Relations among elements of the density matrix. I.

Definiteness inequalities
William Band and James L. Park*
Washington State University, Pullman, Washington 99163
(Received 6 June 1972; revised manuscript received 16 November 1972)

The statistical operator of quantum theory may be determined empirically by computations based
upon the measured mean values of a set of observables we have called a quorum. The requirement
that a statistical operator be positive semidefinite is then used to generate a family of inequalities
connecting these quorum means. Like the simpler uncertainty relations, these inequalities are
universal, valid for all quantum states. In the special case of pure states, the method yields a family
of equalities.

1. THE QUORUM CONCEPT 2. DEFINITENESS OF THE STATISTICAL MATRIX

According to quantum mechanics, every reproducible Three independent defining properties are customarily
state preparation scheme II is characterized by a statis- attributed to the statistical operator:
tical operator p in the sense that
(i) hermiticity,
Tr(pA) = (A), ( 1) (ii) unit trace,
where A is the Hermitian operator for an observable of (iii) positive semidefiniteness.
interest and (A) denotes the arithmetic mean of data for
that observable gathered from an ensemble of systems Characteristics (i) and (ii) have already been incor-
each prepared in the manner II. Recently we have ex- porated into the quorum theory; every matrix representa-
plored 1 ,2,3 the problem of empirical state determination, tion of p which satisfies (1) and whose elements are
formulated as follows: given II and the means to mea- functions of quorum means will automatically be Hermi-
sure any A, how much data is needed in order to deter- tian, and of trace unity.
mine the unknown p? This problem has been attacked in Property (iii) may be derived 7 from the consistency con-
the past by several authors, including Feenberg,4

11 December 2024 21:10:53

dition that a dichotomic observable, represented by a
Kemble,5 and Gale, Guth, and Trammell. 6 projector 1 cp) (cp 1 onto a Hilbert vector cp, must have a
For an N-dimensional Hilbert space, any matrix repre- nonnegative mean value since the eigenvalues of the pro-
sentation of relation (1) contains N2 - 1 independent jector are 0 and 1. Thus,
real unknowns in the statistical matrix p (also commonly
called the density matrix). This is a consequence of the Tr(plcp)(¢I) = (¢ Ipl¢) ~ o. (2)
Hermiticity of p and of its unit trace. The unknowns
occur linearly; hence, if N2 - 1 observables {A} are But ¢ is arbitrary; hence by definition p is positive semi-
ohosen so that the associated N2 - 1 linear algebraic definite, or nonnegative definite.
equations like (1) possess a unique solution set, then the
It follows that the quorum means of which statistical-
elements of the p matrix may be determined in terms of
matrix elements are functions must be interrelated in
the N2 - 1 mean values {(A)} by standard methods for
such a manner that the statistical matrix will be non-
solving linear systems of equations.
negative definite. Such a connection among the quorum
We have elsewhere called a set of observables {A} observables is established by application of the old
whose mean values {(A)} constitute sufficient informa- algebraic theorems which states that all prinCipal minor
tion to deduce the statistical operator p a quorum of determinants of a nonnegative definite matrix must be
observables. nonnegative.
In the present paper it is sufficient to acknowledge To be explicit, consider an N x N statistical matrix p
simply that such quorums exist, that the statistical with (k, 1) element PkZ' An n-dimensional prinCipal
operator p may be expressed as a function of quorum minor matrix is obtained by striking out N - n rows and
means {(A)}. It will be demonstrated below that such their corresponding columns; thus, the common element
representations of p, when considered in the light of an of each struck row-column pair will be in the principal
old theorem in matrix algebra, permit us to generate diagonal of p. The standard proof that the determinants
families of quantal inequalities reminiscent of, but more of these minor matrices are all nonnegative is based on
elaborate than, the uncertainty relations. In the sequel the Hermiticity of the quadratic form (2) and on the in-
(Paper II) we shall investigate a new class of conserved variance of determinants under Similarity transforma-
quantities which are revealed by the study of quorums. tions.
The present authors have developed systematic pro- Since each PkZ is a function of quorum means {(A)}, the
cedures for the construction of quorums for physical nonnegativity of prinCipal minor determinants is ulti-
systems with N-dimensional Hilbert spaces and, under mately expressible as a family of inequalities involving
certain Circumstances, for systems with infinite-dimen- the quorum means. Like the celebrated uncertainty
sional Hilbert spaces. relations, these definiteness inequalities are valid for all
preparations of state.
Directly verifiable illustrations of denSity matrices as
functions of quorum means will be given below; the The family of definiteness inequalities becomes a family
reader interested in the philosophical and mathematical of equalities for n > 1 whenever the preparation is pure.
origins of the quorum theory is referred to the literature For a pure state, p is a projector 11/1) (1/1 I; in terms of
cited earlier. matrix elements,

551 J. Math. Phys., Vol. 14, No.5, May 1973 Copyright © 1973 by the American Institute of Physics 551
552 w. Band and J. L. Park: Relations among elements of the density matrix. I 552

(3) (Pk) is the probability that a ak -measurement will yield

+1.
A typical minor determinant of a pure statistical matrix
Statistical matrix:
will therefore have the form
e b1b2.. • bn (lJIa1IJIb*)(1JI IJI *) .•• ("' IJIbtl*) P) (~) - i) - i(P) - i)j
1 "2 b 2 'l'an (p) =( y

(4) (p x) - t) + i(P') - i) 1- (J~) •

(11)
Definiteness inequalities:
Since the e-system is totally skew-symmetric and
:lJl
IJIb b*2" 'IJI:" is completely symmetric, the minor deter- (1) One-dimensional minors:
minant vanishes, provided n > 1. Hence for a pure
Os (p) s 1. (12)
statistical matrix all principal minor determinants with
n > 1 vanish; we can therefore generate a family of de-
(2) Two-dimensional minor (detp):
finiteness equalities relating the various quorum means
in any pure state.
(Px ) + (p) + (P~») - (Px )2 + (p)2 + (P)2 ~ t. (13)
3. ILLUSTRATIONS
C. Harmonic oscillator with 2-level energy cutoff
Several expressions are given below for matrix elements
of the statistical operator expressed as functions of Quorum theory is readily applicable to systems with
quorum means. As noted above, the procedures 9 used to infinite-dimensional Hilbert spaces whenever it is known
discover quorum observables will not be reproduced that the state preparation II has this property: there is an
here, nor will the straightforward but sometimes lengthy 0?sez:vab.le C (t~e cutoff observable 10 ) whose probability
dlstributlon varushes except for a finite number n of C-
algebra by which the matrix elements are obtained from
systems of linear equations. There is, however, no need eigenvalues. Thus in a representation diagonal in C the
(~inite) statistical matrix will have only n nonzero
for the reader to accept the matrix elements on faith·
diagonal elements. From the inequalityll
their validity may be checked directly by using (1). '
(14)
A. Spin-% system
Quorum: ax' ay, a ~ (standard Pauli matrices). valid for any positive semidefinite p, it then follows that

11 December 2024 21:10:53

all elements of the statistical matrix vanish except for
Statistical matrix: an n x n submatrix.
In the present example, the cutoff observable is energy
(5) Hand n = 2; specifically, only the two lowest energy
levels have nonzero probability.
Quorum: x (position),p (momentum),
Definiteness inequalities:
(1) One-dimensional minors: H = (P2/2m) + (mw 2/2)x 2 •
1 + (a) ~ 0, (6) Statistical matrix;
Let Pc denote the 2 x 2 nonzero submatrix of p. Then
The relations (6), involving only one component of the
polarization vector (a), are uninteresting since they con- 3 - (IQ
vey no information not already obvious from the (X) - i(P») ,
(15)
spectrum {- 1, + I} of a ~ • Pc = t ( (X) + i (p) - 1 + (K)
(2) Two-dimensional minor (detp): where
t(l - (a)2 - (a)2 - (a y )2) ~ 0
(7)
K =(2/tiw)H, X =(2mw/ti)1/2x, P =(2/mnw)1/2p.
or (16)
(ax)2 + (ay)2 + (a)2 s 1. Definiteness inequalities:
Pure state definiteness equality: (1) One-dimensional minors:
(a,)2 + (ay )2 + (0 ~)2 = 1. (8)
or (17)
By subtracting (8) from the familiar equation
tiw/2 s (H) s 3tiw/2.
(ax2) + (a:}) + (a~2) = 3, (9)
Relation (17) is expected for a harmonic oscillator cer-
we obtain the following relation among uncertainties for tain to yield upon energy measurement one of its two
any pure spin-t state: lowest eigenvalues.

(aax )2 + (aa y )2 + (a a .. )2 = 2. (10) (2) Two-dimensional minor:

8(x), (p») s 2(H) - (tiw)-1(H)2 - (3/4)nw (18)
B. Spin-% system (alternative quorum)
where
Quorum: Px ' Py ' ~, where Pk denotes the projector onto
the a k -eigenvector belonging to eigenvalue + 1, etc. 8(x), (p») = (p)2/2m) + (mw 2/2)(x)2.

J. Math. Phys., Vol. 14, No.5, May 1973

553 W. Band and J. l. Park: Relations among elements of the density matrix. I 553

The expression (18) is of interest in connection With the However, in higher dimensional spaces a matrix of unit
classical limit problem since 8 is the classical energy trace may satisfy this inequality and yet fail to be non-
function with quantal means (x> and (p) as arguments; negative definite. Thus our approach will in general pro-
thus, (18) reflects a basic disparity between classical duce additional inequalities which are not derivable from
and quantal energy concepts. We intend to investigate (19).
the quorum theory approach to the classical limit prob-
lem in another publication. D. Spin-1 system
The foregoing illustrations in a two-dimensional Hilbert Quorum: Jx,JpJ z (angular momentum components),J;,
space yielded several inequalities derivable also from J;,JX.)l,Ju ' wnereJab == JaJb + JbJa • A discussion con-
the well known relation cerning measurement of quorum members Jab is given in
Ref.3.
Trp2 s 1. (19)
Statistical matrix (representation diagonal in J .. ; Ii = 1):

1 + !\(J.. ) (1/2v2)[(Jx ) + (Jax ) !\(Jx2) - (Jy2)

- (J}> - (J;» - i «JJ'> + (Jy ",>)] - i(Jxy »

(1/2v2)[ (Jx > + (J u) (1/2v2)[(Jx ) - (Jzx)

(p) = - 1 + (J}) + (J;) (20)
+ i«Jy ) + (Jy ",»)] - i«J) - (Jyz »]

!\ (J2)
x
- (J2)
y (1/2v2)[(Jx ) - (Ju ) 1- !\(J.. >
+ i(Jxy » + i «Jy ) - (Jy ",))] + (J;) + (J}»

Definiteness inequalities:
The expression (25) typifies the complex interconnec-
(1) One-dimensional minors: Consider for instance the tions among the means of quantal observables that may
(2,2) element, which yields be discovered by generating definiteness inequalities.
Moreover, the equality case of (25) is illustrative of the
(21) definiteness equalities which link quorum means for
systems prepared in pure quantum states.

11 December 2024 21:10:53

This is expected, since for spin-l we know that
(J2)
x + (J2)
.)I + (J2)
.. =2 (22) *Work supported by Research Corporation.
lW. Band and J. L. Park, Found. Phys. 1, 133 (1970).
and 2J. L. Park and W. Band, Found. Phys. 1, 211 (1971).
(23) 3W. Band and J. L. Park. Found. Phys. 1, 339 (1971).
4E. Feenberg [thesis (Harvard University, 1933)] considered
(2) Two-dimensional minors: At this level the method one-dimensional wave mechanics only and assumed in effect that p
begins to reveal complicated new relationships among was known in advance to be a projector onto a ray IJi (wavefunction),
the quorum observables that are not antiCipated intui- the problem being to determine 1Ji.
tively. As an example, we compute the upper left minor sE. C. Kemble in Fundamental PrinCiples of Quantum Mechanics
determinant of (20). (McGraw-Hill, New York, 1937), p. 71, attempted to extend
Feenberg's work (Ref. 4) to multidimensional wave mechanics.
6W. Gale, E. Guth, and G. T. Trammell in Phys. Rev. 165, 1434
[1 + ~(J) - (J}) - (Jy2»][- 1 + (Jx2) + (Jy2)]
(1968) corrected an error in Kemble (Ref. 5) and developed an
- H«Jx ) + (Ju »2 + (".)I) + (Jy»2] ;?: O. (24) approach extensible to general density matrices.
7J. von Neumann, Mathematical Foundations of Quantum Mechanics,
The first term in (24) may be simplified by applying (22) trans!. by R. T. Beyer (Princeton U. P., Princeton, N.J., 1955), p. 317.
to obtain 8F. Hohn, Elementary Matrix Algebra (Macmillan, New York, 1964),
2nd ed., p. 353.
!\(J.. ) + (~2})(1 - (J:» 9Refs. 2, 3.
IORef. 3.
- t[«Jx ) + (Jax »2 + «J.)I) + (Jy»2] ;?: O. (25) llRef. 7, p. 101.

J. Math. Phys., Vol. 14, No.5, May 1973