PHYSICAL REVIEW D VOLUME 24, NUMBER 6 15 SEPTEMBER 1981

Pointer basis of quantum apparatus: Into what mixture does the wave packet collapse?
W. H. Zurek
Center for Theoretical Physics, The University of Texas at Austin, Austin, Texas 78712
and California Institute of Technology, Pasadena, California 9II25~
(Received 3 April 1981j
The form of the interaction Hamiltonian between the apparatus and its environment is sufficient to determine
which observable of the measured quantum system can be considered "recorded" by the apparatus. The basis that
—
contains this record the pointer basis of the apparatus —consists of the eigenvectors of the operator which
commutes with the apparatus-environment interaction Hamiltonian. Thus the environment can be said to perform a
nondemolition measurement of an observable diagonal in the pointer basis.

I. WHAT IS MEASURED IN A QUANTUM measurement is completed the quantum system

MEASUREMENT? will in fact end up in one of the states ( r)} rather
~

Von Neumann' has shown that the unitary evolu- than in one of the states ~s)?
tion alone suffices to establish a nonseParable A particularly acute manifestation of this am-
correlation between the state vector ~A) of the biguity in the choice of the preferred apparatus
quantum apparatus 8 and the state vector p) of ~
basis occurs when all the coefficients c, in Eq.
the quantum system 8 which is to be measured: (1.1) happen to have the same magnitude. In that
case, whenever the set (~A„)}is orthonormal, the
g, ) )) ())) = jPa. ~A. &} s(pc. ~s&} set of the relative states (}~))is orthonormal as
well. Then the apparatus by virtue of being corre-
lated with the state of the system contains not
=
Q c, (A, ) S (s) . only all the information about the observable
Here ~A, ) and ~s) are basis vectors for the ap- S =Z, e, ~s)(s ~; it must equally well contain all the
information about many other observables R
paratus 8 and system f, respectively, while ~A, )
is the initial state of the apparatus. =P„f„~r)6 ~, defined on the Hilbert space of the
system S. This is so despite the fact that R and
Equation (1.1) seems, at first sight, to solve
S do not, in general, commute. Yet we know that
the problem of measurement in quantum mecha-
quantum mechanics prevents one from measuring
nics. States of the apparatus ( ~A, )}are now cor-
simultaneously two noncommuting observables
related with the states of the quantum system
with arbitrary accuracy. Moreover, everyday
( ~s)}. To the question "What has been measured experience convinces us that the choice of
on S~" one may be tempted to reply "The observ-
able S =Q, e, ~s)(s ~, of course. " The apparatus
"what ha, s this apparatus measured" cannot be
made arbitrarily, long after the apparatus-system
8, however, is itself presumably described by interaction has taken place, as Eqs. (1.1)-(1.3)
quantum mechanics. Therefore, nothing can pre
would seem to imply. The "real-world" appara-
vent one from expressing the state of (t in terms
of an alternative orthonormal basis ( ~A„)} corn tuses constructed to measure momentum do mea-
posed of superpositions of states ~A, ):
sure momentum and not the conjugate observable,
position.
A question can then be raised: What does, in
the real-world apparatuses, determine this ap-
parently unique pointer basis ( ~A~)}, which re-
In terms of this new apparatus basis the state
cords the corresponding relative states ( ~Pl} of
of the combined 0@ system can be readily rewrit-
the system'P
ten:
Interaction with the environment is the key
feature that distinguishes the here-proposed model
gc, ~A, )@ ~s) = P ~A„)N& gc, (A„A, ) ~s) of the apparatus from the manifestly quantum
systems. We argue that the apparatus cannot be
(1.3) observed in a; superposition of the. pointer-basis
states because its state vector is being continu-
This equation defines a set of Everett's relative ously collapsed. It is the "monitoring" of the
states( ~x)}, i. e. , normalized, but, in general, apparatus by the environment which results in
not mutually orthogonal states of the system & the apparent reduction of the wave packet. Cor-
relative to the arbitrarily chosen basis set ( ~A „)} relations between states of the pointer basis and
of the apparatus. ' Does that imply that when the corresponding relative states of the system are

1516
 POINTER BASIS OF QUANTUM APPARATUS: INTO WHAT. ..
nevertheless preserved in the final mixed-state by-bit measurement (see Fig. 1). This same ex-
density matrix: ample has been recently employed by Scully, Shea,
and McCullen, ' who used a bistable atom as a
microscopic model of "Wigner's friend. »"
During the passage through the first of the
- Z If. l'I&. &&&. l IP&&P I
(1.4)
Stern-Gerlach magnets, momentum and hence
position of the spin become correlated with the
eigenstate of the spin component along the z axis.
Hence, even though below we do not face the in-
In particular, if a spin enters RSG in an initially
soluble question of quantum theory of measure-
ment: "What causes the collapse of the system-
pure eigenstate o) = ( 0&+ 0&)/v 2 of the spin in
I I I

the direction x, then the splitting of the beam can

apparatus-environment combined wave function?"
be represented by
we do determine into what mixture the wave func-
tion appears to have collapsed. lo&e l@(r f)&=CI~&ctl+o(r, t)&+ It&e I&]&&(r, t)&+~2.
In the following section we begin our discussion
describing a simple example of the interaction (2. 1)
between a pair of two-state systems —
"bit-by-bit Here IQ(r, f)), Ip~ ~(r, f)) describe the time-
measurement" —which leads to correlations of the dependent position of the wave packet. In the third
type described by Eqs. (1.1) —(1.3). We will argue Stern-Gerlach magnet, lpi) becomes identical
that the nonuniqueness of the apparatus basis leads with lp~&, or, in other words, the two beams
to apparent nonseparability paradoxes of the same correlated with spin states 0) and 4& finally re-
I I

nature as those encountered in the Einstein- combine. If there were no measurements made
Podolsky-Rosen (EPR) experiment. on the position of the spin or on the spin itself
In Sec. III we will use the von Neumann equation inside RSG, the spin would leave in the pure state
to show that the apparatus-environment interac- I
o).
tion Hamiltonian must couple in a nonperturba- Insert now a bistable atom to serve as an ap-
tive' ' way into that apparatus observable that paratus along the trajectory of the spin If). Sup-
is diagonal in the pointer basis. This in turn pose it is initially in one of its two states I=&.
determines relative states of the system which Suppose, moreover, that the interaction Hamil-
can be considered "recorded" by the apparatus. tonian between atom and spin is given by
Section IV shows how the general results of Sec.
III resolve issues raised for the bit-by-bit mea- If,„=g.(r —r„)(l»~& I+ l»&~ l)~(l=&&-I+ I=&&= I),
surement problem in Sec. II. Section V contains (2.2)
a brief discussion of the most important new
—
concept introduced, in this paper pointer basis- (a)
in the context of measurement theory and practice.
Conclusions of the paper are stated in Sec. VI.

II. A BIT-BY-BIT MEASUREMENT

~ -Q
To explore the physical consequences of the
N z) S N
nonuniqueness of the basis chosen to represent
an apparatus correlated with a quantum system,
it is best to examine a simple thought experiment.
Here we shall consider a two-state quantum ap-
paratus used to "measure" (the word "premea- (b)
surement" would be more exact but less con-
venient) the other two-state system. (Two states
allow one to store one bit of information and
hence a "bit-by-bit measurement. )
Three Stern-Gerlach magnets can be arranged
to first split, and then recombine, the spin-&

setup (RSG)."
beam. This is the usual reversible Stern-Gerlach
We shall supplement it by a bi-
stable atom acting as a (quantum) apparatus.
FIG. 1. (a) Reversible Stern-Gerlach setup. (b) Sche-
Such an atom, inserted along one of the two pos- matic representation of the trajectory of the spin carrier
sible trajectories of the spin-& particle provides in the bit-by-bit measurement. Possible location of the
a possible, if impractical, realization of the bit- bistable atom given by an asterisk.
1518 H. ZURKK

where g is a coupling constant, v is a short-range of the spin, and hence the record of the value of
interaction potential, e. g. , v( Ir I) CC 6(r -r„), -r„ the spin itself.
and r, r„are positions of the spin and the atom, The evolution of the wave function I4') of the
respectively. With this interaction Hamiltonian, combined spin-atom system as it proceeds from
it is not difficult to demonstrate that the final the initial, pure state can be written as
state of the atom contains a record of the path

o& e I
=& e I
y& -{I» e I
y»+ » 3 y, &] ~
I I I
= &/~2. (2.3)
When the spin-atom interaction begins, one must use t;he Schrodinger equation to calculate the effect of the
passing spin on the atom
iII4& =(H, +H„+H») +&. I
(2.4)
This yields"

I
+& ={I~& I«) I=&+b«) I=&] I
&~&+ I
» I
=& 1«&V~2 (2.5)

where the time-dependent coefficients a and b sa- Now whenever the atom is found in the state ), I

tisfy coupled equations the spin is in the state I

», and vice versa. The
state of the spin has become nonseparably cor-
i@a = ea+gby(t),
related with the state of the atom. This final state
i hb = ebg+a-y(t) . (2.6) with perfect correlation, Eq. (2.11), will be used
throughout the rest of this paper.
Here 2& is the difference between the energy of the
At first sight it might appear that the problem of
states =) and I=), while
I
measurement in quantum mechanics has already
been solved. After all, the atom apparatus con-
&N fd'r(( (~=, &)(~(~ —~ )((i(~, &)). (2.7) tains a record of whether the spin (system) has
assumed the » or I » state while traversing the
For simplicity we shall assume & =0, and, for all I

reversible Stern-Gerlach setup. This can be easi-

times, Iv(r —r„) lp&(r, t)) =0. Now the
&P&(r, t)
ly checked by measuring the state of the atom and
solution of Eq. (2.6) is straightforward:
verifying that the spin can be found in the appro-
a(t) = cosA(t), priate, correlated direction. This is represented
(2.8) by projections =&& = and =)&
b(t) = -i sinA(t); I I I

Here A is the action, measured in units of 6:

I=&&=l+&= I» I=&8'I+&/~2
=-il» 3 I=& ~ (2.12)
I-&&= I~& @&/~2.
)-a(~) = g
f r(()a (2.9)
I

There can be no doubt that quantum mechanics pre-

dicts 100/o correlation between the state of the
After the spin has emerged from the reversible atom and the state of the spin.
Stern-Gerlach setup, A becomes a constant:
The question "What has been recorded by the
A=limA(t) . atom apparatus'P" becomes nevertheless apparent
when rather than using =)& = and =)& = one
I I I I

Consequently, the final wave function is given by" employs projection operators corresponding to
the alternative basis:
I+& ={I»8 IcosAI=&-i st~i=&]+ » + I I
=&]

(2.10) I
&=(I=& I=&)/~,
(2.13)
Clearly, the final state is still pure and no irre-
versible measurement, no collapse of the wave Clearly
function I4& could have occurred. However, it is
straightforward to demonstrate that the illusion
I+&&+ +&
I
=-i{I+& ( »+i »] I I I
&&/~2
of a collapse may arise when one considers the
spin and the atom as two independent systems.
=-i{I+&3 I-&] I I
&&/~2 (2.14a)
This is best seen if A=)T/2, i.e. , and, by the same token,

I+& ={I» I
=& -il » I =&] I I
e&/&~ I-&&- I+& =i{I-& I -&/8 @)/» I
2 . (2.14b)
 POINTER BASIS OF QUANTUM APPARATUS: INTO %HAT. . . 1519

Here who consult the pointer of the ideal apparatus

learn that the system is in one of the eigenstates
and of P and not in some arbitrarily chosen "relative
"
state. However, as we have seen so far, quantum
mechanics alone, when applied to an isolated,
define a new basis for the spin spanned by its composite object consisting of an apparatus and a
eigenstates along the y axis. Thus, the wave func- system, cannot in principle determine which ob-
tion ~%& given by Eq. (2.11) can be equally well servable has been measured. Below we shall argue
written as that the possibility of a natural choice of "what has
~4') = i', ~(+&.8 [ . )
( ) S [ )]'S( y&i&2 . (2.16) been measured" arises when one recognizes the
Hence, there is also a 100/q correlation between following: (a) The apparatus 8 interacts with its
the state of the atom and the spin in a basis com- environment 8 via some specific interaction Harn-
pletely different from the one used to correlate iltonian Keg. (b) The observer S consults only the
the value of the spin with its posftion, i.e. , 0), pointer of the apparatus and not the state of the
~

[4). We have used a Stern-Gerlach magnet with a environment.

field gradient in the di, rection g to measure spin in The apparatus-environment interaction can then
the direction y. Moreover, we can choose what we
- be regarded as an additional measurement estab-
lishing nonseparable correlations between the ap-
ceased to interact with it. "
snail measure on the atom long hfter the spin has
paratus and the environment. As a result, in-
formation about the environment (usually regarded
Counterintuitive predictions of quantum mechan-
ics for this bit-by-bit measurement correspond as "noise"} obliterates the information about the
closely to the nonseparability of the "Einstein- just premeasured quantum system g . However,
Podolsky-Bosen paradox. ""
For what is measured when the Hamiltonian 3C+~ commutes with the~ob-
on the state of the atom influences the state of the servable II of the apparatus, then this particular
spin. Moreover, even though the spin has been observable will not be perturbed. Only the basis
split inside RSG into the two well-separated beams, consisting of the eigenstates of II, the pointer
one definitely carrying ~0) and the other ~4), once basis, will contain nothing but the information
these beams are recombined the atom can supply about the quantum system itself. Moreover, the
definite and correct information about the spin's combined 65 system is now represented by a mix-
alignment (i.e. , parallel or antiparallel) with re- ture diagonal in a particular product basis, con-
spect to other directions. sisting of the eigenvectors of the pointer basis
—
Wigner's friend to use the analogy of Bef. 10— of the apparatus and the corresponding relative
not only ends up in a superposition after taking a states of the system.
look at the single spin passing through BSG, he One can anticipate the main result of this section
—
can also be led to admit that he has seen inside by saying that the pointer basis of the apparatus
the RSG magnets with magnetic field gradients in 8 is chosen by the form of the apparatus-environ-
the z direction —
spin aligned along the y axis. ment interaction: It is this basis which contains
This conclusion appears preposterous. For a reliable record of the state of the system Q. This
quantum systems it has been nevertheless con- in turn determines uniquely those relative states
vincingly verified in experiments stimulated by of the system which are correlated with the ap-
the EPR paradox. Furthermore, if one denies paratus. Moreover, apparatus-environment cor-

Neumann, London and Bauer, "

(in disagreement with original proposals of von
' and Wigner") any
relations do not allow one to observe the 85 com-
bination in a superposition. Instead, it becomes
special role to consciousness, there is seemingly a mixture diagonal in the basis constructed from
nothing that could keep one from describing an ar- the pointer-basis eigenstates ~A~& and the corre-
bitrary system, no matter how large, by a state sponding relative states of the system. I et us
vector and SchrMdinger equation. After all, there stress that details of the state of the environment
is nothing in the laws of physics that would make itself are not necessary to determine the pointer
quantum mechanics applicab}. e to a few-body sys- basis. The form of the apparatus-environment
tem but render it invalid for a truly many-body interaction X+~ suffices for that purpose.
system, even'if it contains 10~' or more atoms as To verify the existence and further clarify the
long as it remains isolated. role of the pointer basis f ~AP), consider evolution
of the density matrix pz+ of the three combined
III. POINTER BASIS OF THE APPARATUS systems: the to-be-measured quantum system
It is usually taken for granted that the apparatus 5, the apparatus 8, and the environment 8:
measured — perfectly or imperfectly —a particular
observable P. For, human or nonhuman observers -iripseg =lapses~&s+&~+&g+SCse+Xeg+X~]. (S.1)
 The correspondence between the components of tion of the form of Eq. (1.1) is established. After-
the total Hamiltonian and the element's of the inter- wards the interaction between the system and the
acting combination f85 is evident from the nota- apparatus is effectively nil, i.e. , K+~»3C&z, and
tion. We purposely disregard the interaction be- X+ can be set equal to zero.
tween the environment 5 and the "rest of the (iii) All the vectors of the pointer basis cor-
world. " This interaction is of no importance for respond to the same degenerate energy eigenstate,
the choice of the pointer basis, at least as long i.e., Ku ~A&} =E~A~}, where E is not a function of
as it does not alter the form of BC~g. To write Eq. p. (However, the A~} are not degenerate any fur-
(3..1) we have moreover assumed (1) that all the ther. ) This is equivalent to a physical requirement
interactions are pairwise, i.e. , that 3C&@=0, and that the measurement should not lead to the ex-
(2) that the environment 8 can be regarded as a,
quantum system. Assumption (1) is customary paratus. "
change of energy between the system and the ap-
All of these assumptions are stronger
and clear, even though it may prevent one from than is absolutely necessary. Adopting them in
even an approximate treatment of the gravitational this form makes the following discussion much
interaction beyond its Newtonian pairwise form. simpler and allows us to concentrate on the main
Assumption (2) deserves further scrutiny. One idea rather than on the detailed and cumbersome
should clarify what is meant by the "environment, " review of the subcases.
i.e., which degrees of freedom of the Universe The density matrix of the SQb combination
should be taken into account in determining "what —
evolves then immediately after the correlation
mixture does the wave packet collapse into. " between the system g, and the apparatus 8 has
We define the environment b as consisting of all been established —according to the equation
those degrees of freedom which contribute signi-
ficantly to the evolution of the state of the apparat-
fffPsa8 -=
[Psas Ks+Ke+K8+Ke8]
=
us. For a given experimental setup one could de- [pgss Ka+Ka+Ks]+ [psss. Keg] (3 2)
vise a criterion based on the average strength of The effect of the first commutator bracket can now
the interaction, or, alternativel. y, on the amount of be disregarded. This follows from the fact that
action exchanged between the environment and the the time evolution of the apparatus states ~A~(t)}
apparatus. The objective of such a criterion is to = exp[ —(i/h)Et]~, A&(0)} leaves the diagonal terms
exclude all those degrees of freedom whose total of the density matrix invariant. Therefore, evolu-
contribution to the environment-apparatus interac- tion of the apparatus 8 due to X~ does not obliter-
tion can be, for all practical purposes, disre- ate the information about the system g.
garded. Eventually, one can distinguish between
—
the immediate environment 5 and include it in
The second commutator [,
Xss] introduces cor-
relations between the apparatus and its environ-
—
the density matrix pros of Eq. (3.1) and "the rest
"
of the world, or a "remote environment (R" which
ment. It will leave diagonal terms of the diagonal
terms of the density matrix invariant only if it
need not be taken into account. Interaction between will commute with the very projection operators
5 and S can be in general quite strong. As we ~A~}&A~~ which are to appear on the diagonal. Con-
have already decided we shall not take it. into ac- sequently, if ~A ) is to remain correlated with the
count in Eq. (3.1). This additional term would relative state of the quantum system 5, K~ must
complicate our arguments, and our final conct, u- satisfy
sion does not depend on whether we set it equal to
—
zero "decouple" the environment from the rest of
the Universe — or take it into account and consider
K~, g v, ~A, }&A,
~
=0 (3.3)

p+&+ rather than p+. Having agreed that the en- for any arbitrary choice of coefficients m . Defin-
vironment may be in principle regarded as iso- ing the pointer observable,
lated, we can introduce an appropriate basis sys-
tem ~e} spanning its Hilbert space. II = gv, [A, )&A, [, (3.4)
Three more assumptions of an essentially tech-
nical nature can now be introduced: where we now require m to be strictly real, one
—
(i)K3s =0 the quantum system itself remains can reexpress the above condition by stating that
isolated from the environment. If this assumption the pointer basis (~A~}) is a complete set of eigen-
is violated after the premeasurement has occurred, functions of the operator II that commutes with the
then the apparatus will contain the information Hamiltonian 3C+~. .
about the state in which the quantum system was,
but not necessarily is, any more.
[II, Keg] =0. (3.5)
(ii) Kqu acts only for a, very short period of time. The above condition can be interpreted by analogy
During that time interval 3C~+»X+~ and a correla- with quantum nondemolition measurements, con-
 POINTER BASIS OF QUANTUM APPARATUS: INTO WHAT. .. 1521

sidered recently by Braginsky and his group, ' the ment. This so-cal. led pointer basis retains the in-
Caltech group, ' and Unruh. ' There the to-be-mea- formation about the outcome of the premeasure-
sured quantum system is the Weber bar used also ment despite the imperfect isolation of the ap-
as the detector of a gravitational wave. Here it paratus from its surroundings. In the next sec-
is the apparatus itself. There the measurement tion we return to the example of the bit-by-bit
is performed by a complicated setup designed not measurement to see how the interaction with the
to perturb, say, the eigenstate of the phonon num- environment can select the unique pointer basis.
ber operator or some other suitably chosen ob-
IV. POINTER BASIS IN A BIT-BY-BIT
servable. Here the "measurement" is performed '
MEASUREMENT
by the environment itself, and the apparatus-en-
Consider the state of the bistable atom-spin
vironment interaction Hamiltonian chooses the
pointer observable II as the one which will be
—
system familiar from Sec. II after a perfect —
correlation between the two has been established:
measured by the environment in a nonperturbative
fashion. l~& = (- ~ I'& I
= &+ l » I
= )] ~~ (4 l}
The interaction Hamiltonian can then depend on
only one apparatus observable — '
on II. ' In parti-
A
(Here and below we shall omit the cumbersome
and, for our purposes, irrelevant spatio-temporal
cular, any interaction Hamiltonian of the form
component of the total wave function, lQ(r, t)&.)
The model environment consists of an additional
Keg= A~ A~ I3 ge~. 6 4 + g~~~i~ + 6 E,
p 6&e bistable atom, basis states of which are la& = = ) l

and le') = = ). Note that the states of the environ-

(3.6) l

ment atom are denoted by normal "brackets"

does satisfy condition (3.5). Moreover, both e) l
rather than the usual "bra-kets. This corres- "
and g', ~z may explicitly depend on time due to the ponds to an implicit assumption that the environ-
interaction with the remote environment (R. As ment atom can be distinguished from the apparatus
long as the interaction remains diagonal in the atom. The apparatus-environment interaction
pointer basis, it will not disturb correlations of Hamiltonian
the apparatus with the states of the system relative Res = g( l4& (0 + i) (4 )
l l l
to the pointer basis.
Of course, the environment-apparatus interac- (I
=
tion that allows for the existence of the pointer (I = &( = I
+ I
= &( I
&
(4.2)
basis does not suffice to ensure successful func-
tioning of the apparatus. Premeasurement, which
is nonperturbative with respect to the basis = ), l

~ ) of the apparatus, but does influence the state

correlates the quantum state of the apparatus with l

the state of the system, plays an absolutely essen- of the environment as well as the state of the sys-
tial role. Moreover, if after the measurement we tem in any basis other than s), = ). Res ob- l l

expect the measured system to collapse into one of viously commutes with the "spin operator" o = II
the mutually orthogonal eigenstates of the mea- , as SC„=err
sured observable P, then immediately after the + I=)( I}
premeasurement the combined Qs wave function Let us suppose, for example, that the initial
should be of the form Q~b~IA~& S lp), where (lp&] state of the environment was given by ), so l

constitutes that orthonormal basis composed of the that the state of the combined Q+ Q+ 8 system
eigenstates of the observable P. When the states immediately after the measurement has occurred
relative to the orthonormal pointer basis are not can be written as a direct product
mutually orthogonal, the measurement will be only le&= le)c =) l

imperfect. =(
In this section we have established the main re-
z l~&e l
= &+ l~&e l
=&)I l
= }yw.
(4.3)
sult of this paper: For the quantum systems known Now it is not difficult to show that this initial state
as apparatuses, there exists some basis (lA~&) under the influence of the interaction Hamiltonian
not perturbed by the interaction with the environ- K~g will evolve into

IC& = &- ~ l» l
= & ~c»A(&}
I
= }-~»»(f} l
}] + l~& l
= & ~cosA«} [
* &+ ~ st»«} I
= &]4'~ (4.4)

If we choose the interaction between the apparatus l-L) and lT) as follows:
and environment to terminate when A = v/4, then
the states of the apparatus become correlated with pl~& l=& l~)+ l» l=&g l
}P~2
the two orthogonal states of the environment atom (4.5)
1522 H. ZUREK

Here orientation of the spin in the direction of the y

+i axis can be derived from the state of the appa-
(4. 6) ratus alone.
T) = ( = ) —f = )) /~2 .
~ ~ ~

While calculations leading to the above conclu-

Now, if we act on the apparatus by means of the
= ) ( = or
sions are certainly correct, one might object that
projection operators ) ( = ~, we ~ ~ ~
the assumptions of a particular initial state for the
find, as in Sec. II, spins in the corresponding
apparatus and a specific amount of action ex-
states ~4) or ~&), respectively. This is readily
changed between apparatus and environment are
verified by checking that
unrealistically restrictive. For example, if the
I
=&& = IC'&= ~~&8 (
=&8 IT)/v2 apparatus-environment interaction were not term-
(4. 7) inated after A =0. 125h, but continued forever,
]
= )( = [e) = g [%) 8
) 8 [~)/~2.
[
=
then every once in a while the state of the system-
The environment has become correlated with the apparatus combination would become independent
state of the apparatus without perturbing the eigen- of the state of the environment. This would occur
states of the observable II of the apparatus. whenever A = + (n/4)h, where n is an integer.
Therefore, we can still trust the memory of the Thus, we would again, once in a while, face the
apparatus concerning the value of the spin in the paradox of not knowing what has been measured
direction z. by the apparatus. This objection is certainly
However, if we try to use projection operators serious. One can show, however, as we shall do
I+ &+ or — — t'see Eq. (2. 14)] in an attempt
& & &
in a following paper, that when the size of the en-
I

to establish whether the spin was

I

or we
I

~
-) ~
-), vironment increases, the state of the apparatus-
find'it impossible to accomplish. To argue this we system combination becomes pure more and more
first calculate rarely, and a close analogy between that "memory
recovery time" and the Poincare recurrence time
I+ &&+ IC'&=(-~ l~&8 I»+ l~&8 I~)]81+ &/2 can be made.
IC&=&-il&&8 I»+ I&&8 l~)&81-&/2.
V. DISCUSSION
(4.8)
Clearly, ~l) cannot be written as a sum of two Von Neumann, facing the necessity for the re-
terms in the form duction of the wave packet, rejected the idea that
an additional quantum apparatus 6', coupled to the
IC'& = &. I~.&8 I+ &8 I+.)+ &- l~-&8 I
-&8 I+-&. original 8, can be of any help in resolving difficul-
(4.9) ties of the measurement problem. For, he rea-
Because of the correlations with the environ- soned, the state of the combined Sea' system, after
ment, knowing the state of the apparatus in the all the correlations have been established, would
j+), —) basis does not suffice any more to de-
~ finally evolve into
termine the state of the system. Part of the in-
formation about the state of the spin has been ') 8 8 -~A,') 8
"transferred" from the apparatus to the environ-
~A, ~A, ) ~p) Q b~ (A~) 8 ~p)
ment. And both the environment and the appara-
tus are correlated with ~f) or ~k) states of the
=
g b, (A, & 8 (A, & 8 (p&.
spin. (5 . l)
We can therefore conclude that when the en- Thus 8' stands in the same relation to 6 as 8
vironment atom is present and interacts with the with respect to g. The f inal f88 ' wave function
apparatus via 3CeS given by Eq. (4.2), and the is still pure. No reduction of the wave packet has
amount of exchanged- action is 0. 125 in the units been accomplished.
of h, then the apparatus-spin system will retain The reasoning of von Neumann presented above
perfect correlation in only one product basis is no doubt correct. Yet, the goal of this paper
( [0) 8 ( = ), 0) 8 = )}of the direct-product space.
( [
is to show that when the environment 8, playing
Hence, (~ = ); = )j is the pointer basis of the
~
the role of the additional apparatus, is taken
apparatus, which will eventually appear on the properly into account the question "What mixture
diagonal of the density matrix obtained by tracing does the wave packet collapse into?" acquires a
out "environmental degrees of freedom, i. e. , " definite answer. It may be surprising that one can
the state of the environment atom. Measurements say so much about the collapse without having to
mRde by the appRrRtus on R spin eigenfunction specify where or how it takes place. The aim of
along any other direction are to some degree this section is to argue that the very question dis-
obliterated by the interaction with the environ- cussed in this paper, as well as many other phys-
ment. In particular, no information about the ical)y interesting questions concerning the pro-
 POINTER BASIS OF QUANTUM APPARATUS: INTO %HAT. .. 1523

cess of measurement, can be answered without

having to decide whether, where, when, or how
{
In an arbitrary basis IAg) different from the
pointer basis, this equality will not hold; an ad-
the ultimate collapse occurs. ditional measurement by 8' will destroy the per-
We have to agree with von Neumann that. adding fect correlation between 8 and g established in
more and more apparatuses only delays the mo- the first step:
ment when the reduction of the wave packet would
have to occur. This infinite regress ean be term- IA,') 8 Q c, IA, ) 8 I
s)
inated only if there are some entities in the Uni-
verse which can put an end to the unitary evolu-
tion, be it macroscopic objects of Copenhagen in-
= Qd IA,') s IA, ) 8 Is) (5.4)
terpretation" or conscious beings preferred by
von Neumann himself, ' as well as by other prom-
with
inent physicists. '""
Alternatively, one could re- (5.5)
sign himself to the unitary evolution predicted by
the quantum theory for the Universe. Then it is What has been measured on g, by 8 is no longer up
the individual consciousness alone which appears to an observer to decide. Already at the level of
to evolve in a nonunitary fashion choosing a single the above etluations the pointer basis {IA~)) de-
path in this labyrinth, "many world" Universe.
Both of these opposing views of the collapse prob-
termines the corresponding relative basis Ip))
as the only choice. And the collapse of the wave
{
lem have been criticized and defended. What is function has not yet occurred; both Egs. (5.2) and
important for us is that both of them agree on (5.4) represent, beyond doubt, pure states. In
one absolutely crucial point: To describe the this sense the above considerations answer a
world "as we know it"' there must be two distinct question about the appearance of the collapse
types of evolution —
the reversible, deterministic without having to specify where one draws the
one, which has been confirmed for the microworld, "line. "
as well as the irreversible, random one which In real-world apparatuses the role of the "ad-
must provide for the choices experienced by the ditional apparatus 8'" or, equivalently, the role
consciousness. The two, interpretations differ of the "environment 8"
is usually played by part
only as to where the boundary between the unitary of the physical setup of the apparatus itself. For,
and nonunitary domains should be drawn. The what we have called "the apparatus 8" is just a-
many world interpretation allows nothing but the small part of the complete setup, which can be
individual consciousness in the "random" domain. fully described by a state vector in nondegener-
The Copenhagen interpretation extends that do- ate Hilbert space spanned by a set of basis vec-
main to include "macroscopic objects. "Both in- {
tors IA~)). In contrast to this simplified model,
terpretations do agree that the boundary must be setups of real-world apparatuses are much more
there; "in the empty courtyard many a game can- complicated and demand extensive product spaces
not be a game until a line has been drawn —.
. . , no to allow for a complete description. Out of this
—
matter where to separate one side from the vast product Hilbert space we have singled out
other. "" just one subspace, claiming it describes the
One may divide all the questions that can be "
"pointer, and hence epitomizes the apparatus
posed about quantum measurement into those itself. The rest of the apparatus setup described
which have answers depending on cohere the line by the remaining parts of the product space by —
is drawn, as opposed to the ones whose answers far larger than the subspace used to represent
depend only on sehether the line is drawn. We the apparatus proper —
describes then a natural
claim that the-question "Which mixture does the immediate environment 8. As long as the coup-
wave packet appear to have collapsed into?" be- ling between the apparatus 8 and its "built-in"
longs to the second category. Clearly, there is environment allows for the existence of the pointer
{
just one pointer basis IA~)) in which the addition-
al measurement
basis, the apparatus will be able to record the
corresponding relative states of the to-be-mea-
sured quantum system p. Therefore, part of the
IAo& IA. & I&&
"= IA'& 2 "IA.& I&& apparatus setup, the built-in environment, can be
said to act as an interface between the apparatus
proper 8 and the rest of the world.
"=gh, IA, &e IA, &e Ip&
The function of this interface is not to isolate 8,
(5 .2) as one might have at first guessed. On the con-
allows the equality
trary, it proves advantageous to couple 8 via a
Ib, I'= 15, I' (5.3) well-defined and carefully controlled X6& which
1524 H. ZUREK

leaves the pointer basis of the apparatus undis- 26 is intimately related to the existence of the
turbed. Consequently, the time evolution of the here-discussed pointer basis. We hope to give a
combined $6', 8 object preserves correlations be- more complete discussion of this relation in fu-
tween the pointer-basis eigenvectors ~A~) and the ture publications.
relative states ~P) of the measured system g. The
difficulty of isolating large quantum systems, VI. CONCLUSIONS
"
stressed, by Zeh, among others, and more re- We have shown that the interaction between the
cently by Wigner, '"" emerges as the crucial mo- quantum apparatus 8 and its environment 8 may
tivation: It proves easier to construct a con- single out a preferred pointer basis of the appara-
trolled coupling than to isolate. tus. This will happen always when the interaction
Let us moreover note that in the context of HamiltonianA, K~g commutes with an apparatus ob-
"many world" interpretation, Deutsch ' has re- servable II. Corre. lations betwe'en the set of eigen-
cently postulated existence of a preferred basis, functions of II and the corresponding relative
which he calls the "interpretation basis. " It is states of the system will then remain unperturbed
determined by the requirement that, at the instant despite the evolution of the apparatus generated by
of completion of any interaction, a measurement X@&. The choice of II determines what states of
has indeed taken place. the quantum system g can be recorded. Thus, in
Finally, it is worth adding that situations where a certain sense it is the environment of the appa-
a quantum system acquires a preferred basis be- ratus which participates in deciding what the ap-e@a

cause of its coupling to another system have al- paratus measures: The pointer observable II of
ready been discussed in the context of quantum the apparatus, the one on which the environment
theory of measurement. Simonius" has noticed performs the "nondemolition measurement, "
re-
that quantum systems, interacting with their mains the only one endowed with the maximum
natural environment interpreted as a "background information about the state of the quantum system
of probes, like photons or particles, " will reveal g.
"classical" features, i.e. , localization of mac-
ACKNOWLEDGMENTS
roscopic bodies, localization of atoms within
I would like to thank Professor John Archibald
Moreover, in the "Zeno Paradox"" "
molecules, and stability of metastable compounds.
metastable Wheeler for many stimulating discussions. I am
states of quantum systems are stabilized by the also grateful to Bill Wootters for helpful com-
appropriate coupling with other quantum systems, ments on the manuscript. This work was sup-
playing the role of external observers. ported by NSF Grant No. PHY-7826592 and by the
Each of the developments described in Refs. 19- Center for Theoretical Physics.

*Present address. 9E. P. Wigner, Am. J. Phys. 31, 368 (1963).

J. von Neumann, Mathematische Grundlagen der M. Scully, R. Shea, and J.
D. McCullen, Phys. Rep.
Quantenmechanik (Springer, Berlin, 1932); Mathe- 43C, 485 (1978).
matical Eoundations of Quantum Mechanics (Prince- E. P. Wigner, in The Scientist Speculates, edited by
ton University Press, Princeton, New Jersey, 1955). I. J. Good London, 1961; Basic Books,
(Heinemann,
Relative states were introduced by H. Everett, Rev. 1962), p. 284; reprinted inE. P. Wigner,
New York,
Mod. Phys. 29, 454 (1957). Symmetries and Reflections (Indiana University Press,
3L. Landau and R. Peierls, Z. Phys. 69, 56 (1931). Bloomington and London, 1967) p. 171.
D. Bohm, Quantum Theory (Prentice-Hall, Englewood One might object to this simple equation for «) by
~

Cliffs, New Jersey, 1951). noting that p, ) has undergone a scattering from a
~

V. B. Braginsky and A. B. Manukin, Measurement of o(r —rz) potential, while ~p, ) did not. Hence, even
I
S'eak orces in Physics Experiments (University of after recombination g, ) & P&). This problem can
~ ~

Chicago Press, Chicago, 1977); V. B. Braginsky, be easily remedied by placing a second, identical
Yu. I. Vorontsov, and F. Ya. Khalili, Pis'ma Zh. scatterer symmetrical to the bistable atom along the
Eksp. Teor. F iz. 27, 296 (1978) [JETP Lett. 27, 276 trajectory of the spin l). ~

(1978)]. Delayed choice measurements. were proposed by J. A.

K. S. Thorne, R. W. P. Drever, C. M. Caves, M. Zim- Wheeler, in Mathematical Eoundations of Quantum
mermann, and V. D. Sandberg, Phys. Rev. Lett. 40, Theory, edited by A. R. Marlow (Academic, New
667 (1978); K. S. Thorne, C. M. Caves, V. D. Sand- York, 1978). For the relationship between delayed
berg, M. Zimmermann, and R. P. Drever, in Sources choice and nonseparability, see also W. K. Wootters
of Gravitational Radiation, edited by L. Smarr (Cam- and W. H. Zurek, Phys. Rev. D 19, 473 (1979).
bridge University Press, Cambridge, 1979), p. 49. A. Einstein, B. Podolsky; and N. Rosen, Phys. Rev.
W. G. Unruh, Phys. Rev. D 19, 2888 (1979). 47, 777 (1935).
H. Wakita, Prog. Theor. Phys. 27, 139 (1962). 15
F. London and E. Bauer, La Theoric de l'Observation
 POINTER BASIS OF QUANTUM APPARATUS: INTO %HAT. ..
eg Mecanique Quantique (Hermann, Paris, 1939); OE. p. Wigner, Interpretation of Quantum Mechanics,
English translation in Quantum Theory of Measure- lecture notes of E. P. Wigner in Quantum Theory of
ment, edited by J. A. Wheeler and W. H. Zurek Measurement, edited by J. A. Wheeler and W. H.
(Princeton University Press, Princeton, to be pub- Zurek (Princeton University Press, Princeton, to be
lished). published) .
6This is also a sufficient condition for II =gw& )4&) ('A& [ E. P. Wigner,invited lecture at the IVth Inter-Am-
to be a nondemolition observable lC. M. Caves, K. S. erican Undergraduate Conference in Theoretical
Thorne, R. W. P. Drewer, V. D. Sandberg, and Physics, The University of Texas, 1981 (unpublished).
22
M. Zimmermann, Rev. Mod. Phys. 52, 341 (1980)]. D. Deut8ch, University of Texas report, 1981 (un-
"M. Jammer, in The Philosophical DeveLopment of published).
Quantum Mechanics (Wiley, New York, 1974), gives M. Simonius, Phys. Rev. Lett. 40, 980 (1978).
an excellent discussion of the different views of quan- B. Misra E. C. G. Sudarshan, J. Math. Phys. 18,
and
tum measurement. 756 (1977); C. B. Chiu, E. C. G. Sudarshan,
J. A. Wheeler, Erontiers of Time (North-Holland, 25
and B. Misra, Phys. Rev. D 16, 520 (1977).
Amsterdam, 1979), p. 404. K. Kraus, University of Texas Report No. DOE-ER-
~
H. D. Zeh, in Eoundations of Quantum Mechanics, pro- 03992-390, 1980 (unpublished).
ceedings of the International School of Physics "Enrico A. Peres, Am. J. Phys. 46, 745 (1978); 48, 931
"
Fermi, 1970, edited by B. D'Espagnat (Academic, (1980).
New York, 1971), pp. 263-273.