quant-ph0008097/t.tex
1: %000712 - bellexp - Why do Bell experimemts? - Ian C. Percival
2: %Sent as .ps file to Nature date
3: %after edition 1.35, alterations to account for 
4: %the new detection loophole results.
5: %Check refs: Bohm's book,
6: 
7: \documentstyle[12pt]{article}
8: \begin{document}
9: 
10: 
11: \parskip 2mm plus 1mm \parindent=0pt
12: \def\cl{\centerline}\def\lel{\leftline}\def\rl{\rightline}
13:  \def\hs1{\hskip1mm} \def\h10{\hskip10mm} \def\hx{\h10\hbox}
14: \def\vs{\vskip3mm} \def\vup{\vskip-2mm}
15: \def\page{\vfill\eject}
16: \def\<{\langle} \def\>{\rangle} \def\br{\bf\rm}  \def\it{\tenit}
17:  \def\de{\partial} \def\Tr{{\rm Tr}} \def\dag{^\dagger} 
18: \def\half{{\scriptstyle{1\over 2}}}
19: \def\ne{=\hskip-3.3mm /\hskip3.3mm}\def\tr{{\rm tr}}
20: \def\Pr{{\br Pr}}\def\to{\rightarrow}
21: \def\vb{\vskip20mm}\def\vm{\vskip10mm}
22: \def\cite{}
23: 
24: \def\be{\begin{equation}}\def\ee{\end{equation}}
25: 
26: \vs\bf\cl{Why do Bell experiments?} 
27: 
28: 
29: 
30: \vs
31: 
32: \centerline {by}
33: 
34: \vs
35: 
36: \centerline {Ian C. Percival}\vs 
37: 
38: \centerline {Department of Physics} 
39: \centerline {Queen Mary and Westfield College, University of London} 
40: \centerline {Mile End Road, London E1 4NS, England} 
41: 
42: \vs\vs\vs
43: 
44: \rm
45: 
46: Experiments over three decades have been unable to demonstrate weak
47: nonlocality in the sense of Bell unambiguously, without loopholes.
48: The last important loophole remaining is the detection
49: loophole$^{1,2,3}$, which is being tackled by at least three
50: experimental groups$^{4,5,6}$.  This letter counters five
51: common beliefs about Bell experiments, shows the importance of
52: these experiments, and presents alternative scenarios for future
53: developments.
54: % Simple figure of black box
55: Figure here.
56: 
57: The figure shows the bare bones of a system designed to test Bell
58: inequalities$^7$.  It may be considered as a black box at rest in a
59: laboratory frame, with two input ports $\alpha$, $\beta$ and two
60: output ports $a$, $b$.  The input port $\alpha$ and the output port
61: $a$ are parts of a subsystem $A$ on the left of the box, and $\beta$,
62: $b$ are parts of a subsystem $B$ on the right. The minimum distance
63: between $A$ and $B$ is $L$.
64: 
65: One run of an experiment on the black box starts with inputs labelled
66: $\alpha$, $\beta$ which end before time $t=0$ in the laboratory frame.
67: The outputs are labelled $a$, $b$ start after time $t=T$.  The inputs
68: and outputs are entirely classical, but the system has entangled
69: quantum components.
70: 
71: A sufficient number of runs provides transition or conditional
72: probabilities denoted
73: 
74: \be\label{} \Pr(\alpha,\beta\to a,b)  \ee 
75: 
76: for the outputs $a,b$ of the whole system, given the inputs
77: $\alpha,\beta$.  Following Wigner's formulation$^{8}$, locality
78: implies that they satisfy the Bell inequality
79: 
80: \be\label{} \Pr(1,2\to +,-) +\Pr(2,0\to +,-) -\Pr(0,1\to
81: +,-) \ge 0, \ee 
82: 
83: together with two more inequalities given by cyclic
84: permutations of $0,1,2$.  
85: 
86: The subsystems $A$ and $B$ are so far apart that 
87: 
88: \be\label{} L > Tc, \ee 
89: 
90: where $c$ is the velocity of light, so it is not possible to send
91: signals from $\alpha$ to $b$ or from $\beta$ to $a$.  Without this
92: condition, the experiment has a loophole, sometimes called the
93: locality loophole$^9$.  All early experiments had this loophole, but
94: Aspect's group$^{10}$ and more recently Zeilinger's and
95: Gisin's groups$^{11,12}$ have removed it.
96: 
97: In Bell's original thought experiment, there are three possible values
98: of $\alpha$ and corresponding values of $\beta$, denoted $0,1,2$.
99: They might be three settings of the angle of a polarizer, or of the
100: angle of measurement of a particle of spin one-half.  There are two
101: possible values of each separate output, denoted $+$ and $-$.  They are
102: typically two orthogonal directions of polarization of a photon, or
103: two opposite spins of an atom.  In an ideal Bell experiment, the
104: inputs $\alpha$ and $\beta$ are determined by completely independent
105: random variables.  If they are not random, as for the Aspect and Gisin
106: experiments, there is a further loophole, which was removed by
107: Zeilinger's group in an experiment in which the locality loophole was
108: also overcome$^{11}$.
109: 
110: The detection loophole$^{1,2,3}$ described below has proved to be
111: the most difficult of all, and has never been overcome in any
112: published experiment, though the groups of Fry and Walther$^4$, of
113: Wineland$^{5}$ and of Blatt$^{6}$ are trying.  Fry and Walther aim
114: to close the locality loophole in the same experiment.
115: 
116: Bell experiments, or experiments of the Bell type, are experiments to
117: test the original Bell inequalities, or one of the many other
118: inequalities that follow from locality.  Here are five common beliefs
119: about them.
120: 
121: 1 Their only purpose is to exclude local hidden variable theories,
122: which are of little interest anyway.
123: 
124: \vup 2 Violation of the inequalities follows inevitably from the laws of
125: quantum mechanics and their interpretation.
126: 
127: \vup 3  Experiments have already shown that the Bell inequalities are
128: violated, apart from the detection loophole, which is so unbelievable
129: that it is not worth considering seriously.
130: 
131: \vup 4 Einstein's view that all physical laws are local was his one
132: definite major mistake.  
133: 
134: \vup 5 Bell experiments are therefore no longer worth doing.
135: 
136: 
137: Reasons are presented here for rejecting all of these beliefs, and
138: replacing them by the alternatives:
139: 
140: 1 There are more important reasons for doing Bell experiments, including
141: Bell's weak nonlocality. 
142: 
143: \vup 2 Neither the violation nor the nonlocality follow inevitably from
144: quantum mechanics.
145: 
146: \vup 3 There are at least two good reasons why the detection loophole
147: should be taken seriously.
148: 
149: \vup 4 Einstein has been right before, when many in the physics community
150: were wrong, and we need conclusive experimental evidence of
151: nonlocality before judging him on {\it this} issue.
152: 
153: \vup 5 Bell experiments are among the most important in physics.
154: 
155: 
156: \vs{\bf Why do Bell experiments?} 
157: 
158: Guinness, Bass and Worthington are brands of beer.  It is questionable
159: whether, if Guinness is good for you, Bass is bad and Worthington is
160: worse.  These are matters of taste and prejudice.  Forward, nonlocal
161: and backward causality are brands of causality.  If forward causality
162: is good for you, nonlocal causality is bad and backward causality is
163: worse.  These are matters of experience.  We are particularly
164: concerned with the question of nonlocal causality, in which cause and
165: effect are spatially separated in spacetime, so that a signal from
166: cause to effect would have to go faster than the velocity of light.
167: According to classical special relativity, an event can affect a
168: future event, in or on its forward light cone, but not a spatially
169: separated event, and certainly not a past event.
170:   
171: But apparently, according to quantum theory, classical events that are
172: linked by quantum systems are different.  For them, there is a sense
173: in which causality might act nonlocally, but without any signalling
174: faster than light.  This is Bell's weak nonlocality, which can be
175: formulated in terms of the classical inputs and outputs of a black
176: box $^{13}$.
177: 
178: An electrical engineer's black box consists of a circuit with input
179: and output terminals.  He may not know what circuit is inside, but it
180: is assumed here to be classical.  If there is no noise in the circuit,
181: then the black box is deterministic.  The outputs then depend on the
182: inputs through a unique transfer function $F$, and by experimenting
183: with different inputs and looking at the outputs, engineers can find
184: $F$.  In practice the resistors in the circuit produce noise, which we
185: assume to be classical noise.  The system is then stochastic.  The
186: noisy circuit can be represented by a probability distribution
187: $\Pr(F)$ over the transfer functions $F$, in which the unknown values
188: of supposedly classical background variables, like the coordinates of
189: thermal electrons, determine the particular $F$ that operates.
190: 
191: A physicist's black box contains an evolving physical system, such
192: as a classical electrical circuit, or an entangled quantum state with
193: classical inputs and outputs.  She may not know what physical system
194: is inside, but by experimenting with different inputs and looking at
195: the outputs, she can find out something about it.  
196: 
197: For deterministic systems, special relativity distinguishes between
198: local transfer functions $F$ in which the influence of an input on an
199: output goes at no more than the velocity of light, and nonlocal
200: transfer functions $F$, for which the influences can act over
201: spacelike intervals.  It is possible to determine whether the transfer
202: function of a system is local or not by experimenting with different
203: values of the inputs, and observing the outputs.  There is no need to
204: look inside the black box.  All real classical systems have local
205: transfer functions, as required by special relativity.
206: 
207: When classical or quantum systems are stochastic, special relativity
208: distinguishes between three types of black box system, defined in
209: terms of probabilities $\Pr(F)$ of transfer functions.  The first are
210: local systems, for which the transition probabilities of the outputs
211: given the inputs can be obtained from a probability distribution
212: $\Pr(F)$ in which only local $F$ contribute.  It is therefore not
213: possible to send signals faster than the velocity of light.  For the
214: second type, the transition probabilities can only be obtained from
215: $\Pr(F)$ in which at least one nonlocal transfer function has nonzero
216: probability, so there is an element of nonlocality.  But nevertheless
217: it is not possible to send signals faster than the velocity of light.
218: The system is then said to be weakly nonlocal, or nonlocal in the
219: sense of Bell.  For the third type, which has never been seen, it is
220: possible to send signals faster than the velocity of light.
221: 
222: The stochasticity of classical systems comes from background variables
223: that are not included in the system, but for quantum systems it does
224: not come from any background variables that we can see, so either they
225: are assumed not to exist, as in the Copenhagen interpretation, or they
226: are called hidden variables.
227: 
228: A Bell experiment is a black box with classical terminals and an
229: entangled quantum system inside, where the source of entanglement is
230: inside the box.  For photon polarization the setting of the
231: orientations of the polarizers is an input, and the detection of the
232: directions of polarization is an output.  All the inputs and outputs
233: are classical events.
234: 
235: Real laboratory Bell experiments are treated in the next section.  In
236: this one we treat only ideal Bell thought experiments in which the
237: entangled quantum system is sufficiently close to a pure state, and
238: the measurements sufficiently good, that the black box is weakly
239: nonlocal.  
240: 
241: The classic Bell was proposed to test whether local hidden variable
242: theories are possible$^7$.  But quantum black boxes also tell us
243: something about the world: there are correlations between classical
244: events that can only be produced by quantum links.  These correlations
245: are important in their own right.  They demonstrate weak nonlocality.
246: They also show that the properties of our world cannot be explained
247: using local hidden variables, but that is not their main significance,
248: which persists independently of any theory, local or nonlocal.
249: An experimenter who has never seen the apparatus before can
250: tell by experimenting with the inputs and outputs, and without
251: looking inside, that the black box contains a quantum system.  This
252: property of quantum black boxes comes from weak nonlocality.
253: 
254: So Bell experiments and weak nonlocality are important for all quantum
255: physicists, whether they support hidden variable theories or not.  The
256: weak nonlocality of quantum measurement is unique in modern physics:
257: classical dynamics, quantum dynamics and general relativity are all
258: local.  Today only ideal experiments are weakly nonlocal, though
259: tomorrow they could be real.
260: 
261: In modern quantum computation it is proposed to put quantum correlations
262: to good use.  Violation of Bell inequalities is a benchmark experiment
263: for quantum computers, and for this reason alone would be worth doing
264: even if there were no interest in weak nonlocality.
265: 
266: 
267: \vs{\bf Nonlocality and quantum mechanics}
268: 
269: Real experiments, with or without loopholes, are approximations to
270: ideal experiments without them.  There are possible limits on
271: the approach to the ideal that are explored by attempts to carry out
272: an experiment without loopholes.  Theoretically it appears that it is
273: possible to approach arbitrarily close to the ideal, by improving the
274: efficiency of detectors, collimation, etc., but there appears to have
275: been a `conspiracy' of nature that prevented this.  Such conspiracies
276: in physics have a long history.
277: 
278: For example, the first law of thermodynamics says that heat is a form
279: of energy.  In the early 19th century there appeared to be a
280: conspiracy that prevented anyone from extracting all this energy from
281: a system and using it.  We now call this conspiracy the second law of
282: thermodynamics.  Bell's opinion, `It is hard for me to believe that
283: quantum mechanics works so nicely for inefficient practical setups and
284: is yet going to fail badly when sufficient refinements have made.'
285: may be right, but irrelevant.  Quantum theory does not have to fail.
286: The necessary refinements may not be possible.  Current quantum theory
287: would then be incomplete, just as the first law of thermodynamics is
288: incomplete.  Einstein thought that quantum mechanics is
289: incomplete$^{14}$, but this was not the kind of incompleteness he
290: described.
291: 
292: Santos has suggested earlier that the laws of quantum measurement
293: might be compatible with locality $^{15}$.  This idea can be
294: illuminated by an analogy, comparing the second law of thermodynamics
295: and the possible breakdown of the nonlocality argument.  Classical
296: systems obey the laws of Hamiltonian dynamics, despite the second law,
297: which limits energy transfer from real systems with many
298: particles.  Similarly quantum systems might obey all the laws of
299: quantum dynamics and quantum measurement, despite a supplementary law
300: which excludes weakly nonlocal systems, thus ensuring that physics
301: remains local.  No one yet knows any such supplementary law.
302: 
303: In thermodynamics, the second law was discovered as a result of many
304: trials showing practical limitations on getting useful energy from
305: heat.  Locality holds for all Bell experiments to date, but it is on
306: much weaker ground, as Bell experiments are relatively few, so
307: experiments designed to test it are among the most important in
308: physics.
309:   
310: 
311: \vs{\bf Experiments and the detection loophole}
312: 
313: A Bell experiment without loopholes would be an experiment from
314: which we could deduce weak nonlocality without further assumptions.
315: It is nearly four decades since the inequalities were obtained and
316: experiments tentatively suggested, and three decades since the first
317: experiments.  There is still no published clear experimental
318: demonstration of weak nonlocality, because of the detection loophole,
319: which follows.
320: 
321: Real experiments have outcomes that are excluded in ideal experiments.
322: For example, a photon or an atom may be detected at $A$, but not at
323: $B$, as a result of imperfect detectors, or losses due to absorption
324: or bad collimation.  These outcomes affect the inequalities, and the
325: tests of nonlocality.  There are further assumptions that have to be
326: made in order to obtain the probabilities that appear in the
327: inequalities.  One such assumption is that the detector efficiency is
328: independent of the local `hidden' variables, as discussed by Clauser
329: and Horne$^{16}$ and by Gisin and Gisin$^{17}$.  If the possibility
330: of such a dependence is accepted, nonlocality cannot be demonstrated
331: until the detection efficiency reaches a threshold.  This is the
332: loophole.
333: 
334: At first sight such a dependence seems unlikely, but there are
335: two good reasons why the detection loophole should not be ignored.
336:  
337: The first reason lies in the definition of efficiency, for which the
338: analogy with thermodynamics is useful.  If only the first law of
339: thermodynamics applied, then it would be reasonable to measure the
340: efficiency of a heat engine as the proportion of the total energy that
341: is extracted.  But once the second law is recognized, this definition
342: is inappropriate, and we revise the definition to take temperature
343: differences into account.  Similarly, if there are values of local hidden
344: variables that play a role in determining whether or not there is a
345: response from a particle detector, it is no longer appropriate to
346: measure the efficiency of the detector in the conventional way.  The
347: measure of efficiency should take into account the values of the hidden
348: variables.  With such a new definition, the dependence seems
349: natural$^{17}$.
350: 
351: The second reason lies in the alternative.  The dependence appears
352: unlikely to some people, but the alternative is nonlocality, which is
353: a break with the whole of the rest of physics.  We must not base
354: conclusions about such an overwhelmingly important universal issue on
355: some prejudice about the properties of mere detectors.
356: 
357: We cannot dismiss the detection loophole.  We have to try to close it
358: by improving the experiments.
359: 
360: 
361: \vs{\bf Einstein's mistake?}
362: 
363: Einstein introduced light quanta in 1905, but leading physicists still
364: did not accept them as late as 1913, so we should be careful before
365: rejecting his other ideas, whatever the majority thinks$^{18}$.
366: He believed that nature obeys local laws, and Bell showed that this
367: assumption might be tested experimentally.  It appeared that weak
368: nonlocality followed from the laws of quantum dynamics and quantum
369: measurement, but this is not so.
370: 
371: 
372: 
373: \vs{\bf What now?}
374: 
375: Quantum technology has advanced so much during the last decade that
376: the detection loophole might soon be closed, using spin states of
377: atoms or otherwise.  For the future, there are several possible
378: scenarios, of which two are the most likely:
379: 
380: {\it EITHER}
381: 
382: 1. The inequalities cannot be violated.  The apparent conspiracy is
383: due to a new law of nature, consistent with current quantum theory,
384: but limiting the accessible states of matter to those for which
385: locality reigns.  The common view is wrong and Einstein was right.
386: 
387: {\it OR} 
388: 
389: 2. The inequalities can be violated.  The apparent conspiracy that has
390: prevented the unambiguous experimental confirmation of Bell
391: nonlocality is due to practical difficulties that can be overcome.
392: Experiments will close the detection loophole.  This would be a
393: significant advance, a benchmark on the way to quantum computation.
394: Simultaneous closure of both the detection and locality loopholes
395: would confirm the common view that the laws of nature are weakly
396: nonlocal and that Einstein was wrong.
397: 
398: This issue can only be resolved by experiment.  That is why Bell
399: experiments are so important.
400: 
401:  
402: 
403: \vs{\bf References}
404: 
405: 
406: 1. Pearle, P., Hidden-variable example based on data rejection {\it Phys. Rev. D}, {\bf 2} 1418-1425 (1970).
407: 
408: 2. Clauser,J.F.,Horne, M.A., Shimony,A. \& Holt, R.A., Proposed experiment to test local hidden variable theories, {\it
409:  Phys. Rev.  Lett.}, {\bf 23}, 880-884, (1969).
410: 
411: 3. Garg, A. \& Mermin, N.D., Detector inefficiencies in the EPR
412: experiment, {\it Phys. Rev D}, {\bf 35}, 3831-3835 (1987).
413: 
414: 4. Fry, E.S. \& Walther,T., Fundamental tests of quantum mechanics,
415:   {\it Adv. Atom. Molec. and Opt. Phys.}, {\bf 42} 1-27 (2000).
416: 
417: 5. Private communication, D. Wineland.
418: 
419: 6. Private communication, R. Blatt. 
420: 
421: 7. Bell, J.S., On the Einstein-Podolsky-Rosen paradox, {\it Physics}, 
422: {\bf 1} 195-200 (1964)
423: 
424: 8. Wigner,E.P. , On hidden variables and quantum mechanical
425:  probabilities {\it Am. J. Phys.}, {\bf 38}, 1005-1009 (1970).
426: 
427: 9. Bohm, D., {\it Quantum Theory} (Prentice-Hall, Englewood Cliffs, N.J.,
428: 1951).
429: 
430: 10. Aspect,A. Dalibard,J. \& Roger,G., Experimental test of Bell
431: inequalities using time-varying analyzers, {\it Phys. Rev. Lett.},
432: {\bf 49}, 1804-1807, (1982).
433: 
434: 11. Weihs,G. et al, Violation of Bell's inequality under strict locality
435: conditions, {\it Phys. Rev. Lett.}, {\bf 81}, 5039-5043 (1998).
436:  
437: 12. Tittel,W. et al, Long-distance Bell-type tests using energy-time
438: entangled photons, {\it Phys. Rev. A} {\bf 59},4150-4163 (1999).
439: 
440: 13. Percival,I.C., Quantum transfer functions, weak nonlocality and
441:  relativity , {\it Phys. Lett. A}, {\bf 244}, 495-501, (1998).
442: 
443: 14. Einstein,A. Podolsky,B. \&Rosen,N., Can quantum-mechanical
444: description of reality be considered complete?, {\it Phys. Rev.} {\bf
445: 47}, 777-780 (1935).
446: 
447: 15. Santos,E., Critical analysis of the empirical tests of local
448: hidden-variable theories, {\it Phys. Rev. A} {\bf 46}, 3646-3656,
449: (1992).
450: 
451: 16. Clauser, J.F. and Horne,M.A. {\it Phys. Rev. D} {\bf 10},
452: 526-535, (1974).
453: 
454: 17. Gisin,N. \& Gisin,B. Local hidden variable model of quantum
455: correlation exploiting the detection loophole, {\it Phys. Lett. A}
456: {\bf 260}, 323-327, (1999).
457: 
458: 18. Pais,A. {\it Subtle is the Lord \dots}, (Oxford University Press, 1982) p382.
459: 
460: 
461: \vs{\bf Acknowledgements}
462: 
463: I thank Roberto Basoalto, Ed Fry, Nicolas Gisin, Serge Haroche, Thomas
464: Walther and Ting Yu helpful and stimulating communications and
465: Leverhulme, the ESF and PPARC for support.
466: 
467:  \end{document}
468: 
469: \end
470: 
471: 
472: 
473: 
474: 
475: 
476: 
477: 
478: 
479: 
480: 
481: 
482: 
483: 
484: 
485: 
486: 
487: 
488: 
489: