Professional Documents
Culture Documents
This article has been downloaded from IOPscience. Please scroll down to see the full text article.
(http://iopscience.iop.org/1464-4266/4/4/323)
View the table of contents for this issue, or go to the journal homepage for more
Download details:
IP Address: 217.203.136.188
The article was downloaded on 19/06/2010 at 01:43
E-mail: thw@tamu.edu
For a given two-particle system, if particle 1 is found to In an actual experiment, the goal will be to choose
have spin-up (↑) in some spatial direction, particle 2 must the experimental conditions such that the resulting quantum
necessarily have spin-down (↓) in that direction and vice mechanical prediction, SQM , provides a maximum violation of
versa. Specifically, quantum mechanics predicts that the the BCH inequality. For θ1 , θ2 , θ1 , θ2 equal to 135◦ , 0◦ , 225◦ ,
total spin of this two-particle system is zero, and that for and 90◦ , respectively, we find
measurement of both components in one specific direction 2
the results will be opposite with unit probability. However, ◦ ◦ ◦ ◦1 √ ε−
SQM (135 , 0 , 225 , 90 ) = ηgε+ 1 + 2F (3)
quantum mechanically it is not possible to predict the absolute 2 ε+
orientation of either spin, or to even predict absolute values for
two of the components of the spin of a single particle. where η is the detection efficiency for both detectors, g is
John Bell considered Bohm’s version of the EPR the conditional probability of particle 2 entering detector 2,
experiment. Then, by assuming positive probabilities, he if particle 1 has entered detector 1, ε± are measures of the
showed that any HV theory satisfying a physically reasonable quality of the spin analysis, and F is a measure of the purity
condition of locality (LHV theory, e.g. a classical theory) will of the entangled state. In order to obtain the largest possible
yield statistical predictions which must satisfy restrictions for violation of the BCH inequality, equation (2), by the quantum
certain correlated phenomena involving entangled states (see mechanical prediction, equation (3), the parameters η, g, ε± ,
equation (1)). In other words, classical theories put upper F should be as close to unity as possible.
bounds on measurements of the statistical correlations between Finally, any test of the Bell inequalities is statistical in
the spin components of the two particles. These restrictions nature; specifically, count rates have statistical errors and the
can be cast into the form of inequalities, which are now measured quantities are ratios of coincidence count rates to
generally referred to as Bell inequalities. Thus, at least in singles count rates. These statistical errors must be sufficiently
principle, Bell’s work made it possible for the first time to small so that the resulting error in the measured value of S is
experimentally distinguish between the LHV and quantum much less than the magnitude of the violation (the amount
mechanical pictures. It should be emphasized that no specifics by which SQM exceeds one). Specifically, it is not the
of an LHV theory are involved in Bell’s argument. Thus, a magnitude of the violation, but rather its comparison to the
test of a Bell inequality is general and leads to discrimination error limits of an experimentally determined value of S that
between any LHV and quantum mechanics. is a measure of the significance of the violation. As an aside
Several researchers have since formulated other versions it should be noted that some experimental parameters must
of the Bell inequalities (Clauser and Shimony 1978) that are be measured (and therefore have errors associated with them)
more experimentally amenable. A well known example is in order to evaluate SQM ; therefore, the quantum mechanical
the Bell–Clauser–Horne (BCH) inequality (Clauser and Horne prediction SQM will have errors associated with it.
1974). The BCH inequality is
S(θ1 , θ2 , θ1 , θ2 ) = {R↑↑ (θ1 , θ2 ) − R↑↑ (θ1 , θ2 ) + R↑↑ (θ1 , θ2 ) 2.2. Loopholes
+ R↑↑ (θ1 , θ2 )}{R1↑ (θ1 ) + R2↑ (θ2 )}−1 1, (2) Despite a substantial number of experimental tests of the
Bell inequalities, no experiment to date has been entirely
where a typical R↑↑ (θ1 , θ2 ) denotes the experimental
loophole free (Pearle 1970, Clauser and Horne 1974, Clauser
coincidence count rates when both particles are detected with
and Shimony 1978, Zeilinger 1986, Santos 1996, Gisin and
spin up in the directions θ1 and θ2 , respectively; R1↑ (θ1 )
Gisin 1999). Specifically, one or more of the following
and R2↑ (θ2 ) are singles count rates at detectors 1 and 2,
loopholes were present: (1) the spatial correlation loophole, (2)
respectively. The BCH inequality is especially important
the detection efficiency loophole, and (3) the enforcement
because it provides a direct constraint on the experimentally
of locality (communication loophole). For a more detailed
accessible (observed) detection count rates; it does not require
discussion on the loopholes in previous experiments, we refer
any additional assumptions for experimental implementation.
to the literature (Zeilinger 1986, Santos 1996, Fry and Walther
However, it depends on the ratio of coincidence rates to singles
1999, 2001).
rates and is therefore proportional to the detector efficiency.
Consequently, for low detection efficiency, the inequality
(equation (2)) is always satisfied by the quantum mechanical The first loophole. The spatial correlation loophole relates to
predictions; this is easily recognized by noting that for low the requirement that correlations between the two particles can
detection efficiencies the singles rates in the denominator are only be measured as long as each partner of the entangled pair
large compared with the coincidence rates in the numerator. enters its respective detector. In the ideal case the two-particle
Therefore, equation (2) can only be used for a definitive test if entangled state is generated by a two-body decay that involves
the detectors have very high efficiencies so that the quantum only the two particles that are entangled.
mechanical predictions violate the inequality. Note also, that
it is not necessary to simultaneously measure both projections The second loophole. The detection efficiency loophole
in the case of the BCH inequality. originates from experimental imperfections. Low detection
The principal problem for experimental tests of the BCH efficiencies reduce the observable correlation, since the lower
inequality lies in the fact that experimental imperfections the detection efficiency, the lower the probability that both
generally preclude a definitive test, i.e. they tend to change the partners of an entangled pair will be detected. However, high
quantum mechanical predictions so that they no longer violate detection efficiencies must also be accompanied by a high
the Bell inequalities. probability of actually detecting both particles of an entangled
S377
T Walther and E S Fry
two-photon
dissociation
199 199
Hg Hg
199
two-photon Hg S=0 two-photon
2
ionization ionization
for spin-analysis for spin-analysis
S378
Mercury—the Rosetta stone of physics?
on 199 Hg, which has nuclear spin I = 12 ; it is the first attempt the subsequent analysing polarizer. Zeilinger et al used
to actually test the BCH inequalities (equation (2)). a light emitting diode and a beamsplitter as their random
In our experiment 199 Hg2 dimers are generated in a generator. This experiment represents a very important step
supersonic jet expansion and are photo-dissociated using two in the clarification of the EPR argument.
laser beams in a stimulated Raman transition (see figures 1 The flight time of the particles from the source to their
and 2). The two wavelengths are at approximately 266 and respective detectors does not influence the arguments in
355 nm. The energy difference between a photon from each the discussion of the locality condition. It is solely the
laser beam is distributed equally to the kinetic energies of the communication between the two analyser/detectors that is
two atoms. Due to momentum conservation the atoms will important. However, the time of flight for the particles is
fly apart in exactly opposite directions (180◦ ) in the centre-of- important to a discussion of the lifetime of the entangled
mass frame. However, as a consequence of the initial velocity
state. It is conceivable that the strong correlations predicted
of the dimers in the supersonic jet, the dissociated atoms will
by quantum mechanics only exist for a limited time. This is
separate at an angle of approximately 130◦ in the laboratory
the reason why the experiment by Gisin and co-workers (Tittel
frame.
et al 1998) represents a major step forward. They performed
The preparation of the entangled state given by
equation (1) requires selecting a molecular state with a total a Bell inequality test of the Franson type (Franson 1989) with
nuclear spin I = 0. This is possible due to the specific position–time entanglement, where the two detectors were
symmetry rules for the total wavefunction of a homonuclear separated by more than 10 km. They measured correlations
diatomic molecule consisting of two fermions. Thorough consistent with quantum mechanics after taking into account
analysis of these symmetries (Walther and Fry 1997b) shows the losses in the fibre due to imperfect transmission. The
that only the entangled nuclear spin singlet state is dissociated large separation requires a minimum entangled state lifetime
if the excitation transition at 266 nm starts from states whose of around 30 µs; this is very long compared with the 40 ns
rotational quantum number N is even. in Zeilinger’s experiment. Herein lies one of the benefits of
Determination of the correlation in the spin components using atoms. Atom speeds are generally much less than the
of the two entangled atoms as well as detection of the velocity of light and even with relatively small separations
atoms is achieved via a state-selective two-photon excitation– of the detectors a very long lifetime (milliseconds) of the
ionization scheme (see figure 3). Based on the selection entanglement is tested via the Bell inequalities.
rules for electronic transitions, one can photo-ionize only The first entanglement of atoms was generated by
those atoms which are in the state m F = +1/2, i.e. spin up Haroche and co-workers (Hagley et al 1997). The specific
along the quantization axis by using left-circularly polarized
implementation of this extremely challenging experiment,
resonant radiation at 253.7 nm. Since the quantization axis is
however, did not allow a test of the Bell inequality in the context
determined by the propagation direction of the left-circular
described here.
polarized laser beam, the spin component in any direction
is measured by orienting the laser beam to propagate in that Wineland and co-workers (Rowe et al 2001) are the
direction. The ionized atoms are detected in two independent first group to close the detection efficiency loophole. They
ways, i.e. by observing the photo-ions as well as the photo- prepared two Be+ -ions in an ion trap in an entangled state
electrons. between two hyperfine Zeeman substates. The two substates
In our experiment the enforcement of Einstein locality, can be identified with spin-up and spin-down and are therefore
also known as the communication loophole, can be equivalent to an entangled state as depicted by equation (1).
implemented by employing electro-optic modulators (EOM). The determination of the spin components is, however,
Specifically, the EOM together with a beamsplitting polarizer considerably more subtle than in the usual case of spin- 12
can, in a couple of nanoseconds, change the propagation particles. The detection efficiency loophole is closed by
direction of the excitation laser beam and hence the component resonantly scattering photons in a cycling transition. Due to
of nuclear spin angular momentum being observed. A 12 m the large number of photons observed, the detection efficiency
separation between the detectors will be necessary in order to can be very large. A drawback, however, is that this detection
rigorously close the locality loophole. This estimate includes scheme takes time. And due to the proximity of the two ions
allowances for the selection of a random number, switching in the trap, it is impossible in these experiments to enforce
the EOM, firing the nanosecond-detection lasers, the passage- the locality condition. In fact, Vaidman has argued that the
time of the electrons through the detection system including situation is even worse in these experiments (Vaidman 2001).
the channeltron, and the time to digitize the detector output.
Specifically, due to the detection of the state by the scattering
of many photons, both detection events and the choice of
2.4. Relationship to other experiments both analyser orientations overlap throughout a relatively long
Recently, Zeilinger and co-workers were the first to close the period of time (orders of magnitude longer than the time
detection efficiency loophole (Weihs et al 1998). They utilized required for light to travel between the two ions). Another
a down-conversion source and coupled the entangled photon drawback is that the parameters in the experiment such as
pairs into two fibres that led to two detector arrangements that fidelity of the purity of the entangled state, the detection
were separated by 400 m. At each detector, fast EOM rotated efficiency and the analyser efficiency cannot be determined
the plane of polarization between two randomly selected independently. Thus, an independent comparison with the
values, 0◦ and 45◦ . This is equivalent to randomly rotating quantum mechanical expectation value cannot be performed.
S379
T Walther and E S Fry
2 3
Level 3 F=1/2 (6p ) P0
Ionization limit
Ionization (197.3 nm)
F=3/2 3
(6s6p) 6 P°1
Level 2 F=1/2
22 GHz
−
σ−
Analysis (253.7 nm)
2 1
Level 1 F=1/2 (6s ) 6 S 0
Figure 3. Photo-ionization scheme used for spin analysis of the Hg atoms. The quantization axis is determined by the propagation direction
of the left-hand circular (σ − ) polarized 253.7 nm laser beam. The atom has been analysed as ‘spin-up’ (m F = +1/2) if it is ionized in the
two-photon process, 253.7 + 197.3 nm (via the 6 3 Po1 (F = 1/2) state). Excitation/ionization transitions from m F = −1/2 correspond to a
‘spin-down’ analysis; by the selection rule for σ − , m F = −1, this transition must go via the 6 3 P1o (F = 3/2) state and is highly
suppressed due to its 22 GHz detuning from the F = 1/2 state.
S380
Mercury—the Rosetta stone of physics?
λ C=435.8 nm
γ C=7.19 MHz
c +
λ D=546.1 nm
γ D=9.14 MHz D λ E =404.7 nm kB
C 0
d γ E = 3.32 MHz
a kD
e −
o
B +70.7
λ B =253.7 nm
γ B = 1.3 MHz − 74.7 o
b b kC
(a) (b) (c)
Figure 4. (a) Relevant energy levels of mercury to produce lasing at 253.7 nm. (b) Dressed states for a generic four-level scheme, i.e. levels
a–d from part (a). (c) Wavevector orientations that cancel the Doppler effect for the three-photon transition |d ↔ |b; the x-axis is
perpendicular to the figure, and ki is the wavevector for the transition with wavelength λi .
The dressed-state picture generated by the driving fields are proportional to the wavevector k, this is equivalent to
C and D and bare states |b, |c, and |d is discussed elsewhere kB + kC − kD = 0. Thus, if kB + kC − kD = 0 then E will be
(Lukin et al 1999). The results are summarized in figure 4(b). constant regardless of the atomic velocity. This relation can
For the case of all three fields on resonance, the three dressed always be satisfied, if the three k-vectors satisfy the triangle
states are given to first order in D by condition. Consequently, it is possible to observe the sharp
gain spike even in a Doppler-broadened medium, provided we
1 D
|+ = √ |a + |c + |d (4) properly orient the three laser beams. This is an important
2 C result. Furthermore, such a cancellation of the Doppler effect
can never be achieved in a three-level LWI system with only
1 D
|− = √ |a − |c − |d (5) two transitions.
2 C
Figure 5 shows the result of averaging the imaginary part
D of the susceptibility over a 3D Maxwell velocity distribution
|0 = |d − |a. (6)
C of Hg atoms at a temperature of 300 K. The required
In the limit D → 0 the two dressed states |+ and |− wavevector orientations (laser propagation directions) are
correspond to the usual Autler–Townes components split by given in figure 4(c). Figure 5(a) shows the usual Doppler
2C and the dressed state |0 corresponds to the bare state |d. In broadened absorption profile for the |b ↔ |a resonance
the presence of the perturbing drive D, the state |0 contains an transition when both C and D are off. Turning on C has
admixture of |a and thus has a non-zero dipole matrix element a negligible effect on this Doppler broadened profile. The
with ground state |b. As a result of this coupling, there are usual Autler–Townes doublet is washed out by the different
transitions between |0 and |b, corresponding to three-photon Doppler shifts. To see effects at zero detuning, figures 5(b)–
resonances from |d to |b that exhibit interference effects. (d) have an expanded abscissa (×10). Although the y-axis
It is straightforward to solve the equations of motion of units are arbitrary, they are the same for all four subfigures.
the density matrix (Scully and Zubairy 1997) for the system Figure 5(c) shows the characteristic, although relatively small,
of figure 4. The most relevant result for the present discussion sharp absorption spike at resonance when the perturbing drive
is that the gain spike is very sensitive to the energy separation D is turned on. Finally, figure 5(d) shows a strong gain spike
between states |d and |b; if it varies, the detuning at which when an incoherent pump rab excites some atoms to the state
the gain spike appears varies correspondingly. These facts |a. As discussed above, a second incoherent pump rec excites
are understandable physically by noting that, in the bare state atoms from |e to |c to avoid optical pumping into state |e. It
basis, the gain arises from three-photon Raman-like transitions is noteworthy that the peak gain actually exceeds the depth of
from |d to |b. For example, consider bare atomic energy the Doppler-broadened absorption line. Because of the angles
levels ordered as in figure 4(a). The energy separation of |d chosen for the three lasers such that kB + kC − kD = 0, all
and |b is then E = hν B + hνC − hν D , where ν B , νC , and ν D velocity groups participate in the gain process. Every atom in
are the resonance frequencies of the corresponding transitions. the broad Doppler profile contributes to this narrow gain spike!
Clearly, if E changes and the drive frequencies νC and ν D are For the conditions shown, this peak is 800 kHz, FWHM.
fixed then the frequency ν B of the gain spike must change. It is also interesting to investigate the behaviour of this
For a moving atom, the three photons will have different gain spike as a function of the deviation in the direction of B
Doppler shifts; thus, the effective energy separation of |d from the ideal direction, specified as the +y-axis in figure 4(b).
and |b will generally vary with atomic velocity and the gain Since the cancellation of the Doppler effect is no longer perfect
spike will be washed out by velocity averaging. This is just for such deviations, a rapid reduction of the gain should be
the four-level version of previous three-level LWI studies. observed with increasing deviation. Indeed, the calculated
However, for the atomic energy levels in figure 4(a), the FWHM of the gain is only 0.13◦ = 2.3 mrad. This results
energy separation E is independent of velocity, if the Doppler in a very narrow gain spike as a function of angle. Thus,
shifts satisfy ν B + νC − ν D = 0. Since Doppler shifts regardless of the diameter or length of the lasing medium, the
S381
T Walther and E S Fry
-0.02 -0.026
(a) (c)
-0.03 -0.0265
-0.025 0.05
D
C C
rab
0.025
(b) (d)
-0.026 -0.025
-10 -5 0 5 10 -10 -5 0 5 10
Figure 5. Doppler-averaged results for an ensemble of Hg atoms with a Maxwell velocity distribution at a temperature of 300 K. The
imaginary part of the susceptibility χ (in arbitrary units) is shown versus the detuning of the |b ↔ |a transition (in units of γ D ). Subfigures
(a)–(d) represent the results for cases when the different radiation fields are applied as depicted in the insets (double-headed arrows). The
parameters used are (a) C = 0.0 or C = 6.0, D = 0.0 (expanded abscissa), (b) C = 6.0, D = 0.0, (c) C = 6.0, D = 0.45 and (d) C = 6.0,
D = 0.45, rab = 0.05 (dashed arrow). Furthermore, an auxiliary incoherent pump on the |e ↔ |c transition with rate rec = 0.3γ D is
included in the calculation.
gain is unidirectional and is positive only at very small angles (iii) the vapour pressure can be easily adjusted over a wide
to the axis. This means that the laser output will always be range;
well collimated; it is lasing without mirrors! This is especially (iv) mercury possesses long-lived metastable states (3 P0 -
interesting for lasing in the vacuum ultraviolet where it is and 3 P2 -state), one of which (3 P2 -state) can be trapped
difficult to make mirrors. magnetically and/or optically;
Numerical calculations with the energy level scheme of (v) due to the absence of hyperfine or fine structure in the
figure 4(a), including all magnetic sublevels and appropriate ground state no repumping lasers are required in order to
angular coupling coefficients, show that coherent population prevent optical pumping into states not accessible to the
trapping occurs on the |d ↔ |c transition and that gain is trapping laser.
reduced because population is optically pumped into these
trapped states. To destroy the trapping, an incoherent pump
rcd must be added on the |d ↔ |c transition. But, to avoid 5. Summary
destroying the four-level coherence, the polarizations must be
carefully chosen. Mercury continues to be the choice for many exciting
The experiment analysed here for lasing at 253.7 nm experiments in atomic physics. The quest for a final answer in
is readily generalizable to the VUV. For example, the same the test of Bell inequalities and thus the answer to the question
scheme can be modified to lase at 185 nm using the 6 1 P1 ↔ whether or not quantum mechanics is a complete theory
6 1 S0 transition of Hg. Practical cw lasers at other wavelengths creates new ideas for more and more refined experiments.
in the deep and/or vacuum ultraviolet are also possible. Simultaneously closing the existing loopholes has proven
Furthermore, the technique described for producing Doppler- difficult. However, the atom-based experiments that have
free gain by angular compensation of the Doppler shift is emerged in the last few years are especially promising.
readily applicable to any multi-step process involving more
than two photons.
Acknowledgments
4. Trapping The research described in this paper was supported by
the Robert A Welch Foundation grant no A-1218, by the
In still another application of Hg to fundamental experiments,
National Science Foundation grant PHY-9732459, and by the
it is noted that mercury is well suited for magneto-optical
Office of Naval Research. The authors thank John Clauser,
trapping experiments. Specifically,
Marlan Scully, George Welch, Chris J Bednar, Suzanne Yelin,
(i) the trapping occurs on a forbidden transition resulting in and Michael Fleischhauer for many helpful discussions on
small Doppler-limited temperatures; the theory and encouragement as well as Robert Kenefick,
(ii) mercury has bosonic and fermionic isotopes, which all can Seiichirou Yokoo, Ling Wang and Joe Musser for their
be trapped; experimental contributions and active participation.
S382
Mercury—the Rosetta stone of physics?
S383