Lecture 8. From nonlinear optical effects to squeezing.
Squeezing obtained from nonlinear optical effects: PDC and FWM, Kerr effect. Passing from
quadrature to twin-beam squeezing and vice versa. Obtaining sub-Poissonian statistics through
feedforward technique and through second harmonic generation. The role of loss.
1. Quadrature squeezing.
In this lecture we will continue discussing how nonclassical light can be produced through
PDC and FWM. Unlike in the previous lecture, now we focus on high-gain effects and bright
light. Therefore the criteria for nonclassicality will be different now: squeezing, sub-
Poissonian photon statistics, and sub-shot-noise photon-number correlations (noise reduction).
Parametric down-conversion. Consider the degenerate Hamiltonian,
Hˆ i( a )2 h.c. (1)
In Lecture 4 we have derived that this Hamiltonian makes the quadratures evolve as
qˆ eG qˆ0 , pˆ e G pˆ 0 ,
(2)
G 2t.
Through this evolution, an input vacuum state becomes squeezed vacuum, see Fig.2 in
Lecture 4. One can describe it by acting on the input vacuum state by the evolution operator,
1 G
exp{ dtHˆ } exp{ ( a ) 2 h.c.} Sˆ (G ) ,
i 2
where we introduced the squeezing operator Sˆ (G ) . Then the squeezed state can be written as
Sˆ (G ) vac . (3)
To observe this squeezing, one should perform homodyne tomography and measure the
Wigner function (Lecture 5). The nonclassical feature is then the uncertainty of a quadrature
being below the shot noise. A measure of this nonclassicality is quadrature squeezing in dB,
defined as
p 2
10 log 2 10 log[e 2G ].
p0
(We assume here that it is the p quadrature that is squeezed; in the general case it can be any
quadrature.)
The record quadrature squeezing achieved now is -15 dB [H. Wahlbruch et al., PRL 17,
110801 (2016)]. It means that the variance of the squeezed quadrature was 101.5 32 times
smaller than the shot-noise limit, and the uncertainty was 32 5.6 times smaller than the
shot-noise limit (Fig.1 schematically shows the corresponding Wigner function). The mean
number of photons in this squeezed vacuum was sinh 2 G , with G ln(5.6) 1.7, which makes
about 7 photons. This is a relatively weak state
p
of light, but its detection was made possible by
using a strong local oscillator. W ( q, p )
It is worth mentioning that in such experiments, q
PDC occurs in a cavity, which sets the spectrum
of the squeezed vacuum (should be resonant
with the cavity) and provides a single frequency
and angular mode.
Fig. 1
At the same time, one can obtain, instead of a
squeezed-vacuum state, a squeezed coherent state, which has the same squeezing but a larger
number of photons. There are 2 ways to do it.
1. To send to the nonlinear crystal not
the vacuum (no input beam, except the p
pump) but a coherent state . What W ( q, p )
we obtain then can be understood from
the evolution (2); see also Fig.2 of q
Lecture 4. The state will then be written
as
Sˆ (G ) .
Fig. 2
The corresponding Wigner functions are
shown in Fig.2: in blue for the initial coherent state and in red for the resulting displaced
squeezed state.
2. The other way is to overlap the squeezed vacuum with a coherent state on a beamsplitter. In
order not to destroy the squeezing, the transmission of the beamsplitter for the squeezed
vacuum should be very high; the coherent light will then be attenuated but remain coherent.
This procedure will then displace the squeezed vacuum state along the amplitude of the
coherent state. Figure 3 shows the case where the displacement is along the squeezing
direction. Blue circle and ellipse show the initial states (the coherent state is shown already
attenuated after the beamsplitter) and the red ellipse shows the displaced squeezed state. One
can say that this state has a ‘coherent carrier’ (the red ‘stick’).
This situation is interesting because it allows one to prepare
bright sub-Poissonian light. It is useful for any measurements p
involving direct detection because the accuracy of photon-
number measurement can be very high. And the state shown
in Fig.3 (red ellipse) indeed corresponds to sub-Poissonian
light, as it is squeezed in the amplitude -> in the number of W ( q, p )
photons.
Kerr squeezing. Consider now a material with cubic q
nonlinearity. Is it possible to observe quadrature squeezing in
this case? Clearly, Hamiltonian (1) can be only realized if the
pump is at the same wavelength as the squeezed light. At first Fig. 3
sight, it is a problem because one cannot get rid of it. But this
is only a problem if one needs photon pairs, not bright light. In the case of bright light, one
can consider the pump as the ‘coherent carrier’ for squeezed vacuum, as in the previous
example.
In a material with a large cubic nonlinearity, the Kerr
effect will indeed lead to squeezing (Fig.4). To see this, p
consider the evolution of a coherent state in the course of
its propagation through such a material (for instance, a
fibre). Points corresponding to higher amplitudes will have
a larger refractive index and larger phase delays. As a
result, the uncertainty region will evolve and become
compressed (grey ellipse) in some direction, corresponding q
to a certain quadrature (red arrows). Note that it is not the
amplitude, because the Kerr effect leaves the total
Fig.4
amplitude uncertainty constant. Rather, there appears some
correlation (coupling) between the amplitude and the
phase, - that is why the shape is tilted.
To obtain amplitude squeezing, one can overlap this beam with some coherent one and shift
the ellipse in such a way that the squeezing becomes amplitude squeezing, as in Fig.3.
2. Two-mode squeezing and entanglement.
Consider now the two-mode Hamiltonian
Hˆ ia1 a2 h.c. (4)
We know that it leads to sub-shot-noise photon-number correlations, with the nonclassical
feature
Var ( N1 N 2 )
NRF 1. (5)
N1 N 2
But there is another nonclassical feature one can observe for the state at the output of such a
nonlinear material (two-mode squeezer). Recall the two-mode Bogoliubov transformations
(Lecture 4),
a1 ua10 va20 ,
a2 ua20 va10 ,
u sinh G , v cosh G.
By introducing quadratures, a1, 2 qˆ1, 2 pˆ 1, 2 , we get
q1 ip1 u( q10 ip10 ) v ( q20 ip20 ),
q2 ip2 u( q10 ip10 ) v ( q20 ip20 ).
This leads to 4 equations for the quadratures (we take the real and imaginary parts separately):
q1 uq10 vq20 ,
q2 uq20 vq10
p1 up10 vp20 ,
p2 up20 vp10 . q2 p2
Let us see what happens with the sums F ( q1 , q2 ) F ( p1 , p2 )
and differences of the quadratures.
Then we get p1
q1 q2 eG ( q10 q20 ), q1
q1 q2 e G ( q10 q20 )
p1 p2 e G ( p10 p20 ),
p1 p2 eG ( p10 q20 ). Fig. 5
It means that the sum of position
quadratures, q1 q2 , and the difference of the momentum quadratures, p1 p2 , will get anti-
squeezed, while the difference q1 q2 and the sum p1 p2 , will get squeezed:
( q1 q2 ) ( p1 p2 ) eG ( q10 q20 ) eG ( q10 q20 ) 2 eG 2 q10 eG 2 / 2 1,
( q1 q2 ) ( p1 p2 ) ... 1.
At the same time, taken separately, each quadrature will not be squeezed. But there will be
correlation between the quadratures. This exactly means entanglement for quadratures, and
we can draw the probability amplitude distributions F ( q1 , q2 ) and F ( p1 , p2 ) (Fig.5) similar to
how we drew F ( k1 , k2 ) at Lecture 7.
3. Transformations between two-mode and quadrature squeezing.
One can pass from quadrature to two-mode squeezing and vice versa. Indeed, if we introduce
new modes, characterized by new annihilation operators,
1 1
as [a1 a2 ], ad [a1 a2 ],
2 2
and the corresponding creation operators, then Hamiltonian (4) can be rewritten:
Hˆ i [( as )2 ( ad ) 2 ] h.c.
2
It means that for the new modes s,d there will be quadrature squeezing; moreover, if 0 ,
the squeezed quadratures will be pˆ s , qˆd , while qˆ s , pˆ d will be anti-squeezed.
4. Obtaining sub-Poissonian pulsed light.
Feedforward technique. Very similar to the heralded preparation of single photons is the so-
called feedforward technique where one of the bright pulsed twin beams (2 in Fig.6) is
registered by a detector and then the other
one (beam 1) is transmitted only if the
detector measures a certain number of 1
photons, within some bounds. This
detector of course does not need to count
photons, but just integrates their number
(one can say, measures the total energy).
2
gate
For beam 1, sub-Poissonian statistics will
then be observed: N 2 N , or the Fano
Fig.6
factor will be F 1 .
Sub-Poissonian light from SHG. The same effect can be achieved through the second
harmonic generation (SHG). Indeed, due to SHG, pairs or groups of photons will be removed
from the pump beam, and the pump will become anti-
bunched. The same effect can be illustrated by a simple x2
diagram on the phase plane (Fig.6). The initial pump
state (blue) becomes squeezed in the amplitude, because
intensity peaks are suppressed.
5. The role of losses.
All these bright-light nonclassical effects (quadrature x1
squeezing, sub-shot-noise photon-number correlations,
sub-Poissonian statistics) are very sensitive to losses. Fig.6
As an example, consider quadrature squeezing. Losses,
or non-unity quantum efficiency of a detector can be represented as a beamsplitter having
transmissivity t and reflectivity r 1 . Then, the annihilation operator a at the
output will be expressed through the annihilation operator a0 at the input as
a a0 1 a v ,
where av is the vacuum annihilation operator. By taking the Hermitian and anti-Hermitian
parts, we see that the quadratures will change as
q q0 1 qv ,
p p0 1 pv .
For the variances, one can then write
q 2 q02 (1 ) qv2 ,
p 2 p02 (1 ) pv2 .
Suppose that initially, the p-quadrature was squeezed. After the loss, its squeezing will be
given by
p 2 / pv2 p02 / pv2 1 .
Even if the squeezing was infinite initially, p02 0 , after loss it will be 1 , limited by the
quantum efficiency.
This behavior of nonclassical features is different from the case of ‘faint’ light and
nonclassicality signs based on the normalized correlation functions, which we considered in
Lecture 7.
Home task: using the same model of loss, calculate the effect of 50% quantum efficiency on a
beam of light with the Fano factor 0.1.
Books:
1. Bachor, Ralph, A Guide to Experiments in Quantum Optics.
