Академический Документы
Профессиональный Документы
Культура Документы
We seek a quantum-theoretic expression for the probability that an unstable particle prepared initially in a
well defined state p will be found to decay sometime during a given interval. It is argued that probabilities
like this which pertain to continuous monitoring possess operational meaning. A simple natural approach
to this problem leads to the conclusion that an unstable particle which is continuously observed to see
whether it decays will never be found to decay! Since recording the track of an unstable particle (which
can be distinguished from its decay products) approximately realizes such continuous observations, the
above conclusion seems to pose a paradox which we call Zeno's paradox in quantum theory. The relation
of this result to that of some previous works and its implications and possible resolutions are briefly
discussed. The mathematical transcription of the above-mentioned conclusion is a structure theorem
concerning semigroups. Although special cases of this theorem are known, the general formulation and
the proof given here are believed to be new. We also note that the known "no-go" theorem concerning
the semigroup law for the reduced evolution of any physical system (including decaying systems) is
subSllmed under our theorem as a direct corollary.
756 Journal of Mathematical Physics. Vol. 18. No.4. April 1977 Copyright © 1977 American Institute of Physics 756
The concept of continuous observation would indeed The mathematical transcription of Zeno's paradox is
be bereft of any physical meaning if it could be estab- a structure theorem concerning a class of strongly con-
lished that the fundamental constituents of the real world tinuous semigroups. This theorem is formulated and
and the interaction between them are such as to exclude proved in Sec. 3 of this paper and may possess some in-
the possibility of arbitrarily frequent observations. But, trinsic interest apart from its application in the present
on the one hand, we cannot claim as final our present context of a theory of continuous observation. As a by-
knowledge of the constituents and interactions of the product of this investigation we find that the known re-
real world. On the other hand, to agree that there is a sult 3 concerning the incompatibility of the semibounded-
limitation on the frequency of observation amounts to ness of the Hamiltonian H with the requirement that
claiming the existence of an elementary and indivisible E exp(iHt) E form a strictly contractive semigroup can
unit of time. Though the existence of an elementary be subsumed under the above- mentioned theorem as
interval of time is an exciting possibility, it is not part one of its direct corOllaries.
of the currently accepted and tested physical theories.
Some of the implications and possible resolutions of
We, thus, feel that the notion of continuous observa- the quantum Zeno's paradox will be briefly discussed in
tion should be accepted, at least for the present, as the concluding section of the paper.
physically meaningful and quantum theory should be
pressed to yield an answer to questions relating the Finally it may be mentioned that the conclusion called
probabilities pertaining to such observations. here the Zeno's paradox in quantum theory has been
noted in some previous works, 4_6 but the present
Continuous observation processes seem to be realized analysis of the problem is carried out in a more general
in practice also, at least apprOXimately, by the tracks of and mathematically rigorous setting than the previous
unstable charged particles in bubble chambers and other works. This, we feel, is not merely a dispensable
detecting media. The observation of the track amounts luxury, but is necessary to locate the preCise assump-
practically to a more or less continuous monitoring of tions on which the "Zeno's paradox" rests.
the existence of the unstable particle and thus a mea-
surement of E during the period of the particle's flight
2. QUANTUM THEORETICAL EXPRESSIONS FOR
through the detection chamber. We are therefore led to
P(O, t;p) AND RELATED PROBABILITIES
accept as operationally meaningful the P(D, l;p), Q(D, t;p),
and R(O, ft, t;p). To be a complete theory, quantum The three probability functions P, Q, R introduced
theory must provide an algorithm for computing these in the previous section relate to the results of continu-
probabilities. ous observation throughout an interval of time. By
In the next section we describe what appears to be their very definitions they must obey the relations
the natural approach to determining quantum-theoretic
expressions for these probabilities. Our investigation
P(O, t; p) + Q(O, t; p) = 1
leads to the paradoxical result mentioned at the begin- and
ning of this section: An unstable particle observed con-
R(O, tt, t;p) = Q(a, tt; p) pea, t - it; Pt),
tinuously whether it has decayed or not will never be
found to decay! Since this evokes the famous paradox of where Pt is the state in which the system (prepared
Zeno denying the possibility of motion to a flying arrow, initially in the state p) finds itself at tt after being con-
we call this result the Zeno's paradox in quantum theory tinuously observed and found to be undecayed throughout
[a, tt). We may therefore concentrate our attention on
In fact, if E is taken to be the projection to the set
calculating Q and Pt.
of localized states of a particle (or, a quantum arrow)
in a given region D of space, then one concludes that We start with the system in the state p and make a
the particle will never be found to arrive in a disjoint series of n + 1 measurements, which are idealized to be
region D' provided it is continuously observed whether instantaneous, at times a, tin, 2t/n, .•• , (n - 1)t/n, and
it has entered D' or not: The" arrow" cannot move to t. We seek the probability Q(~, n;p) that it be found
where it is not! undecayed in each of these measurements. It is natural
This result acquires an even more picturesque and to assume that Q(~;p)=Q(O,t;p) can be evaluated as
paradoxical formulation when it is applied to the "hell- the limit of Q(~, n; p) when n - "", provided the limit
ish contraption" considered in the Schrodinger's cat exists.
paradox. 2 It may be recalled that the contraption con- Let us denote by pen, i) the state in which the system
sists of an unstable (quantum) particle placed in a box finds itself after the (n + 1) measurements at 0, tin,
equipped with an efficient counter and a cat inside a 2t/n, •.• , t and being found to be undecayed in each of
steel chamber. If the particle decays, the counter these measurements. Now, according to the orthodox
triggers and, in its turn, activates a tiny hammer which theory of measurement, if a measurement of E on the
breaks a container of cyanide in the steel chamber. system is carried out yielding the result "yes" (that is,
Monitoring the vital functions of the cat amounts to ob- "undecayed"), then the state of the system collapses to
serving if the particle has decayed or not. In view of a new (unnormalizedl) state p' of the form
the Zeno's paradox formulated above, should we con-
clude that the particle will never decay? Will the cat ,
p' =6 AjpA, (3)
escape the cruel death awaiting it, against which it has
no defense, provided its vital signs are constantly with
watched with loving care? 6, Aj A,=E. (4)
757 J. Math. Phys., Vol. 18, No.4, April 1977 B. Misra and E.C.G. Sudarshan 757
The collapsed state p' given by (3) is, in general, not Returning to ideal measurements we have to proceed
uniquely determined by the measured observable E and to the limit for n - 00. We define
the observed outcome but depends also on the details of
the measuring apparatus. This circumstance is re- pet) = s-lim pen, f), (9)
n~ ..
flected in the nonuniqueness of the operators A J satis-
fying (4). (10)
The mapping (3) of the density matrices is very provided the limits on the right-hand side exist. Hence,
closely related to the "completely positive maps" de- if the limit
fined by
s-lim Tn(t) "" s-lim [EU(t/n) E] n = T(t)
n_1IO n .. .,
p-L; V"pV!=A(V,,)p,
" exists for t>- 0, then we may make the identification
pet) = {Tr[T(t) pT *(t)W 1 • T(t) pT *(t) (11)
The "state collapse" caused by "nonselective" measure- for the resultant (normalized) state obtained as a result
ments of E is described by such maps. They will be of continuous observation and verification that the sys-
considered in a future publication in the context of re- tem remained undecayed throughout the interval. The
peated and continuous nonselective measurements. probability Q(~; p) for this outcome is given by
Quantum theory envisages also the possibility of Q(~; p) =lim Tr[Tn(t) pT!(f)]
ideal measurements under which the collapse of the n~"
Moreover, it is also easy to show that the standard 3. ZENO'S PARADOX IN QUANTUM THEORY
interpretation of the quantum theoretical formalism
entails the formula In the preceding section we arrived at formula (13)
for the probability P(~; p) that the system prepared
(8) initially in the undecayed state p will be observed to
decay sometime during the interval ~ = [0, t]. Despite
In fact, (8) is a special case of a more general formula the natural derivation of (13) we now show that the
for the probability connections between several succes- probability P(~; p) vanishes for all finite intervals ~
sive observations. 8 It is important, however, to bear provided that the initial state was undecayed,
in mind that the general formula discussed in8 [and, a
fortiori, formula (8)] holds only under the assumption Tr(pE) = 1. (15)
that the successive measurements under consideration
are ideal in the sense described above. For nonideal Weare thus led to the paradOxical conclusion that an
successive measurements these formulas do not yield unstable particle will not decay as long as it is kept
correct probability connections. under continuous observation as to whether it decays
758 J. Math. Phys., Vol. 18, No.4, April 1977 B. Misra and E.C.G. Sudarshan 758
or not. The mathematical transcription of this state- s-lim [EV(tln) E)" = EV(t) E = T(t)
"_00
ment is the following theorem.
for all t ~ 0, but
Theorem 1: Let U (t) == exp( - iHt) , t real, designate a
strongly continuous one-parameter group of unitary T*(t) T(t) =EV(t*) EV(t) E
operators in the (separable) Hilbert space H. Let E
=V*(t)EV(t)*E for all t>O.
denote an orthogonal projection in H. Assume that:
(i) The self-adjoint generator H of the group U(t) is Thus the conclusion of the theorem is violated, though
semibounded. the assumptions in its formulation except the semi-
boundedness of the self-adjoint generator V(t) are met.
(ii) There exists an (antiunitary) operator 8 such that
[Strictly speaking, assumption (ii) about the existence
8E8-1 =E, of 8 is also not satisfied, but it was necessary only to
prove the existence of T(- t) and T(- t) is trivially veri-
8U(t) 8- 1 =U(- t) for all t. fied to exist in the present example. ]
(iii) s-lim" _ 00 [EU (tin) E)" == T(t) exists for all t ~ 0. We now turn to the
(iv) s-lim,,_o.T(t)=E. Proof of Theorem 1: The existence of
Then s-lim,,_oo[EU(tln)E)"==T(t) exists for all real t
T(t) == s-lim
._00 [EU(tln) E]' (17)
and possesses the following properties:
(a) The function t - T(t) is strongly continuous and for all real t follows immediately from the assumed
for all real t and s satisfies the semigroup law: existence of T(t) for positive t and assumption (ii). In
T(t) T(s) = T(t + s), fact for t ~ 0
(b) and T(- t) == s-lim
,_00 [EU(- tin) E]'
T*(t) = T(- t).
= s-lim 1
._00 8[EU(tln) E]' 8-
Remarks: (1) The conclusions of the theorem imply
the relation: = 8T(t) 8-1• (18)
T* (t) T(t) =E for all realt (16)
To prove assertion (b) we observe that
so that P(~;p)=Tr[p(I-E»)=O for all p satisfying (15).
[EU(- tin) E]' = ([EU(tln) E] "}*
(2) With 8 interpreted as the time-reversal (or CPT)
operation, the assumption (ii) of the theorem turns out - T*(t) weakly as n - co.
759 J. Math. Phys., Vol. 18, No.4, April 1977 B. Misra and E.C.G. Sudarshan 759
(3) Proof of Lemma 2: To prove (1) we start with the
1
21Ti
f"
_..
Fn(t)dt - 0
(t+i)2(t-Z)-'
Imz < O. (22)
representation (21) for Fn(z). By assumption, Wet)
=s-limn _ .. F n(t) for all real t and
1
Proof of Lemma 1: The assertion (1) follows from II (t +~>"2~: _ z) II ~ (~I::: )1- for all n.
the positive self-adjointness of H and its proof is
We can therefore apply again the Lebesgue theorem on
standard. To prove assertion (2) we start with Cauchy's
dominated convergence and conclude that
integral formula for the function Fn(z)/(z + i)2 which is
holomorphic in the open upper half-plane, W(z) = s-limFn(z)
n- ..
+ 1_!
F (z) __
(z i)2 - 21Ti
c
Fn(z') ,
(z' + i)2(z' _ z) dz, Imz> 0,
exists and has the representation
(z + i)2 f" Wet)
closing the point z and contained entirely in the open From the well known Vitali's theorem l1 we can conclude
upper half-plane. A similar integral representation that W(z) is holomorphic in the open upper half-plane.
holds of course for the holomorphic function F n(z) itself.
But with the choice we have made the integrand vanishes To prove the semigroup property of W(z) we show
faster than Iz'1-1 as Iz'l - 00, Hence if we choose the first that this law holds for pure imaginary values,
closed contour C to be the axis running from - 00 + iE W(is) W(it) = W(i(t + s»
to + 00 + iE and an infinite semicircle we could rewrite
the contour integral as an open line integral for all positive t and s.
To this end, first consider the case where f and s are
Fn(t + iE)
(21 ') rationally related so that there exist positive integers
p, q for which
Imz > E> O.
s +t s t
The (operator) norm of this integrand is dominated by + q) = rp = rq
r( p
the integrable function for all integers r. For such s, t we can deduce
(1 + f)-l (Imz - 0 EO)"l
1
21Ti
f c
Fn(z') d ' f
(Z'+i)2(z'-Z) z
0
or Imz< .
tinuous for s> O.
To prove assertion (3) we observe that the operators
W(is) = s-lim [E exp(- Hs/n) El n
Lemma 2: With the same notation as in Lemma 1 let 1'1 .~ ...
(1) W(z) =s-liIDn_ .. Fn(z) exists for all z with Imz> O. BG=GB=B
760 J. Math. Phys., Vol. 18, No.4, April 1977 B. Misra and E.C.G. Sudarshan 760
Lemma 3: Under the assumptions of Lemma 2, we The proof of assertion (a) of the theorem may now be
have the weak limit for operators along the real axis easily completed by combining the conclusions of the
preceding lemmas,
w-lim W(s + i11);: W(s) for almost all real s. (26)
~. o. W(s) = w-lim W(s + i1) = w-lim G exp(iB(s + i11)] G
" .. 0. 1J .. 0+
Proof of Lemma 3: To obtain this weak limit let us = G exp(iBs) G for almost all real s. (24')
start from the integral representation (25) rewritten in
Thus W(s) l¥*(s) = G for almost all real s. According to
the form
assumption (iv) in the statement of Theorem 1,
. _(s+i+i11)2
W(s+z11)- 27Ti
f'" W(t)
(t+i)2(t-s-i1)) dt, 11> O. w-lim W(s) W*(s)=w-lim T(- s) T*(- s)
.'" s .. o+ s-o+
On the other hand, from (22) and the Lesbegue domi- =w-lim T*(s) T(s) =E. (28)
nated convergence theorem s .. 0+
o= (s + i + i11 )2 f~ W(t) dt Thus G=E and we may rewrite (24) in the form
27Ti .'" (t+i)2(t-s+i1)) , 11>0.
W(s)=Eexp(iBs)E for almost all s,
Therefore, (29)
EB=BE=B.
.
W(S+Z11)=
(s + i + i11 )2
1T
f'"
.",
W(t) 1)
(t+i)2 (t_S)2+ 11 2 dt. But we can now strengthen this relation for W(s) to read
W(s) = E exp(iBs) E for all s (30)
For any two vectors zJ;, cf> in H we may write
in view of the strong continuity of W(s). Combining (29)
(zJ;, W(s +i11) cf»
and (30) we immediately deduce the validity of
_ (s + i + i11)2 f~ (zJ;, W(t) cf» 11 assertion (a).
- 7T (t+i)2 (t_s)2+ 11 2 dt •
.~ Although not of primary interest for the discussion
in this paper, we recall the known result 3 that if His
Since the quantity (zJ;, W(t) cf»/(t + i)2 considered as a
semibounded, the operators EU(t) E cannot form a
function of t is integrable, it follows that
semigroup for t>-- 0 except in the event of E commuting
lim (zJ;, W(s + i'ry) cf» = (zJ;, W(s) cf» (27) with U(t) for all real t. We may subsume this result
TJ~ 0+ as a corollary to Theorem 1.
for almost all s. Corollary: Let the self-adjoint operator H be semi-
bounded, let E be an orthogonal projection and let U(t)
To complete the assertion of Lemma 3 a technical stand for exp(- iHt). If {EU(t) E It >-- O} form a semigroup,
difficulty is to be resolved. For a given pair zJ;, cf> of then
vectors, the assertion (29) has been shown to hold for
almost all s. The exceptional set (of measure zero) EU(t)=U(t)E for all real t. (31)
where this result may not hold may appear to depend on Proof: The semigroup property for EU(t) E, i. e., the
the pair zJ;, cf> chosen. To show that there is at most a relation
single null set outside which (27) holds for all pairs
zJ;, cf> we proceed as follows: Let f) be a countable dense EU(t)EU(s)E=EU(t+s)E for all real s,t (32)
subset of the separable Hilbert space H and let N be will imply
the union of the countable family of exceptional null
sets corresponding to all pairs zJ;, cf> with zJ; E f), cf> E f). EU(t)EU*(t)E=E for all real t (32')
This set IV is a set of measure zero and the weak and hence
limit (26) holds everywhere outside this set for zJ;, cf> in
f), but then (27) will hold in the complement of N for E ~ U(t) EU* (t) for all real t.
all pairs zJ;, cf> not necessarily in f). In fact, writing
Multiplying this equation from the left by U(- t) and the
A(s, 11) = W(s + i11) - W(s) right by U*(- t) will yield
U(- t)EU*(- t) ~E for all real t.
we may obtain
Together these two inequalities imply
(zJ;,A(s, 11) cf» = (zJ; - zJ;n, A(s, 11) cf» + (zJ;n, A(s, n)(cf> - cf>n»
E = U(t) EU*(t) for all real t
+ (zJ;n, A(s, 11) cf>n)'
or, equivalently,
We see that for s outside the exceptional set N the first
EU(t)=U(t)E for all real t.
two terms on the right-hand side tend to zero as n _00,
since we may choose The proof of the corollary is thus reduced to the
proof of (32) or (32').
s-lim zJ;n = zJ;, s-lim cf>n = cf>.
PI .. to ,. . . .
Since the operators EU(t) E are assumed to form
The third term, by hypothesis, goes to zero as 11 - O. a semigroup for t>-- 0 and [EU(t) E]* =EU(- t) E, for all
since zJ;n, cf>n are chosen to lie in f). (The proof of this positive integers n and all real t we have
lemma incorporates a suggestion due to K. Sinha.] (EU(t/n) E]n = EU(f) E.
761 J. Math. Phys., Vol. 18, No.4, April 1977 B. Misra and E.C.G. Sudarshan 761
Hence (i) B(O, t);" 0,
=T(t)
"~ .
s-lim[EU(t/n) E]" = EU(t) E (ii)B(O,t)pB(O,s), tps,
(iv) P(O, ti p) - Tr[ pE~] for t - 0 •. Since so far no such principle has been derived from
or incorporated into quantum theory, this is not a
In addition, P(O, ti p) may be assumed to be continuous satisfactory way of resolving the paradox at the present
as a function of t. If we were to succeed in finding a time.
formula (2) Zeno's paradox is based on the assumption that the
P(D, Ii p) = Tr[ pB(O, t)], (34) continuous measurements are ideal measurements. But
measurements (or, observations) involved in the re-
then the operator B(O, t) would have the following cording of the track of an unstable particle in a detect-
properties: ing medium are nonideal in the sense of (3).
762 J. Math. Phys., Vol. 18, No.4, April 1977 B. Misra and E.C.G. Sudarshan 762
This is a tenable view and it would deny the validity also be reemphasized that the probabilities such as
of Theorem 1 as stated and proved in this paper. It has Q(A; 1/, p) [or Q(A; p)J that pertain to the outcomes of
the somewhat unsettling side effect that P(O, t; p) and successive measurements (or continuous measure-
hence the "observed lifetime" of an unstable particle ments) depend on the law according to which "state
is not a property of the particle (and its Hamiltonian) collapses" occur at the time of measurements. Thus
only, but depends on the details of the observation one may say that the "collapse of state vector" caused
process. At the present time we have no indication that by measurement, which has haunted the foundation of
this is so. quantum mechanics like an invisible ghost becomes visi-
ble through probabilities such as Q(A; 1/, p), etc. The
(3) The record of the track of a particle is not a con-
probabilities pertaining to the outcomes of several suc-
tinuous observation that the particle has not decayed,
cessive (as well as continuous) measurements there-
but only a discrete sequence of such observations; while
fore deserve further theoretical as well as experimental
Zeno's paradox obtains only in the limit of continuous
study than they have received so far.
observations.
While this is tenable, the sufficiently repeated moni- ACKNOWLEDGM ENTS
toring of the particle should again lengthen the life-
time. There is, however, no indication that the life- It is a pleasure to thank Professors C. Chiu, M.
time of a (charged) unstable particle (say, a muon) is Guenin, K. Gustafson, Ph. Martin, C piron, and
appreciably increased in the process of its track forma- K. Sinha for their critical comments as well as con-
tion through bubble chamber. To shed additional light structive suggestions. We are also grateful to the
on this question a quantitative investigation of the effect referee for some useful comments, in particular, for
of repeated monitoring on the lifetime of particles (in bringing to our notice Refs. 5 and 6. One of us (B. M. )
specific models) is in progress. 13 was partially supported by Swiss NSF during summer
1975.
(4) Natural though it seems, it is wrong to assume
that the temporal evolution of a quantum system under
continuing observation can be described by a linear
operator of time-evolution such as T(t). It can be de- *Work supported in part by Energy Research and Development
scribed only in terms of a persistent interaction between Administration Contract E(40-1)3992.
I Another possible suggestion is to interpret
the quantum system and the classical measuring ap-
paratus. When this is done the quantum Zeno's paradox
will either disappear or if it survives, at least, it will flot Tr[pU" (s)E U(s) 1ds
be understandable as the drastic change in the behavior as the desired probability Q (0, t;p). Apart from the fact that
of the quantum system caused by its continuous inter- there is really no convinCing reason for this interpretation
action with a classical measuring apparatus. this expression is not generally a monotone (decreasing)
function of time t, a property which Q(O,t;p) should possess.
This pOint of view is at present only a program since 2E. Schrodinger, Naturwissenschaften 23, 807 (1935).
there is no standard and detailed theory for the actual 3D. Williams, Commun. Math. Phys. 21, 314 (1971); L.
coupling between quantum systems with classical mea- Horwitz, J. P. Marchand, and J. Lavita, Rocky Mount. J.
suring apparatus. A beginning in this direction is made Math. 1 225 (1975).
4G.R. Ailcock, Ann. Phys. (N. Y.) 53, 251-348 (1969).
in a forthcoming paper. 14 5W. Yorgrau in Problems in Phil osophy in Science, edited by
1. Lakatos and A. Musgrave (North-Holland, Amsterdam,
Having been forced into such unusual points of view
1968), pp. 191-92.
by the quantum Zeno's paradox one is prompted to draw 6H. Ekstein and A. Seigert, Ann. Phys. (N. Y.) 68, 509 (1971),
also some parallels between it and certain empirical 7B. Misra and E. C. G. Sudarshan, in preparation.
findings in the study of human awareness. We shall 8E. P. Wigner, in Foundations oj Quantum Mechanics, edited
present such close parallels between the quantum Zeno's by B. d'Espgnat (Academic, New York, 1971), especially
paradox and the findings of sensory deprivation and formulas 14 and 14(a), p. 16.
9P.R. Chernoff, Mem. Am. Math. Soc. 140 (1974).
other experiments pertaining to the study of conscious-
IOC.N. Friedman, Indiana Univ. Math.J. 21, 1001-11 (1972).
ness in a separate publication. l1E. Hille and R. Phillips, Functional Analysis and Semigroups
In conclUSion, it seems to us that the problems posed (Am. Math. Soc. Colloq. Publ., PrOVidence, R.I., 1957).
12Ref. 11, Sec. 22.3.
by Zeno's paradox have no clean cut resolution at the 13C. Chiu, B. Misra, and E.C.G. Sudarshan, in preparation.
present time and deserve further discussions. It may 14E. C.G. Sudarshan and B. Misra, in preparation.
763 J. Math. Phys., Vol. 18, No.4, April 1977 B. Misra and E.C.G. Sudarshan 763