Squeezed Light

Thomas Rotter
University of New Mexico, Department of Physics
and Astronomy, Albuquerque, NM 87131, USA
(Dated: January 28, 2000)
Abstract
This paper gives an overview on squeezed light with an emphasis on its generation. Squeezing
means, that the uctuation in one of two conjugate components (e.g. amplitude and phase) is
suppressed while enhanced in the other. Consequently squeezing can be generated by phase sensi-
tive processes. Such processes are known from nonlinear optics, e.g. second harmonic generation,
parametric down{ and upconversion and four wave mixing, and they have been successfully used to
generate squeezed light. Application of squeezed light is still rare because it is dicult to maintain
the squeezed character.

PACS number: 42.50

1
 I. INTRODUCTION

A classical electromagnetic eld consists of waves with well dened amplitude and phase.
However, in a quantum treatment, uctuations are associated with both conjugate variables.
Equivalently the eld can be described in two conjugate quadrature components and the
uncertainties in the two conjugate variables satisfy the Heisenberg uncertainty principle. A
coherent state, that most nearly describes a classical electromagnetic eld, has an equal
amount of uncertainty in the two variables, and the product is the minimum uncertainty.
This case is usually called the shot noise limit.
While the product of the uncertainties in the two conjugate quadrature components
cannot be minimized further, it is possible to reduce the uctuations in one of them well
below the quantum limit. The canonically conjugate quadrature component must then
display enhanced uctuations in order to fulll the Heisenberg uncertainty principle. Those
states are called squeezed states.
Squeezed light with uctuations below the standard quantum limit in one of the quadra-
ture components has many attractive applications, e.g. in optical communication, in pre-
cision and sensitive measurements such as gravitational wave detection, or in noise free
amplication. Squeezed light is however extraordinarily fragile and may be degraded even
by a beam splitter or a mirror, as those admit the vacuum uctuations from outside to enter,
which exceed the squeezed uctuations.
In this paper I will brie y describe the theory of squeezed states and their properties.
The main focus will be on generation of squeezed light by a variety of processes, including
 - and  - nonlinearities. I will present experiments that have been successfully carried
(2) (3)

out, as well as brie y describe theoretical approaches to investigate the generation processes.
Comparisons between theoretical and experimental results will be given.
 II. SQUEEZED STATES
A Theoretical considerations

Consider a quantized single-mode electric eld of frequency  [1]:

E~ (t) = E~(^ae,it + a^y eit ) (1)
where a^ and a^y are the non-Hermitian annihilation and creation operators, respectively,
obeying the commutation relation [^a; a^y] = 1. The eld can also be written in terms of
Hermitian quadratures X^ and Y^ ,
E~ (t) = 2E~(X^ cos t + Y^ sin t): (2)
Here X^ = (^a + a^y) and Y^ = i (^a , a^y) are dimensionless versions of the position and
1
2
1
2

momentum operators x^ and p^. The uncertainty relation

X^
Y^  41 (3)
follows from the commutation relation [X; ^ Y^ ] = i . If the equality in the uncertainty relation
2

(3) holds, then the minimum amount of uctuation, the shot noise limit, is reached. Further
reduction of uctuation, squeezing, is possible only in one of the two canonically conjugate
variables
(
X^i) < 41
2
(i = 1 or 2) (4)
at the expense of increasing uctuation, antisqueezing [11], in the other, so that (3) is
still satised. An ideal squeezed state is obtained if in addition to (4) the equality in the
uncertainty relation (3) also holds.
Figure II A illustrates a coherent state in the phasor plane. The expectation value of the
annihilation operator a^ = X^ + iY^ has amplitude  and phase . The hatched areas indicate
the probability distribution for an event yielding a phasor terminating in one of the points
in the phasor plane, the circle indicating the root-mean-square deviation of the distribution.
This is a schematic illustration of the Wigner distribution.
FIG. 1: Representation of a coherent state in the complex phasor plane with X along the real axis
and Y along the imaginary, and below Y versus phase  = !t. [2]

Figure II A shows a squeezed state. The Wigner distribution is represented by an ellipse.

There are two phases to be considered, that of the expectation of the phasor  and that of
the orientation of the ellipse of the Wigner distribution (or  in Fig. II A). The area of the
ellipse is the same as the area of the circle of the phase-independent zero-point uctuations,
since the uncertainty principle must be obeyed. As shown in Fig. II A, the major and minor
axes of the ellipse depend on each other as e s and e,s, where s is a squeezing parameter.
1 +
2
1
2

s = 0 corresponds to a coherent state with no squeezing, as depicted in Fig. II A.

For the general case of Fig. II A and Fig. II A the unitary squeeze operator [7] is given by

S^( ) = e 2  a^ , 2  a^ ; (5)
1  2 1 y2

where  is the complex squeeze parameter, which depends on s and  from Fig. II A,

 = sei ; 0  s < 1; 0    2: (6)

The vacuum coherent state j0i corresponds to Fig. II A with  = 0, likewise squeezed
vacuum j0si = S^( )j0i corresponds to Fig. II A with  = 0, or to Fig. II A. In contrast to
FIG. 2: Representation of a squeezed state in the complex phasor plane. X along the real axis
and Y along the imaginary, and below Y versus phase  = !t. [2]

squeezed vacuum, the situation  6= 0 is called bright squeezed light. Every squeezed state
can then be described as
j;  i = D^ ()S^( )j0i (7)
where
D^ () = ea^ , a^ (8)
y 

is a coherent displacement operator [7].

The squeeze operator (5) transforms the annihilation operator according to the
Bogoliubov-transformation
a^ ! S^,1 ( )^aS^( ) = a^ cosh s , a^yei sinh s: (9)
The mean of the quadrature operators hX^ i, hY^ i correspond then to the real and imaginary
part of , respectively, as shown in Fig. II A. The variances are
h(
X ) i = 41 (e, s cos 12  + e s sin 12 );
2 2 2 2 2
(10)
h(
Y ) i = 41 (e, s sin 12  + e s cos 21 );
2 2 2 2 2
(11)
FIG. 3: Uncertainty ellipse of the ideal squeezed vacuum state. [7]

corresponding to Fig. II A.
Detection of squeezed states requires a phase sensitive scheme that measures the variance
of a quadrature (4) of the eld. In homodyne detectors the squeezed input state is superposed
by a beamsplitter on the mode from a local oscillator with the same frequency, and this
superposition is measured by a photodetector. The principle is somewhat similar to that of
a lock in amplier, which is familiar to every experimentalist. Squeezing in the input mode
is then revealed by sub-Poissonian photocount statistics. In a balanced homodyne detector
both output ports of the beamsplitter are measured and subtracted from each other. This
technique removes the noise contributions that are made by the input signal and the local
oscillator alone. A thorough discussion of these and other detection schemes can be found
e.g. in [7] or [1].

III. GENERATION OF SQUEEZED STATES

To generate squeezed light from coherent light is to reduce uctuation in one quadrature
component, while enhancing it in the canonically conjugate component. As can be seen
from (2), this is a phase sensitive process. Such phase sensitive processes are known from
the nonlinear interaction of light with matter, e.g. second harmonic generation, parametric
down- and upconversion and four wave mixing. All of these processes have in fact been
successfully employed to generate squeezed light.

A Second harmonic generation

Although most of the earlier experiments on squeezing have concentrated on the frequency
downconversion process, the reverse process, namely second harmonic generation (SHG),
has also attracted considerable attention. It is one of the simplest  - or second order
(2)

nonlinear optical processes and it can create nonclassical light for both the fundamental
and the harmonic elds [3, 4, 5]. Substantial squeezing arises if the nonlinear crystal is
suitably long or if it is placed in an optical cavity to resonate the fundamental or the
second-harmonic mode or both. Although noise reduction as high as 52% was observed with
an actively stabilized doubly resonant cavity [10], it is quite dicult to maintain doubly
resonant condition for a suciently long time. Therefore singly resonant cavities, in which
only the fundamental mode is conned, have been proposed [9] to generate squeezed light.
A theoretical analysis by Paschotta et al. [9] predicted that such a system can produce
9.5 dB of squeezing in the second harmonic output and demonstrated this in an experiment
featuring a monolithic nonlinear device. The analysis by Paschotta et al. is based on a
two-photon loss model. Here the intensity of the fundamental is treated so that it remains
constant during interaction with the second harmonic. This assumption is reasonable if
the single-pass conversion eciency is quite small. To estimate the maximum amount of
squeezing, which is reached at high pump powers and high conversion eciencies, Maeda
et al. [4] has performed an analysis under consideration of fundamental depletion in the
crystal. It is found that a singly resonant system can produce an arbitrary amount of
squeezing in the second harmonic. This high noise suppression is explained by an optical
negative feedback mechanism realized by the fundamental cavity mode. Because of the large
conversion eciency, the fundamental eld is strongly correlated with the second harmonic.
Since the fundamental is re ected at the cavity mirrors, it carries information on uctuations
of the output second harmonic over to the next circulation, thus providing the negative
feedback. The re ectivity of the cavity mirrors is therefore a limiting factor of squeezing.

FIG. 4: Schematic representation of the experimental setup of Ralph et al. [6]. A diode pumped
Nd:YAG laser was used to pump the monolithic doubler. The Faraday isolator prevents signicant
backre ection from entering the laser. A half-wave plate (=2) permits variable attenuation. The
electrooptic modulator (EOM) phase modulates the driving eld at 88 MHz. An error signal for
locking the laser is derived from the 88 MHz signal driving the EOM and the fundamental light
re ected from the monolith by use of a double balanced mixer. The second harmonic beam exits
the front face of the monolith, where it is separated from the fundamental by use of two dichroic
beam splitters. It is then incident upon a balanced detector, the summed and dierenced outputs
of which are monitored with a spectrum analyzer. [6]

As an example of an experimental realization of squeezing by SHG, consider the experi-

ment performed by Ralph et al. [6]. Here squeezing from a singly resonant second harmonic
generating crystal was observed. The experimental setup is depicted in Fig. III A. The
summed photocurrent of the two detectors yields the intensity noise of the light, and the
dierenced photocurrent yields the quantum noise level.
Figure III A shows comparisons of the experimental results with theoretical calculations.
In the theoretical analysis a technique was employed that permits the noise characteristics
of the pump source used in the experiment to be explicitly modelled. The pump laser and
the SHG crystal were treated as single quantum systems. As can be seen from Fig. III A,
FIG. 5: left: Comparison of experimental and theoretical noise spectra of the second harmonic
light. Right: Comparison of squeezing spectra with (solid curve) and without (dashed curve) pump
noise. [6]

excellent quantitative agreement was obtained between theory and experiment. Squeezing
emerges well below 20 MHz, because the fundamental cavity acts as a noise lter, reducing
the amount of laser noise that enters the SHG crystal. From the graph on the right hand
side of Fig. III A it is clear, that a major factor limiting the squeezing is the noise of the
laser pump.
In another experiment performed by Bell et al. [8], bright tunable ultraviolet squeezed
light was generated. Squeezing of ultraviolet light is more dicult than at longer wave-
lengths, because beta-barium borate (BBO), which has a small nonlinear coecient, must
be used for SHG instead of the LiNbO or KNbO . This wavelength region is however
3 3

of particular interest for atomic spectroscopy, as many atoms have strong absorption lines
here. The experimental setup is in principle not very dierent from that of [6], Fig. III A.
The SHG crystal was also placed in a singly resonant cavity in order to generate maximum
squeezing on the second harmonic beam rather than the fundamental. The squeezed light
was detected by a balanced homodyne detector. Squeezing was detected here at much lower
frequencies than in [6], 3:5 : : : 18 MHz. These are smaller frequencies than the repetition
rate of the laser, which means that not every pulse is squeezed, instead the integration over
a group of pulses is considered.
This experiment was theoretically modelled by the above mentioned theory of Paschotta
et al. [9], because the small nonlinear coecient of BBO justies the neglection of funda-
mental depletion. The results show that for all frequencies the measured noise spectrum
lies above the theoretical spectrum, but there are minima in the experimental spectrum
that approach the theoretical values. The extra noise can at least partially be attributed
to the argon laser that pumps the Ti:Sapphire laser, which results in extra noise at the
fundamental. Furthermore, the dependence of the squeezing on the laser power and on the
input coupling of the fundamental eld into the SHG cavity was investigated. In agreement
with the theory, the observed squeezing increases with increasing laser power. As for the
dependence on the input coupling, the theory predicts that the amount of squeezing will
increase continuously as the input coupling is decreased, since small input coupling reduces
the uctuation of the fundamental in the SHG crystal. This applies for the case of no other
losses than from the SHG. However, for a nite loss there will be an optimum input coupling
to achieve maximum squeezing. This behavior was conrmed by the experiment.

B Degenerate parametric amplication

The degenerate parametric amplier (DPA) was the earliest candidate suggested to pro-
duce squeezed light. As SHG, parametric amplication is also a  -process and generates
(2)

downconverted light at one half of the pump wave frequency. (Here signal and idler beam
have the same wavelength, they are degenerate.) The strong quantum correlations between
the two downconverted photons are responsible for the squeezing. A theoretical analysis [1],
treating the pump eld classically and neglecting pump depletion, shows that the output of a
DPA can in principle be squeezed to 100% (which means zero uncertainty in one component
and innite in the other) and is in an ideal squeezed state, as described in section II. This
makes it a particularly important source of squeezed radiation. However, this result holds
only for a perfectly coherent, monochromatic pump with a stabilized intensity. Amplitude
and phase uctuations in the pump eld due to noise in the laser radiation will degrade the
squeezing [1, 7].
 C Optical parametric oscillator

The parametric amplier becomes an optical parametric oscillator (OPO), when placed
in an optical cavity. This is the preferred method to generate squeezing since the optical
cavity helps increase the interaction of the eld with the nonlinear medium, as mentioned
in section III A for the case of SHG. A theoretical analysis [1] shows, that the maximum
amount of squeezing of the cavity eld is only 50%. This is due to vacuum uctuations that
enter through the out-coupling mirror. The eld emitted from the OPO can however be
almost perfectly squeezed, since the cavity eld and the eld due to the vacuum uctuations
entering the cavity through the out-coupler become correlated over time, thus making it
possible for the residual uctuations in the out-coupled cavity eld to cancel out with the
corresponding uctuations of the vacuum eld, which is re ected on the out-coupling mirror.
This leads to almost perfect multi-mode, or spectral squeezing at an appropriate frequency.
In an experiment performed by Breitenbach et al. [11] a frequency doubled Nd:YAG laser
pumps a monolithic degenerate type-I OPO. Its output is analysed in a balanced homodyne
detector. The OPO consists of the parametric amplier, being a LiNbO -crystal doped with
3

MgO, whose end faces are polished and appropriately coated, thus providing the cavity.
The monolithic design has the advantage of good mechanical stability and small losses.
Furthermore, because of the short length of the cavity, the linewidth of the resonator is
broad, which implies a broad bandwidth in the noise reduction. In the frequency range of
1 : : : 30 MHz the squeezing spectrum as well as the antisqueezing spectrum was recorded.
The maximal squeezing level was 5.5 dB below the vacuum noise level at a frequency of
2 MHz.
From the measurements of pairs of rotated quadrature components the Wigner function
of the system was reconstructed. This technique is called the optical homodyne tomography.
It represents the rst step in a complete experimental characterization of a squeezed state.
 D Four-wave mixing

In contrast to the previously discussed processes four-wave mixing is a  - or third

(3)

order nonlinear process [1]. Here two planar counterpropagating pump waves interact in the
nonlinear medium with a probe eld entering at an arbitrary angle to the pump waves and
yield a fourth (output) wave. The latter is proportional to the complex conjugate of the
probe wave. This possibility of generating phase conjugate waves has many applications in
adaptive optics. Four-wave mixing is also an important source of squeezed light, as seen in
the considerable number of experiments that have been successfully performed.
Squeezed states of light were observed using both near-resonant and nonresonant optical
nonlinearities. For a theoretical analysis of the latter case, the medium can be characterized
by a classical susceptibility . This is the case for four wave mixing in glass ber and for
many nonlinear crystals, e.g. MgO:LiNgO , which can also be used as a parametric amplier,
3

as described in section III C. In four wave mixing experiments in atomic beams and vapors,
however, the pump is tuned close to an atomic resonance. Here losses are inherent and
spontaneous emission must be considered. In this case the medium must be quantized and
can no longer be characterized by a classical .
For near-resonant experiments there are still two choices to be made: degenerate or non-
degenerate four-wave mixing. In the degenerate case the pump and signal elds are at the
same frequency. Squeezing is measured here only at a weak eld at the pump frequency.
Nondegenerate four-wave mixing provides a great advantage as for minimizing the eects
of spontaneous emission [12]. Nondegenerate in this case means that each of the two weak
intensity modes is equally and oppositely detuned in frequency from the central high intensity
pump mode.
Slusher et al. [12] have investigated squeezed light generated by four wave mixing near the
Na atomic resonance, both experimentally and theoretically. The experimental apparatus
is shown in Fig. III D.
It consists of a cw ring dye laser, that is tuned to near the weaker hyperne group
FIG. 6: Schematic diagram of the experimental arrangement for generating squeezed light by
four-wave mixing near the Na atomic resonance, used by Slusher et al. [12]. Explanation see text.

of the D Na resonance. The laser crosses an atomic beam of Na and is re ected back
2

into the same Na interaction region, resulting in a standing wave pattern. This drives a
polarization in Na to generate pairs of photons at a frequency shift of  595 MHz from the
pump frequency. These pair frequencies are are resonant in a cavity oriented at a small
angle (0.86) with respect to the pump beam. To detect the squeezed light from this cavity,
a balanced homodyne detector is used. The theoretical description involves a fully quantum
mechanical treatment in which the medium is modelled as N two level atoms, and which
includes losses and dephasing due to spontaneous emission near the atomic resonance. The
theory predicts large squeezed-noise reductions of the order of 10 at pump intensities of
0:1 : : : 0:5 times the saturation intensity of the atomic resonance. The experimental results
compare well with theoretical predictions at small pump intensities and large detuning
of
the pump frequency with respect to the center of the atomic resonance. At small detuning
and large pump intensity an extra phase insensitive noise is observed that is not predicted
by the theory at hand. At high pump intensities, the four wave mixing process reaches the
threshold for oscillation, in agreement with the theoretical model.

IV. SUMMARY

The processes mentioned in section III are but a small sampling of a great variety of
processes that can be employed for the generation of squeezed light. Other successfully ap-
plied techniques include photon number squeezing in semiconductors [13], amplitude squeez-
ing in semiconductor diode lasers [18], utilizing the Kerr-nonlinearity as another  - or(3)

third order nonlinear process, e.g. in optical bers [14, 15], using a Q-switched laser to
pump a parametric downconverter, which exhibits squeezing [16], producing amplitude-
squeezed optical solitons [17], using phase-sensitive parametric ampliers to compensate for
the propagation losses incurred by solitons in optical bers, which squeezes uctuations in
the soliton frequency, thus reducing the Gordon-Haus jitter [19], etc. However, the applica-
tion of squeezed light in any of the areas, that were mentioned at the beginning, is still rare.
This is due to the fragility of squeezed states, that makes technological application dicult.

REFERENCES

[1] Marlan O. Scully and M. Suhail Zubairy, Quantum optics. (Cambridge University
Press, 1997)
[2] Hermann A. Haus, J. Opt. Soc. Am. B 12, 2019 (1995)
[3] L. A. Lugiato, G. Strini, and F. De Martini, Opt. Lett. 8, 256 (1983)
[4] Joji Maeda and Kazuro Kikuchi, Opt. Lett. 21, 821 (1996)
[5] Ruo-Ding Li and Prem Kumar, Opt. Lett. 18, 1961 (1993)
[6] T. C. Ralph, M. S. Taubman, A. G. White, D. E. McClelland, and H.-A. Bachor, Opt.
Lett. 20, 1316 (1995)
[7] R. Loudon and P. L. Knight, J. Mod. Opt. 34, 709 (1987)
[8] A. S. Bell, E. Riis, and A. I. Ferguson, Opt. Lett. 22, 531 (1997)
 [9] R. Paschotta, M. Collet, P. Kurz, K. Fiedler, H.-A. Bachor, and J. Mlynek, Phys. Rev.
Lett. 72, 3807 (1994)
[10] H. Tsuchida, Opt. Lett. 20, 2240 (1995)
[11] G. Breitenbach, T. Muller, S. F. Pereira, J.-Ph. Poizat, S. Schiller, and J. Mlynek, J.
Opt. Soc. Am. B 12, 2304 (1995)
[12] R. E. Slusher, B. Yurke, P. Grangier, A. LaPorta, D. F. Walls, and M. Reid, J. Opt.
Soc. Am. B 4, 1453 (1987)
[13] Malvin C. Teich, Frederico Capasso, and Bahaa E. A. Saleh, J. Opt. Soc. Am. B 4,
1663 (1987)
[14] G. J. Milburn, M. D. Levenson, R. M. Shelby, S. H. Perlmutter, R. G. DeVoe, and D.
F. Walls, J. Opt. Soc. Am. B 4, 1476 (1987)
[15] M. D. Levenson and R. M. Shelby, J. Mod. Opt. 34, 775 (1987)
[16] Orhan Aytur and Prem Kumar, Opt. Lett. 17, 529 (1992)
[17] Dmitriy Krylov, Keren Bergman, and Yinchieh Lai, Opt. Lett. 24, 774 (1999)
[18] Yong-qing Li and Min Xiao, Opt. Express 2, 110 (1998)
[19] I. H. Deutsch and I. Abram, J. Opt. Soc. Am. B 11, 2303 (1994)