You are on page 1of 5

Volume 76.

number 2 OPTICS COMMUNICATIONS I 5 April I990

INFORMATION THEORY APPROACH


TO QUANTUM NON-DEMOLITION MEASUREMENTS

A. VOURDAS
Departmeni qfElectrical EngInewing and Eleclronics, L’nrvers~tyof’liverpool. P.O. B~.Y 147. Llwrpool L69 J’BX. L’K

Recewed 18 August 1989: revised manuscript received I3 December I989

Probabilities and entropies are introduced for the description of two successive measurements. They are used to define the
concept of low-demolition measurements which is weaker but also less restrictive than the concept of non-demolition measure-
ments. These ideas are exemplified in the case of the double Stern-Gerlach experiment.

1. Introduction than zero then according to their value we charac-


terise the measurements as low or high demolition
The study of the quantum theory of measurement measurements. In this way we extend the strong but
[ 1 ] is important for theoretical and practical rea- very restrictive concept of quantum non-demolition
sons. From a theoretical point of view it offers an into the weaker but less restrictive concept of low
insight into the fundamentals of quantum theory. demolition, which allows us to choose measurements
From a practical point of view it is important in sub- with as low demolition as it is required. within the
jects like quantum optical communications [ 21 and accuracy of a particular experiment.
quantum electronics. Applications of new types of
quantum light (e.g. squeezed states or number ei-
genstates) into those areas require the use of the 2. Quantum probabilities
quantum theory of measurement. Particularly im-
portant for optical communications is the language Let A,. A2 be two observables (hermitian opera-
of entropy and information [ 3,4] and we have re- tors) and i,!/ ). 2.i;’ their eigenvalues and / uk!’ ).
cently presented some calculations in this direction / u,G) ) there eigenstates correspondingly. n;. ,M can
in quantum context [ 5 1. But from a theoretical point be discrete or continuous variables: here we assume
of view also, the quantity of entropy has provided that they are discrete variables. We call zA! ), n.G) the
new insight into deep problems like the uncertainty projection operators j 14.i:) ) (u,i! ) 1, 1uij) ) ( u ii’ /
principle [ 61 and the Einstein-Podolsky-Rosen ex- respectively. We consider a quantum system de-
periment [ 71. It is therefore clear that an under- scribed by a density matrix p (Trp= 1 ). The mea-
standing of the properties of entropy in the context surement il, on this system will give the result $,’
of the quantum theory of measurement is essential. with probability
In this paper we shall consider two successive mea-
P(i~~‘)=Tr[n,1!)P7C~~‘]=(U6~)IPIU~~)) (1)
surements and introduce the entropy based quan-
tities R, R, that provide a measure of the unpre- If the measurement A, gives the value A,&“, the
dictability of the second measurement, in the case system after the measurement will be described bq
that the result of the first measurement is known. If the density matrix 71,’) This is the well-known “col-
they are equal to zero then the second measurement lapse of the wave-function” and plays an important
is completely predictable from the result of the first role in von Neumann’s theory of measurement. Al-
measurement i.e. we have a quantum non-demoli- ternative formulations of the quantum theory of
tion measurement (e.g. [ 81 ). If they are different measurement have been discussed in refs. [ 91. A

164 0030-4018/90/$03.50 0 Elsevier Science Publishers B.V. (North-Holland)


Volume 76, number 2 OPTICS COMMUNICATIONS 15 April 1990

general measurement /1, will change the density ma-


trix of the system from p to p’

(2)
=p’(n,,n~:‘)=p(~~:‘)-6,(1,2). (8)

We see that the right hand side of eq. (8) is equal to


We next consider the following two experiments: p’(A,, ,I$)) and is not (in general) equal to
In the experiment A, only the measurement A, is p(Ag) ). The physical interpretation of this is that in
performed at time tz, on a system described initially experiment B we can make the statement that the
by the density matrix p. The probability that the re- value of the observable /ii at the time t, is either
sult is Ati), is AI’) or LJ’) or J.4’) etc. in experiment A, due to
quantum interference, we cannot make this state-
P(~G’)=(~G’IPl~M’> . (3) ment. This situation is similar to the well-known two-
In the experiment B the successive measurements slit interference experiment in quantum mechanics.
A,, A, are performed at times t,, t2 (tl < tz) on a sys- If we do not observe the electron paths we have
tem described initially by the density matrix p. The quantum interference and we cannot make the state-
probability that the outcome of the second mea- ment that the electrons either go through the first hole
surement is A$), is or they go through the second hole; alternatively we
can look at the paths and destroy quantum interfer-
ence. Our p(lG)), p’ (A,, A$:‘) describe quantita-
tively these two cases in our own context.
(4)
We should point out that the concepts of joint and
conditional probability, defined above in a quantum
We have taken here into account the fact that the first
context, are slightly different from those in a clas-
measurements transform the initial density matrix p
sical context. We see for example in eq. ( 8 ), that the
into the density matrix p’ of eq. ( 2). It is easy to see
xNp(Ati’, ng’) is equal to p’(A,,Ag’) but is not
that if [A,, A,]=0 (or in a classical system)
(in general) equal to p(Ag)). The p’ (A,, ,tg)) has
p(A$‘)=p’(A,, ,ig)). In general these two proba-
been introduced above in order to create a mathe-
bilities are different from each other. We can easily
matical structure for the quantum joint and condi-
prove that
tional probabilities, similar to the one of classical
s,(l,2)=P(~~:‘)-P’(/1,,1~,2’) systems. In this way we can easily generalise in the
next section, the existing entropy theorems for clas-
= ,c, (U~“l~fc’)>(~~l’IPl~J!‘)(~~‘I~~2’) > sical systems, into quantum systems. The distinction
K;-T betweenp(llG)) andp’ (A,, AZ,“) is fundamental for
quantum systems.
~6,(1,2)=0. (5)
We next consider experiment B and introduce the
joint probability p(Ag), A$)) that the results of the 3. Entropy approach to successive measurements
measurements A,, /1* will be Lg’, ng) and also the
conditional probability p(A$) Id#‘) that the result The entropy of a quantum system [ lo] described
of A2 will be Ag’ provided that the result of L!, is by the density matrix p is given by ( kB = 1, ZI= 1)
ny,.
S= -Trplnp. (9)
However, other entropies can also be used to de-
scribe certain probability distributions characteriz-
(6)
ing the system. For example the

S(2) = - C p(Ag)) lnp(Ag)) (10)


in experiment A provides a measure of the uncer-

165
Volume 76, number 2 OPTICS COMMUNICATIONS 15 .April 1990

tainty in the result of the ii1 measurement. Similarly


[A,,A,]=OeR=O, (19)
the
[A,, A21 =OsR, =0 , (20)
S’(l,2)=- C~‘(il,,i.~?))Inp’(il,,~~~)) (11)
[/I,.il,]=o+R~=l. (31)
in experiment B, provides a measure of the uncer-
tainty in the result of the A2 measurement. We also The S( 2 I 1) is a measure of how unpredictable the
define the entropies result of the second measurement is, if the result of
the first measurement is known. We use it here for
Iv,.!4 the study of quantum non-demolition measurements.
Two successive measurements are quantum non-
(12)
demolition measurements. if the result of the second
measurement is completely predictable from the re-
sult of the first measurement, i.e. if S( 2 I 1 ) = 0. Eq.
.A( ( 14) shows that this happens if and only if [ill,
(13) ,4,] =O. The ratio R compares the unpredictability
of the result of the second measurement in the case
Taking into account our discussion in the previous
that the result of the first measurement is known (i.e.
section. we easily show that
the S( 2 11 ) ). with the unpredictability of the result
[il,,iiz]=O~S(2] l)=O, (14) of the second measurement (given by S’ ( 1. 2 ) ). It
can be used to introduce the concept of low-demo-
[~,,~2]=o~S(2)=S’(1,2). (15) lition measurements where the result of the second
In the case of eq. ( 14) the inverse statement is true measurement is “almost predictable” from the first
i.e. the S( 2 11) = 0 implies that all S( 2 ],I$ ’ ) = 0 and measurement. By “almost predictable” we mean that
use of eq. (7) shows that I u,(,” ) = (u,$?)) which the value of R is near to zero. Successive measure-
proves that the ii,, ,4, commute. In the case of eq. ments with value of R near to one are high-demo-
( 15 ) the inverse statement is not true and in the ex- lition measurements.
ample of section 4 we shall see that it is possible to The ratio R, which compares the S( 2 I 1) (for the
have S(2)=S’(l. 2) although the /i,, ,4, do not experiment B) with the S(2) (for the experiment
commute. A) could also be used to characterize the demolition
It is known that in classical systems (e.g. [4] ) of two successive measurements. According to
whether R, takes values near zero or higher values
S,,(2ll)~S,,(2)~S,,(1,2). (16) we characterize the measurements as low or high
demolition measurements correspondingly.
The proof makes use of the fact that Ip(/l.j,!),
Note that the base of the logarithms in the entro-
?.s)) is equal to p(;l,iF)) in classical systems. In
pies, defines the units (2 for bits; e for nats). The
quantum systems it is equal to p’ (/I,. l.,g’) and the
ratios of eq. (18) are however. independent of the
same proof easily gives
base.
S(2]l)~S’(1,2)~s(1,2). (17)

We introduce the quantities 4. Example


R =S(2Il), R =S’(L2)
I We now consider the double Stern-Gerlach exper-
S(2) 2 S(2) -
iment [ 1 1 1. Particles with spin f travel in the Jj-di-
&+wL (18)
rection through a non-uniform magnetic field in the
2 S’( 1.2) . :-direction. The beam splits into two parts and we
K, <HZ. O<R,il .
let one of those parts through a second non-uniform
magnetic field in a new direction which we call z’.
Eqs. (14), (15) can be rewritten as We assume that Z’ lies on the .YZplane and we call

166
Volume 76, number 2 OPTICS COMMUNICATIONS 15 April 1990

6 the angle between z, z’ . In this experiment Ai, /12


are the successive measurements of the spin com-
ponents s,, s: correspondingly. We call 11 + ) , I 1 - )
the “up” and “down” eigenstates of S, and 12 + ),
12- ) the “up” and “down” eigenstates of s:. Let us
assume that the incident beam is described by the
pure state Is) and let us call s+, s_ the amplitudes

s+=(sll+), S_=(S(l-). (22)


We also call y the argument of the complex number
S-S: (Y=arg (s-s*+)). It is clear that

P(l+)=l~~11+)12=I~+12,
p(1-)=l(S11-)12=IS_12, (23)
Fig. 2. R, R,,R2 as a function of 0 for Is+I*=Is_I’=O.5 and
-is+ 12+ IS_ 12=1 .
y=45.
A standard quantum mechanical calculation gives

p(2-Il+)=l(l+l2-)12=sin28/2,

p(2+ll+)=l(1+12+)12=cos2e/2,

p(2-~1-)=1(1--12-)12=c0s%/2,

p(2+11-)=l(l-12+)j2=sin28/2, (24)
p(2+)= I(s12+)I2=Is+c0se/2+s_ sin8/212,

p(2-)=l(sl2-)[*=I--S+sinB/s+s_c0se/21Z.

Wenowuseeqs. (ll), (13) toget

s(21 i)=-[C0s2(e/2)lnC0s2(e/2)
0.2 0-L 0.6 0.8 ‘*O I S,12
+ sin2(e/2) In sin2(e/2)] ,

S’(1,2)=-K,lnK,-K,lnK,, (25)
K,= Is,I2sin28/2+ Is_ I2 cos28/2,

K2= IS_ I2 sin28/2+ IS+ I2 cos2t?/2.

In figs. 1, 2 we plot the R,R,,R2 as a function of


8inthecase ~s+~2=l~_12=0.5andy=0 (fig. 1) and
y=45 (fig. 2). In fig. 3 we plot R,R,,R2 as a func-
tionof ls+12fory=45and8=20.1nfigs. 1,2wesee
that for f&O we have non-demolition measure-
ments; for low values of 8 we have low-demolition
measurements; and for values of 6 near 90” we high-
Fig. I. R, RI,RZ asa function of 8 for Is+ 12= Is_ 1*=0.5 and demolition measurements. In fig. 3 we see that in the
y=o. region 0.3 < Is+ I 2< 0.7 we have low-demolition

167
Volume 76. number 2 OPTICS COMMUNICATIONS I5 April 1990

measurements. We also see that it is possible to have [ 41 A.M. Yaglom and I.M. Yaglom, Probability and information
(D. Reidel Publ., Dordrecht, 1983):
non-commuting measurements with R2= 1 and
F.M. Reza. An introduction to information theory (Mc-
therefore the inverse of the statements of eqs. ( 15 ), Graw-Hill, New York, I96 I ).
(2 1) is not true. [ 51A. Vourdas, Phys. Rev. A37 ( 1988) 3890; Phys. Rev. A39
We have introduced in this paper the quantities R, ( 1989) 206; IEEE Trans. Info. Theo.. to appear.
R, that provide a measure of the demolition in two [ 61 D. Deutsch, Phys. Rev. Let 50 ( 1983) 63 I;
M.H. Partovi. Phys. Rev. Let 50 ( 1983) 1883:
successive measurements and can be used to define
I. Bialynicki-Birula and J. Mycielski. Corn. Math. Phys. 44
low-demolition measurements. (1975) 129.
H. Maassen and J.B.M. Uffnk, Phys. Rev. Let. 60 ( I988 )
1103.
[ 71 S.L. Braunstein and CM. Caves, Phys. Rev. Let 61 (1988)
References 662.
[S] V.B. Braginsky. Y.1. Vorontsov and K.S. Thorne. Science
209 ( 1980) 547:
[ I ] J. Von Neumann, Mathematical foundations of quantum C.M. Caves, KS. Thorne. R.W.P. Drever, V.D. Sandberg
mechanics (Princeton Univ. Press, Princeton, 1955); and M. Zimmermann, Rev. Mod. Phys. 52 ( 1980) 34 I.
G. Ludwig, Die grundlagen der Quantenmechanik [ 91 H. Margenau, Phil. Sci. 30 ( 1963 ) 1; Ann. Phys 23 (1963)
(Springer, Berlin, 1954); 469;
A.S. Holevo, J. Multivar. Anal. 3 ( 1973) 337; E.B. Davies and J.T. Lewis, Corn. Math. Phys. 17 (1970)
A.S. Holevo, Probabilistic and statistical aspects of quantum 239;
theory (North Holland, Amsterdam, 1982); E.B. Davies, J. Functional. Anal. 6 (1970) 318;
G. Lindbland. Corn. Math. Phys. 33 (1973) 305; P.A. Benmoff, J. Math. Phys. 13 ( 1972) 1347.
J.A. Wheeler and W.H. Zurek, eds., Quantum theory and [ lo] A. Wehrl, Rev. Mod. Phys. 50 ( 1978) 22 I.
measurement (Princeton Univ. Press, Princeton, 1983 ). [ 1 I ] D. Bohm, Quantum theory (Prentice-Hall. New York.
[2] C. W. Helstrom, Quantum detection and estimation theory 1951);
(Academic Press, London, 1976); R.P. Feynman. R.B. Leighton and M. Sands, The Feynman
R.O. Harger, ed., Optical communication theory (Dowden. Lectures on Physics, Vol. 3 (Wesley, Reading Mass.. 1965).
Hutchinson, Ross, Pennsylvania 1977)
[3] C.E. Shannon and W. Weaver, Mathematical theory of
communication (University of Illinois Press, Urbana, 1949 )

168

You might also like