Beruflich Dokumente
Kultur Dokumente
1
2 Bell’s Inequalities 2
2 Bell’s Inequalities
Now we derive Bell’s theorem, following his original for-
mulation based on Bohm’s thought experiment [2]. We
consider two spin one-half particles in the singlet state,
moving freely in opposite directions. Measurements can
be made on any spin components σ1 and σ2 . If a mea-
surement of σ1 ·~a, where ~a is an arbitrary unit vector,
yields A = +1 (in units of ~/2), then σ2 ·~a yields B = −1.
According to the EPR argument, any measurements per-
Fig. 1: A violation of the CHSH inequality is found when
formed on particle 1 cannot influence those made on 2 to
φ = π/4.
comply with locality. Since we can deduce any spin com-
ponent of particle 2 by measuring the same component
of 1, the results must be predetermined, as implied by
realism. Quantum mechanics, which describes the sys- ˆ
tem with a wavefunction ψ, cannot predict these results. P (~a, ~b) − P (~a, ~c) ≤ dλ ρ(λ) 1 − A(~b, λ)A(~c, λ)
Therefore, we assume that there is a more complete spec-
(5)
ification of the state, given by a set of parameters λ.
The second term on the right-hand side is P (~b, ~c).
According to this hidden-value theory, the measure-
Therefore, we obtain
ment of particle 1 is determined by the direction of de-
tector 1, ~a, and the set of parameters λ, and analogously
1 + P (~b, ~c) ≥ P (~a, ~b) − P (~a, ~c) (6)
for particle 2, i.e. A(~a, λ), B(~b, λ). The vital assumption
here is that the result A for particle 1 does not depend Inequality (6) is the first of the so-called “Bell inequal-
on the setting ~b, the direction of detector 2. Any theory ities”. One case in which the quantum-mechanical pre-
that denies the existence of action-at-a-distance must be diction (2) contradicts (6) is when (a, cc) = 2π/3 and
local in this sense. The expectation value of measuring (a,
cb) = (b, cc) = π/3, for which the quantum-mechanical
σ1 ·~a on particle 1 and σ2 ·~b on particle 2 is
expectation value is P (~a, ~b)ψ = P (~b, ~c)ψ = −1/2, and
ˆ
P (~a, ~c)ψ = 1/2. It follows that 1 + P (~b, ~c)ψ = 1/2 and
P (~a, ~b) = dλ ρ(λ)A(~a, λ)B(~b, λ) (1)
|P (~a, ~b)ψ − P (~a, ~c)ψ | = 1, in conflict with the inequality
hold by hidden-value theories.
where ρ(λ) is the probability density function of λ. This
The formulation of Bell’s theorem is unrealizable in
result should equal the quantum mechanical expectation
real experiments. One of the key assumptions to derive
value,
(6) is that there is a particular configuration of the polar-
izers with perfect correlation, A(~a, λ) = −B(~a, λ). How-
P (~a, ~b)ψ = hψS | σ1 ·~a σ2 ·~b |ψS i = −~a·~b (2) ever, real polarizers attenuate and leak some of the sig-
However, it will be shown that this is not the case. nal on the orthogonal channel. An alternative proof that
When ~a = ~b, the quantum mechanical predictions imply does not require perfect correlation, A(~a, λ) = −B(~a, λ),
A(~a, λ) = −B(~a, λ). Presuming this, (1) can be written was made by Clauser, Horne, Shimony and Holt (CHSH)
as [7], who obtained the following inequality,
ˆ
P (~a, ~b) = − dλ ρ(λ)A(~a, λ)A(~b, λ)
(3) S = P (~a, ~b) − P (a~0 , ~b) + P (~a, b~0 ) + P (a~0 , b~0 ) ≤ 2
(7)
Now, we subtract P (~a, ~c), where ~c is another unit vec- Parameters a, a0 represent different settings on detec-
tor, tor 1, and b, b0 for detector 2. A violation is found for
the configuration shown on figure 1, with φ = π/4. How-
ever, it is not obvious how to apply this expression in a
P (~a, ~b) − P (~a, ~c) real experiment. To obtain the expected values we need
ˆ the total number of events, but detectors have efficien-
= dλ ρ(λ)A(~a, λ)A(~b, λ) A(~b, λ)A(~c, λ) − 1 (4) cies less than 100 %. Furthermore, analyzers are usually
single-channeled, i.e. they block the photon if it is in the
We take the absolute value, recalling that A(~a, λ) = orthogonal state. CHSH propose to assign a value of -1
±1, to those cases in which the photon is blocked. However,
4 Bell Inequality Tests 3
√
state, |ψi = (1/ 1 + r2 )(|HV i + r |V Hi). For example,
an efficiency of the 70% is enough is the background noise
is 0.02%. In that case, r = 0.136.
Fig. 5: Experimental setup of Freedman (top) and As- 4.1.3 Second Generation
pect (bottom). After [10, 1].
The freedom-of-choice loophole was closed in 1998 by
Weihs et al. [22]. This experiment used a high-intensity
source of polarization-entangled photon pairs, produced
discussed the experimental requirements to perform such
by degenerate type-II parametric down-conversion [15].
an experiment [7], and analyzed previous experiments
Although it has to make the “fair-sampling” assumption,
using entangled photon pairs. One of them, using en-
it goes one step further from the Aspect experiment by
tangled gamma rays from positronium annihilation, was
making a truly random choice of the measurement set-
not considered valid due to the difficulty of performing
tings. Weihs emphasizes that a pseudo-random-number
polarization measurements on high energy photons. The
generator cannot be used, since its state is predetermined
experiment of Kocher and Commins [14], which used vis-
by an algorithm. Instead, a physical random number
ible photons from a calcium source, was more promising,
generator was used, based on a light-emitting-diode mon-
but did not measure at enough polarization angles. They
itored by photomultipliers. Another difference from pre-
also pointed out that the efficiency of the polarizers was
vious experimental setups is that both measuring sta-
too low for a conclusive result.
tions are completely independent, to avoid any common
The experiment used the atomic cascade conceived by context that could influence results. Instead of using a
Kocher, and performed polarization measurements us- coincidence circuit, events are recorded independently on
ing a high quality single-channel analyzer, with efficiency each side, and compared only once the experiment has
= 0.97 on the parallel axis, ⊥ = 0.037 on the orthog- finished. An atomic clock is used at each measurement
onal. The experimental setup is illustrated in figure 5 station to provide an accurate time bases, and are syn-
(top). The photomultiplier detectors were cooled to re- chronized only before the experiment.
duce dark rates, but the efficiency of the detectors was To ensure the locality condition, observers are sepa-
pretty low, η = 1.5·10−3 . Coincidence circuits were used rated 400 m, which requires that the measurement pro-
to obtain the coincidence rates. Due to the low intensity cess takes less than 1.3 µs. The whole measurement
of the entangled photon source, the coincidence rate was process involves the selection of a random analyzer di-
very low, on the order of 0.1 counts per second. It was rection, the setting of the analyzer, the detection of a
necessary to make very long experimental runs to obtain photon, and the recording of results. In this experiment,
good statistics (they report 200 hours). Results show a the measurement process take much less than 1.3 µs,
violation of Bell’s inequality, and good agreement with thanks to the use of high-speed random number genera-
the predictions of quantum mechanics. tors and electro-optical modulators to change the polar-
The experiment of Freedman, however, did not make ization accordingly. Two different computers are used at
any attempt to close any loopholes. The locality loop- each measurement station, and a third one is used after
hole was evidently open due to the table-top setup, as the experiment to analyze data. The experiment shows
well as the efficiency loophole. The first solid attempt to a violation of Bell’s inequality and good agreement with
4 Bell Inequality Tests 6
quantum-mechanical predictions. However, the authors Fig. 7: Space-time diagram of the experiment performed
recognize that it is still necessary to close the efficiency by the Zeilinger group in 2015. After [12].
loophole to have a conclusive result. In this experiment,
the efficiency of the detectors is 5%, far below the re-
quirements to close the loophole. NIST [21] and the other from Vienna [12]. The exper-
iments are essentially an extension of those of 2013 to
close the locality loophole. In the experiment of Vienna,
4.1.4 Third Generation the observation stations were separated about 60 me-
The detection loophole was closed for photons in 2013, ters, while those of NIST by 200 m. Fig. 7 shows the
in two independent experiments that still left the local- space-time diagram of the Vienna experiment. The blue
ity loophole open [5, 11]. This achievement was made cone illustrates the time at which a luminal signal could
possible by the use of superconducting transition-edge arrive at the measurement stations. The choice of mea-
sensors (TES), which offer very high efficiency (in the surement setting (green line), performed by a physical
order of 90%) and extremely low levels of dark counts. random number generator, must be completed before the
These experiments used CH inequality (eq. (8)), which blue line to close the freedom-of-choice loophole. The en-
does not need the fair-sampling assumption. Instead of tangled photons arrive somewhat later, since they propa-
maximally entangled states, they used Eberhard states to gate through optical fibers at v < c. To close the locality
reduce the efficiency required, as explained in section 3.1. loophole, the measurement of A must be completed be-
The choice of the parameter r is a compromise between fore a luminal signal from B is sent at the setting choice
the level of dark counts and the efficiency threshold. In (red line). This ensures, by Einstein’s causality, that the
both experiments, the efficiency is on the order of 75%, measurement process at A is independent of B.
once all losses are taken into account.
The experimental setup is almost identical to the ex- 4.2 Experiments with Solid-State Qubits
periment discussed in section 4.1.3, the most significant
difference being the detection mechanism. Christensen A research group from Delft closed the three major loop-
[5] criticizes that the experiment of Giustina suffers from holes a few months before the experiments with pho-
the coincidence-time loophole. This loophole refers to tons reported before, by using distant electron spin en-
the definition of coincident events between A and B. tanglement of solid-state qubits [13]. The experimental
These have usually been defined with respect to detection setup is quite different from the experiments reported be-
events, using a small coincidence window. To close this fore. The detection loophole is closed by using an event-
loophole, it is necessary to define the coincidence events ready experiment. Entanglement of electrons separated
with respect to a system clock, instead of a detection by more than one kilometer is achieved by entanglement
event. Since the jitter of TES detectors (500 ns) is much swapping via an intermediate location.
longer than the time between single photon emissions (8 Entanglement between distant solid-state qubits was
ns), Christensen pulses the laser, creating event intervals demonstrated by this group in 2013 [3]. The qubit con-
much longer than the detector jitter. The experiment of sists of a nitrogen-vacancy (NV) defect centre in dia-
Giustina solved this problem in 2014 [16]. mond, which combines a long-lived electronic spin (S =
The detection, freedom-of-choice, and locality loop- 1) with a robust optical interface. Measurements report
holes were closed with photons for the first time in 2015. a spin coherence time exceeding 10 ms, which is one of
Two groups reported simultaneously this result, one from the longest spin coherence times achieved in solids.
4 Bell Inequality Tests 7
[13] B. Hensen, H. Bernien, A. E. Dréau, A. Reiserer, [22] Gregor Weihs, Thomas Jennewein, Christoph Si-
N. Kalb, M. S. Blok, J. Ruitenberg, R. F. L. Ver- mon, Harald Weinfurter, and Anton Zeilinger.
meulen, R. N. Schouten, C. Abellán, W. Amaya, Violation of Bell’s Inequality under Strict Ein-
V. Pruneri, M. W. Mitchell, M. Markham, D. J. stein Locality Conditions. Physical Review Letters,
Twitchen, D. Elkouss, S. Wehner, T. H. Taminiau, 81(23):5039–5043, 1998.
and R. Hanson. Loophole-free Bell inequality viola-
tion using electron spins separated by 1.3 kilometres.
Nature, 526:682–686, 2015.