You are on page 1of 35

Bell Inequalities –

A Historical Perspective

4 February 2013
Physics 135b, Winter 2013
Is the moon there
when nobody looks?
Reality and the quantum theory
Einstein maintained that quantum metaphysics entails spooky actions
at a distance; experiments have now shown that what bothered Einstein
is not a debatable point but the observed behavior of the real world.

N. David Mermin

Quantum mechanics is magic1 ition expects to be there. in part of space A; it should also be
Einstein didn't like this. He wanted independent of whether or not any
In May 1935, Albert Einstein, Boris things out there to have properties, measurement at all is carried out
Podolsky and Nathan Rosen published2 whether or not they were measured4: in space A. If one adheres to this
an argument that quantum mechanics We often discussed his notions on program, one can hardly consider
fails to provide a complete description objective reality. I recall that dur- the quantum-theoretical descrip-
of physical reality. Today, 50 years ing one walk Einstein suddenly tion as a complete representation
later, the EPR paper and the theoreti- stopped, turned to me and asked of the physically real. If one tries
cal and experimental work it inspired whether I really believed that the to do so in spite of this, one has to
remain remarkable for the vivid illus- moon exists only when I look at it. assume that the physically real in
tration they provide of one of the most The EPR paper describes a situation B suffers a sudden change as a
bizarre aspects of the world revealed to ingeniously contrived to force the quan- result of a measurement in A. My
us by the quantum theory. tum theory into asserting that proper- instinct for physics bristles at this.
Einstein's talent for saying memora- ties in a space-time region B are the Or, in March 1947,
ble things did him a disservice when he result of an act of measurement in I cannot seriously believe in [the
declared "God does not play dice," for it another space-time region A, so far quantum theory] because it cannot
has been held ever since that the basis from B that there is no possibility of the be reconciled with the idea that
for his opposition to quantum mechan- measurement in A exerting an influ- physics should represent a reality
ics was the claim that a fundamental ence on region B by any known dynami- in time and space, free from spooky
understanding of the world can only be cal mechanism. Under these condi- actions at a distance.
statistical. But the EPR paper, his tions, Einstein maintained that the The "spooky actions at a distance"
most powerful attack on the quantum properties in A must have existed all (spukhafte Fernwirkungen) are the ac-
theory, focuses on quite a different along. quisition of a definite value of a proper-
aspect: the doctrine that physical prop- ty by the system in region B by virtue of
erties have in general no objective Spooky actions at a distance the measurement carried out in region
reality independent of the act of obser- Many of his simplest and most explic- A. The EPR paper presents a wave-
vation. As Pascual Jordan put it3 it statements of this position can be function that describes two correlated
Observations not only disturb found in Einstein's correspondence particles, localized in regions A and B,
what has to be measured, they with Max Born.5 Throughout the book far apart. In this particular two-parti-
produce it. . . . We compel [the elec- (which sometimes reads like a Nabokov cle state one can learn (in the sense of
tron] to assume a definite posi- novel), Born, pained by Einstein's dis- being able to predict with certainty the
tion. . . . We ourselves produce the taste for the statistical character of the
results of measurement. quantum theory, repeatedly fails, both David Mermin is director of the Laboratory of
Jordan's statement is something of a in his letters and in his later commen- Atomic and Solid State Physics at Cornell
truism for contemporary physicists. tary on the correspondence, to under- University. A solid-state theorist, he has
Underlying it, we have all been stand what is really bothering Ein- recently come up with some quasithoughts
taught, is the disruption of what is stein. Einstein tries over and over about quasicrystals. He is known to PHYSICS
being measured by the act of measure- again, without success, to make himself TODAY readers as the person who made
ment, made unavoidable by the exis- clear. In March 1948, for example, he "boojum" an internationally accepted scienti-
tence of the quantum of action, which writes: fic term. With N. W. Ashcroft, he is about to
generally makes it impossible even in start updating the world's funniest solid-state
That which really exists in B physics text. He says he is bothered by Bell's
principle to construct probes that can should . . . not depend on what theorem, but may have rocks in his head
yield the information classical intu- kind of measurement is carried out anyway.

38 PHYSICS TODAY / APRIL 1985 0031-9228 / 85 / 0400 38-10/$0,1.Q0 © 1985American Institute of Physios
Downloaded 04 Jan 2013 to 131.215.225.9. Redistribution subject to AIP license or copyright; see http://www.physicstoday.org/about_us/terms
I.8 CAN QUANTUM-MECHANICALDESCRIPTIONOF V ( r 1 ,1 2 ) :
PHYSICAL REALITY BE CONSIDEREDCOMPLETE?
where
9 '(x') =
BontsPoporsry,ANDNnrHnNRosBN
ArseRrErxsrErN,
This *, however ir
operator
p=(

correspondingto tl
momentumof the se
hand, if .B is the coc
it has for eigenfunct
trr(f,r
In a complete theory there is an element corresponding quantum mechanics is not complete or (2) these two
t o e a c h e l e m e n to f r e a l i t y . A s u f f i c i e n tc o n d i t i o n f o r t h e q u an t i t i e sc a nn o t h a v es i r n u l t a n e o urse a l i ty . C o n s i d e r a t i o n
r e a l i t yo f a p h y s i c a q
l u a n t i t f i s t h e p o s s i b i l i t yo f p r e d i c t i n g of the problem of making predictionsconcerninga system
correspondingto
it with certainty, without disturbing the system. ln o . nt h e b a s i so f m e a s u r e n r e n tnsr a d eo n a n o t h e rs y s t e mt h a t 6(rr-r) is thewell-i
q u a n t u m m e c h a n i c si n t h e c a s eo f t w o p h y s i c a lq u a n t i t i e s h a d .p r e v i o u s l yi n t e r a c t e dw i r h i t l e a d st o t h e r e s u l tt h a t i f Eq. ( 8) in t his case
d e s c r i b e db y n o n - c o m n r u t i n go p e r a t o r s ,t h e k n o w l e d g eo f ( l ) i s f a l s et h e n ( 2 ) i s a l s o f a l s e .O n e i s t h u s l e d t o c o n c l u d e
o n e p r e c l u d e st h e k n o w l e d g eo f t h e o t h e r . T h e n e i t h e r . ( l ) t h a t t h e d e s c r i p t i o no f r e a l i t y a s g i v e n b y a w a v e f u n c t i o n
t h e d e s c r i p t i o no f r e a l i t . vg i v e n b y t h e w a v e f u n c t i o n i n . ; is not conrplete. V(rr, re):

where
l. ; Whatever the meaning assigned to the term
complele, the following requirement for a com- P* (:
) A NY _ serious consideration of a physical ,p,(xr) :
I s t 2 t i th )
,( \ thedry must take into account the dis- plete theory seems to be a necessary nery J*
I elemenl of the physical reality must houeo counter-
( tinction betweenthe -obiectiverealitv. which is
port in the physical lheory. \\ie shall call this the
{ independentof any theory, and the physical
i conceptswith which the theory operates.These condition of completeness.The second question This g", however,
conceptsare intended to correspondwith the is thus easily answered,as soon as we are able to operator
objectivereality, and by meansof theseconcepts decide what are the elements of the physical
we picture this reality to ourselves. reality.
judge The elementsof the physical reality cannot corresponding to tl:
In attempting to the successof a
coordinate o[ the sec
physicaltheory, we may ask ourselvestwo ques- be determined by o priori philosophicalcon-
tions: (1) "Is the theory correct?"and (2) "Is siderations,but must be found by an appealto PQ-
the descriptiongiven by the theory complete?" results of experimentsand measurements. A
It is only in the casein which positiveanswers comprehensive definitionof reality is, however, we have shown that
may be given to both of thesequestions,that the unnecessary for our purpose.We shall be satisfied 9* and 9, to be eigr
conceptsof the theory may be said to be satis- with the followingcritericn,which we regardas muting operators,
factory. The correctness of the theory is judged reasonable.If , uilh,oul in ony uay disturbing quantities.
by the degreeof agreementbetween the con- syslem, wc con predicl wilh certainty (i.e., wilh Returning now tc
clusionsof the theory and human experience. probobi,lityequol to unily) the aolue of o physical' plated in Eqs. (7) a
This experience, which aloneenablesus to make quontity, then there existson elementof physicol. and g, are indeed e
inferencesabout reality, in physics takes the reality corresponding lo this physicalquontity. lt commuting operatorl
form of experimentand measurement.It is the seemsto us that this criterion.while far from the eigenvaluespr al
secondquestionthat we wish to considerhere,as exhaustingall possibleways of recognizinga measuring either .4 r
appliedto quantum mechanics. physicalreality, at least providesus with one predict with certaint

Originallypublishedin PhysicalReview,47, 777-80( I 935).


140 EINSTEIN, PODOLSKY, ROSEN

sight this assumption is entirely reasonable, for infinite series (7) is reduced to a single term
the information obtainable from a wave function ,l,rk)ur@).
seems to correspond exactly to what can be The set of functi onsa^(rr) i s determi nedby
measured without altering the state of the t h e c h o i c eo f t h e p h y s i c aql u a n t i t yA . l f , i n s t e a d
' thi s,
system. We shall show, however, that this as- of w e had chosenanotherquanti ty,say I,
sumption, together with the criterion of reality havi ngthe ei genval ues br, bz,D r, ' ' ' and ei gen-
given above, leads to a contradiction. functi onsur(rr), r,z(rr), aa(rr), w e shoul d
haveobtai ned,i nsteadof E q. (7), the expansi on
2.
V (r,, * r): i ,p.(r:)un(rr), -(8)
eho -,i*64, a
For this purpose let us suppose that we have r-l <- .r 4.{f...-i' ba&b
two systems, I and II, which we permit to inter-
where s,'s are the new coefficients.If now the
aqt from the time l=0 to l=7, after which time
q u a n t i t y B i s m e a s u r e cal n d i s f c l u n dt o h a v e t h e
we supposethat there is no longer any interaction
v a l u e 0 , , w e c o n c l u d et h a t a f t e r t h e m e a s u r e m e n t
between the two parts. We suppose further that
the first system is left irr the state given by u,(rr)
the states of the two systems before l:0 were
and the second system is left in the state given
known. We can. then calculate with the help of
by p,(rr).
Schrddinger's equation the state of the combined
We see therefore that, as a consequenceof two
system I+II at any subsequent time; in par-
different measurements performed upon the first
ticular, for any l>7. Let us designate the cor- system, the second system may be left in states
responding wave function by !V. We cannot, rwith two different wave functions. On the other
however, calculate the state in which either one hand. since at the time o[ measurementthe two
of the two systems is left aftdr the interaction, systems no longer interact, no real change can
This, according to quantum mechanics,can be tike place in the second system in consequence
done onlytwith the help of further measurements, of anything that may be done to the 6rst system.
by a process known as the reduction o! the woae This is, of course,merely a statement of what is
pochet. Let us consider the essentials of this meant by the absenceof an interaction between
process. the two systems. Thus, fu$ possible to ossjgtptllo
'some
L e t a r , a z ,& s , . . . b e t h e e i g e n v a l u e os f dfuryfit wouefunclions.(in our eNample p1 and
physical quantity .4 pertaining to system I and p,) t_othe samee{\(the second system after
zr(rr), uz(x), u3(fir), t h e c o r r e s p o n d i n g t h e i n t e r a c t i o nr v i t h t h e f i r s t ) .
eigenlunctions,where rr stands fclr the variables Now, it may hapllen that the two wave func-
used to describe the 6rst system. Then V, con. t i o n s , 9 * a n d g r , a r g e i g e n f u n c t i o n so f t w o n o n -
sidered as a function of ff1,cdo be expressedas commuting operators corresponding to some
physical quantities P and Q, respectively.That
e."* ',,tl+lt 1' _
+ !t,(r1, f,z)= |n(xiu^(x), (7) this may actually be the case can best be shown
F, by an example. Let us suppose that the two
systems are two particles, and that
where rz stands for the variables used to describe
the secondsystem. Here f.(rs) are to be regarded A@

V(rr, *i = I g?tith)(zr-tttts\vr1,p, (9)


merely as the coelficientsof the expansion clf V
,J __
into a series of orthogonal lunctions zn(rr).
Suppose now that the quantity A is measured where xs is some constant. Let A be the momen-
and it is found that it has the value ar.. It is then tum of the 6rst particle : then, as \ryehave seen
concluded that after the measurement the first i n E q . ( 4 ) , i t s e i g e n f u n c t i o n sw i l l b e
system is left in the state given by the wave
u ,(x) = 6l2tilhl ott ( 10)
function u*(xt), and that the second system is
left in the state given by the wave function correspondingto the eigenvaluep. Since we have
g*(rr). This is the process of reduction of the here the case of a continuous spectrum, Eq. (7)
wave packet; the wave packet given by the w i l l n o w b e w r i t t e n
,,, < nui:*I*:!dlr"

'+l

i.Lf, rf.) *
r
Jtl,+pr)
I ' 8 PHYSICAL REALITY 141
/
IN OF f- disturbingthe secondsystem,either the value
V(rt, xr): ,l*')' o(x')d|, ( r1 ) of the quantity P (that is po) or the valueof the
;TE? J*t
quantity Q (that is g,). In accordance
where with our
g , ( X z ): g - t 2 t i t h(): t - e uP
). ( t 2 ) criterion of reality, in the first case we mus!
EN congidE_llhe qugntitv-P jrs beine an elementof
This po however is the eigenfunction of the reality, in the secondcasethe quantity
Q is an
operator ilrenrentof reaIty.. tsut, ulg_hlve seen,ptll
element
P=(h/2ri)0/dxz, ( 1 3 ) wave functions *r and
wave_functionlt! and s,
9, bglong
belong to the,samq
the,same
re{ity.
corresponding to the eigenvalue - p of the Previously we proved that either (1) the
momentum of the second particle. On the other quantum-mechanical descriptionof reality given
hand, if B is the coordinate of the first particle, by the wave function is not completeor (2) when
it has for eigenfunctions the operators correspondingto two physical
quantitiesdo not commute the two quantities
,lete or (2) these two, t',(rr) : d(rr -.r), (l 4 )
cannothave simultaneous reality.Starting then{f
s reality. Consideration
ns concerninga system corresponding to the eigenvalue r, where with the assumption that the wave function! E
r-
on another system 6 ( r t - r ) i s t h e w e l l - k n o w n D i r a c d e l t a - f u n c t i o n . doesgive a completedescriptionof the physicalI
ads to the result that i Eq. (8) in this case becomes reality, we arrived at the conclusion that two '
: is thus led to concl physical quantities,with noncommutingoper-
ven by a wave function', ators, can have simultaneousreality. Thus theb'
v(r,, .r, : , ( 1 5 ) negationof (1) leadsto the negationof the only
f:p,(xz)a,(xt)dr,fot?.
where o other alternative (2). We are thus forced to
signed to the term
rement for a
t t' Xo- concludethat the quantum-mechanical descrip-
ecessary one: anryi p,(xz) : f* s't itht:-',::il- tion of physicalreality given by wavi functions
J* is not complete.
must lnae o counler-,
One could object to this conclusionon the
e shall call this the. : h 6 ( x- rr+ f,g). (16)
groundsthat our criterion of reality is not suf-
he second question
This s,, however,is the eigenfunctionof the fi'cientlyrestrictive.Indeed,one would not arrive
)n as we are able to at our conclusionif one insistedthat two or more
operator
ts of the physical ( 1 7 ) physical quantities can be regardedas simul-
Q=x,
taneouselementsof reality only whenlhey con be
ical reality cannot corresponding to the eigenvaluer**o of the simullaneouslymcasuredor predictel.-On this
philosophical con- coordinateof the secondparticle.Since point of view, sinceeitherone or the other, but
rd by an appeal to
PQ-QP=h/Zd, (18) not both simultaneously, of the quantitiesP
measurements. A !,and Q can be predicted,they are not simultane-
'eality we have shown thai it is in generalpossiblefor ously real. This makes the reality of P and Q
is, however,
ile shall be satisfied p1 and 9, to be eigenfunctionsof two noncom- dependupon the processof measurementc?rrr?d
which we regard as muting operators, correspondingto physical out on the first system,which doesnot disturb
y way disturbing o quantities. the secondsystem in any way. No reasonable
cerlointy (i.e., utilh Returning now to the gerreralcase contem- definition of reality could be expectedto permit
aalue of a physicol plated in Eqs. (7) and (8), we assumethat 9r this.
element of physical and ,p, arc.indeedeigenfunctions of some non- While we have thus shown that the wave
'\lsical guantity. lt commutingoperutorsP and Q, coiresponding to functiondoesnot providea completedescription
on, while far from the eigenvalues 2r and 0,, respectively. Thus, by of the physicalreality,we left openthe question
s of recognizing a measuringeitherA or B we are in a positionto of whetheror not sucha descriptionexists.We .
rvides us with one prtidictwith certainty,and without in any way believe,however,that such a theory is possible.

q nf" -Qr,f.) -erl^.nrr.l


J.-0o$.o* r^f*-r-,"^ f,d...l,'tte'rc{
-
/ 5..'r?r f t
o:x1.\& --(drrlr) Xt
), rtg /er.lra",."
l, xr
t!
s . ,.'fru!,1 r',-: , I
a-1 f
'1.
,-4 .\e" a
.lr"r"-'o +'.', r,Y'i" "';lc'
theory. ft seer
III.2 THE PARADOXOF EINSTEIN,ROSEN,AND PODOLSKY
physicaltheorv
(1) Every e
Dnvp Bonu
----- completcphysicr
--:a

-4 l..Xr;r...rirr,J . As to n'hat a
I
F physicaltheory
o.
't decidedfinaliy r
.y.o--9
t u r.:.irr.] r.r rn ) neverthelesssug
f,ln'*tt-''
of reality, rvhicl
! l, a,'4 ,.lrq.lq
(2) If, rvithc
t
certainty (i.e., '
quantity, then
physicalquanti'
The authors
recognized in ot
restricted onese
criterionalone,
results.
''I The useof tl
assumptions,u'.
i authors, but rrl
(3) The rvo:
16. The Paradoxof Einstein, Rosen,4nd Podolsky. fn an article in separatelyexist
the PhysicalReview,t Einstein,Rosen,and Podolskyraisea seriouscriti- (4) Every o
cism of the validity of the generally'accepted interpretationof quantum definedmathem
theory. This objectionis raisedin the form of a paracioxto rvhichthey We shall"te.
are Ied on the basisof their analysisof a certain hypotheticalexperiment, order to permit
which we shall discussin detail later. Their criticism has, in fact, been authors, but in
shownto be unjustified[] and basedon assumptionsconcerningthe nature appliedat the c'
of matter which implicitly eontradict the quantum theory at the outset. Ilo\r', let us
Nevertheless,theseimplicit assumptionsseem,at f;rst sight, so natural that all relevan
and inevitable that a careful study of the points which the authors 1v&vefunotion,
raised affords deep and penetrating insight into the differencebetween differ hy at rnos
classicaland quantum conceptsof the nature of matter. quanl,umstate.
The authors first undertook to definecriteria for a completephysical reality rvasto sl
theory is untcnr
a completedesc
J Phys.Rca.,47r777 (L935). of reality") ex
(l N. Botrr,Phys,Rev,48r696 (1935);W, H. Furry, Phys.Rev.49,393,476 (1936). proved, then or
ArA \a ef'-'
t This criterior
Podolsky do not r
reality ahvays has
Originallypublishedai sections15-19,Chapter22 of QuantumTheory,David Bohm, pp. 6t l-23, that this counterp,
.t See Chap. 9,
EnglewoodCliffs (195l).
Prenticc-Hall,

⎧ ↑↓ − ↓↑ ⎫
⎪ 1,2 1,2

⎪ ↑↓ + ↓↑ ⎪⎪
⎪ 1,2 1,2
Belli 1,2 = ⎨ ⎬
⎪ ↑↑ 1,2
+ ↓↓ 1,2 ⎪
⎪ ⎪
, i,ili,ti-.*i;

iil.'ii+r I-
. ;:j:;l::
'l:!';:";n,i.

.rlrf;j:,i

⎪ ↑↑ − ↓↓ 1,2 ⎪
'
"::,,.r

⎩ ⎭
.' i.r.-

1,2
NI.2 EPRPARADOX 361
,e value. This Theee are
. of atom No. 1
9o : u+(1)u+(2) *, = u1(1)u-(2)
nt of the same
rthesis,the two h : u-(1)u-(2) 9a : u-(1)u+(2)
f measuring a,n where za and u- are the one-particlespin wave functions representing,
rut in any wa,y respectively,a spin h/2 and - h/2, and the argument (l) or (2) refers,
I an element of respectively,to the particlewhich has this spin. Norv 9. and parepresent
rve measuredc, the two possiblesituations in rvhieh each particle has a definite z com-
s an element of ponent of the spin in a direction rvhich is oppositeto that of the other.
ris is true, how- The wave function for a system of total spin zero is the following linear
neaf
r I
Jr
icle No. 2 even combinationof 9. and *a (seeChap. 17, Sec.9): ,t
ace. For since L( l(
I
+
"*i^7!d
>f measurement 4.L I *?cDotJes 0o:ft<*,-fii
et us remember
t the apparatus
AJ"-i7-Il
^JT)-
The particular sign with which 9. and {a $e combined is of cmcial
ht, and thus to
importancein determining the combinedspin, for if they are combined
nponent in any
n'ith a * sign, one obtains an angular momentum of ll (but with a ?r;to
,hed without iu
vnlue of the z componentof the angular momentum). We denote this
criterion (2) of
result below:
r must exist in
ition of all three gt = fa) ? (27)
pecify, at most, hw,*
recision,we are It is clear, then, that the total angular momentum is an interference
r not provide a property of 9, and 9a. On the other hand, the only statesin which each
hesecondatom. particle has a definite spin oppositeto.that'of the other are represented
look for & new either by 9. or by 9a separately. Thus, in any state in which the value
on w&srpossible. of a, for each particle is definite, the total angular momentum must be
y ERP involves indefinite. Vice versa,rvhen'everthe total angularmomentum is definite,
hat the world is then neither atom can correctlybe regardedas having a definitevalue of u ⎧ ↑↓ − ↓↑ ⎫
fined t'elements its own spin, for if it did, there could be no interferengebetweenf. and ⎪ 1,2

1,2

lifferent picture
9a, and it is just this interference which is required to prodrrcea definite ⎪ ↑↓ + ↓↑ ⎪⎪
' ⎪ 1,2 1,2
. This picture total angular momentum.' !t Belli 1,2 = ⎨ ⎬
on of the hypo- Besidesleading to a definite value of the combined spin, however, ⎪ ↑↑ 1,2
+ ↓↓ 1,2 ⎪
k of the theory. definite phaserelationsbetweenf, and 9ahave additionalphysicalmean- ⎪ ⎪
rg to Quantum ing, for they also imply that if the samecomionent of the spin of each ⎪ ↑↑ − ↓↓ 1,2 ⎪
,hat the present atom is measured,the resultswill be correlated. Such correlationscan
⎩ 1,2 ⎭
Einstein, Rosen be demonstrated,for example,in a processin which the z componentof
to be described the spin of eachatom is measuredby allowingeachatom to passthrough
a separateStern-Gerlach apparatus(seeFig. 1). For the sakeof,simpl,;^-
lour basic w&ve ity, rve can supposethat both spins are measuredat the same time,
re constructed.r although no results rvill dependsignificantly on this assumption. The
by multiplyingthe Hamiltonianat the time of measurement is then [seeeqs.(10a)and (f0b)J:
:pend on tbe sD&ce
lV : p(Ko * z'gc!o)ot,
* p(JCo* ztB!o)o2,,

11'l +
,
a (.1(,: . il_,:) ,i i' .l+t t i\
rnt;

enA r I
/
<t polar,%il ceTatt4;otr S
aF ,/
J
.rG t
- '
4,
>
2
PHYSTCAL REVIEW VOLUME 17, NUMBER T , AN UAR Y l. lell
;l
ty*
tf
:i
contaminations. Tbe active Cus wagpackedin a smallil capsulcC
I'
Letters to the Editor 8-mm diameterand 8-mm lcngtb. Thi annihilationradiatioue1 iF
collimatedby a lead block 6X6X6 in. tith s l-in.channel drilld
through the cent€r of the block, such thrt thc spread of the bcaq
UBLICATION o! bricl reports of hnportanl discoacricsin was found to be lessthail'3o.The aluii&um scatteren wercI iq
i;
plrysicsmay bcsccurd by ddressing lhcm lo lhis dcperlmal. in dia,meterand f-in. tong. Tley *erc dcsigntd to abrorb eba
i The closingdalafor thisdepartmailis foc wceksprim lo lhc datc o! 40 percent of the annihilatiort"fadiatirtri'l€ntthwisc and to lioit {
t issue.No prooJwill besentlo thc oulhors.Thc Board of Edilors doa '6i
I the'multiph ir:atterinf ih€ aUiaUon ircfit er 90o tob
! nol hold itse$ rcsponsiblcjm thc opinions c*prasd by lhc corre- than 15 percent. The crystal of the @unFr subten& rn angle
.,T {
spondails.Communications shouldnol cxccd 6ffi uords in lcnglh. 43o at the point in the tieiai rhaii 20 ptrcent of the incidot
rl

t,
' a ;
radiation has beenabsorbed:that is, et thc ibrcrption r.idpoilt
lll r of the scatterer. The me1n scattCring lnglc is very closc to t?,
the predictedmaximumof anisotntpy.Under thescconditiou, $;
i{,, The Angular Correlationnf Scattered
scatteredradiation tekin as the iourtting diftcrencedetectedb;
the scintillation counttr rith and without thc scatterer in pbo
6s3ih ilation Radiation*
rt;,
.11
jr C. S. V/u ero l. Snarxov
is three times the over-all background.
In taking the coincidince niea3utefithtt, one detector waslq
'dCtrctor
I'ufin Physics Labordtorict Yoth.Nao Yorh fired in poaition, and tbc sccond ras briented to foq
di:
Iril
: I'
!:J:*r"ri.\r;i;,tr.Nc., different positions with iiinuth dificnincct (p) of 0", 9f, lf,
,dt r, and 270nbetweenthe dctittor iuis. AJiEr ine-t,the scconddettctl
'li'
ilr{l :
I S earty as 1946,J. A. Wheelertproposedan experimentto
,( r verify a predictionof pair theory, that the two quanta was kept 6.ted and thc i*t 6nt r6t{ttd. The total petiodd
i1; emitted in the annihilationof a position-electron pair, with zero
measurementlastedabout 30 continuctuihouB. On accountoftL
high coincidenccratt obctrvad (tle tfie toincidchcerata forfr
relative angular momentum, are polarized at ryh!_g!glgg_!9,
perpendiorlat poaition et thl btginnini rfrs 6f the order of ffl
each other. This suRgestion involvescoincidence measuremcntg
per minute), thc autisticel d*iitioiia etc much impmfd r
ot-ih-'ffiitering oiloth the anhihilation photons at various
compared to the rc3ulB from G-M couhtln. The asyranctl
aaimuths.l'he detailedtheoreticalinvestigationswere reported
ratio from our best nrn is
by Pryceand lVardr and by Snyder, Pasternack,and Hornbostel.t
The predictedmarimum asymmetryratio of coincidence counts Coinddencl counting tete (.11.)
r2'04+0'0E'
iii ,: when tle two countersare at right anglesto each other to coin- f€
lflri cidencecountswhen the countersare co-planaris as large as 2.85
and occursat a scatteringangleof tl=82". Bleuler and Bradtr
where*0.06 is tbc pfobfblt incatrcrior. Thc calnrlatederyi
l{ili usedtwo end-windowG-M countersas detectorsand observedan
metry ratio for our gtadttricd lrffracihclrt ii 2.00.Th
asymmetryratio not inconsistentwith the theory. Nevertheless, the agreement is vcry satisfcctorfiFilftha rvorkfubeingph$C
the margin of error associatedwith their resultsis so large that a
to ett€nd thc invcstigttjdnstd rliifd'idcd gpomctrical conditic'
detailedcomparisonbetweenthe theory and experimentsis made We wish td erprus our eppreciation to Professon I. I
l,jiii rather difficult. In the meantime,Hannar perforn,edsimilar
Dunning, W. W. Hrvcirs, Jr., and t. t, Rrinriter for theird
'tncourig;neot, Wc rbd,{ish to rhrnl tl
and found stant iritercst and
ftlii experiments with moreefrcientcounterarrangements
the asymmetry ratio observedto be consistently srnallcr than
cyclotron$oup for pttpdnB thi Cq.. souici ind thc U. S.AE
l!*ri tJrosepredicted.Therefore,it appearedto be highly desirableto wbich aidedmateriellyin tlc pcrloitiiiacg6f thir rescarc'h. .

rffiii
'ij;;1,
reinvestigatethis problem by using more efrcjeut dctectorsand
more favorableconditions.
The recentlydevelopedscintillationcounterhas beenproved
to be a reliableand highlye6cientgamma-raydetector.With this
. Partielly supportcd by thc AFC.
t J. A Wheeler, Ann. New Yorh Acrd. Sd. {t, 2t9 (1946).
t M. H. L. Pryce ard J. C. Wa.-. Nrtun 16o,435 (terz1.
r Snyder. Paltern cL eqd Ho.abctcl. Phyr. Rcv.6t, {{o (194t).
I E. Bleuler and H. L. 8redt. Phyr. Rev. 73. l39t (l9lt).
I R. C. Hanrra, Nature f62. 332 (t94t).
i . : lli l improved efficiency,which is around ten times tbat of G-M
r |
": counters,there will be an increasein the coincidence..counting
i l ; ,J .
r l r t ' r l
rate of one bundredtimes. In our experiments,two RCA 5819
i
photo-multipliertubesand two anthracenecrystals tXlXl in.
dl
lini
i,'l
rvereused.Tbe efficiencyfor the annihilation radiation obtained
with tbesc anthracenecrystals is seven to eight percent which
comparesfavorably with the calculatedvalue. The geometrical The OpticalDetectionof dadiofrequency
ir ll ;i l I arrangementis schematicallyshownin Fig. 1. Resonance
: l,'if;l' ' r
' ' ll' The positronsourceCus was activated by deuteronbombard' M. t{. L PiYcr
ment on a coppcr target in the Columbiacyclotron. The electro. Ckrcndor Labqatory, Otfqd, Lnlbnd
plating methodwas employedto separateCu activity from other Octobcr31, l9i9

T N a recent DaDerundei thk titte, Bittcri iliecrrgrestlc 44


tnI I a radiofrequin.y 6eld on thE optical Zctmann efiect. Ecl
trat6 the question by treating er etoaic syltetrr thcq,4
t
state is rS, making optical.transitionsto r ? rtate. Thc rtoo.
in a stead'ymagnetit 6eld Er on Hhich is.supcrposcd t rory
'errgular
t. radiofrequencyficld:fl6, in thc rf plahfr of frcquc'g
I According to Bitteq wben therc is a rclonance betrccl r
uo- tpofl,/1, the precessionfrequcncy of thc 4in Eoqco::
!.9 ground state, certein obscrvableghengerleppcu to lbe'tp
a a
Al lnirlcrr
c,trfth

o | 2 3. ! 3'

F'r6. t. Schematic diagram of experiment.


Vol.4l, 1978.printedin GreatBritain
Uo Prog.'Phn., !k
I {#^
I
I
I
1 B.ll't theorem: experimentaltests and implications
II
I
lj ('1,.\ustiRt
andABl{ERst-III,IoNyI
$
I t|.ttt\
Laboratorv-L-'t37, ivlagneticFusionEnergvDivision' Livermorc'
| :il::;:S*ilril3rc
BostonUnivcrsity,Boston,Massachusctts 022i5'
I iii.prri-.";;;l Ph1'sics
andPhilosophy,
I us.r
I
I
I Ahtract
I B.ll'.*rlrcorcnrrcprcscnts a significantadvanccin understanding the conccptual
I found,,tinns of quantum mechanics. The thcorem shorvs that essentiallyall -lgl
f ,rcri.:t of natural phenomena that are formulated rvithin the framewgrk of .5gaE1 ' PCr)2, a
*r U.grsing a singleexperimentalarrangement.Moreover, the prediaions
|
,!5..-rh.ffiies must significantlydiffer from those by quantum mechanics.
! Erpcrirn"ntal rcsults cvidcntlyrefutethe thebrem'spredictionsfor thesctheoriesand
'I'hc..conclusions
thosc of qurtntum mcchanics. are-philosophically
I hrorr startling:
onc
,ifisr must totallyabandon thc-rcalisticphilosophy of most workin! scicntisi,
eldr:urutic:tllyrcl'iscdur conccptof spitcc*tirne.

I'his rcviervwasrcceivedin Fcbruarv 1978.

<' t ar<rfoa' J.'v'-'c" to i|: '} '*t' 4o"tl o N


Fr*r.
7.1,7"1;'' f--
, o* (' tell'lot !on'\h.e-io.t^
{?ruc. ,!+-. 1"'11
0'O1)tt'i"7"'ll *
,-orl not.^tJ
L'./-
'l,r
'c L;el"- c

I -"[
,|
",iotctq-
( ::11.i*..
)tuf:ttcs,-/
\

f \\:orksupp'rtcd in parr by the National science Foundation.

lJ#835i7tj,'0012-1881s05.00 @ 1928 1'he Institutc of physics


g a bI[r>, lt>,-lu?,Irt"1
Bei
1886 J F Clausr andA Shimony

somchowin a singletstate,for example,by dissociation ol the spin-Osystem.V ncc of t t on'loc"t


spin components of eachof theseparticles
may tlicn be mcasurcd indcpendentlyat t o aslt r vlr st hsrt
optionof the experimenter.The spin part of the statevectoris givenby: c t c r i s t i c, r f l r i d '
quanttlln lttecltat
'n: I ua-(Z)-rr,1-(1)
@ ua'(z)1.
lrl I
le cl'.rssrll' tlctctr
\r.
,
)rfur*(I) sYSt clns.St r ont
I'lereo.fiu5!(l)= +u6*(1),so that zr*(1) quantummechanic:rlly describcs a stateir re i:rtcr provctl b'
rvhichparticlet has spin 'up' or 'dolvn', r e a l i s t i cl u c l l t h c
alongthe directionii; rr;'
respectively,
hasan analogous meaningconcerningparticle2. Sincethe singletstare\l' is spheric' trangenrcnt lrqlrins
ally symmetric,i canspecifyany direction. Supposethat one measurcsthe spinot ha i" obscr r ". r blv'
particle1 alongthe .t axis. The outcomeis not predeterminedby the description 'I', I n t h i s s c c t i u r r
'Bel
But from it, one can prcdict that if particle1 is found to havcits spin parailelto collect ivcll'ls
.t axis,thcn particlc2 rvill be found to haveits spln antiparallclto the i :rxisif thct heor cr trtr ir it 'llsr r t i
i componentof its spin is alsomeasurcd.'I'hus, thc cxperimcntcrcan arrange hisf orany, , ncr vlr ois
apparatusin such a lvay that hc can predictthc valucoI the .t corrrponcnt of spinoi t ' d i s t : r ' c t " ( i i 1 I t i
particle2 presumablyrvithoutintcractingrvith it (if thcrc is no acrion-at-a-distancd' l e a s to n c s i t t t , r t i c t l
Likcwise,hc can arrangethc apparatusso that hc can prcrlictany otlrcr colltponcnl Y e f y c l ,r s t ' t , ' i t : t v
of thc spin of particlc2. Thc conclusion of thc :rrgtrmcnt is ttratrll cornponcnts 0f P a t h o l o g i t ' : t li r t s t r t
spinof eachparticlcarc dcfinite,whichof courscis not so in ttrcqtranttrm-rncclrlnicl a c h i c v c t l: l t rr r rr : r ' c
description.l{ence,a hidden-variables theor,yscen}sto be rcquirccl. perforrrrctl. .\ tltl
Some commentsare in order coircerningEPR's prcmisesin the light of Deil'srelied ul)on t{) :l
'
tl:eorem. If prernise(i) is takento assertthat all of the quantum-mcclranical predic' cxper ir lr cntlrs u
tions are correct,then Bell's theoremhas shorvnit to be inconsistcntrvith prcmises loopholcs ( discu.
(ii) and (iii). Actually,in the body of their argumentEPR usedonly a fcrvpredictionst heseloopholcs
rvith probabilityone, rvhich are atypicalin quantum nrechanics, wircreasthc dis' Bell'st ir cor cmP
erepancies which Beli exhibitedbetrveenlocal realistictheoriesand qtrantumnrech'
atricsinvolvedstatisticalpredictions.If it was EPR's intentiorrto aim at a iridden'
3.1. I)cterninist
variablestheorywhich is localand realistic,and rvhichagrcesrvith all thc statisticd
predictionsof quantum mechanics-asmany readershave understoodthem-then, In his l):tPcr
'['h'.rtsvs'
of course,Bell'stheorcmshorvsmathematically that this aim cannotbe achievcd.\\'e in \2.
shallnot try to ansrverthe historicalquestionof thcir intcnt. 'f'wo statcments, horvevur,cal sinqlct st - lt c
can bc madewith confidence.First, the argumcntfrom their premisesis valicl,once t h e s p i n c o m o (
'. r l
the above-mentioncd arnbiguityis clearcdup. Sccond,thc physicalsituationrvhich of per t iele2
'l'ltc
they envisaged is of immensevaluefor examiningthe philosophicaI implicationsoi 1'rodu
qttantummechanics and (via Bell'swork) for exploringthe limitationsof thc familvof t wo dist ir r ctol
localrealistii physicaltheories. nr ech: r niculll;
SVSt ct l'1. I ; or t i
p r c t l i c t i o nf o r
3. Bell's theorem
'fhere
is a vast literatureconccrningthe consistcncy
of hidd.;ariablcs theorics ,\ slteci"rl
rvith the algebraicstructure of the observablesof quantum mechanics.'fhc major s v i t c n t . \ \ ' h t
resultsof this literatureare summarisedin appendix2, but they are not inclispcnsablc
for underltandingthe content and implicationsof Bell's theorem. Fleuristically, 'fl
howcver,this literaturewasvery importantfor Bell's ryork. In thc courseof preparing for all ri.
\
a revierv article on 'impossibility' proofs of hiddcn-variablesinterpretationsc,[ 'l'lrc r'ot
l'
quantum mechanics,Beil studied the theoriesproposedby de Broglie (1928)anci r o d e l r o t ct h c
Bohm (1952). He noticed,as Bohrn had alreadyrealiscd,that in orrlcr to rcproduce c l a s so f t h c o r
the quantum-theoreticpredictionsfor a systcrnof DPII typc, thcy postulatcdthc causc confttsi
Bell's theorem:erpaimeatal testsand implicatiotts 1887 -

of non-localinteractionsbetrveenspatiallyseparatedparticles.Pcll wasthus
e.ristence L".). rs .
ledto ask rvhetherthe peculiarnon-localityexhibitedby thesemodelsis a generic
irr "cteristicof hidden-variablestheoriesthat agreewith the statisticalpredictions Aoa./".)

h*te-l
va-fr.rb( e-
%eo"ie !

tlcscriptionY. I In this scctionrve reviov some of thesederivations,which rve shall rcfer to


parrllclto the lcollcctivcly as'Bell's thcorcm'. Our purposehere is to arrivc at vcrsionsof Bcli's
ri. .t ,,*is if the lrhcorcm rvhich satisfythc follorvingcriteria.(i) The hypotheses seemto be inescapable
:rrn irrransc his ltorenyone rvho is cornmitted to physical realismand to the non-existence of action-at-
ncnr of spin of lr-distancc.(ii) Discrepancies rvith the predictions by quantum mcchanics occurin at
-ar-rr-clistance). .fhrstonesituationrvhichis cxperimer.tally realisable.Criterion(i) is, in our opinion,
her componentlierycloseto having been achieved,althoughthe hypotheses are violatedby some
conrponcnts of llrrhologicelinstances of local realistictheories. Criterion (ii) has cssentiallybecn
rrm-rncciranical lrhieved; horvcver, the experinrent rvhichit specifiesis dificult, and hasnot yet been
2crformcd. Additionalassumptions, not implicit in localityand reaiism,havebecn
lishr of Bell's ;elied upon to allow easierexperimentsto be considercd.(The assumptionsand
:irrrnicirlpredic. ,-rpcritnents
are discussed in $$5and 6.) Unfortunately,this fact leavesopenvarious
t rrith prcmisesnopholes (discussed in $$5-7). It must be strcssed,horvever,that the existenceof
fervpredictions
'h,:rcasthe dis- iell's theorempresented in thissection. .0rur-e.
Ae-I..t o r Io. -?
rlrtllntuln mech-
inr a t l h idc len- C-'. " f)-
J.r.Detennitistictocattidden-uariabtes ttrcories
;#4r;;r$ {/.t1./\> D
rll thc stirtistical
.otl thcrtr-then, In his pxPerof 1965Bell consideredBohm'sGedanhenexpainenl, describedabovc
'fhat ,riL
r c rch i n 'c d. W e r!2. systemconsistsof trvospin-|,particles,preparedin thc quantum-mechani- }l
tn('nts,hrlwever, ;l singlctst:.rtc Y/ givenby cquetion(2 . 1). T,etAa be the resultof a mcasur€rrrcrt of
aa
comPo,rent of particle1 of the pair alon5the directiond,"andlct 135bc that tr
; c si s vu l ir l, onc e :rcspin
2:rlongdircction6. We takethe unit of spin as hlT; hcncc,As, !:r. t l.
+l r ,
si rrr:rti o rrtv hic h rfprrticlc
'l'hcprotluct.rl,i.Bris
I i n r p l i c a t i o n s
of a singlcobscn'ableof the two-particiesystcm (evcnthough
s of thc family of distinctopcrations arc needed in order to measure it). It is rcprescnted quantum
by a sclf-edjointoperatoron the Hilbert spacc associatedwith the
*chanically
For this Gedankenexl>ninenl
,rstenr. onecanreadilycalculatcthe quanturn-mechanical
n/\>
for tl'reexpectationvalue of this observablef:
lcdictio,r o-\"
-f .l' *
l E \ , 6 ) l o = ( Y l o r . 4 c z . 6 l y )-=A . 6 . ( 3 . 1 ) ,.^,..F1" 6jr,rt,' ,.
l'
.{ spccialcascof cqurtion(1.1) containsthe detcrminisrn
irnplicitin this iclcalisccl
n. \\'hcn thc analyscrsarc prrallcl,we have:
ir o t i rrrl i s pc ns able
lEG, d)lv= - I (3.2)
n. I l ctrri s t ic ally , '['hus,
r 'rsc o l p rc par \ lll ri. one can prcdict rvjth qert4tnlylhq resuh B, reviouslyobtaini
ntcrl )rctltl ions
i Thc rrotation of this rcview is to use the rvavefunction or the lettcrs QM as a subscript
rlrlic (1928)aod Cdnotc thc qurntum-mcchanical prediction. We omit thc subscript for predictions by thc
rtl:r ttl rcp of tl:corics irrclur..:.1 by thc posrulates of Bell's theorem, when this convcntion clocs 1ot

A.t= tl
-')' i)0stulated confusion.
- , ^, - qJc</ c J1ttA .1r"1 a -> ,"dl c.A,\e-
t[ - r T wnc
ctJ(4't
/ J ^
fortuCLe-

Turf:c le 2-
-
,_
t'
"
''
, -o " ei = . t (
fgl /-r , L'o-,rA .-n^<.Lt, ir.l
1888 "tr. o*ea,i'\o,t (+rocJ F Clauserand A Shinmy
1
\
thercstrltl(I'@.Sincethequantum.lncchanicalstatc\|,.docsnot U s i n ge t l r t , t t t
t---;------l::_.
t!' ' dctermrncthe rcsultof an individualmcasurcment, this fact (via EPIt's argument)
. ',rr0oJn.:f sttggests that there existsa more completcspecification this
of tlre state in rvh-ich
nG e r^1..)l' dctcrminismis nunifcst. fu. .l.note tiiit ,toi. by the rirrgltsmil'), although ir
Inequalirv
le-' r'.;01'-;. .( may rvellhavemanvdimensions, discreteand/orcontinuoutporir, and diffcrentparE ' B e l l ' s t
t\ 'l
of. it interactingrvith cithcr apparatus,etc. Presumablythe quantum stateY' ist irtcrlu
. 1 F . 1. - { ,\ sirrrPl.
' , c q ! . 1- r relatcdpartialspecificationofihis't't.l@ie.terministichidclen.
yegAblgt-t hp ty -g!-gly-e hsisal$eo tr whi .h p or tsl. r OE* irGn.lo lnd inetpr;ili
.t t-E,+ ee9j i a n g l cr i ( 7 ;
t/s :l€$,Jot tti'tigbthdbjelgblgsstqrrailun-*ssl;,rlrs-_r
,lr{iyt hayc{.fia'rc o"lues.
4 Let A bc thcspace of thc.Gt.s trior anenscmbrecompris.,t of a ueryiorg.nunrbe?
: l*,*.ltn, of thc observed systems.We nrakeno restrictic,ns ast; rvhattypeoi ,pi.. thisis, For thcsc..li
vQefebl'? nor to its dirncnsionality,nor do rvcrequirclinearityfor np"roiitnsrviihit, butof
'$.c.ri''-]
courscrvererluircthat a set of Borelsubscts of A be dcfincd,so that irrobabilitv il
measutcs cnnbc clcfincduponit. Wereprescnt thedistribution functionforthestato 't

A on thc spaee A by thesymbolp. Forlhisensemble rvctakep to haybnormone; r n e s c v i . t t L l (


Fta r

tion and th:


r r 1\
I n d P =1 . or icnt : ltions
'l'ite'\'cl
In a dcterministic
hidden-variables
theorythe obsenable
A6.85 hasa definiu
det er r ninist
}value@i.!s.)(i)forthestatei.Fortheseth.ori.s
I a deterministic tlrcoryis localif for all d anrlf andatl A e A ae hat't: c o n d i t i , r n( .
hiddaz-ztariables
F' t h e s p i n sr r
\t G lo aol.t (Au.Bs)
(l): lo(^).Br()). /'t 4l

l:
Iii
(i
JThat is, onccthc statetr is specificcl
f
t upon i and f. Any reasonable
and thc particlcshavcscparatcd,
.4 candepcndonly uponI and d but not 6. Likcwisem.nsui"mentsof B clcpcnd
mcasurcrncntsof
onh'
physicaltheory that is realisticand deterministicand lit t le r ', r lr r c
that dcniesthe existcnceof action-at-a-distancg
3.2. I"rreu

AnY :tr'

.; is localin this sense.(i\{oregeneral o c c u r i n


definitionsof 'local' will bc considercdin $3.3.) Ifor such theoricsthe expeitation Pt
! 3 . 1 , is th,
valueof A6.Bs,is then given
v by @)ncatisrn
rl^\-
['
|
.)
,
(.-#t -. ,'--ir.Ll.a, : ,
E(e,- 1 \ r , t t \ \ r r
b): t6 Aa(A)Br()),
di: , : , i r - .
i
t r t - ] \ *
_)^6n1lgh
l . i n + ,
slsterrls rvl
l^-rn
\f,.I is i
artfrGtS'
Bcll's(1965)proofof the theoremconsistsof showingthat if the locality-coirclirion
13.{) hinges st r
and the condition(3.2) for partial agrccmcntwith quantum mcclnnicsirrc both t his cclr r at
satisfied,then the expectationvaluessatisfya s:irqpleinequallty..T'his incquafitvis S e c o n du ,
thenanalternativepredictiontothatbyqu;niffieeXpectationv,lu. Finallv, [ ,
of,46.86,. The predictionsmadeby this inequalityare quantitativelytliffcrcntfrom Unf or
thoseof equation(3.1). r eal dct et
The demonstration is straightforward.Eciuation(3.2) can hold if anclonly if somc a[t(
at t cnr pt ir
Ar(l)= -Br(I) (3.6) t hat in sr
holdsfor all i eA. Using cquation(3.6) we calculatethc followingfunction,rvl:ich thc nlc.t:
invoIvesthrcediffercntpossiblcorientationsoftheanalysers: The pro
consicl,:r,
\ EG,il- E(6,i)- - Jn[lr(r) As(A)-Aa(i)Au(l)],tp 'l'lrc
( 3. 2 ; w r
= -.fn 4o1A)A4l)[1
- At(A)Ae(
))] dp. thct.rrr:tll
'fhcy
SinceA, R= t l, this lastcxprcssion
canbc rvritten: d'
prcdictit
lEG,b- EG,d)| < In [l = As(A)A6e)]
dp, but rvhir
Eeg'sthenan: erpairnttal testsandimplicatiuts 1889 -f:
rl'
rte docs not (3.3),(3.5)and(3.6)wehave:
Usingcquations + k?l
ft's argument) l\ (3.7)
in which this v(a, 61-n1a,6)| < t + E$, ey
l, atthoughit Inequality(3.7) is the first of a family of inequalitieswhich are co{ectivelycalled
di{fcrcntParts 'Bell'sinequalities'. \
\[" ,yl r,iaQlg
n state is a A simpieinstanceof the disagreement be^trveen the predictionsof cquation(3. 1)
inistichicldcn- and to be coplanar, makingan iE s
an,lincquality(3 .7) is provided by takingd; 6 i .rvithi f\er.
e of statesof a angleof hp trith ,i, a-ncJ 6 rnaking an angleof nl3 rvith both d and d. Then:
:linitevllues.
'y largenumber d,6=6.t:L d , e= - l b,ArioL
rf spaccthrs ls, For thesedirections:
u'ith it, but of !3 Q-ll
:1
ft84,6)l*-lEG,6)lvl while 1+[E(6,i)]u,:l.
:hat probabilirY
on for the states
vc norlnone: Thesevaluesdo not satisfyinequality(3.7). Hencethe quantum-mechanical predic-
tionand that by inequality(3 .7) are incompatible,at leastfor somepairsof analyser
( 3. 3 ) oricntttions.
'6 hrs a definite The version of Beli's theoremjust proved can be summarisedas follows: no
theory satisfyingequation (3,2) and the locality
':rlityls follorvi: deterministic4hidden-variables
G .h canagreervithall of the predictionsby quantummcchanics
conclition concerning
/ ,\ e i\ ue haoe:
thespinsof a pair of spin-2tParticlesin the singletstate.
(3.4)

of 3.2. Foratord to tlrc non'idzalisedcase ^ j.,'- 'J. lo


n'tcilsurcn'Icnts '
rf B dcPcndonlY Any argumentwhosescopeis strictly limited to a discussionof-idgl ryr,.,l6ofc' ;'* tt.
lctcrnrinistic and iittlevaltreto rvorkingphysicists,rvhoendeavourto describesystemsthat can and do ' .t-(* tt--.S
j. ()lorc gcneral gccurin practice.The immenseheuristicvalueof Bell's(1965)argument,outlinedin
; thc exPectation 13.1,is that it leadsto formulationsthal prglide direct experimentalpredictionsfor
ryrt.mswhich canactuallybe producedin a laboratory.By itself,the derivationgivcn
to do this, because of its reiianceupon thc cxistence of a pair of
(3.s) in$3.1is insufficientfor 'I'hat
is, thc ubovcproof
lnllyscroricntations which thereis a llerfcctcorrclation.
t v c o t r t l i t i o(n3 . 4 ) hirrgcs stronglyupon thc conditionthat cquation(3.2) hold'cxactly.Usc is tnadcof
. cl, rt''i .t u rc, b< lt h rhiscquationin thrce lvays. First, it allowsthc ;roof to go throughmathernatically.
l'his i n cq u ' alit Yis Second, determinismis derivablefrom it and doesnot haveto bc postulatcdscparately.
expcctltion vaiUe Finally, reasons for to be discussed, it assumes that the localitypostulateis reasonable.
civ .iilTcrcnt frorn Unfortunately, equation (3 .2) cannot hold exactlyin au actual cxpcrirnont. Any
rc:rl'Jctcctor rvill have an efficiency less than 100o/o, and,any rcal analyscrwill havc
attcnuation as well as some leakage into its orthogonal channel. Sincc lvc arc
rl if :rn tl .r'n lY if iomc
!$cmpring to dcal rvith not just one but a whole class of thcorics, it is quit" possiblc
( 3 . 6 ) rhatin sonlcof thcscthcoriesthe aboveimperfections'arcinhcrcntly correiirtcdwith
rhentc:tsurcment and detectionprJcessesin a way that dependsupon the state l.
rg iurrction,rvhich problems which arise when tr.€s€three implicationscannot be drawn will be
The
considc'rcd in turn.
'I'he problem concerningthe derivation'slnathematicalrcli'-.^ie upon equatiorr
dp
i3.2)rvasfirst solvedby CHSH. They demonstratedthat a diffc.;nt(.3.2) proof of the
I d P ' ,,Ieorem follows from the above formalism, without requiring cquatio. to hold.
fhey dcrived a diflerent inequality that is violated by the qu'.,,frt^.^-111e3hanical
oredictions for systemsrvhichneverachievethe perfectcorrelatio.-of €r1121i.n. (3.2),
;out 't1.
rvhichdo achieve a necessary minimum correlation. The incq-1 which rcsults
ueu's thcortm: expaimtntal testsazd impliuations 1891

A secondapproach,that usedby Bell (I97t) (althoughnot specificallystatedbut


clearfrom the conte.xt),is to employ an auxiliaryapparatus('event-ready'detcctors)
(3.8)
to Ineasurethe number of pairs emitted by the source. This possibilityis shown
pccial ;chernatically in figure 1. For this scherne,one can simply take the ensembleto
'1971)
coosistof the particlesrvhichactuallytrigger the 'event-ready'detectors.Whether or
nt the not.atriggcringoccursclearlycloesnot dependupon the analyserorientationi No
pioblcrn rvithlocalitylriscs from the presenceof thl signalrprop.grtingto the remote
t ' t ' i r , 1 1 3 rpplrltttscs, since thesesignalscan be simply consideredas part of the state l.
rrncl cr L'nfortun:rtcl1',in practicemost conceivablJievent-ready'detictors dcpolariseor
rtc'Ctltl icsrroythc particles.The valueof this approachis thuslimitea.
ruories An aitogethcrdiffcrentapproachlyas-employecl by Clauserand Horne (Ig7+).
1972) Theyderivedrn inequalityfrom the hypothesisof locrllty and realismin rvhichonly
iltrary, ratiosof the obsen'edparticle detectionprobabiiitiesrpp.rr, and the normalisation
rlurl i tv condition equation(3 .3) is not requiredfor its derivation.The inffuenceof the size
'l'he
'.
I :rs to Spin{2} 'up' detector

Irr the
8' :, '1
\
\\
'
E v e n l -r e o d y ' d e l e c t o r s 'n
S p i n( 1 1' u p ' d e t e c t o r
A ^ =. 1 r .! i.nt
- $

ociltcd '. Anolyscr ( ,/


t'-'-
nc,.[..-r
x Oxrs ,Y
{t :
I
, scn rb l c
.is otrly
f>' i^,
-',"f
Source
r1.: \e'l
\j \ '-'+'
-\0n the N e r t h e cr e t e c t o r (l.-4qh.:
bar r- 1 .,t<)-- N e i t h e rd e l e c t o r S
/ Anclyser2 Ao=0
:ltl ti'len
(-/Q'
>/
''u rti cl cs \/
t
S p r n( 2l ' , J o w n d' e t e c t o r Coincidence
a sp i n - 9r: -1
-
't i ti o n e d Apporctus2 Detector
gotcsrgnols
lr scrvcd , 'Event-rcady'
fiElt"j. Apparatus configuration uscd for Bell's 1971 proof. clercctorssignal
p urti cl e = lloth arms that a pair of pnrticlcshas bcen emittecl. For a given gatc signal, the rclult
: l)cl l (l cn t ot: citltcr arm is assignedthc valuc + I if the correspondingspin-up dctcct.lr rcsponds,
- I if thc spin-down dctcctor rcsponds,ancl
. ir, since 0 if ncirhcr dctcctor rcsp<.rnds.
na l vse rs .
riburtion, cithe cgscmblc thusvanishes.Their apparatus arrangement doesnot havethe 'event-
rv l 'ca rl e sdY'dctcc.tors of figurc 1, nor docsit havetwo detectorsfor eachapparatus but only
r. :i .ll rcrl ry rnc.lt is tlrus nrr.rch sinrplcr,and is shownschcmatically in figuri 2.
u ts COn- In thc rcrnaindcrof this section,rve will shorvholv thcselitter two approachcs
.svstcrn.) Uirst, horvcvcr,we will discussthe aforementioncd
;rocccd. generalisation
of the
;i'rcctthat lcrlir,v postuhteto includeinhcrentlystochastic theories.
r ci trtl cth e
'c, i t i s n o
j.J.Geneiolisutiott
of the locality concept
r i r rt r : n t .
, rrscd b v Cc'rrsiclcr citherof the experimentalconfigurationsfor Bohm'sGedankedexbaiment,
1111r rr spiO :cscribcdin $3.3. Actually there is nothingin the proof which requiresthe systems
r i u n t t t i o l r bespin-! prrticle-s-They may b9 any discrete-state quantum-mechanicrllycor-
r r', rcl l i sq rhtedcrrrissions. (l{orvever,not all quantum-mechanicilly correlatccl,y.t.i., .r.,
r r u t k r -.st ;redictcdtrr violltc the.resultinginequalities.A careful choiceis requlecl tc find
rrtlcs froq :rervhiclris ln appropriatetest case.) In Bohm's Gedanhenexpainezr ihc symbolsd
id 6 lrc t:rkcnto rcpresentthe orientationsof the Stern-Geilachmagnetsused for
Bell's theorcm:erpaimental testsand inpHcatioru 1893

rvhichthe following valueswere assignedby Bell:


'spin-up' detectortriggeredby particie 1
(* t,
.4,(I): _ t, 'spin-down' detectortriggeredby particle 1 (3.e(a))
{ \
l . 0 , particle 1 not detected \

detectortriggeredby particle2
(*t,'spin-up'
'spin-down'detectortriggeredby particle2
Ba()): - 1, (3.9(D))
./--/ {
I O, particle 2 not detected.
bv (:lt. .\ source Fora givenstateI of the emittedcompositesystem,we denotethe expectationvalues
r2rztusc0nsistsof an for thJsequantitiesby the symbolsArlll and 86(I). In the generalcasetheseaverage
jnlctcrsalnd6reS-
valuesrvili differ from the valuesassignedby equations(3 .9). Sincethe valuesfor I
cleaandllrepresent
and.B are boundedby 1, it follows that:
s.
<1
l.4"(r)l < 1.
lBo(l)l (3.10)
. rt lfld b rruy rep- Usingthe generaldefinitionof localityof $3.3,we can write the expectationvaluefor
cxp e ri n rent er .A s rheproduct A686 as:
faratuses 1 and 2, E(a, b)= Jr,,{"(,\)80(f) dp. (3 . 11)
res,as neccssary.
to include systems Sincewe are including in our ensembleonly those particleswhich have previously
'event-ready'detectors,we are assuredthat the distribution p and the
tc'rministicsYstems
riggeredthe
rnd th a t n r lv loc ally rangedf integrationA are independentof a and 6. Now considerthe cxpression:
vstcn ts,rr' hic hhav e Ao1\186,(r)l
E(o,b)- E(a,b')=J,rlAa(A)Bo(l)- dp
, s i n a n i n her ent ly
; for itrt-vplrticular rvherewe take a' and b' to be alternativcsettin$ for analysers1 and 2, respectivcly.
,b:r[riliticsrn:tv each Thiscan be rewrittenas:
'cspcctivr:lv,and of E(a,b)-E(o, /''): Ja Aa(^)Bb(l)[lt A,'1t14o,(l)]dp
.'iseindependent of
- J" A^(A)Bb'(lXl
t Aa,Q)Er,())l
dp.
. 'stl r:rtth c out c om e Usinginequalities(3.10),we then have:
t 1(.Il trrti l l ' t t c or npo-
rt tlrccspcritrtcnter lE(o,b)- E(a,D')| dp+J"[1t Aa'$)Ba(l)]
< In [t t AE( )8l,,,(l)] dp
bt :ri n i n gknor v ledge U I
) c o n t r r i n isn l o r m a - lE(o,b)- E(a,b')l< tlE(a',6')+E(o',6)l + 2 [n de.
-
LNI rl I
pre su rn l blvr ev eals Hence' \. -'rll-'
yfif C{)lll})OnCnt Of a
. (3.r2) 1"\t/,"
t it drrcs prcscribe, rs a, a',0 and b' inthe centralexpressionof (3.12),
Bvre-deiinitionof the paramete
tcrrr is rtrlt c:tusally
theminussign may bc permutedto any one of the four terms. Inequaliry(3.12)and
rce ti rc ttto s Y s t em s
itspcrmtrtationsarc one form of Bell's inequality,and reprcscnta gencralprediction
('rrl s:l l 'uIr ' r f ( lr r nc d.
the aboveassumptions.
iorthc theoriescovcredby
In order to completethe proof of thc theorem,it is sufficientto show that in at
leastonesituationthe predictionsby quantum mechanicscontradictinequality(3.12).
Thcquantum-mechanical predictiontEG,6)]c* for the two spin-| particleexample,
t Lrc:rl i tv. I c f init ion.
rvhendue aecountis taken of imperfcctiorrsin the analysers,detcctorsand state
rsscdi n $. 1. 2: r nd is will bc of the form:
prep:rration,
c rrsso ci :t t ct ppar
tl a-
l s s i l l l c o t t t C t ) t T t C S t, O =Cd.6
lE@,6)lo" (3.13)
--

V o r u v r 2 8 .N u u g e n l 4 PHYSICAL REVIEN(/ LETTERS J A e n r rI 9 7 ; Y o t u u t 2 E .N u . v e t

Fraser and Rudolph Hwa. He is indebted to the den- var iablesub


400 following members of Group A at the Lawrence Eechanics' r elY
Berkeley Laboratory for generously allowing hiu ag thecharacter
N )da to participate in the analysis of the K' exposure: hesargued that t!
a
M. A l ston-Ganj ost, A . B arbaro-Gal ti eri , P . J. strictive- Howev
D avi s, S . M. Fl att6, J. H . Fri edman, G. R . L ceseof two sPatir
: M. J. Mati son, J. J. Murray, M. S . R abi n, F. T. chanicallycorrel
z - ) d 6 S ol mi tz, N . J. U yeda, V . W al uch, and R . W i nd- showthat anY hid
molders. oolythe natural a
leadsto Predictic
0ict with quantun
.Work supported by the U. S. Atomic Energy Comnnis. BelI 's
Pr oof wa
sioq urder Contract No. AT(04-3)-3{ PA f91. Clauser , Hor r
rK. G. Wilson, Coroell Uoiversity Report No. CLIiS- by
pinted out that t
131, 1970(to be prblished).
equalitYcan be t€
I 200
2w. R. Fraser et al., to h published.
E
3H. D. I. Abarbanel, Phys. Rev. D 3, 2227 (19?l). bg all local hidd'
c
tR. c. Hwa, to be prbUshed eristing exPerim'
s D , Z . F r e e d m a n , C . E . J o n e s , ' F . E . I . o w . a n d J , E . tbr this Purpose.
:
Y o u n g , P h y s . R e v . L e t t . 2 9 , 1 1 9 7( 1 9 ? i ) . oI an exPeriment
6c. E. DeTar, Phys. neT. o 3, 128 (19?1).
to nrle out local :
?A. Bassetto, M. Toller, and L. Sertorie, Nucl. Phyl.
hlgb statistical a,
834, r (1971). tn the Present v
-4 '? 2 '
,,1r, T. Mueller, Phys. Rev. D 4, 150(19?l). .'
$W. Xo and R. L. Lander, P:hys. Rev. Lctt. ze , tOOt tloo in linear pol:
FIG. 4. (a) Gr and (b) G3as defiqediq t}tetext. (l97r). I, emitted ln a J
loJ. Erwia, lV. Ko, R. L. Laoder, D. E. Pellett, aad lle decaying ato:
P . M . Y a g i r , P h y s . R e v . L e t t . 2 ? , t S ? / .( 1 9 ? 1 ) . rlcally placed opt
The statistics on the eight-prong data are not ItThe correlattoa lenglh of aboG I is euen shorter o( tro lenses, a \
god but show characteristiessimilar to those than the short-range Mucller-Regge-theoretlcal value aadremovable pc
fo r si x- pr ong. lR. C. Arnold, ANL Report No. ANL-HEP ?139, 19?1 tcctor (see Fig. 1
We present this dramatic behrvior of the two (unpublished), and Ref. 5J. In that tbeory a corrdatioo
measured: R(<g),
n - ' s a s f u n c t i o n so f t h e i r r a p i d i t y s e p a r a t i o na s a length of l is only achiwed for the center-center cor?F
pboton detection,
ch a l l e nge t o any t he o ry o f i n c l u s i v e re a c ti o n s . laiion if the iatcrcept of eptlc trajectories (as the r'-
r- channcl hes exotic qua.otum nusrbcrs) is - I lW. Ko, tleen the planes ,
The author wishes to thank Richard Lander for be orientation lf
R. L. Lander, and C, R i s k , Phys. R c v . k t t . 2 7 , 1 4 7 6
his support and encouragement. The author also (1971)1. Thc correlallon lcngllt of I or 2 ls usually prr cotncidencerate '
a p p rec iat es t he m an y u s e fu l d i s c u s s i o n s w i th h i m, dicted for fragment-ccbter or fragmeot-fratmeot cor- he coincidence r
David Pellett, and Philip Ya{er. He also benefit' relations. 4, t h. coincidenr
te d fro m s t im ulat ing c o n v e rs a ti o n s w i th Wi l l i a m tzH. T. Nieh and J. M. wang, to b€ published.

Experirnental Theories*
Testof LocalHidden-Variable

Stuart J. Freedman and John F. Clauser


Defitrtment of Physics and Launence Eerheley Laboralory, (Jniuersity of Catlfonia, Bet*el.ey, California 91720
(Recelved 4 February L972't

We have mearured the llnear polarizatton correlatlon of the photons emtttr.d ln an atom-
ic cascade of calctum. tt has bcen shown by a generalizatlon of Bcll's lneqnali..:'that the
eristence of local htdden variables imposes restrictions on thls corelatlon ln confllci
with the predictions of quantum mechanics. oltr data, in agreement withquantum me-
chanics, vlolate these restrlctions to high stattstical accuracy, thus provldlng strong evl- nG. l. schematir
dence against local hidden-varlable theorles. i clectronlcs. Sca
Ftr sl the discrimi
Since quanh.rmmechanics was first developed, features, tlr€nr. arise because a quantum state \ each 100-sec co
there have been repeated suggestions that its sta- represents a statistical ensemble of "hidden- rcaiera and
the e..rpr
d ou paper tape anL
tistical features possibly might be described by variable states." Proofs by von Neumann and
Ee r.
an underlying deterministic substructure. Such others, demonstrating the impossibility of a hid'

938
J A P R r Ll g 7 2 \ , o L U M E2 E .
NUMBTRI{ PHYSICAL REVIEW LETTERS J A p n tI l 9 - l

with quantum
den-variablesubstructureconsistent
indebted to the 1 moved. Quantum mechanics predicts that R(q)
t
I the Lawrence rely On various asSumptions concern- and R o are rel ated as fol l ow s3' 5:
rnechaniCS,
rsly allowing hlq{ ingh. character of the hidden variables't Bell
R (p)/R o = f (ert + €,r)( €" " * e,2) + t ( . r t - €, "t )
these assumptions are unduly re-
he K' e:eosure: ir-r -S,t.d that
by considering an idealized x(€r2-e^2)F,(0)cosLp, (1a)
Saltieri, P. J. t strictive. However,
nal, G. R. Lync\ caseof two spatially separatedbut quantum-me- while
S. Rabin, F. T. chanically correlatedsystems, he was able to
and R. Wind- that any hidden-variabletheory satisfying R L / R o =| . ( e, t + € , r ) , (tul
t show natural assumptionof "locality" also
't
ody the and
leadsto predictions("Bell's inequalitY")in con-
I
flict with quantummechanics.r \/ Ro = t(e ,t * e,'). (l c)
c Energyao.trt Bell's proof was extendedto realizable systems
PA 191. by Clauser, Horne, Shimony, and Holt,s who also Here eri (e,n')is the transmittanceof the fth po-
leport No. CLIN- larizer for light polarized parallel (perpendicu-
pointedout that their generalizationof Bell's in-
equditycan be tested experimentally, thus test- lar) to the polarizer axis, sd Fr(6) is a function
:d.
ingatl local hidd'en-variabletheories, but that of the half-angle 6 subtendedby the primary lens-
I
rt 2227 $9711.
existingexperimentalresults were insufficient es. It representsa depolarizationdue to noncol-
9. Low, aod J. E. fbr this purpose. This Letter reoorts the results. linearity of the two photons,and approachesunity
7l). Ofan eXDerirnpnt shinh ^rp srrfjinientlyDfeCiSe., for infinitesimal detector sotid angles. IFor this
(r9?1). 1onrle out locil hidden-variable theoties '!,i!h_ e x p e r i m e n t ,I = 3 0 n ,a n dF , ( 3 0 1 ) . 0 . 9 9] .
.torie, Nucl. highstatistical accuracv- We malre the following assumptionsfor any lo-
In the presentwork we ineasuredthe correla- cal hidden-vrritrble theory: tl) The two photons
: 9 ?1 ) .
tion in linear polarization of two photons 7, and propagaters separatedlocaiizea particles. (2) A
v. ktt. 26,
r, e m i t t e di n a J = 0 - . I = 1 - J = 0 a t o m i ee a s e a d e . binary selection process ocgurs for each photon
. E. Pellefr, a.od The decaying atoms w€r€ viewed by fwo sffllrn€t- at eachpolartzer (transmissionor no-transmis-
,34 (19?l). ricdly placed opticd systems, each eonsisting sion). Tlris selection does not dependupon the
i even short€r d tro lenses, a wavelengthfitter, a rotatable orientation of the distant polarizer.
theoreticd vdtr andremovable polrrlzer, and a slngle-photon de- [n actclition,we make the following assumption
HEP 7139, 19?1 tector(see Flg. 1). The following quantities were to all0ry a comparison of the generalization of
lory a corr
mea$Eed: R(El, the coincidencerete for two- Bell's inequality with out erperiment: (3) eU
nter-ceater
:ories (as tbe t'- Srotondetectioq as a frrnctlon of the angle g b€- photonsinctdent on a detector have a probability
cs) is - 1 [w. Ko] treea the planes of linear polartzatlon defined by of detectlonlhat is independentof whether or not
t. lalt. 27, L176 he orientaHonof the insertcd polrrteersl R,, the thC photon has passed through a polarizer.c
r 2 is usGlly coincldencerate wtth polerlzet 2 removed; r!, The ebovetssumptions constrain the coinci-
nt-fragment thecolncidencerate wlth polarizer I removed; dence rates by the following inequal,ities?:
&, the coincidencerate with both polarizers re- - I { A ( 9 )- ' 0 , (2)
ublished.

where
@a.o'tx
+R-
-a{c)
\7'
_ Sn(ge) _ R,
_ R(3<p)

ril q '
Ro Ro Ro
i
For sufficiently small detector solid angles and
highly efficient polarizers, these inequalities (2)
are not satisfiedby the quantum-mechanical pre-
rtom- diction (1) for a range of values of g. Maximum
. the
violatlonsoccur et g =22*' [a(p) > 0J and I =67*o
:t
]- I [a(p) < - 1]. At these anglesof ma:rimumviola-
{ evl-
t
FIG. 1. Schematlcdtagram o{aenarahri and associat- tion, inequdities (2) can be combinedinto the
d electrrnlcs. Scalers (not shiwd monltored tjre out- simpler and more convenientexpression_
--5---- .. . t(
I
{ pttr of the dtsertmlnators and colncidencecircutts dur- )'-&ttt [, o*J
rantum state t{ each100-eeccount pertd. Ttre contents of the -6=IR(zz*"r/Ro
- fi(6?
*,)/nol-+ ig. (3)
of "hidden- lcrlere and the er<perlmental;onftguratlon were record-
leumann and eoon papertape and analyzedon an tBM f620-II com- which does not involve ,?, or &.
,ibility of a hid-' ruer. The experimentalarrangementwas similar to

t$76alf7r-? -l I 4 / | t' o'oo? > O 939


deltn-'-''o5o
,l ((

6r
V o l u u e 2 8 .N u v s r n l 4 PHYSICAL REVIE\fl TETTERS J Apnrr t972
V O T U M E] F

-3
l a r i z e r s r e m o v e d w e r e 4 l = 1 . ?x 1 0 a n d 4 2 = l . s
x 10'3.s
A di agram of the el ectroni cs i s i ncl ude d in Fig.
1. The overall system time resolution was about
1.5 nsec. The short i ntermedi ate state l i jet im e
(- 5 nsec) permitted a narrow coincidence windon
4p4srP,
(8.1 nsec). A second coi nctdence channe l dis-
pl aced i n ti me by 50 nsec moni tored the n um ber
of acci dental coi nci dences, the true coi ncidence
rate bei ng determi ned by subtracti on.t0 A t im e-
ot.'so to-amplitude converter and pulse-height analyzer
measured the time-delay spectrum of the two
F I G . 2. Lev e l s c h e m eo { c a l c i u l . D a s h e dl i nes show photons. The resulting exponential gave the in-
the route for excitationto the initial state 4p?tSo. termedi ate state l i feti me.rt
FIG.3. t
The coincidence rates depended upon the beam
plerizers,
and lamp intensities, the latter gradually decreu. nced, Pio
that of Kocher and Commins.E A calcium atomic i ng duri ng a run. ?he tl pi cal coi nci dence r at e tbr Predicti
beam effused from a tantalum oyen, as shown in with polarizeis removed ranged f rom 0.3 to 0.1 tle ueaeur
Fig. 1. The continuum output of a deuterium arc countx/sec, and the accidental rate ranged from .qlee of th
lamp (ORIEL C-42-72-12) was passed through an 0.01 to 0.002 counts/sec. Long runs requ ir ed by
interference filter IZSOA fult width at halJ-maxi- the low coincidenee rate necessitated automatic
nons for i
mum (FWHM), 2ffi ttansmission at 2275 A ] data colteetions
focused on the beam. Resonance absorption of "naa The system was cycled with 100-sec counting
f
2215-i photon excited calcium atoms to the 3d4p periods. Periods with one or both polarizers in.
t P , s t at e. Of th e a to m s th a t d i d n o t d e c a y di rect- rWork sup
serted alternated with periods in which both po-
''t11e best'
ly to the ground stlte, a,bout 7% decayed to the larizers were removed. Both polarizers rotated
frkclv G
1 4P".So state, from whictr they cascaded through
the 4s4PrP, intermediate state to lhe ground
according to a prescribed sequence. For a giveu
run, R(q)/Ro was calculated by summing counts
Ectltn, 193

I state with the emission of fwo photons at 5513 A for all configUrations corresponding to angle g
Ieclrrlnics t
1955)1.For
I (r,) and 4227i tAl (see Fig. 2). At the interac- and dividing by half the sum of the counts in the rcr J. S. B€
tion region (roughly, a cylinder 5 mm high and 3 l.l.
I adjacent periods of the sequence in which both s. geu
I mm in diameter) the density of the calcium was polarizers were moved. Data f,or Rr/ \ and R2/ [re4.
It

J about 1x 10r0atoms/cmt. To avoid spherical Ro were analyzed in a similar fashion. The val- . rJ. oause
t tya. Rev.
I
't
1 aberrations which would have reduced counter ef-
ficiencies, aspheric primary lenses (8.0 cm
ues given here are averages over the orientatioo
of the inserted polarizer. This cycling and aver'
{A hldden-
a , 3r tc tndepe
i diam , , f = 0. 8) w e re u s e d . P h o to n s h w e re sel ect- aging procedure minimized the effeets of driIt
t t ed by a filter with 10 A fWnU ana iO% transmis- and apparatus asymmetry.
Ittlzer, and
ib molysl-.
t
I
i sion, *6 7z by a filter with O A fWnM and 2ffi The results of the measurements of the corre- tr rd R2 be:
6cy are so c
lation R(q)/Ro, corresponding to a total integra'
I
transmission. The requirement for large effi- '
cient linear polarizers led us to employ "pile-of - tion time of - t00 h,. are shown in Fig. 3. All er' hd. Horne,
hprrUtrshed
plates" polarizers. Each polarizer consisted of ror l i mi ts are cons6rvati ve esti mates of I st an-
ftrdum Me
ten 0.3-mm-thick glass sheets inclined nearly at dard deviation. Using the values at 22|" and tcbool of ph
Brewster's angle. The sheets were attached to 67*", we obtain-L:9059-_q0..0!8 in clear viola[on b, Neryyor
hinged frames, and could be folded completely of i negual i ty (!).r2 fu" [ne" * or., w i oU Jer ve no'
out of the optical path. A Geneva rr€chi,nism ro- evidence for a deviation from the predictions of
fated each polarizer through increments of 22+'. quantum mechani cs, cal cul ated from | ' re mea-
The measured transmittances of the polarizers sured polarizer efficiences and solid angles, utd
w e r e € r t = 0 . 9 ?* 0 . 0 1 , € , r t= 0 . 0 3 8* 0 . 0 0 4 , €, ' show n as the sol i d curve i n i i g. 3. W e consider
= 0. 96 r 0. 01, a n d € o ,'= 0 .0 3 ?t 0 .0 0 4 . T h e photo- these results to be strong evidence against iocal
m ult iplier de te c to rs (R C A C 3 1 0 0 0 E , q u antum ef- h idden-variable theories.
f i c i e n c y = 0 . 1 3 a t 5 5 1 3A ; ^ a n d R C A 8 8 5 0 , q u a n t u m The authors wish to express their sincerest aP'
ef f ic ienc y = 0.2 8 a t 4 2 2 7A) w e re c o o l e d , reduci ng preci ati on for gui danceand hel p from P ro f essor
dar k r at es t o ? 5 a n d 2 0 0 c o u n ts ,/s e c , re specti ve- E ugene C ommi ns, to P rofessor C hari es TowneS
Iy. The measured counter efficiencies with po- for hi s encouragementof thi s w ork, and to M . Sio'

940
] A p n r r1 9 7 2 PHYSICAL REVIE\Y' LETTERS j Apnrl l97l
V o l u r q e2 8 ,N u u a e n l ' {

0 ' 3 a n dt , = 1 . 5 6Tnis assunrptlon is much weaker than the assunrption


n t a d eb y L . R . K a s d a y , J . O l m a n , a n d C . S . W u , B u l l .
included in Fig. A m e r . P h y s . S o c . E , 5 8 6 ( 1 9 7 0 ) ,i n t h e i r d i s c u s s i o no f
the two-y decay of posltronium; see L. R. Kasday, in
ution was about
"Foundations of Quantum Mechanics, Proceetliqgs of
st ate l i fe ti m e 'Enrico
the International School of Physics Fermi,'
rcidence window o Course IL" (Academic, New York, to be published).
channel dis- ?The inequatity A(rp)< 0 is derived in Refs.3 and 5.
+.
ed the number The other forms of the hidden-variable restriction are
ue coincidence obtained by similar arguments; see S. Freedman, Ph. D.
. ion.ro A ti m e- thesis, University of California, Berkeley, Lawrence
-height analyzer Berkeley Laboratory Report No. LBL-391, 19?2 (un-
published).
n of the two o zz, 45, 67i so EC.
A. Kocher and E. D. Commlns, Phys. Rev. Lett.
rl gave the in- ANGLE
9 IN OEGREES 1 8 , 5 ? 5 ( 1 9 6 7 ) iC . A . K o c h e r , P h . D . t h e s i s , U n i v e r s i t y
FIG. 3. Coincidence rate with angle v between lhe of California, Berkeley, Lawrence Berkeley Labora-
upon the beam plarizers, divided by the rate wlth both polarlzers re- tory Report No. UCRL-1758?, 1967 (unpublished).
'adually decreas. elte counter efflciencies are given by rii = (Qr/4r)T,
moved, plotted versus the angle g. Tbe solid ltne is
ncidence rate r. x€*t, where O; ls the solid angl€, ?i is the transmis-
tle prediction by quantum mechanics, calculatcd ualng
rom 0.3 to 0.1 \ tbe measured efficiencies of the polarizers and solid slon of the fllter, e I is the quantum efficiency, and L,
aaglesof the exPeriment. accountg for other losses. The measurement of 42 was
te ranged fmm
made, employing the propertles of tie calcium cascade,
uns required by r
by compartng the coincldence rate and the 71 slngles
eted automatic :, rate altcr euttable baclqground correctlon; 4r was then
mons for helpful suggestions.
irrferred from tle lnowa qrrantum efliclencies and frlter
'|
)-sec counting transnlrElong aegumtng that Oi and L1 were the same
h polarizers h- i for both detcctor ayetema.
'Wori< aupported by U. S. Atomic Energy Commiscton.- l0An crttmete of thc accidentai rate was also obtained
which both no- -1
tlte best-known proof is by J. von Neumenn, Mattv- from tle etrglcs ratcs. lte two estlmates gave consis-
.arizers rotated i
tutis che Grun dbgea dc r Qunt e m ec lu nih (SprlrBe r, teat relrltg. In fact, our concluelons are not changed
ce. For a grveoi Berlln, 1932) [Mathematical Fanfuttotts of Qaantum lJ acctdcntals arc ncglectcd entlrely; the signal-to-ac-
um mi n g co u nt s ' Meclunics (Princeton Univ. Press. Princeton, N. J., cidental ratio wtth potartzer removed is about 40 to I
ing to angle g 1955)1.For a critical review of this and other proofs for the data presented.
e counts in the ,d cceJ. S. Bell, Rev. Mod. Phys. 38, 417 (1966). ttRcgonence trapping, encountered at high beam densi-
2J.
in which both I S. Belt, Physics (Long Is. Ctty, N.Y.) 1, 195 ties, resulted in a lengtlcntng of the obsened lifetime
' Rr/\ nd Rz/ (1e64). and a sllght decrease tn the polarlzatloncorrelationam-
tJ. Clauser, M. Horne, A. pllhrde, ree J. P. Berrat, J. Pbya. Radium 20, 541, 633
;hion. ttre val-,f Shlmony, and R. Holt,
' the orientation P h y sR. ev. Lett.23, 880 (1969). (1959). At low beam densltles the meesuredlifetime
{A
hidden-variJie theory need not requlre tbat R1 and is constgtent wlth prevlously measured values. See
:l.ing and aver{ Rl be independent of the orientation of the inserted po- W. L. Welse, M. W. Smlth, and B. M. Miles, Atomic
:cts of drift hrlzer, and we do not assume thle Lndependencein our Traasltion Probabllltbs, U. S. Nattonal Bureau of Stan-
dataana[ysig. Hou'ever, the results are consistent rvith dards Reference Data Serie- ,z (U.S. GPO, Washing-
, of the cort€- | Rl andR2 being independent of angle, and for simplicity t o n , D . C . , 1 9 6 9 ) ,V o l . 2 .
theyane so denoted. l21te rcsultg thct are of interest tn comparison with
toral integra'I lltt.
Fig. 3. All er't Hor:re, Ph. D. tbesle, Boeton Unlversity, 1g?0 the htddea-varlable inequalltles are R /R0 = 0.49? r 0.009,
funpubllghed).see aleo A. shlmoay, in,.Foundations of R z / R o - 0 . 4 9 9t 0 . 0 0 9 , R e 2 + 1 / R o = 0 . 4 0 0* 0 . 0 0 ? , a n d
rratesof 1 staa'
Qualtu.utMechaalce, Prpceedtngs of the lnternational R$7+"|/Ro=0.100 t0.003. We thus obtain r(22f')
at 22|'and Schoolof Phyalcr 'Enrlco Ferml,'Courge 11," ({cadem- = o . t o + a 0 . 0 2 6 . o 6 a 1 6 ? { ? = - 1 . 0 9 ?r 0 . 0 1 8 , i n v i o l a t t o n
in clear violatioo. lc, Ne* Yort, to be publtahe4. of tnequeltttes (2).
, we o b se rv e no
r predictions of :
rom th e me a-
;olid angles, and'
3, We consider.
rce against locd

reir sincerest alt


from Professor
Charles Townes
lrk, and to M.

941
VOLUME 36 24 MAY 1976 NUMBER 21

Experimental Investigation of a Polarization Correlation Anomaly*


John F. Clauserj'
Materials and Molecular Research Division, Lasurence Berkeley Laboratory, University of California, Berkeley,
California 94720, and DePartment of Physics, University of California, Berkeley, California 94720
(Received 24 November 1975)
An experiment was performed to search for the anomalous two-photon polarization cor-
relation observed earlier by Holt and Pipkin using a cascade of atomic mercury. The ex-
periment is a sensitive test of various aspects of the foundations of quantum mechanics.
Although the present experimental arrangement differed only slightly from theirs, the
anomalous results were not observed.

Following the suggestion by Clauser et al, ' in certain atomic cascades. Atoms undergoing
which in turn was inspired by Bell's theorem, ' cascade decays are viewed by two symmetrical-
two experiments were performed. Their purpose ly placed optical systems, each containing a ro-
was to distinguish between the predictions by the tatable linear polarizer and a single-photon de-
whole class of local hidden-variable theories and tector (see Fig. i). The rate of coincidence
conventional quantum mechanics. Moreover, counts R(qr) for two single-photon detections is
Clauser and Horne' have since shown that these measured as a function of the angle y between
experiments also test the more general (not nec- the orientations of the inserted polarizers. It is
essarily deterministic) class of objective local compared with the coincidence rate Ro measured
theories. This class includes any theory contain- with both polarizers removed.
ing objectivity and naive locality, and thus has a Objective local theories and local hidden-vari-
strong intuitive appeal. Unfortunately, the re- able theories require that the following constraint
sults of the two experiments are in conflict. The governs these rates'"':
results obtained by Freedman and Clauser4 at ")/R, —R (67—,")/R J ——,' 0.
5 = ~R (22 —, -
the University of California, Berkeley are in ex-
cellent agreement with the predictions of quan- An appropriate choice of a cascade is required
tum mechanics, and appear to exclude general so that the photons, though spatially separated,
objective local theories. ' On the other hand, the are strongly quantum-mechanically correlated,
unpublished results obtained at Harvard Univer- with the individual photon polarizations retaining
sity by Holt and Pipkin' distinctly favor objective a mutual nonlocal interference effect. ' With such
local theories (and/or local hidden-variable the- a choice, as well as with some rather stringent
ories) and, as such, are in disagreement with minimum requirements on the polarizer efficien-
the quantum-mechanical predictions. This Let- cies and collimator solid angles, the quantum-
ter describes a third experiment attempting to mechanically predicted correlation violates the
repeat, at least in part, the conditions of the above constraint. ' ' These specifications were
Harvard University experiment. achieved in the experiment of Freedman and
The experiment consists of measuring the po- Clauser by generating the photons in a =0- J J
larization correlation of optical photons emitted = 1-J=0 cascade of atomic calcium. The cas-

I223

δ exp = 0.0385 ± 0.0093  0


VoLuxr {7, Nutr(rrr 7 PHYSICAL REVIE!T LETTERS l 7 A ucusr l98l

ErpcrlncatrlTcsr of RerllgticLocrl Thcoriervie Bell'rTheorem


Alatq Aspect, Phf[ppe Grangter, Md G€rard Roger
rU, h*Jad,
InstJtytd'Optl4ueTh1odqu ct A\pllqdc. Unh)el' F41406Otsay, Fmncc
(Raceived30 Merch 196f)
We hevemeasuredthe ltnerr polerlzatloneorrelatlonof the photongemitted in a radle-
tt"eaAhigh-efficiencytourceprovidedanimprovedstetictical
sccurlcy end an abtlity to pcrform new t€!t!. Our rcrultt. in excellentagreementwtth
the qusntummechanicelprediction!, strootly vtolatethe generalizedBoll's inequalttict,
and nrle out tbe wholeclasa of reeliatic local theories. No eignificantchangein resulta
rg obrerved wltb lourc+polertzc acprreioae of rp to 6.5 m.

P A CSn u m b e rs : 0 3 .6 5 .8 2 ,3 2 .5 0 .+ d ,32.E 0.K f

Slnce the developmentof quurtum mechtrlcs, but requlres tlntlet aeeumptlon^1.


there have been repeated suggesdonsthat lts Palrs o( low-energy photoru emltted in ccrtaln
stattgtlcal features poeaibly might be described atomlc radlattve cascadegale candldatesfor
by an underlytng debermtnlstlc gubstructure, a better teEtE.z Wtth a reasonablcasrumption
quantumatate repreEentlnga stattgtlcal ensem- aboutthe detector e(ftcienclet,&t ths tctutl ex-
ble of "hidden-varlablestateE.' In 1905Bellr perlments consdtute e valuablc tert of local
ehowedthat any such "hldden-vartable gubatruc- redietlc theorler vta Bell's thcorcn. So far,
ture, " if local, ytelde predtctions that dllfer tour e:cperlmentgrrof thie type hrve betn carted
significantly from those ol quantummechanics out; tluee of them have agreed with QM predic-
(QM) in some special situatlom. Bell's theorem tions. In the most recent such e:geriment by Fry
wag extendedln 1969by Claurer, Hornc, Shtmony, and Thompson (upholdtngQD{), a hlgh pumping
and Holt2to cover actuat systemc, providing an rate of e (J = 1)- (J = 1)- ("r = 0) cascadeqna
erperimental test lor all local htdden-varlable attatned ustng a tunable laser, allowtrt shorter
theorlea. Frrrther generallzatlonrt.{ haye potnted periods of data collectlon (80 min).
out that determlnlso lr not e cruclel leature In thlg Letter we report thc regult! d Er€l!-
leadtng to the conlllct wtth QnA That Lr, Bell's urementg of the polarizatlon comehtlong of visi-
lnequalitles have bcen ahown to apply to e broad- bte photongemttted ln e (J= 0) - (J = t) - (J= 0)
er clase of tfieorler: "obtcctlvc local th€orl€!,"s atomlc radletlvc oa^lelde. Thc ercitetlon rate ig
or 'tealtetic locet tlporlcs'e that carl be tagted more than tcn Smcr greater then thet o( Fry and
ln actual eryerimentl. $rch c:Fcrlmentr rre Thourpcolr. A 3rcet vlrstoty of tcrtr wcrc thug
reallzatlons ol Bohnr'l "Gcdlnhsnclgcrtnentr " t performcd. In tcltg anelogurr to prevlour onel,
insplred by the frmour p.Fr by Elrutctq Podol- wc havl ettelncd a gtatlaticel eccura€y never
sky, and Roseu? A gourcc cnttr pdm d lult- hcretdort rchlcvd. Dltl are dlrectly compared
ably correlated photoru (or rpln-| prrEclo tn I to thc QM prctllctlonr for the full 300'range of
singlet state) that ceparltc. Onc pcrfornr cor- relattvc orlenteEoru of thc polarizers, Moreover, t
related measurementl d tlpir polrriaeHom rtUl we havc cerrled out e morc general tegt of Bell's
use of remote polartzem (or Stern-Gcrlach Ett- inequalttle! that docg rot require the asaumptlon
netg) in varioua ortenteEotlr. For puUculer rets o( rotatlonal lnvrrl&cc.
of polarlzer orientattonr, thc tvo pariiclc cort!- The qrperta-ent war pcrlormed lor varlous dis-
latloqe predlcted tr QM do not obcy Bell'r theo- tences bctvecn'.he lource utd the polarizers.
rem. For lerge tepiullforutr our retultr ere able to 't
The ftrst such lnveetlgatloru employedthe 2-z rule out varioul hypotheselrr'tr according to
decayr produced by ground-attte pocltronlum whlch e nonlectorlztngpure state for two Fat-
annlhllatlon. Except lor onc o( thcm, the expcri- ticler (such a^re singlet stete) evolves towards
Esnt!''t agre€ rtth the QM predlcttonr, Hoe- a mlxhrre of fectorlzlng stater when the two pilr-
€yer, beceuseof thg leck ol eltlctent polarlzerr ttcles separatc. Accordtngly, ruch a locdtza-
for 0.S-MeV photonr, atrong rupplementary
areumpttons ere n€celrary to lnterpret thege
regult! vlr Bell'! thgorem.t $rallarly, to 8r-
tlon process then occur! over dlgtancecof the
order of the coherence length of the wave packets
assoctatedwttlr the emltted photonr. $ch a hypo-
iI
perlmentrothat u6e! palru ol protoru wlth corrc- thesis has already been testedpreviouelyusing I
lated eplru allordr rerults ln agreementwtth QM, patrs of y ray!, wlth conllictlng results,t but *

460
'!
r982 Vorurrr {9. Nuuarr 2 PHYSICAL REVIEV TETTERS l2Jurv 1982
3

ific Erpcrlmcntrl Rerllzrtlon of Elnsteln-Podolsky-Rosen-Bohm


of
Gedankcnexpeilmenti
,tgs A New Vlohtlon of Bell's Inequrlltles
Alain Aspect, PhitippeGrangier,andG€rardRoger
Institut d'Optique Th|orique et ApplQu|e, Laboratoire assocf au Centre Natbnat de Ia Recherche
Scienlifique,
Linit'ersit| Pads..Sud, F-91406 Arsay, Fmnce
(Recetved 30 December 19E1)
87
57 Tte llnear-polarlzaEou corelalloa of pairs of photoas emttted la a radlatfue cascade of
34 calclrrn bas been measured. lbe new er(perlmentd echeme, uelng two-cha.mel polarlzers
37 (1.e., opHcal a.oalogsof Stern-Gerlacb ftltcra), la a atretgbtfor.wardflqnqpogldon of
Ela-
et€lr-Podolslcy-Roeeu-Bohm gefunhenerperiment. Tbe preaentresult€, lu excelled
agreement witl tbe guaDhrmmechaatcal predlcfr,one, lead to tbe greatest vlolad,ou of gen-
erallzed Bell's laegrraltties ever achleved.
PACSnurobers: 03.65.82, 35.80.+a
37

In the well-known Einstein-Podolalcy -Rosen- r1; for photons, a meaaurementalong i yielde


Bohm gefunhenexpnimentI (Fig. 1), a source the reault + I il the polarization is found parallel
ds emits pairs of epln-i particles, in a singlet state to 4 ad -1 if the polartzetion ie found perpen-
d (or pairs of photonsln a similar nonfactorizing dieulet. For a ringlet atate, quenfummechanics
I
state). After the parttclee have eeparated, one prcclictr tona corrclr$on betweeaguch uteas-
I
performe correlated measurementgof their apin urcasntr on thC t?o partlcles. Let us deoote by
be componentaalong arbitrary directons i anO6. P. . (i $ tm psob.btlitsas oi obtaining tlre result
Each measurementcan yield hpo results, denoted, rl along i brrtfclc 1) rnd tl along 6 (particle 2).
lTrc grrenuty
E(;,D =p.r(i,O *r-_(;,6')
-p.-(i, D -p_. (i il (1 )
is the correlation coefficient of the mea€urementl
I
on the two particleg. Bellz considered theorler Itonl (or d protonr). Bst no efticient analyzers
elglaining such correlationa iu due to coEmon rrc rvrllrblc rt rueb ernrgles, and the results
r o m-
propertseg of both particles of the rame pdr; tiet rould havr bccn obhined wlilr ideal polar -
3:
adding a locality aseumptioq he showed that they , laerr rrr dcdrccd tndlrectly lrom Compton gcat-
reor.
tl
.l
a.re constrained by certatn inequalitieg that arc t ri$ cectlncrrt!. The validity of such a pro-
't
.t not always obeyedby the predic$ons of qumtum ccdrrc la tsc eoatcxt d Bell'g theorem has been
I
c
a
mechanict. Suchtieorieg are called3 ,'reali!flc
local theorieg" and they lead to the generellzcC
crltscllcd,t"
Thcrc rrc rlao c4erimnta rlth pairg of low-
..
1! Bell'a inequalitles' eocrgy photoaecnltted ln atomic radiative calr-
t -2sS<2, \S.!|t crder. Tnrc pohrtzers arc available in the viei-
!,, (2)
ble range. Horevet, all previous e:rperiments
where lnvolved slngle-churnel rnalyzers, transmttting
onc polarizatlon (ior F) *O blocki'g th€ orthog-
-+ s = E ( i , O_ e ( i f , ) * e ( i , , D* E ( i , ,6 , )
\ 5 ! , * . . i I--)b)0 ..
1 involves lour Eearurements in four various ori-
tl-'-"1 entationg. On tlre otlrer hand, for guitable sets I
{), t rl
of orientationE,rthe *raltrrm mecl4lgal predic -
I
tions can rea$ the values S=r2r'2, in clear con-
. tradiction with (2): Quantummectianicgcannot I
tl be canglele4 by an underlyingstructure sucn asll
'l -
"realigtic local theories." _t -t
Several experimentewith increasing accuracy
FG. 1. EtDstetD-Podolsky-Rosen-Bohm gedantenex-
havebeen performed, and they clearly favor quan-
periment. lTeapla-l partlctes (or phoroos)in a siag-
.fum mechanics.3,3 Unlorirnately, nonealloweda let atatc {or stnllar) aeparate. lte spln cornponeats
direct test using inequalit.es(2), since nonefol- (or ltnear polertzetioas) of I aad 2 ate mea-curedaiong
5 lowedthe schemeof Fig. I closetyenough. Some i aad 6. Que.uAtra mecbaaicspredlcts strong correia-
. experimentswere performedwith pairs of pho- dons betweeathese measrrrements.

S oci ety
@ 1982The A meri canP hysi cal 9l
rt-

V o r u v e i 9 , N u v s e n2 PHYSICAL REVIE\Y/ LETTERS l2Jurv t982

onal cne. The measuredquantitieswere thus


o n l y t h e c o i n c i d e n c rea t e s i n + 1 c h a n n e l s :R , , ( i ,
61. SeveraldiJficuttiesthen arises as a result of
the very low efficiencyof the detectionsystem
(the photomuitiptiershavelow quantumefficien-
cies andthe angularacceptance is smaU). The
n n e 3 . s u r e m e -nrtfsp o l a r i z a t i o na r e i n h e r e n t l yi n -
cornplete:l!'hena pair has beenemrtted, iJ no Coincrdonca
countis obtainedat one of the photomultipliers,
there is no way to know whetherit has been
FTG. 2. E.xperimental setup. T3o polrrimetecs I and
m i s s e db y t h e ( l o w - e f f i c i e n c yd) e t e c t o ro r w h e t h - E, b orlentatloas i aad b, perform true dichotomic
er it has beenblockedby the polarizer (onlythe measuremeots of llnear polarization on photons u1 c.od
latter casewouldbe a real polarizationmeas- vt. Each polarlmeter is rotata,ble :ror.rad tire rds of
urement). Thus, coincidencecountingrates such tbe iactdeat beam. The cor.ratirg electronics moaitors
as R.-(;, b-)or R--(i,6) cannotbe measureddi- tbe si.6les aad tbe colaeidences.
rectly. It is neverthelesspossibleto derive
from the erperimental data numerical quantities
which can (accordingto quantummechanica)pos- schemeof Fig. 1. True dichotomicpoiarization
sibly violate Bell-type ineqr:alitiea. For thir measurementson visible photonshavebeenper -
purpose, one has to resort to auxiliary experi- formed by replacing ordinary polarizers by two-
menli, where coincidencerateg a.te measrrred channelpolarieers, separatingfwo orthogonal
with one or both polarizer! removed. Some!€t- linea.rpolarizetions, followedby fwo'photomul-
soning, with a few additional-end very nabrfal tipliers (Ftg. 2). The potarizationmeasure-
-ggumptions (such as thc "no-€tillaneement" mentgthen beeomevery simil,ar to usual Stern-
assumptionof Clauser and Horne?),then dlows Geriachmeasurementsfor spin-r!particles.s
one to obtain actually operationel inequaliti€g. Using e fourfold esincidencetechnique,w€
ln this Letter, we report the results of an ex- measurein a single run the four coincidence
p{riment following mueh more closely thc ideal r rateg R..(i,il, yietdingdirectly ttrecorrelation
I

eoefficient for the measurementsalong i and 6:

(3)

It. is then sufficient to repeat thc lane E€lsut€-


menLsfor three other choices of orientations, lphotomuttiplieis on the other side is congtant.
and inequalities (2) can direcily be used as a We have also observed that the sum of the four
test of realistic local theorles versus quantrn coinctdencerates R..(;,0 is constantwhen
mechanicg. This procedureig soundiJ the meas- changingthe orientations; thus the siee of the
ured values (3) of the correlation coeflicientscan selectedsample is found constant.
be tal<enequal to the delinition (1), 1.e., iI we We haveused the high-efficiencysource pre-
assumethat the ensembleof actually detected v i o u s l yd e s c r i b e d . ! A ( J = 0 ) - ( J = l ) - ( J = 0 ) - c a s -
pairs is a faithful sample of all emitted pairs. cade in calcium-40 iiEGiiiGtf eibited by npo-
This assumptionis highly reasonablewith our photonabsorption, with use o{ two single-mode
very symmetrical scheme, where the hro meas- lasers. Pairs of photons(at wavelengthsA,
urement results + 1 and -1 ate treated in the = 5 5 1 . 3n m a n d A , = 4 2 2 . ?n m ) c o r r e l a t e di n p o -
same way (the detectionefflencies in both chan- Iarization are emitted at a typical rate of 5 x 10?
nels of a polarizer are equal). All data are col- s-t. The polarizers are polarizingcubes(Fig. 2)
lected in very similar experimentalconditions, madeof two prisms with suitabledielectric thin
the only changesbeing rotations of the polarizers. films on the sides stuck together;the faces are
Sucha procedureallowg us not only to nuppresg antirefelctioncoatad. Cube I transmits light
possiblesystematicerrors (e.9., changesoccur- polarized in the incidenceplane onto the active
ring whenremovingthe polarizers) but also to hypotenuse(parallet polarization, atongi) *nite
controi more experimentalparameters. For in- i t r e f l e c t s t h e o r t h o g o n apt o l a r i z a t i o n( p e r p e n d r c u -
stance,we havecheckedthat the sum of the coin- l a r p o l a r i z a t i o n ) .C u b eI I w o r k s s i m i l a r l y . F o r
cidencerates of one photomultiptierwith both a c t u a lp o l a r i z e r sw e d e fi n e t r a n s m i s s i o na n d r e -
92
47 sigma
E

Vorumr {9, Nu}rsen2J PHYSICAT REVIE\v LETTERS 20Decrvoerl9t2

tor trapping as b decreases is reflective of the Our results strongly support the conclusionsof
lncorporationof periodic componentsinto the Grossmannand Thonrae.
sequenceof numbers generated. This research was supportedby the Oftice of
To summarize the motivation and principal con- Basic Enerry Sciencesof the U. S. Department
clusion of this Letter, we restatet that lor values of Enerry.
of b where numerically generatedsequenc$ ap-
Pear to be chaotic, it has not been settl,edwheth-
er those sequences"are truly chaotic, or wheth-
€r, in fact, they are really periodic, but wtth l t )P erml nent address: trl i l es l aboratori es, E l khar t ,
exceedinglylarge periods and very long tran- Irid. 46652.
sients required to settle down." On the one hand, (hbreseot address: Departmentof Chemlstry,
Grossmanand Thomaer't have suggestedthat StanfordUalverslty, Stanford, Cal. 94305.
tE . ott, R ev. I\l od.P hys. 53, 655 0980.
(only) the parameter value D= 1 generatespure
?C.A. Walsb andJ. J. Kozak, Phys. Rev. Irtt.
chaos [see the discussionlollowing Eq. (31) of !!,
1 5 0 0( 1 9 3 1 ) .
Ret, 5 and the correlations plotted in their Fig. lE. S. lloutroll, Proc. Symp..Appl.I\fath.Am. Math.
9l On the other hand, for certain othn values Soc.!!, 193 (196{); E. W. Moitroll aqd C. W. rffelss,
of D, numberical results of Lorenz (reported in J. trtath.Phys. !,, teZ (1965); E. W. IUontroll, J. Math.
Ref. 1) "strongly suggest that the sequencesare P bys. 10, ?53 (f969).
{K. Tomta , ln futtern Fornntion by Dlnanlc Sts -
truly chaotic." The purposeof this communica-
tion was to use an independentand exact result tems and futtern Reeognition,cdltcd by H. lfakca
(Sprtnger-Verlag,Hetdelberg,19?9),pp..90-9?..
from the statistical-mechanicaltheory of d=l 55. Tbornaeand S. Grossmana, J. Stat. Phys. 26,
random walks to test the randomnessof the p?rtr- 1 E 5( 1 9 6 1 )
bollc map for parameter values where thc exis- 65. Gr.ossmaaoaod S. Tlromae, 2. l{aturforsch. 32a,
tence ol "bue chaos" is still an op€nquestion. 1 3 s 3( r 9 ? ? ) .

ErperlmentrlTestof Belt'sInaqurliticsUsint Time.YrryintAnrlyzers


Alain Aspect, Jcan Dnlibard,(" lnd Girard Roger
Institut d'O1tiqueThioriquc ct Appliqute, F-91406 Orsay Cldcx, Fmnce
(Recctvcd2? Septcmhr t9E2) il.*r-,,r.
Correlatlous of llqcar polerlzatloas of pairs of photoos bavc bcca measurcd rvith .-..4\ai*11'1,
tlrne-varylag eaelyzcrs. Tbc anelyzer ln aacb leg of ths apparstus lc .n scqrrto-opti-
cd switch follssed by two llscrr polarizers. Ttc ssitcber opcralc et lacomtn.aeuratr [. .l:t \
fregueacler D.la 50 ltHz. Each anlyzrr a.Bou.atsto a polerlzcr wtrtcb Junrps bctwcco
two orleatetlons la e tlmc short comparcd s'tth thc pboton tralslt tlme. lte resulB , L . .l.til.
ara la good rgrccaeat wtth quelhrm ruechanical predlctlons hrt vtolatc Bcll's iacgrrd-
ttres by 5 st:d3rd devlatloos.
PACSuumbcrs:03.55.82,35.60.+s

Bell's inequalities apply to any correlated meas-


urement on two correlated systems. For in-
stance, in the optical version d the Einstein- t(i) Ilrtr
Podolsky-Rosen-Bohm Ge&nhene.rperhn ent,Le
source emits pairs cf photons(Fig. 1). Measure- COINCIDENCE
ments of the correlattons of linear pohrizations TTONITORING
are performed on two photonsbelongingto the
FIC. l. Opdcal verslon of the Elnsteln-Podolslg-
same pair. For pairs emitted in suitablestates, Rosen-BohmGedanhenexperiuenl.The pair of photons
the correlations are strong, To accountlor these u, a,ndr ls anr,lyze_dby llnear potarizers I aod II (ta
correlations, Bel[2consideredtheorieswhich in- orientatioosi and 6) ana photomultlplters. Thc cotn-
voke commonproperties of both membersof the cl dencerate i s nroni tored.

I E04 @ l 9 t 2 T h e A m e r i c a nP h y s i c aSl o c i e t y
'J"n*
't,fc1
.:l
Vor-uue
61. NuMsr.n
I PHYSICAL REVIEW LETTERS 4 Jury 1988

Violation of Bell's Inequality end Clesslcd Ptoblbility


in a Two-PhotonCorreletionExperiment
Z. Y. OuandL. Mandel
Department andAstronr:":r"';i:?:y-:#i;!i"" Rochestcr'
of Phvsics NewYork1627
ll
.l- tr-
l"
{.
'J.'

Correlationmeasurements of mixed signaland idler photonsproduccdin the procq$ oflgg@ic -l


--t" - l. l
d-o_wn-corucrsjos havebcenpcrformedas a functionof two linearoolarizerscttings.It is foundthrt the
Bell inequalityfor two separatedparticlcsis violatcdby about6 standarddeviations,and that clasrical ,. ..
probabilityfor light wavesis violatedsubstantially
alrc.

PACSnumbers:42.50.Wm.
42.50.DY

" 1 l

Althoughthe quantum-mechanical paradoxrelatingto nm are produccd.When the conditiontor d.g.nrr.,.


Einsteinlocalityhasnow beeninvestigated in severalop- phascmatchingis satisfi€d, down-converted, linearlypo-
tical correlation expcriments,l-e someof whichexhibited larizedsignaland idler .photonsemcrte at anglesof
an explicit violationof Bell's inequality,3.s-r there still aboutt 2o relativeto thew Tumpbeamwith the elec.
appcarsto be interestin new observations of this type. tric veton in the planeof the diagrrm. The idlcr pho-
In the most successful of the past expcrimentsthe two tonspassthrougha 90o polarieetion rotrtd, vhile the
correlatedphotonswere producedin the cascadedecay signalphotonstravercca compcn$tingglrrs plate C1
of calciumatoms.6'7More recentlyseveralobscnations producingequaltinc dchy. SiIFd rud irlhr photonsrrc
of nonlocalquantumcorrelations werereportedin exper- thendirestedfrom oppociteridcr torardr r bcamrplitter
iments in which the corrclatedparticlcs were derived (BS), ag ghorn. Thc light bcrm eft€rgitrgfiom thc
from thc interference of signaland idler photonscreated beamsplittet,conri*ing of mircd.l'lnrt rod idlcr pho-
in the process of parametric down.conversion.e-ll tons,passthroughlinearpolrrizenrt rt adjustable an-
ln the followingwe reportanotherphotoncorrelation gles01 and C:, and throughrimilar interference filters
experimentof this type,similarto onc first performedby (IF), and finallyfall on two photodetecton D1 and Dz.
Alley and Shih,ein whichthe observcdviolationof B€llb The photcl€€tric pulscsfrom the tro dct€cton,after
inequalityfor two scparatedparticlesis as large as 6 amplificationandpukc rhaping,tr" fed to the startand
standarddeviations.At the samctime we alsoshowthat stopinputsof a time-todigitalconvertcr(TDC) under
classicalprobabilityrelatingto the wave propcrticsof computercontrsl,that functionrrs a coingidencc coun-
light is violatedto a significantextentin this correlation tcr. The coinEiden€countingrate,aftr subtraction of
experiment. accidcntals,provida e meesure of thc irint probability
An outlineof the experiment is shownin Fig. l. Light P@t,ez)of detectingtwo photonrfor varioursctting
from the 351.1-mmlineof an argon-ionlaserfallson a 4,02of thetws linearpolarizen.
nonlinearcrystal of potassiumdihydrogenphosphate, Theusuallocalityargumcnt, with the assumption that
wheredown-converted photonsof wavelengthabout 702 any two-particlecorrelatioruare attributableto some
hiddcnvariable,with the no-enhanccmcnt hypothaisof
ClauscrandHorne,12 andsith thc rsrumptionthat the
detectorgdo not depend on the biddcnvrriabler,leadsto
thc wclt'koownBcllincquality 12-11

- pGr,eL)+
s -P(ai,e2) PGi,gT
l
+P.Gi,e)-PGi,-)- P(-,ol)=0. (l)
pbi, - ) andPG ,h) arc thecorrcsponding probabili-
ties with oneor the otherlincar polrrizerrtmoved. We
now calculatcthc joint probability P(.f1,hl 5,:t by
quantummechanics andthenby clad<rl waveoptics.
Quantum theory.-lf thc inputto thc dhlectricbeam
splitterBSconsists of an r-polc:i2s6signalphotonanda
rotatcdypolanzd idlerphoton,thentheoutputstatere-
F I G . l . Ou tl i n co f th e a p p a ra tu s . ferredto thetwochannclsI and2 illustratcdin Fig. I is

50 @ l9EEThe AmericanPhysicalSocicty
Voluue 6l, NuusenI PHYSICAL REVIEW LETTERS 4 Jur-v1988

TABLE l. Raultr of coincidcncc@unting mcasuFemcnts


for ccrtain coobinrtinor of polerizcr anglcs0sand ft. -)o
-
a

q
Coincidcnccratc
pcr minutc fr 5so
h
o -.-'-.}.
67. 50 45c 2E.3t 0.E A
22.5e 450 29.8t 0.t 3c
67.5c 0c 29.9t 0.E an
22.50
67.50
0o
No polarizcr
5.6t
34.7t
0.7
0.9
E
l)
No polarizcr 450 36.2t 0.9 a
tr
..a-'-:
€ro
'8
tr l'f--a
whcrewe havewrittcn/, - lEr(+)l', I,- lgf +l | 2 for
thc signaland idlcr intcnsitics.C is a constantcharac- ,3
tcristicof thc dctcctor.Comparison with Eq. (5a)shows 0
,16 t0 ?3 m rG Un t!6
similariticr,but thc tcrmgin (Ii) and Ol) arc abscntin
thc quantum.mcchanical casc,bccauscwc arc dealing
0,furccd
with singlcrignelandidlcr photons. FtG. 2. Mcrsurcd coincidencccountingratc sr a functionof
In thc rpccial carc Rr - l -Tr, Rt - t -Tr, and thc polarizeren3lc 01,witb A fucd rt 45'. Thc full curvc ir
thc quentum prcdiction br$d on Eq. (t t) rcC thc dashdottcd
whenh-r14, E4. (14) rcduccdto
cunc is thc clrsricrl prdhtiot brr€d oo Eq. (lE). The drshcd
PGn*14) - Trl, (tS)
- # Ct(O* I)2)+.*,sin(2or rnd dottd curycs .rc obteinsd by multiplicetioo of thc
rinuridel functionrin Eqr. (i i) en'd(lt), rccFsrively,by 0.?5
whcre to allow for rcdrccd modulrtim ceurcd by irnperfcct align.
rft
m:nt.
- frll I 1 4l,h
I,)2)|h,
"{ t(00'--l 1 (16)
7r Erstrnt(0/) -U]lr/zIl)|. From E4. (16) the phasc 7 of thc modulation depcndson
If wc matc thc rimplifyingassumptionr aboutthc in- thc intcnrityratio Url/U)-R. Althoughf -0 whcn
tcnsityfluctrntbor thEt Ur2)- Ur)2(I + t ), Ul) - 11,1z,t R-1, for verysmsllor vcry largcR thc phasc7 tends
+t), Ul'I.t(ri{4xl +t), thcnthc rclrtivcmodulation towrrds X,rl2. Tbis i8 again in contrast to thc
or "viribility't'y i! E& (15) bcomcs qurotum-mccbanirnl rcsultgivcaby Eq. (l l), for which
the pha$ of thc modulatioair indcpcndcnt of thc ratio
v - (0r)iiti.) rlltitr)* 0,))2, (l?)
of signalto idhr photons.
which is alwryr lcsr than unity and has thc value * TablcI showg thc rcsultsof coincidencc
countingmea-
whcn (Ir) -Oi). ThL ir in cootraEtto thc quantum- surcmcnts for ccrtaiacombinations 0r0z of thc polarizcr
mcchsnicel rcruld'cmbodicrl io Eq. (l l). In thc spccial anglcs,andaftcrgubtraction of accidcntals.As tbc coin-
casc(/i) -(Ir), Eq.(15)rcduccs to cidcnccratc is proportionalto P(0r02), wc can calculate
thc combinatign S givenby rclation(l) up to a scalc
P@n*14)-! C0,)2[t+ * gia2orl. (re) comtant. tf S ir thc quantityaaalogoust; S but cx-
| prcsscdin tcrmsof thc coincidcnccratcs,wcfrnd
S-frQ2.5.;ii.l-h22.i.,0.)+t(6?.sc,4ic)+fi(6?.5o,0.)-nbl.5o,-)-fi(-,45o)
-(l1.5 (le)
HcnccS ir poritivcwitb la sccuracyof about6 gtandard l
deviationg,in violrtiooof thc Bcll iocqualily(l). Unlikc ' diction given by Eq. (tg), rcspcctivcly,with thc scalc
Allcy andShih,ewc prcfcrto bascourconclusions on thc constants K and C adjustcd for bcst fit. The mean
quaatityS rrthls tlrn on tlrir rimplcrcouating-ratc ra- counting ratcs of dctectors I and 2 wcre 2600/sccand
tio d, bccaurcthc Bcll incquality(t) ds not dcpcndon 2t00/scc when 4-02-45o. It will b€ sccn that
symnctry with rcrpct to 81,02,aod thc thrcsholdfor P@nxl4) docs indccd cxhibit thc expcctcd sinusoidal
violationoccunat S -0. modulationwith thc anglc 0s, with half pcriod of 90o.
Whcn0zis fixcdat 45oandthe e,rglc01is varicd,the Thc obscrvedrelativc modulationV obtainedfrom thc
resultsof thc coincidcncc countingm€asurcmcnts arc il- bcst fitting curveis about 0.76, which is grcatcr than thc
lustratcdin Fig. 2. Thc solidcurvcandthc dashdotted 0.50 figure predictedby thc classicalrelation (lE), but
curvc in Fig. 2 corrcrpondto tbc quantum-mcchanical bclow the 1009bvaluc given by the quantum relation
prcdiction$vco by Eq. (tt) andthc clarsicalwavcprc- (t t). Wc bclicvcthat thc rcasonfor thc ubscrvcddepth

52'
PSYSICAL REVIEW
LETTERS
VotuuE8l 7 DECEMBER1998 NutrlsER23

Violationof Bell's InequalityunderStrict EinsteinLocality Conditions


GregorWeihs,ThomasJennewein, Simon,HaraldWeinfufter,andAntonTsilinger
Christoph
Institutfiir Experimentalphysik,Universittit Innsbruck,Technikerstra,Be
25, A-6020 Innsbruck, Austria
(Received 6 August1998)
We observestrongviolationof Bell's inequalityin an Einstein-Podolsky-Rosen-typc experimentwitlr
independentobservcrs. Our cxperimentdefinitely implementsthe idcas behind the well-known work
by Aspectel al. We for thc first time fully enforcethe condition of locality, a central aslglplgiln
thc derivatignof Bell's theorer,! The
-by d
sufficientphysicaldistanccbetweenthe measurement stations,by ultrafastand randomsettingof the
analyzcrs, and by complctelyindependent dataregistration. tS003l-9007(98)07901-01

PACSoumbers:03.65.82

The stronger-than-classical conelations between en- the directionsof polarizationanalysiswereswitchedafter


tangled quantum systems,as first disarvered by Ein- the photonsleft the source. Aspectet al,, however,used
stein, Podolsky,and Rosen (EPR) in 1935 [], have periodicsinusoidalswitching,which is predictable into
ever sinceoccupieda centralposrtionin the discussions the future. Thus communication slowerthan the speed
of the foundationsof quantummechanics.After Bell's of light, or even at the speedof light [8], could in
discovery[2] thatEPR's implicationto explainthe corre- principleexplain the resultsobtained. Thereforethis
lationsusinghiddenparimeterswould contradictthe pre- secondloopholeis still open.
dictionsof quantumphysics,a numberof experimental The assumption of localityin the derivationof Bell's
testshavebeenperformedt3-51. All recentexperiments theoremrequiresthat the individualmeasurement pro-
confirmthe predictionsof quantummechanics.Yet, from cessesof the two observersare spacelikeseparated
a strictlylogicalpoint of view, they don't succeedin rul- (Fig. l). We define an individualmeasurement to last
ing out a localrealisticexplanation completely,because of from the firstpointin time whichcaninfluence thechoice
two essentialloopholes. The first loopholebuilds on the of the analyzersettinguntil the final registrationof the
fact that all experimentsso far detectonly a small subset photon. Suchan individualmeasurement then has to be
of all pairs created[6]. It is thereforenecessaryto as- so quick that it is impossiblefor any informationaboutit
sumethatthe pairsregistered area fair sampleof all pairs to travelvia any (possiblyunknown)channelto the other
emitted. In principlethis could be wrong and once the observerbeforehe, in turn,finisheshis measurement [9].
apparatus is sufficientlyrefinedthe experimentalobserva- Selection of an analyzerdirectionhasto be completely un-
tions will contradictquantummechanics.Yet we agree predictable, whichnecessitates a physicalrandomnumber
with Bell [7] that ".,. it is hard for me to believethat generator.A pseudo-random-number generatorcannotbe
quantummechanicsworks so nicelyfor inefficientpracti- used,sinceits stateat any time is predetermined.Further-
cal set-upsand is yet going to fail badly when sfficient more,to achievecompleteindependence of both observ-
refinementsare made. Of more hnportance,in my opin- ers,one shouldavoid any commoncontextas would be
ion, is the completeabsenceof the vtal time factor in conventional registrationof coincidences asin all previous
existingexperiments,The analyzersare not rototeddur- experiments tl0l. Rather the individual eventsshould
ing theflight of theparticles." be registeredon both sidesindependently and compared
This is the secondloopholewhich so far hasonly been only after the measurements are finished. This requires
encountered in an experimentby Aspectet al. [4] where independent andhighlyaccurate time baseson bothsides.

0031-9007
/98/81(23)/5039(5)$15.00@ 1998TheAmerican
Physical
Society s039
Vot utvc 81,NutrlgeR23 PHYSICAL REVIEW LETTERS 7 DeceN,tsen
1998

A telescope wasusedto narrowtheUV-pumpbeamU2J,


in orderto enhance the couplingof the702nm photons
intothetwo single-mode glassfibers.On thewayto the
fibers,thephotons passed a half-wave plateandthecom-
pensator crystalsnecessary to compensate for in-crystal
birefringence and to adjustthe internalphaseja of the
entangled statel'P) : | /JT(lH)r ly)z + e'elV)rlH)),
whichwechoseg : n.
o The single-mode opticalfibershadbeenselected for a
3.
o cutoffwavelength closeto 700nm to minimizecoupling
E losses.Manualfiberpolarization controllerswereinserted
F
at the sourcelocationinto both arrns to be able to
compensate for anyuniury polarization transformation in
thefibercable.Depolarization withinthefiberswasfound
to be lessthan l% andpolarization provedto be stable
(rotationlessthanl') withinI hour.
Eachof the obscruers(scc Fig. 2) switchedthe di-
.200 200
rectionof local polarizationanalysiswith a transverse
Spae [m] electro-optic modulator.Its optic axis was sct at 45c
FIG. l. Spacetimcdiagramof our Bell cxperiment. Selecting with respect to thesubsequent polarizcr.Applyinga volt-
a random analyzer dircction, scning thc analyzcr, and finally agecauses a rotationof the polarization of light passing
detectinga photon constitutethc measurcmcntprocess. This throughthe modulatorby an angleproportional to the
processon Alice's side must fully lie inside the shadcdregion
voltage[3]. For themeasurements themodulators were
which is invisible to Bob's during his own measurcment.For
oursetup this EFsls that the decision about the scaing has switched fastbetween a rotationof 0oand45o,
to be rnadc rftef point "X" if the conrespondingphotons are Thc modutation systems(high-voltage amplifierand
detectedu ryoc*imc points "I" a,nd'll', respectivcly. In our electro-optic modulator)had a frequencyrangefrom dc
expcrirnentrhe rncrsurcmentproccss(indicatedby a shon black to 30 MHz. Operating the systems at high frequencies
bar) incltding drc choice of a random number took lcss than
only one-tanfi of the maximum allowed dme. The vertical
weobserved a reduced polarization contrast of 97%(Bob)
parts of thc kinked photon world lines emerging from the and987o(Aliee).This,however, is no realdepolarization
sourcc repsescntthe fiber coils at thc sourcb location, which but merelyrefleetsthe fact that we are averaging over
arc obviouslyirrclevantto the locality argument. the polarization rotationinducedby an electricalsignal
from thehigh-volageamplifier,whichis not of prfectly
rectangular shape.
In our exirerirtrnt"fol the fine dme,any mutualinflu- Theactualorienedonfor localpolarization analysiswas
encebetweenSrctwo observadons is excludedwithin the ' determined independently by a physicalrandomnumber
realmof Einsrcinlocality. To achievethis conditionthe
observers "Alice" and"Bob" wercspatially$eparated by
4fr) m acrossthe InnsbruckUniversitysciencecampus, Deteclors
whichin turnmeansthattheindividualmeasurements as Polarizer
definedabovehadto be shorterthan'1.3,rs, thetimefor
directcomrnunication atthespeed of light. Weusedpolar-
izationentangled photonpairswhichweresentto rheob-
FromSource
serversthroughopticalfibers[l]. About250m of each
500m longcablewaslaid outandtherestwasleft coiled AmBfifier
.at thesource(seeFig. l). Thedifference in fiberlength
waslessthanI m, whichmeans thatthe photons werereg-
isteredsimultaneously within5 ns. Theduration qf anin-
dividualrneasurernent waskeptfarbelowthe 1.3prslimit
usinghighspeedphysicalrandomnumbergenerators and
fastelectro-opticmodulators.Independent dataregistra-
tion wasperformed by eachobserver havinghis owntime
intervatanalyzerandatomicclock,synchronized onlyonce FIG. 2. One of the two observerstations. A random num-
beforeeachexperiment cycle. ber generatoris driving the electro-opticmodulator. Silicon
avalanchephotodiodesare used as detectors. A "timc tag" is
Our sourceof photonpairsis degenerate type-Ilpara- storedfor eachdetectedphotontogetherwith the conesponding
metricdown-conversion [5] where we pump a BBOcrys- randomnumber"0" or "l" and the code for the detector"+!'
tal with400mWof 351nm lightfromanargon-ion laser. or "-" conesponding to the two outputsof the polarizer.

5040
:
Vot ulvc 81, Nulrgnn 23 PHYSICAL REVIEW LETTERS 7 DnceMsen
1998

generator. This generator has a light-emitting diode series,but did not in any way degradetime resolutionor
(coherencetime lc = l0 fs) illuminating a beam splitter accuracy.
whose outputs are monitoredby photomultipliers. The Each observer station featured a personalcomputer
subsequentelectroniccircuit sets its output to "0"("1") which stored the tables of time tags accumulatedin
upon receiving a pulse from photomultiplier "0"("1"). a measurementcycle. Long after measurementswere
Events where both photomultipliersregister a photon finishedwe analyzedthe filesfor coincidenceswith a third
within Al < 2 ns are ignored. The resultingbinary ran- computer. Coincidences were identifiedby calculating
dom numbergeneratorhas a maximumtogglefrequency time differencesbetweenAlice's and Bob'stime tagsand
of 500 MHz. By changingthe sourceintensitythe mean comparingthesewith a time window(typicallya few ns).
interval was adjustedto about l0 ns in order to have a As therewerefour channelson eachside-two detectors
high primaryrandombit rate[4,15]. Certainlythis kind with two switch positions-this procedureyielded 16
of random-number generatoris not necessarilyevenlydis- coincidencerates,appropriatefor the analysisof Bell's
tributed. For a testof Bell's inequalityit is, however,not inequality. The coincidencepeak was nearly noise-free
necessaryto have perfectlyeyen distribution,becauseall [signal-to-noiseratio (SNR) > 100] with approximately
corelation functionsarenormalizedto the total numberof Gaussianshapeand a width (FWHM) of about 2 ns.
eventsfor a certaincombinationof the analyzers'settings. All data reportedhere were calculatedwith a window
Still, we kept the distributionevento within ZVoin order of 6 ns.
to obtain an approximatelyequal numbcr of samples Therearemanyvariantsof Bell's inequalities.Herewe
for each setting by changingthe internal photoelectron usea versionfirst derivedby Clauseret al. tl6l (CHSH)
amplificationof the two photomultipliers.Becauseof sinceit appliesdirectlyto our experimentalconfiguration.
the limited speedof the subsequent modulationsystemit The numberof coincidences betweenAlice's detector
was sufficient to sample this random number generator i and Bob's detector1 is denotedby Cii@, F) with
periodicallyat a rate of l0 MHz. i, j e {+, -} where a and p are the directionsof the
The total of the delays occurring in the electronics two polarizationanalyzersand "f" and "-" denotethe
and optics of our random number generator,sampling two outputsof a two-channelpolarizer,respectively.If
circuit, amplifier,electro-opticmodulator,and avalanche we assumethat the detectedpairs are a fair sampleof
photodiodeswas measuredto be 75 ns. Allowing for all pairs emitted,then the normalizedexpectationvalue
another25 ns, to be sure that the autocorrelationof the E(q,F) of the correlationbetweenAlice's and Bob's
random number generatoroutput signal is sufficiently local resultsis E(a, F) : [C++(o, F) + C--(a, p) -
low, it was safe to assumethat the spcific choice of C*-(a,p) - C-+(e,P))/N, whereN is the sumof all
an analyzersctting would not be influencedby any event coincidencerates. In a rather generalform -the CHSH
more than 100 ns earlier. This was much shoner than ygrylitl,ea$
the 1.3 ps thatany informationaboutthe otherobseryer's
measurement would havebeenretarded. S(o,dt, F, F') - lE(o,p) - E(o',B)l ) Pt\.\"-x
The photonswere deiectedby silicon avalanchepho- = z. (l)
+ lr(a, P')+ E(o',9')l
todiodeswith dark count rates(noise) of a few hundred
per second. This is very small comparedto the 10.000- Quantumtheory predictsa sinusoidaldependence for
15.000signalcountsper secondper detector.The pulses the coincidence iati eT*@,P) o sin2(F a) on rhe
of eachdetectorwere fed into electroniccircuits,respon- differenceangleof the analyzerdirectionsin Alice's and
sible for disregardingeventsthat occunedduring transi- Bob's experiments.The samebehaviorcan alsobe seen
tions of the switch signaland encodingthe positionof in the correlation f u n c t i o nE s ^ ( a , 9 ) : - c o s [ 2 ( F -
the switchin an extrasignal. Finally,all detections were Thus, for variouscombinations of analyzerdirec-
time-tagged in specialtime intervalanalyzers with 75 ps t")].
i o n sd , F , o ' , 9 ' t h e s ef u n c t i o n sv i o l a t eB e l l ' s i n e q u a l -
resolutionand 0.5 ns accuracyreferencedto a rubidium ity. Maximumviolationis obtainedusingthe following
standardtogetherwith the appendant switchposirion.The setof anglesSflfi'*: Sq'(0o,45',22,5,67.5';- 2'n:
overalldeadtime of an individualdetectionchannelwas 2.82> 2.
approximatelyI ps. If, however,the perfectcorrelations(a - B :0o or
Using an auxiliaryinput of our time intervalanalyzers 90o) have a reducedvisibility V s I then the quantum
we synchronized Alice's andBob's time scalesby sending theoreticalpredictionsfor E and S are reducedas well by
laserpulses(670 nm wavelength,3 ns width) througha the samefactorindependent of the angle. Thus,because
secondopticalfiber. While the actualjitterbetweenthese the visibilityof the perfectcorrelationsin our experiment
pulseswaslessthan0.5 ns,the auxiliaryinputof thetime was about97Vowe expectS to be not higherthan2.74
intervalanalyzershad a resolutionnot betterthan 20 ns if alignmentof all anglesis perfectand all detecrorsare
thus limiting synchronizationaccuracy. This nonperfect equallyefficient.
synchronization only limited our ability to exactlypredict We performedvariousmeasurements with thedescribed
the apparenttime shift betweenAlice's and Bob's data setup. The datapresented in Fig. 3 arethe resultof a scan

5041
Voluun 81,Nulasen23 PHYSICAL REVIEW LETTERS 7 DeceNasen
1998

Rotation realisticor semiclassicalinterpretationsare still possible,


but we would thenhaveto assumethatthe sampleof pairs
registeredis not a faithful representative of the whole
800
ensembleemitted. While we shareBell's judgmentabout
600 the likelihood of that explanation[7], we agreethat an
v, ultimateexperimentshould also have higher detection/
lc)
'Ec 4 0 0 collectionefficiency,which wasSVoin our experiment.
Further improvements,e.g., having a human observer
.oo
o choosethe analyzerdirections,would again necessitate
c major improvementsof technologyas was the casein or-
€o der to finally, after more than 15 years,go significantly
o
c
'6 beyondthe beautiful 1982expcrimentof Aspectet al. l4l.
ooo Expectingthat any improvedexperimentwill also agree
o
with quantumtheory,a shift of our classicalphilosophi-
400
cal positionsseemsnecessary.Among the possibleim-
plicationsare nonlocalityor completedeterminismor the
r A+0/B+0
abandonment of counterfactualconclusions.Whetheror
o A+1/B+0 not this will finally answerthe eternalquestion:"Is the
.100 .50 0 50 100 moon there,when nobodylooks?" [8], is certainlyup to
BiasVoltage(Alice)[V] the reader'spersonaljudgment.
This work was supported by the Austrian Science
FIG. 3. Fourout of sixrccncoincidence ratesbetweenvarious Foundation(FWF), project56502,by the U.S. NSF Grant
detectionchannelsas functions of bias voltage (analyzer No. PHY
97-2261{ and by the APART programof the
rotation anglc) on Alice's modulator. A + l/B - 0, for
example,are thc coincidences bctweenAlice's "*" dctector Austrian Academy of Sciences.
with swirchhavingbecnin position"1" andBob's"-" dctector
with swirchposition"0". Thedifferenccin hcightis explained
by diffcrentefficicncies
of the dctectors.
ttl A. Einstein, B. Podolsky,andN. Rosen,Phys.Rcv.47,
777(te3s).
t2l J.S. Bell, Physies(tong IslandCity, N.Y.) 1, 195(1965).
of the dc biasvoltagein Alice's modulatibnsystemovera
t3l S.J. Freedman and J.F. Clauser,Phys.Rev.Lett.2E,938
200 V rangein 5 V steps.At eachpointa synchronization (1e72).
pulsetriggcreda measurement periodof 5 s on eachside. t4l A. Aspect,J. Dalibard,and G. Roger,Phys.Rev. Lett, 49,
From thetime-tagserieswe extractedcoincidences afterall 1804(t982).
measurements had beenfinished. Figure 3 showsfour of t5l P.G. Kwiat, K. Mattle,H. Wcinfurter,A. Zeilinger,A. V.
the 16 resultingcoincidenceratesas functionsof the bias Sergienko, and Y. Shih,Phys.Rev.Len. 75,4337(1995).
voltage. Eachcurvecorresponds to a certaindetectorand t6l P. Pearle,PhysRev.D 2, l4l8 (1970).
a certainmodulatorstateon eachside. A nonlineary2 fit [7] The observationthat the photonsin a pair, as usedby us,
showedperfectagreementwith the sine curve predicted are alwaysfound to have different polarizationcan not as
by quantumtheory. Visibility was 97Voas one could easily be understood as the fact that thc socksin a pair,as
worn by Bertlmann,are always found to have different
haveexpectedfrom the previouslymeasuredpolarization
colors. J. S. Bcll, J. Phys. (Paris),Colloq., Suppl.3,
contrast. No oscillationsin the singlescount rateswere C2, 4l (1981), reprintcd in Spealcable ard Unspeakable
found. We want to stressagainthat the accidentalco- in Quantum Mechanics (Cambridge University Press,
incidenceshavenot beensubtractedfrom the plotteddata. Cambridge,1987),pp. 139-158.
In orderto give quantitativeresultsfor the violationof t8l A .Zei l i nger,P hys.Len. l l E A , I (1986).
Bell's inequalitywith betterstatistics,wc performedex- t9l G. Weihs, H. Weinfurter,and A. Zeilinger, in Experi-
perimentalruns with the settings0o,45dfor Alice's and mentalMetaphysics, editedby R.S. Cohenet al. (Kluwer
22.5o,67 .5' for Bob's polarizationanalyzer.A typicalob- Academic Publishers,Dordrecht, 1997), pp. 271-280;
servedvalueof the functionS in sucha measurement was G. Weihs, H. Weinfurter,and A. Zeilinger,Acta Phys.
S : 2.73 t 0.02 for 14700 coincidence eventscollected Slovaca47,337 (1997).
in f0 $ mis coiiespondsto a violation of the CHSH in- [l0] P.V. Christiansen, Phys.Today3E,No. ll, ll (1985).
tl[ W. Tittel, J. Brendel,B. Gisin, T. Herzog,H. Zbinden,
equalityof 30 standarddeviationsassumingonly statistical
and N. Gisin,Phys.Rev. A 57,3229(1998).
erors. If we allow for asymmetriesbetweenthe detec-
tl2l P.R. Tapster,J.G.Rarity,andP.C.M. Owens,Phys.Rev.
tors and minor enors of the modulatorvoltagesour result Lett. 73, 1923(1994);C. H. Monken,P. H. SoutoRibeiro,
agreesvery well with the quantumtheoreticalprediction. and S. Pddua,Phys.Rev. A 57,R2267(1998).
While our results confirm the quantum theoretical [3] Precisely speaking,the modulator introducesa phase
predictions[7], we admit that, howeverunlikely, local shift betweenthe linearly polarizedcomponentsparallel

\. :\ \-\
50/'2 r.'J. ;{.c* 1.t.-liV' ''. I .rr A.j.rq.
t
,.T',''r.
fi ll
l'.r
\l
letters to nature
Nru,,'.(dr,Q) - Natr.r.n,(dr,dr)
(1)
Experimental
uiolationofa Bell's Q(4',Q): Nr".. * Ndff.r.n,

is measuredby repeatingthe experimentmany times. Here N.u-.


inequality
with efficiefitdetection and N46.r.n,arethe number of measurements wherethe two results
werethe sameand different,respectively.The CHSH form of Bell's
M. A Rowe*, D. Kielpinski*, U. ffieyer*, G.A Sackett*r W, ill.ltano*,
inequalitiesstatesthat the correlationsresultingfrom local realistic
G. Monroet & D. J. Wineland*
theoriesmust obey:
* TimeandFrequenryDivisioraNationallnstituteof Standards
and Technology, B(a,, 6r,!r,tr) :14(6t,^lz)- q(qv^y)l
Boulder,Colorado80305,USA
f Departmentof Physia,Universityof Michigan,Ann Arbor,Michigan48109, I lq(6,,Br)+ q(av P)l < 2 (2)
USA
where d1 and 6t (0, and 72) are specificvaluesof dr (dz). For
Local realism is the idea that objects have definite properties example,in a photon experimentls,parametricdown-conversion
whether or not they are measured,and that measurementsof preparesa pair of photonsin a singletEinstein-Podolsky-Rosen
thesepropertiesarenot affectedby eventstaking placesuffrciently (EPR)pair.After this, a variablerotation of the photon polarization
far awayr.Einstein, Podolsky and Rosen2used these reasonable is appliedto eachphoton. Finally,the photons'polarizationstates,
assumptionsto concludethat quantum mechanicsis incomplete. verticalor horizontal,aredetermined.
Starting in 1965, Bell and others constructed mathematical Our experimentpreparesa pair of two-levelatomic ions in a
inequalities whereby experimental tests could distinguish repeatableconfigpration(entangledstate).Next, a laserfield is
between quantum mechanics and local realistic theoriesr'3-5. appliedto the particles;the classical manipulationvariablesarethe
Many experimentsr'6-rs have since been done that are consistent
with quantum mechanicsand inconsistentwith local realism.But
theseconclusionsremain the subjectof considerableinterest and
debate, and experiments are still being refined to overcome
'loopholes'that might allow a local redistic interpretation. Here I Classical

we have measured correlations in the classical properties of


massiveentangledparticles (eBe+ions): thesecorrelationsviolate
a form of Bell'sinequality.Our measuredvalueof the appropriate
Bell's 'signal' is2.25 -t-0.03,whereasavalue of 2 is the maximum
allowed by local realistic theories of nature. In contrast to
previous measurementswith massiveparticles,.this violation of
Bell's inequality was obtained by uie of a complete set of
9(Qr'02)
measurements.Moreover, the high detection efficiency of our
apparatuseliminatesthe so-called'detection'Ioophole
Early experimentsto test Bell'sinequalitieswere subj,Sctto two
primary, although seemingly implausible, loopholes. The first
might be termedthe localityor'lightcone' loophole,in rvhichthe
correlations of apparently separateevents could result from
unknown subluminalsignalspropagatingbetweendifferentregions
of the apparatus.Aspectt6has given a brief history of this issue,
starting with the experimentsof ref. 8 and highlighting the strict
relativistic separationbetweenmeasurementsreported by the Classical Detection laser
Innsbruckgrouptt.Similarresultshavealsobeenreportedfor the
Genevaexperimentr4'r7. Thesecondloopholeis usuallyreferredto as
the detectionloophole. All experimentsup to now have had
detectionefficiencieslow enoughto allow the possibility that the
subensemble of detectedeventsagreeswith quantum mechanics
eventhoughthe entireensemblesatisfies Bell'sinequalities. There-
foreit mustbe assumed that the detectedeventsrePresent theentire
ensemble; a fair-samplinghypothesis. Severalproposalsfor closing Q(Qr'Qz)
this loopholehavebeenmaders-z4; webelievethe experimentthat we
reporthereis the firstto do so.Anotherfeatureof our experimentis
that it usesmassiveparticles.A previoustestof Bell'sinequalitywas
carriedout on protons25, but the interpretationof the detected
eventsrelied on quantum mechanics,as symmetriesvalid given
quantummechanicswereusedto extrapolatethe datato a complete
setof Bell'sangles.Herewe do not makesuchassumptions.
A Bell measurementof the type suggestedby Clauser,Horne, Figure1 lllustration experiments
ofhowBell'sinequality work.Theideaisthata 'magic
Shimony and Holts (CHSH) consistsof three basic ingredients box'emitsa pairofparticles. Weattempt to determine thejointpropertiesofthese
(Fig.la). Firstis the preparationof a pair of particlesin a repeatable particles various
byapplying manipulations
classical tothemandobserving the
'magic' of
box in Fig. 1a). conelations themeasurementoutputs.a, A general CHSH typeof Bell's
inequality
startingconfiguration(the output of the
Second,a variableclassicalmanipulation is appliedindependently experiment. Themanipulation
b,0urexperiment. isa laserwave withphases
applied dr
to eachparticle;thesemanipulationsarelabelleddr and {2. Finally, and02toion1 andion2 respectively, Themeasurement is the ofphotons
detection
in the detection phase, a classicalproperty with two possible emanating fromtheionsupon ofadetection
application Twopossible
laser. measurement
outcomesis measuredfor eachof the particles.The correlationof outcomes arepossible, offewphotons
detection (asdepictedforion1inthefigure)orthe
theseoutcomes ofmany
detection photons
(asdepictedforion2 inthefigure).

unrr lnn lvot. aos I r s FF.BRI'ARY2001lwww.nature.com M,etloot 791


letters to nature
state,lry'),to measurementwhere a measurementoutcornewas recordedfor
1 everyexperiment.Our detectioneffi.ciencywas high enoughfor a
ItL ): - { ( 1 4 r r ( 0 t+ 0 2t);1 t) 1- e -(o ' + d ,)l l )) Bell'sinequalityto be violatedwithout requiringthe assumptionof
z\/z
fair sampling,therebyclosingthe detectionloopholein this experi-
- i ( l - r ; @ * 4 2 ) ) ( e - ' t 'Il \l + r - ' c , l l l ) ) l (6) ment. The ions wereseparated by a distancelargeenoughthat no
known interactiorrcould affect the results;however,the lightcone
Using the measurement operatorsN,u*.: N.,[l t tXt t l+ loopholeremainsopenhere.Further detailsof this experimentwill
l l l x l l l l a n d N o , * , . , , , : N , ol x, ti l tl l + l l t x l f l l , t h e c o r -be publishedelsevrhere. I
relationfunctionis calculated
to be
Methods
- 211
q(Qr,6r)-*tzt I I g4sr+6il1' - f@r+tDl2l
= cos(d,* Qr) Phase
calibration
8''
The experirnent was run with specific phase differences of the Raman laser beam fields at
(7) eachion. In order to implement a complete set of laserphases,a calibration of the phaseon
each ion as a function ol. axial trap strength was made. We emphasize that the calibration
The CHSH inequality (equation (2)) is maximally violated by
method is classicalin na.ture.Although quantum mechanics guided the choice of
quantum mechanicsat certainsetsof phaseangles.One suchsetis calibration method, no ,quantum mechanics was used to interpret the signal. General
e, : -(zrl8), 6r - 3118,B, - -(trl8) and h = 3118.With these arguments are used to describe the signal resulting from a sequenceoflaser pulses and its
phaseanglesquantum mechanicspredicts dependence on the classical physical parameters of the system, the laser phase at the ion,
and the ionb position.
r 3r r 3zr\
6^l'[ - 8 ' 8 ' - d ' a ) : z t 1 z ^ r: (8) In the calibration procedure, a Ramsey experiment was performed on two ions. The first
z'l2 Rabi rotation was pr:rformed identically each time. The laser phasesat the ions'
positions for the second tl2Rabi rotation were varied, d1 for ion I and {2 for ion 2. The
This violates the local realism condition, which requires that detection signal is the total number of photons counted during detection.With an
B<2. auxiliary one-ion experiment we first established ernpirically that the individual signal
dependsonly on the laserphaseat an individual ion and is C +,4,cosd. Here C and A are
The correiation function is measured experimentally at four
the offset and amplitude of the one-ion signal. We measure the detector to be linear, so that
sets of phase angles,listed in Table l. The experiment is repeated the detection signal is th.esum of the two ions' individual signals.The two-ion signal is
N,o, : 20,000times at eachof the four setsof phases.For eachset of therefore
phasesthe correlation function is calculated using
* C, 1 - A c o s d=,2 C * z a c o s [ ( d ,+ d , ) ' ] . o r B f t ,- * , i l
C+Acos@ (ll)
(No*N')-N, LLJILI
q- (e)
N,o, By measuring the fringe amplitude and phase as {, : (4, + 6)/2 is swept, we calibrate
6, - 6, as a function oftrap strength and ensure that {, * {, is independentoftrap
Here N6, N1 and AI2are the number of eventswith zero, on,: and two strength.
ions bright, respectively.The correlation values from the four setsof We use the phase conrention that at the ion separation used for the entanglement
plrase angles are combined into the Bell's signal, B(ar,6r,0r,"lr), preparation pulse the rnaximum of the correlation function is at {, : dz = 0 (or
Ad : d,o, = 0). Our measurementprocedure begins by experimentallyfinding this
using equation (2). The correlation values and resulting Bell's
condition of d' - 6z: 0 by keeping Ad : 0 and scanning the synthesizerphaseto find
signals frorn five data runs are given in Table 2. the maximum correlation. The experiment is then adjusted to the phase anglesspecified
So far we have described. the experiment in terms of perfect above by switching the axial trap strength to set A{ and incrementing the synthesizer
implementation of the phase angles. In the actuai experiment, phaseto set @,o,.

however, cyr, 6r, 0z and !2 are not quite the same angles both
times they occur in the Bell's inequality. In our experiment the tocalityissues
dominant reason for this error results from the phase instability of The ions are separated by a distance of approxirnately 3 pm, which is greater than 100
the synthesizer, which can cause the angles to drift appreciably times the size of the wavepacket of each ion. Although the Coulornb interaction strongly
couples the ions'motionr it does not affect the ions' internal states.At this distance,all
during four minutes, the time required to take a complete set of
known relevant interactions are expected to be srnall. For example, dipole-dipole
measurements.This random drift causesa root-mean-squarederror interactions between the ions slightly modifr the light-scattering intensity, but this effect is
for the correlation function of +0.03 on this timescale, which negligible for the ion-ioh separationsusedt'. Also, the detection solid angle is large
propagatesto an error of -f 0.06 for the Bell's signal. The error for enough that Young's interference fringes, ifpresent, are averaged out30.Even though all
knovrn interactions would causenegligible correlations in the measurement outcomes, the
the Beil's signal from the five combined data sets is then +0.03,
ion separationis not large enough to eliminate the lightcone loophole.
consistent with the run-to-run variation observed. Averaging the We note that the experiment would be conceptually simpler if, after creating the
five Bell's signals from Table 2, we arrive at our experimental result, entangledstate,we separatedthe ions so that the input manipulations and measurements
which is were done individually. Flowever, unless we separated the ions by a distance large enough
to overcome the lightcone loophole, this is only a matter of convenienceof description and
r3r 3o\
==2.25{-.0.03 does not change the conclusions that can be drawn from the results.

fr 2U'
'(-
S,T u T) Received
25 Octobcr;acceptr:d
30 November2000.
If we take into account the imperfections of our experiment l. Clauser, ). F. & Shimony, A. Bell's theorem: experimental tests and implications. Rep. Prog. Phys. 4L,
(imperfect state fidelity, manipulations, and detection), this value 1883-1927 (t978).
2. Einstein, A., Podolsky, B. & Rosen, N. Can quantum-nrechanical description of reality be considered
agreeswith the prediction of quantum mechanics.
conrplete? Phys. Rev. 47,777-780 (1935).
The result above was obtained using the outcomes of every 3. Bell,],S.OntheEinstein-Podolsky-Rosenparadox.Physicsl,195-200(1965).
experiment, so that no fair-sampling hypothesis is required. In 4. Bell, LS.inFoundations of Quantun Mechanics(ed. d'Espagnat,B.) l7t-l8l (Academic,NewYork,

this case,the issue of detection effrciency is replaced by detection r97r).


5. Clauser, |. F., Horne, M, A., Shimony, A. & Holt, R. A. Proposed expcriment to test local hidden-
accuracy. The dominant cause of inaccuracy in our state detection variable theories . Phys, Rcv. Lett. 23,880-884 ( 1969).
comes fi'om the bright state becoming dark because of optical 6. Freednran, S. i. & Clauser, J. F. Experinrental test of local hidden-variable theories. P]ys. Rev.Lett. 28,
pumping effects. For example, imperfect circular polarization of 938-941 (1972).
7. Fry E. S. & Thompson, F.. C. Experimental test of local hidden-variable theories. Phys. Rev.Lett. 37,
the detection light allows an ion in the ll) state to be pumped to lf),
465-468 (1976).
resulting in fewer collected photons from a bright ion. Becauseof 8. Aspect, A., Grangie6 P, & Roger, G. Experimental realization ofEinstein-Podolsky-Rosen-Bohm
such errors, a bright ion is misidentified 2o/oof the time as being Gedankenexperirnent:a new violation of Bell's inequalities. Phys. Rev.Lett. 49,91-94 (lggD.
dark. This imperfect detection accuracy decreasesthe magnitude of 9. Aspect, A., Dalibard, J. & Roger, G. Experimental test of Bellt inequalities using time-varying
analyzcrs.Phys. Rev.lztt.49, 1804-1807 (1982).
the measured correlations. We estimate that our Bell's signal would
10. Ou, Z. Y. & Mandel, L. Violation of Bellt inequality and classical probability in a two-photon
be 2.37 with perfect detection accuracy. correlation experinrent. I'hys. R ev. Lett, 61, 50- 53 ( I 98 8).
We have thus presented experimental results of a Bell's inequality I 1. Shih, Y. H. & Altey, C. O. New type of Einstein-Podolsky-Rosen-Bohnr experiment using pairs of light

NATUREIVOL 409I 15FEBRUARY


2001|wwwnature.com #O zoot MacmillanMagazinesLtd 793
PHYSICAL REVIEW
LETTERS
Volur're67 5 A U G U S TI 9 9 I Nuusen6

tI,._
QuantumCryptographyBasedon Bell'sTheorem tr,'.. .l.-r--.'ts"
I
-
't.
ArturK' Ekert "c')'I'
."''!-
Merton Collegeand PhysicsDepartment,Oxford Uniuersity,Oxford OXI 3PU, UnitedKingdom )u, -'t-- i.
(R ecei ved
l E A pri l l 99l ) " ,

Practicalappticationof the generalized


Bell'stheoremin the so-callcdkeydistribu,ionpro."rs in cryp-
tographyis reported.The proposedschemeis bascdon thc Bohm'sversionof thc Einstein-Podolsky-
Rosengedankenexperimentand Bell'stheoremis usedto tcst for eavesdropping.

PACSnumbers: 42.80.Sa.
03.65.82. 89.70.*c

Cryptography, despitea colorfulhistorythat goesback sct up to test Bcll's theorem. Beforc I prmeed any fur-
to 400 8.c.,only bccamepart of mathematics and infor- ther let mc first introducesome'baricnotiongof cryptog.
mationtheorythis century,in the late 1940s,mainlydue raphy. ' \'
to the seminalpapersof Shannonttl. Today,one can Originallythe securityof a cryptdaxt depcndedon the
brieflydefinecryptography as a mathematical systemof scerccyof the entire encryptint'and.'decrypting pro-
transforminginformationso that it is unint€lligibleand cedures;however,today we use cipherr for which the al-
thereforeuseless to thosewho are not meantto haveac- gorithm for cncryptingand decrypting. csuld be revealed
cessto it. However,as the computationalpro€ess associ- to anybodywithout compromisingthe sccurityof a par-
ated with transformingthe informationis alwaysper- ticular cryptogram. In suchciphcn a c€t of spccificpa-
formed by physical means,onc cannot separatethe ramctcrs,callcda.key,is suppliedtogctherwith the plain-
mathematical structurefrom the undcrlyinglawsof phys- tcxt asan input to the encryptingalgorithm,and together
ics that governthe process of computation[2]. Deutseh with the cryptogramas an input to thc decryptingalgo-
has shownthat quantumphysicsenrichesour computa- rithm. Thc cncrypting and decryptingalgorithmsare
tional possibilitiesfar beyondclassicalTuring machines publicly announcrd;the security.of,the eryptogramde-
[2J,andcurrentwork in quantumcryptography originat- pendscntirely on the sccrecyof the.kcyr and this key,
ed by Bennettand Brassardprovidesa goodexampleof which is very important,may consistof any randomly
thisfact [31. ehosen, sufficientlylongstringof bits. Oncethe key is es-
In this paperI will presenta methodin whichthe sccu- tablished,subsequentcommunicationdinvolves sending
rity of the so-calledkey distributionprocessin cryptogra. cryptogramsoyer a publicchannelwhich is rulnerableto
phy dependson the completeness of quantummechanics. total passiveinterception(e.g.,public announcement in
Herecompleteness mcansthat quantumdescription pro- massmedia). However,in orderto establishthe key,two
videsmaximumpossibleinformationabout any system users,who shareno secretinformationinitially, mustat a
underconsideraticn.The proposedschemeis basedon certainstageof communicationusc a rcliablcand a very
the Bohm'swell-knownversionof the Einstein-Podolsky- securechannel. Since the interceptionis a set of mea-
Rosengedankenexperiment[4]; the generalizedBell's surementspcrformedby the eavesdroppcr on this chan-
theorem(Clauser-Horne-Shimony-Holt inequalities)t5l nel, howeverdifficult this might be from a technological
is used to test for eavesdropping.From a theoretical point of view, in principle any classicalchannelcan al-
pointof viewthe schemeprovidesan interesting and new waysbe passivelymonitored,without thc legitimateusers
extensionof Bennettand Brassard'soriginal idea, and being aware that any eavesdropping ,hastaken place.
from an experimental perspectiveoffersa practicalreali- This is not so for quantumchanncls[31. In the follo.ing
zationby a small modificationof experiments that were I describea quantumchannelwhich distributesthe key

@ t ggI The AmericanPhysicalSociety 661

You might also like