scales. At. very small or very large scales
and quantum information. of length, mass and time, however, this intuit.ion is as
likely as not to be misleading or even wrong. This is
indeed the main message of the revolutionary advances
in t.he physical sciences in t.he 20th century. The dis-
covery of quant.um mechanics is the centre-piece of t.hat
revolut.ion.

By the end of the 19th century, the edifice of physics

seemed to be on a firm foundation. The basic equa-
tions of classical mechanics and elect.romagnetism were
in place. But, at the same t.ime, the atomic or discrete
nat.ure of mat.ter (as well as radiation) was slowly but.
surely becoming clear. This was accompanied by a grad-
ual realization that classical physics was seriously inade-
quate in the realm of atoms, and t.hat atomic phenomena
Keywords were perhaps governed by a 'quantum' theory waiting to
Uncertainty principle, matrix be discovered and tested. Originat.ing in the suggestions
mechanics, quantum measure- of Max Planck and Einst.ein , this theory (now called the
ment.
'old quantum t.heory ' ) was developed by Niels Bohr and

4-6------------------------------~------------R-E-S-O-N-A-N-C-E-I--AU-9-U-s-t-2-0-04-
GENERAL I ARTICLE

Arnold Sommerfeld, among others, and showed promise The quantum

initially. But many of its features were quite clearly ad world required the
hoc, and it was evident that a deeper theory lay hidden unambiguous
beneath. language of
These investigations climaxed with the advent of quan- mathematical
tum mechanics in the 1920s. Under the leadership of equations, matrix
a group of remarkably young physicists, the intricacies algebra, differential
of the quantum world were unravelled - each step leav- equations and
ing behind an audience stunned by Nature's ways. High group theory.
drama was involved, as also a clash of personalities pri-
marily from Gottingen, Copenhagen and Berlin in Eu-
rope, who were dogmatic in their views on what could
be the correct theory. The lessons learnt were many.
The workings of the quantum world defied explanations
(however elaborate) based on intuition. A much more
precise language was required - the unambiguous lan-
guage of mathematical equations, matrix algebra, dif-
ferential equations and group theory.

It was crucial, however, to identify the fundamental

physical quantities whose properties, static and dynamic,
constituted the study of quantum physics. And there
were really no significant pointers in this regard from
classical physics (Box 1). The approach to the quantum
world was therefore based, to a large extent, on what
each researcher believed should be the guiding principle
in investigating the unknown.

Werner Heisenberg, one of the founding fathers of quan-

tum mechanics (as opposed to the 'old' quantum theory)

Box 1. Planck's Constant, II,

Ouce the standard framework of qllantulll lllechanics was well established, it becal1le
clear tllat classical physics call, in general, be recovered frOlll qllalltuIll Illechanics as a
limitiug case: Olle lllllSt formally let a certain COllstant II. tend to zero in all expressions.
This cOllstaut, uamed after Planck, is the trademark of quantum lllechallics. It is one of
the three flllldalllental constants of nature, alollg with (" (the speed of light ill vaculllll)
and G (Newtoll's gravitatiollal COllstant).

-R-ES-O-N-A-N--CE--I-A-U-9-U-st--2-0-0-4------------~--------------------------------47
GENERAL I ARTICLE

Box 2. Early Reactions

Eillsteill. for illstance, was cOllvillced that "... it is the theory which decides what we
call observe~', and not the other way around. Erwin Schrodinger, the creator of wave
lllecilallics. all alternative forIllulatioll of quantuIll lllechauics, opilled that he ..... felt
discollraged. uot to say repelled, ... by the lack of visuali:t:ability" in matrix mechanics.
(SchrodillgeT~s formalism deals with the nature and dynamics of a fundamental object
called the wave function associated with a physical system.) Nevertheless, Schrodillger
himself weut ahead soon thereafter to establish the equivalence between wave mechanics
alld lllatrix mechanics, in 1926.

sought to establish a basis for the theoretical aspects of

the quantum mechanics of a system. This was founded
exclusively upon relationships between quantities pertain-
ing to that system which are, in principle, observable.
By 'observables' are meant quantities which are experi-
mentally measurable, either directly or indirectly. This
approach led to a successful formulation of quantum me-
chanics based on the theory of matrices (matrix mechan-
ics), though it initially met with vehement objections
(Box 2).

As a first step, Heisenberg had to identify the relevant

observables. In atomic physics, observational data re-
lated to atomic transitions arising from interactions of
the atoms with light quanta. Heisenberg therefore ar-
gued that these transition-related quantities were the
basic relevant objects. A detailed investigation on these
lines led him to propose the first coherent mathematical
structure f(H the quantum theory of atoms, in 1925. To-
gether with Max Born and Pascual Jordan, who recog-
nized that these quantities obeyed rules prescribed by
matrix algebra, Heisenberg developed the essentials of
Together with Max matrix mechanics later that year.
Born and Pascual
Jordan, Heisenberg
Matrix Mechanics
developed the Generalizing the lesson learnt from the transition-related
essentials of matrix quantities mentioned above, the matrix formulation was
mechanics in 1925. built on the premise that all physical observables m,u.st

-48-------------------------------~-------------R-ES-O-N-A--N-C-E-I-A-U-9-u-st--2-0-04-
GENERAL I ARTICLE

be represented by matrices. The set of eigenvalues of If the result of a

the Inatrix representing an observable is the set of all measurement is a
possible values that could arise as outcomes of experi- certain eigenvalue,
lnents conducted on a system to measure the observable.
the act of
Since the outcome of an experiment to measure a real
measurement is
observable IllUSt be a real number, Hermitian matrices
taken to 'collapse' the
would represent such observables (as their eigenvalues
state of the system to
are real). If the result of a measurement is a certain
that eigenvector
eigenvalue, the corresponding eigenvector represents the
state of the system immediately after the measurement.
The act of measurement is taken to 'collapse' the state of
the systenl to that eigenvector (or eigenstate). Exam-
ples of such eigenstates are those of position, momen-
tum, energy, etc. It may be possible sometimes to make
simultaneous measurenlents of two or more observables.
In that case the system will collapse to a common eigen-
state of these observables right after the measurement.

It was soon realized that the matrices representing the

observables are really operators in an appropriate lin-
ear vector space, and that their eigenstates are column
vectors in that space. This space could even be infinite-
dimensional (Box 3).

Box 3. Linear Vector Space

Tbe space ill which the state of a quantulll system 'lives' is not to be confnsed with the
tlll·ee-dilllellsiollal physical space ill which we live. To get the idea right, COIl sider the
silllple example of a point particle that is free to Illove OIl a lille. An observable ill this
case conld be the position of the particle, represented by the matrix X. Sillce the particle
conld be allywhere 011 the line, the possible outcome of a measnrement of its positioll
conld be allY Olle of an infinite set of eigenvalues of X, denoted by ;1:. Thus X IIlllSt ue
all illfillite-dilllellsional matrix, and hellce so is the corresponding linear vector space.
Tlllls evell one-dimensional Illotioll couM have an infinite-dimensional linear vector space
associated with it!

III fact. this example is not as trivial as it appears to ue. The set of eigellvalues .1: is
a ('oufin'/1,o//,8 iufinity of valnes, in contrast to a discrete iufinite set of values sHch as
1. 2, 3, 4, .. '. This poses certaill technical problems: for illstance, we llluSt extelld the
cOllcept of matrices to objects with a (:ontinwm81y inji:rAite llllIuLer of rows and COlUlllllS:
llCtlllely, opC'fntO'f'8. Operators, fUllCtiollS spaces, and so OIl, thus lllake their appearal1ce
ill qllallt lllll 111ccltauics quite llaturally.

-R-ES-O--N-A-N-C-E-I--A-U9-U-s-t--2-0-04-------------~~-------------------------------4-9
GENERAL I ARTICLE

From matrix theory Further, from matrix theory we know that eigenvectors
we know that corresponding to distinct eigenvalues of a Hermitian ma-
eigenvectors
trix are orthogonal to each other. In this sense, they are
corresponding to
analogous to the x y z axes of the Cartesian coordi-
nate system - except that now there could even be an in-
distinct eigenvalues
finite number of distinct eigenvaiues, and hence as Inany
of a Hermitian matrix
mutually orthogonal eigenvectors 'pointing' along differ-
are orthogonal to
ent independent directions in the linear vector space.
each other.
Again, just as we have unit vectors ex ey ez along the
Cartesian axes, ,we can normalize each eigenvector to
have unit magnitude.

From ex and ey we can form the linear cornbination

f = a ex+ bey, where a and b are any two real nurn-
bers. In much the sanre way, linear superpositions of
the normalized independent eigenvectors can be forn1ed,
with coefficients that are complex numbers. It is iIl1pOr-
tant to note that, in general, such superpositions are not
eigenstates of the matrix concerned.

We now face an interesting situation. Recall that, by

performing an appropriate measurement on the system,
we know the state of the system just after the mea-
surement. Was this the state of the system before the
measurement? Not necessarily! For, prior to the lnea-
surement, the system could have been in a linear su-
perposition of different eigenstates, with unknown (and
unguessable) coefficients. It is like saying that in a coin
toss experiment whose outcome is a 'head' the coin
could have been in a state which was a combination of
head and tail before it was tossed! Of course, this would
never be the case for actual coins, governed as they are
The state of the by the laws of classical physics.
system just after
the measurement
But then, what was the precise state of the quantuln
is not necessarily
system before the measurement? The answer is: we can-
not know. The Copenhagen interpretation is concerned
the state of the
only with outcomes of experiments. Deep philosophi-
system before the
cal questions, peculiar to quantum mechanics, now arise
measurement.
(Box 4).

-50-------------------------------~-~----------R-ES-O-N--A-N-C-E-I-A-U-9-u-st--2-0-04-
GENERAL I ARTICLE

Box 4. Is Quantum Mechanics Complete?

If we call lleVel' kllow the pre-Illeasurement state of a system, is llOt the theory illade-
qllate. or at least illcomplete? For, after all, the systelll snrely has all existence of its
OWl!. illdepelldeut of the act of measurement! (This qnestioll is also applicable to wave
lllechallics, for it too call1lot predict the pre-measurement state,) N ll111erous proposals,
illclndillg a variety of so-called hidd(~n 1Ja1'iable theories, have been made to overcome this
illadeqllacy. bllt Hone of these is fully satisfactory. The last word has probably Hot beeu
said yet ill this regard,

We will not digress into these here.

We know that, to obtain the average value of any ob-

servable, a large number of trials have to be conducted,
i.e., repeated lneasurements have to be made. But, for
a quantuln system, a single measurement of any observ-
able A yields one of the eigenvalues of A as the out-
come, and collapses the state of the system to the cor-
responding eigenstate. Subsequent measurements made
inlInediately thereafter would continue to yield the same
eigenvalue. Does that mean that a single measurement is
enough? No! We have tampered with the pre-measure-
ment state of the system and collapsed it during the first
measureluent. Therefore the post-measurement state is
not the saIne as the initial state, and so it would not
luake sense to repeat measurements on a single system.
The correct thing to do would be to prepare a collection
of a very large number of identical copies of the system Certain pairs of
(an ensem,ble) , and conduct a single trial on each copy. observables simply
The arithmetic mean of all the results thus obtained is cannot be
the average value we seek, denoted by (A). This is in Simultaneously
sharp contrast to what we do in usual experiments on a measured to an
classical system, arbitrarily high level of
precision. In general,
So, can we now claim that ensemble averages should give
two matrices A and B,
us precise answers for all the observables of a system,
unlike mere numbers,
lneasured siluultaneously? (We will allow for instrumen-
tal precision and errors of observation.) The answer is do not commute with
an elnphatic 'No!' It turns out that certain pairs of ob- each other.

-RE-S-O-N-A-N-C-E--I-A-U-9-U-st--2-0~O-4------------~-------------------------------~-
GENERAL I ARTICLE

servables simply cannot be simultaneously measured to

an arbitrarily high level of precision. The source of the
problem is neither observational error nor the least count
of the instruments used, but quantum mechanics itself!
It stems from the fact that, in general, two n1atrices A
and B, unlike mere numbers, do not com,m,ute with each
other: the comm,utator A B - B A == [A, B] =f. O. This
fact has far-reaching implications. Chief among them is
the uncertainty relation due to Heisenberg, subsequently
called the Uncertainty Principle (apparently for the first
time by A Eddington in his Gifford Lectures in 1928),
because of its profound consequences.

The Uncertainty Principle

The Uncertainty Principle was discovered by Heisenberg

when he attempted to understand the so-called 'Trans-
formation Theory' proposed by P A lVI Dirac and Jor-
dan. This theory gives a cogent picture of quantum me-
chanics using linear vector spaces. It clarifies the roles
of both wave mechanics and matrix mechanics, and is
essentially the modern formulation of quantum mechan-
ics.

Heisenberg investigated the question of measurenlent in

Heisenberg realised this framework. He realised that genuine, intrinsic, un-
that genuine, intrinsic, certainties or imprecisions cropped up in the sin1ultane-
uncertainties or ous measurement of the position and linear momentum
imprecisions cropped of a moving particle, and also in the simultaneous mea-
up in the surement of energy and time. He attempted to explain
simultaneous this novel feature through a gedanken or thought experi-
measurement of the Inent, which uses a hypothetical gamma-ray microscope
position and linear to observe electrons. His original argument, however,
momentum of a is not part of our current understanding of the actual
moving particle, and Uncertainty Principle, for it treats interactions between
also in the quantum objects somewhat unrealistically, analogous to
simultaneous mechanical collisions of classical particles. Nevertheless.
measurement of it was Heisenberg's seminal paper in 1927, written after
energy and time. serious discussions with Bohr and Wolfgang Pauli, that

-52------------------------------~-------------RE-S-O-N-A-N-C-E-I--A-U9-U-s-t-2-0-0-4
GENERAL I ARTICLE

Box 5. Infinite-dimensional Matrices

For allY two ('II x 'II) matrices Rand S, Trace (R S) = Trace (S R). This property (called the
'cyclic illvariallce of the trace') Ileed Hot hold good for infinite-dimensiollal matrices like
X a lld Px. Takillg the trace OIl both sides of eqllatioll(l), Trace (X Px) - Trace (P x X) =
'i ii, Trace (IT). B11t the RHS does HOt vanish ill fact, it is infinite in magnitude!

disclosed to the world the existence of the quantum un-

certainty relations.

The first mathematically exact derivation of the Uncer-

tainty Principle for position and momentum seems to
have been given in 1927 by E H Kennard. It hinges
crucially on the fact that

[X Pxl = i17JI, (1)

where ri, == 17,/ (27r) and i == J=T. The x-components

of the position and momentum of a particle are eigen-
values of the operators X and Px respectively. TI is the
identity operator (Box 5). T~e commutator algebra in
(1) is associated with a rich underlying group-theoretic
structure. This has wide applications, most notably in
quantuln optics.

Kennard's derivation pertains to the variances in the

simultaneous measurements of the observables X and Px'
These variances are defined as

(~X)2 == ((X - (X) )2) and (~P x)2 == ((P x - (P x) )2) The commutator
(2) algebra in equation (1 )
is associated with a
Suppose these Ineasurements are made exactly once on rich underlying group-
each copy of the system in an ensemble. Kennard showed theoretic structure.
that it follows from (1) that the product ~X ~P x satis- This has wide
fies applications, most
notably in quantum
~X ~P x >- ~2 (3) optics.

-R-ES-O-N-A-N-C-E--I-A-U-9U-s-t--2-00-4------------~------------------------------5-3
GENERAL I ARTICLE

Box 6. Variance and Higher Moments

TIle variallce is a measure of the departure from the average value of the outcomes ill a
series of llleasurements. Its positive square root (~X or ~ P x ill our case) gives the spread
of the ontcomes auont the average. Clearly, if we conduct N trials ill which we llleasnre
an observable A and the OutCOlllE; is (J, in (~(J,rh of them, the average val1le is just a, alld
~A = O. Tllis happens if the system is in an eigenstate of A. Now (~A)2 call also be
writtell as (A 2) - (A)2. Higher-order couIlterparts of the variance involve the 1ll0l1lellts
(A/,;) where k > 2. Generalizations of equation (3) have been derived s1lbsequently for
these qualltities. We note that any distriuution of outcomes is completely known only if,
apart from the mean and variance, all non-vanishing higher moments are al:-;o knowll.

for every possible normalized state of the particle. This

is the position-momentum uncertainty relationship (Box
6). Analogous relations hold good for the other two
Cartesian components of the position and momentum
of the particle.

The actual value of the uncertainty product (the LHS

in (3)) depends on the state of the system. Minimum
uncertainty states (states for which .6.X.6.P x = 17/2) are
especially interesting, because they are closest to classi-
cal states. (Recall that the classical limit corresponds to
n --of 0.) A 'coherent state' is an example of a minimum
uncertainty state. Ideal single-mode laser light provides
a physical realization of such a state.

The Kennard inequality was generalized in 1929 by

H P Robertson to any pair of 0 bservables A and B of a
system. He showed that

6.A 6.B ?: ~ I ( [A, BJ) I (4)

A 'coherent state' is
an example of a for any state of the system. The relation was generalized
minimum uncertainty further by Schrodinger in 1930 to
state. Ideal single-
mode laser light
provides a physical
realization of such a
state.
(5)

--------------------------------~-------------R-E-SO--N-A-N-C-E-I-A-U-9-U-S-t-2-0-0-4
54
GENERAL I ARTICLE

Box 7. Uncertainty Principle and Cells in Phase Space

TIle state of a classical particle lIloving freely inside a box is flIlly known if its positioll
allel UlOlllentmll are given. But then all infinite lllllllber of sHch states are available~ from
wllich it follows tllat the (], pn:()',.i probability that the particle is ill any particular state
vallishes! This is absnrd, as the particle is definitely inside the box. The resolutioll COUles
through the Ullcertainty Principle which associates spreads like ~X alld ~p x witl! every
cOlllponellt of the position and 1ll0lllelltlllll. This reduces the lllllllber of accessible states
to a finite Illllllber, so that the correspondillg probabilities are 1l0ll-~ero. This is the
origin of the concept of cells in phase space llll<.1erlying the cOlInting procedures Ilsed ill
statistical physics.

Here {A, B} denot.es t.he anti-commutator (A B + B A),

and all averages are with respect to a given state. For
the special case in which A is the position and B is the
lllomentum, it can be checked that, for a coherent state,
the second term in the RHS of (5) vanishes, and we
obtain the Ininimum uncertainty relation ~X ~p = 17,/2.

Equations (3 )-( 5) are mathematically unambiguous. They

are therefore the precise statements of what Heisenberg
attelnpted to establish semi-quantitatively through his
gedanken experiment. They have far-reaching implica-
tions. For instance, the position-momentum uncertainty
relation has played a pivotal role in statistical physics
(Box 7). Similarly, the energy-time uncertainty relation
has played a central role in quantum field theory - a
subject that deals, among others, with the interactions
between elementary particles (Box 8).

Box 8. Energy-Time Uncertainty Principle

The ellergy-tillle ullcertainty relation states that the product of the energy spread ~E
aut! the spread ill tillle ~t of a process also satisfies ~E ~t 2: 17/2. However, ulllike po-
sitioll, 11l0l1lentlllll, euergy, etc., time is a parallleter alld uot all observable ill quaIlt1Ull
lllecliallics. Theinterpretatioll of the energy-time lIIlcertaillty relatioll is therefore SOllle-
what different from that of other 1I1lcertaillty relations. Whell applied to an ullstable
particle, tllis ll11certaillty relation yields illforlllatioll on the half-life of the particle. III
qualltulll field theory, it allows for the so-called virtual processes that are required to
explaill experimental results OIl reactions involving elementary particles.

-R-ES-O-N--A-N-C-E-I-A-U-9-U-st--2-0-0-4------------~~------------------------------"-
GENERAL I ARTICLE

Heisenberg has thus played a major role in the march

of modern physics. To quote an extract from the Nobel
presentation speech by H Pleigel, ". Your quantum
mechanics has created new concepts, and has led physics
into fresh trains of thought, which have now already
proved of fundamental importance for our know le.dge
of the phenomena of physics. The Royal Academy of
Sciences has awarded you the Nobel Prize for Physics
for 1932 in recognition of these studies."

Heisenberg's fellow campaigners in the march, Schrodin-

ger and Dirac, shared the Physics Nobel Prize for 1933.

Suggested Reading

[1] G Gamow, Thirty Years That Shook Physics: The Story of Quantum
Address for Correspondence Theory, Dover Publications, New York, 1966.
5 lakshmibala [2] J Mehra and H Rechenberg, The Discovery of Quantum Mechanics,
Department of Physics Springer-Verlag, New York, 1982.
Indian Institute of Technology [3] M Jammer, The Philosophy of Quantum Mechanics, Wiley, New York,
Madras, (hennai 600 036 1974.
India.
Email:[email protected]

To what extent are we bound by tradition in

the selection of our problems? ... Looking
back upon history in this way, we see that we
apparently have little freedom in the selection
of our problems. We are bound up with the
historical process, our lives are part of this
process, and our choice seems to be restricted
to the decision whether or not we want to
participate in a development that takes place
in our time, with or without our contribu-
tion .... One may say that a fruitful period is
characterized by the fact that the problems are
given, that we need not invent them. This
seems to be true in science as well as in art.

-Werner Heisenberg

5-6---------------------------------~~~-----------R-E-S-O-N-A-N-C-E--I-A-u-g-u-st--2-0-0-4