Squeezed light

A. I. Lvovsky Institute for Quantum Information Science, University of Calgary, Calgary, Canada, T2N 1N4 and Russian Quantum Center, 100 Novaya St., Skolkovo, Moscow region, 143025, Russia∗ (Dated: July 29, 2016) The squeezed state of the electromagnetic field can be generated in many nonlinear optical pro- cesses and finds a wide range of applications in quantum information processing and quantum metrology. This article reviews the basic properties of single-and dual-mode squeezed light states, methods of their preparation and detection, as well as their quantum technology applications.

I. WHAT IS SQUEEZED LIGHT? If R > 1, the position variance is below that of the vac- uum state, so |sqRi is position-squeezed; for R < 1 the A. Single-mode squeezed light state is momentum-squeezed. In other words, if we pre- pare multiple copies of |sqRi, and perform a measure- In squeezed states of light, the noise of the electric field ment of the squeezed observable on each copy, our mea- at certain phases falls below that of the vacuum state. surement results will exhibit less variance than if we per- This means that, when we turn on the squeezed light, formed the same set of measurements on multiple copies we see less noise than no light at all. This apparently of the vacuum state. paradoxical feature is a direct consequence of quantum More generally, we say that a state of a single harmonic oscillator exhibits (quadrature) squeezing if the variance nature of light and cannot be explained within the clas- 2 sical framework. of the position, momentum, or any other quadrature The basic idea of squeezing can be understood by con- in which the state exhibits variance below 1/2. In ac- sidering the quantum harmonic oscillator, familiar from cordance with the uncertainty principle, both position undergraduate quantum mechanics. Its vacuum state and momentum observables, or any two quadratures as- wavefunction in the dimensionless position basis is given sociated with orthogonal angles, cannot be squeezed at by1 the same time. For example, in state (3) the product ∆X2 ∆P 2 = 1/4 is the same as that for the vacuum 1 2 ψ (X) = e−X /2, (1) state (1). 0 π1/4 Squeezing is best visualized by means of the Wigner which in the momentum basis corresponds to function — the quantum analogue of the phase-space probability density. Figure 1(c,d) display the Wigner +∞ 1 Z 1 2 ψ˜ (P ) = √ e−iP X ψ (X)dX = e−P /2 functions of the position- and momentum-squeezed vac- 0 0 1/4 2π −∞ π uum states, respectively. The squeezing feature becomes (2) apparent when these Wigner functions are compared (so the vacuum state wavefunction is the same in the with that of the vacuum state [Fig. 1(a)]. Figure 1(e,f) position and momentum bases). The variance of the po- shows squeezed coherent states, which are analogous to sition and momentum observables in the vacuum state the squeezed vacuum except that their Wigner function equals h0| ∆X2 |0i = h0| ∆P 2 |0i = 1/2. is displaced from the phase space origin akin to the co- The wavefunction of the squeezed-vacuum state |sqRi herent state [Fig. 1(b)]. with the squeezing factor R > 0 is obtained from that of The state shown in Fig. 1(f) is particularly interest- the vacuum state by means of scaling transformation: ing because it exhibits, as a consequence of momentum √ squeezing, phase squeezing — reduction of the uncer- R −(RX)2/2 tainty in the phase with respect to a coherent state of ψR(X) = e , (3) arXiv:1401.4118v2 [quant-ph] 28 Jul 2016 π1/4 the same amplitude. Because the Schr¨odingerevolution and under the standard harmonic oscillator Hamiltonian cor- responds to clockwise rotation of the phase space around ˜ 1 −(P/R)2/2 ψR(P ) = √ e (4) the origin point, the phase squeezing property is pre- 1/4 π R served under this evolution. In the same context, the in the position and momentum bases, respectively. In state in Fig. 1(e) is sometimes called amplitude squeezed. this state, the variances of the two canonical observables According to the quantum theory of light, the Hilbert are space associated with a mode of the electromagnetic field is isomorphic to that of the mechanical harmonic oscil- ∆X2 = 1/(2R2) and ∆P 2 = R2/2. (5)

∗ 2 ˆ ˆ ˆ [email protected] The field quadrature is observable Xθ = X cos θ + P sin θ, where 1 We use convention [X,ˆ Pˆ] = i for the quadrature observables. θ is a real number known as quadrature angle. 2 √ lator. The role of the position and momentum observ- the position operator Xˆ = (ˆa +a ˆ†)/ 2 in this state is ables in this context is played by the electric field magni- zero while its variance equals tudes measured at specific phases. For example, the field (ˆa +a ˆ†)2 1 at phase zero (with respect to a certain reference) cor- h∆X2i = hψ| |ψi = − s, (7) responds to the position observable, that at phase π/2 2 2 to the momentum observable, and so on. Accordingly, so state |ψi is position squeezed for positive s. phase-sensitive measurements of the field in an electro- a) b) magnetic wave are affected by quantum uncertainties. For the coherent and vacuum states, this uncertainty is pump pump phase-independent and equals p ω/2ε V (the standard photon ~ 0 crystal crystal pair quantum limit, or SQL), where ω is the optical frequency photon pair and V is the quantization volume [1]. But squeezed op- tical states exhibit uncertainties below SQL at certain FIG. 2. Spontaneous parametric down-conversion. a) De- phases. generate configuration, leading to single-mode squeezed vac- Dependent on whether the mean coherent amplitude of uum. b) Non-degenerate configuration, leading to two-mode the state is zero, squeezed optical states are classified into squeezed vacuum. squeezed vacuum and (bright) squeezed light. Squeezed coherent states form a subset of bright squeezed light This result illustrates one of the primary methods states. of producing squeezing. Spontaneous parametric down- a) b) conversion (SPDC) is a nonlinear optical process in which a photon of a powerful laser field propagating through a P XP P XPsecond-order nonlinear optical medium may split into two 2 -200 24 6 -2 2 2 -200 24 6 -2 2 photons of lower energy. The frequencies, wavevectors X Dj X -2 246 -2 246 and polarizations of the generated photons are governed by phase-matching conditions. Single-mode squeezing, -2 -2 such as that in the above example, is obtained when SPDC is degenerate: the two generated photons are in- c) d) distinguishable in all their parameters: frequency, direc- P XP P XPtion, and polarization. The quantum state of the optical 2 -200 24 6 -2 2 2 -200 24 6 -2 2 mode into which the photon pairs are emitted exhibits X X squeezing [Fig. 2(a)]. -2 246 -2 246 Aside from being an interesting physical entity by it- -2 -2 self, squeezed light has a variety of applications. One of the primary applications of single-mode squeezed light e) f) is in precision measurements of distances. Such mea- P P surements are typically done by means of interferometry. 2 XP 2 XPQuantum phase noise poses an ultimate limit to inter- -200 24 6 -2 2 -200 24 6 -2 2 ferometry, and the application of squeezing (in particu- X Dj X -2 246 -2 246 lar, the phase squeezed state discussed above) permits -2 -2 expanding this limit beyond the fundamental boundary defined by the SQL. For example, squeezing is employed in the new generation of gravitational wave detectors — FIG. 1. Wigner functions of certain single-oscillator states. GEO 600 in Europe and LIGO in the United States. a) Vacuum state. b) coherent state. c,d) Position- and momentum-squeezed vacuum states. e,f) Position- and momentum-squeezed coherent states with real amplitudes. B. Two-mode squeezed light Panels (b) and (f) show the phase uncertainties of the respec- tive states to emphasize the phase squeezing of state (f). In- sets show wavefunctions in the position and momentum bases. A state that is closely related to the single-oscillator squeezed vacuum in its theoretical description and ex- perimental procedures, but quite different in properties How can one generate optical squeezed states in exper- is the two-mode squeezed vacuum (TMSV), also known iment? Consider the state as the twin-beam state. As the name suggests, this is a state of not one, but two mechanical or electromagnetic s |ψi = |0i − √ |2i , (6) oscillators. We introduce this state by first analyzing the 2 tensor product |0i⊗|0i of vacuum states of the two oscil- where |0i and |2i are photon number (Fock) states and lators. In the position basis, its wavefunction [Fig. 3(a)], s is a real positive number. We assume s to be small, so 1 −X2/2 −X2/2 Ψ (X ,X ) = √ e a e b (8) the norm of state (6) is close to one. The mean value of 00 a b π 3 can be rewritten as both Alice’s and Bob’s observables:

1 2 2 2 2 1 2 2 ˜ −(Pa−Pb) /(4R ) −R (Pa+Pb) /4 Ψ (X ,X ) = √ e−(Xa−Xb) /4e−(Xa+Xb) /4. (9) ΨR(Pa,Pb) = √ e e . (11) 00 a b π π We see that for R > 1 Alice’s and Bob’s momenta are Here, X and X are the position observables of the √ a b anticorrelated, i.e. the variance of the sum (P + P )/ 2 two oscillators which are traditionally associated with a b is below the level expected from two vacuum states fictional experimentalists Alice and Bob. The meaning √ [Fig. 3(d)]. of Eq. (9) is that the observables (X − X )/ 2 and √ a b The two-mode squeezed vacuum does not imply (X +X )/ 2 have a Gaussian distribution with variance a b squeezing in each individual mode. On the contrary, 1/2. This is not surprising because in the double-vacuum Alice’s and Bob’s position and momentum observables state Alice’s and Bob’s position observables are uncorre- in TMSV obey a Gaussian probability distribution with lated and both of them have variance 1/2. The behavior variance of the momentum quadratures in this state is analogous to that of the position. 1 + R4 h∆X2i = h∆X2i = h∆P 2i = h∆P 2i = . (12) a b a b 4R2 a) XB PB that exceeds that of the vacuum state for any R 6= 1. 4 4 In other words, each mode of a TMSV considered indi- vidually is in the thermal state. With increasing R > 1, 2 2 the uncertainty of individual quadratures increases while XA PA that of the difference of Alice’s and Bob’s position observ- -4 -2 2 4 -4 -2 2 4 ables as well as the sum of their momentum observables -2 -2 decreases. In the extreme case of R → ∞, the wavefunctions of -4 -4 the two-modes squeezed state take the form

ΨR(Xa,Xb) ∝ δ(Xa − Xb) (13) b) XB PB Ψ˜ R(Pa,Pb) ∝ δ(Pa + Pb) (14) 4 4 Both Alice’s and Bob’s positions are completely uncer- 2 2 tain, but at the same time precisely equal, whereas the

XA PA momenta are uncertain but precisely opposite. This state -4 -2 2 4 -4 -2 2 4 is the basis of the famous quantum nonlocality paradox in -2 -2 its original formulation of Einstein, Podolsky and Rosen (EPR) [2]. EPR argued that by choosing to perform -4 -4 either a position or momentum measurement on her por- tion of the TMSV, Alice remotely prepares either a state with a certain position or one with a certain momen- FIG. 3. Wavefunctions (not Wigner functions!) of two-mode tum at Bob’s location. But according to the uncertainty states in the position (left) and momentum (right) bases. a) principle, certainty of position implies complete uncer- The two-mode vacuum state is uncorrelated in both bases. tainty of momentum, and vice versa. In other words, by b) The two-mode squeezed state with position observables choosing the setting of her measurement apparatus, Alice correlated, and momentum observables anticorrelated beyond can instantly and remotely, without any interaction, pre- the standard quantum limit. pare at Bob’s station one of two mutually incompatible physical realities. This apparent contradiction to basic The wavefunction of the two-mode squeezed vacuum principles of causality has lead EPR to challenge quan- state |TMSVRi is given by tum mechanics as complete description of physical reality and triggered a debate that continues to this day. 1 2 2 2 2 −(Xa+Xb) /(4R ) −R (Xa−Xb) /4 ΨR(Xa,Xb) = √ e e , Experimental realization of TMSV is largely similar π to that of single-mode squeezing. SPDC is the primary (10) method; however, in contrast to the single-mode case, where R, as previously, is the squeezing factor [Fig. 3(c)]. it is implemented in the non-degenerate configuration. In contrast to the double-vacuum, TMSV is an entangled The photons is each generated pair are emitted into two state, and Alice’s and Bob’s position observables are non- distinguishable modes that become carriers of the TMSV classically correlated thanks to that√ entanglement. For state [Fig. 2(b)]. R > 1, the variance of (Xa − Xb)/ 2 is less than 1/2, In order to understand how non-degenerate SPDC i.e. below the value for the double vacuum state. leads to squeezing, consider the two-mode state The wavefunction of TMSV in the momentum basis is obtained from Eq. (10) by means of Fourier transform by |Ψi = |0i ⊗ |0i + s |1i ⊗ |1i , (15) 4 i.e. a pair of photons has been emitted into Alice’s and If this evolution continues for time t, we will have Bob’s modes with amplitude s. Now if we evaluate the √ Xˆ(t) = Sˆ†(r)Xˆ(0)Sˆ(r) = Xˆ(0)e−r; (24a) variance of the observable (Xa − Xb)/ 2, we find Pˆ(t) = Sˆ†(r)Pˆ(0)Sˆ(r) = Pˆ(0)er, (24b) 1 1 1 h∆(X − X )2i = hΨ| (ˆa +a ˆ† − ˆb − ˆb†)2 |Ψi = − s, 2 a b 4 2 which corresponds to position squeezing by factor R = (16) er and corresponding momentum antisqueezing (Fig. 4). i.e. Alice’s and Bob’s position observables are correlated If the initial state is vacuum, the evolution will result akin to TMSV. A similar calculation shows anticorrela- in a squeezed vacuum state; coherent states will yield tion of Alice’s and Bob’s momentum observables. squeezed light [3]. Both the single-mode and two-mode squeezed vacuum As a self-check, we find the factor of quadrature squeez- states are valuable resources in quantum optical informa- ing in state (18), in analogy to Eq. (7): tion technology. TMSV, in particular, is useful for gen- s s erating heralded single photons and unconditional quan- h0|∆X2|0i 1/2 R = = ≈ 1 + r tum teleportation. h0|Sˆ†(r)∆X2Sˆ(r)|0i 1/2 − r which is in agreement with R = er for small r. II. SALIENT FEATURES OF SQUEEZED The corresponding transformation of the creation and STATES annihilation operators is given by aˆ(t) =a ˆ(0) cosh r − aˆ†(0) sinh r; (25a) A. The squeezing operator aˆ†(t) =a ˆ†(0) cosh r − aˆ(0) sinh r, (25b) We now proceed to a more rigorous mathematical de- known as Bogoliubov transformation. scription of squeezing. Single-mode squeezing occurs un- der the action of operator P Sˆ(ζ) = exp[(ζaˆ2 − ζ∗aˆ†2)/2], (17) where ζ = reiφ is the squeezing parameter, with r = ln R and φ being real numbers, upon the vacuum state. Phase φ determines the angle of the quadrature that is being squeezed. In the following, we assume this phase to be zero so ζ = r. Note that, for a small r, the squeezing operator (17) acting on the vacuum state, generates state √ ˆ 2 †2 S(r) |0i ≈ [1+(raˆ −raˆ )/2] |0i = |0i−(r/ 2) |2i , (18) X which is consistent with Eq. (6) for s = r. The action of the squeezing operator can be analyzed as fictitious evolution under Hamiltonian

2 † 2 Hˆ = i~α[ˆa − (ˆa ) ]/2 (19) ˆ for time t = r/α (so that Sˆ(r) = e−i(H/~)t). Analyzing this evolution in the Heisenberg picture, we use [ˆa, aˆ†] = 1 to find that i aˆ˙ = [H,ˆ aˆ] = −αaˆ† (20) ~ FIG. 4. Transformation of quadratures under the action of and the squeezing Hamiltonian (19) with α > 0. Grey areas show ˙ † examples of Wigner function transformations with r = αt = aˆ = −αa.ˆ (21) ln 2. Now using the expressions for quadrature observables Two-mode squeezing is treated similarly. The two- √ √ Xˆ = (ˆa +a ˆ†)/ 2 and Pˆ = (ˆa − aˆ†)/ 2i, (22) mode squeezing operator is ˆ ˆ ∗ †ˆ† we rewrite Eqs. (20) and (21) as S2(ζ) = exp[(−ζaˆb + ζ aˆ b )]. (26) ˙ Xˆ = −αX; (23a) Assuming, again, a real ζ = r, introducing the fictitious Hamiltonian and recalling that the creation and annihila- ˙ Pˆ = αP. (23b) tion operators associated with different modes commute, 5 we find Decomposing the exponent in right-hand side of the above equation into the Taylor series with respect to α, aˆ(t) =a ˆ(0) cosh r + ˆb(0)† sinh r; (27a) we obtain ˆ ˆ † b(t) = b(0) cosh r +a ˆ(0) sinh r; (27b) ∞ r ∞ m X αn 2R X  1 − R2  α2m hn |sq i √ = . R n! 1 + R2 2(1 + R2) m! and hence n=0 m=0 (33) ˆ ˆ ˆ ˆ ±r Xa(t) ± Xb(t) = [Xa(0) ± Xb(0)]e ; (28a) Because this equality must hold for any real α, each term ∓r Pˆa(t) ± Pˆb(t) = [Pˆa(0) ± Pˆb(0)]e . (28b) of the sum in the left-hand side must equal its counter- part in the right-hand side that contains the same power Initially, Alice’s and Bob’s modes are in vacuum states, of α. Hence n = 2m and and the quadrature observables in these modes are un- r m p correlated. But as the time progresses, Alice’s and Bob’s 2R  1 − R2  (2m)! h2m |sq i = . (34) position observables become correlated while the momen- R 1 + R2 2(1 + R2) m! tum observables become anticorrelated. Since R = er, we have

2R 1 1 − R2 B. Photon number statistics = and = − tanh r, (35) 1 + R2 cosh r 1 + R2

An important component in the theoretical descrip- so Eq. (34) can be rewritten as tion of squeezed light is its decomposition in the photon number basis, i.e. calculating the quantities hn |sqRi for ∞ p 1 X m (2m)! the single-mode squeezed state and hmn |TMSVRi for the |sq i = √ (− tanh r) |2mi . (36) R 2mm! two-mode state. Due to non-commutativity of the pho- cosh r m=0 ton creation and annihilation operators, this calculation turns out surprisingly difficult even for basic squeezed We stop here for a brief discussion. First, we note that vacuum states, let alone squeezed coherent states and that for r  1, Eq. (36) becomes the states that have been affected by losses. Possible √ 2 approaches to this calculation include the disentangling |sqRi = |0i − (r/ 2) |2i + O(r ), (37) theorem for SU(1,1) Lie algebra [4], direct calculation of the wavefunction overlap in the position space [5] or consistently with Eq. (18). Second, note that the transformation of the squeezing operator [6]. Here we squeezed vacuum state (36) contains only terms with derive the photon number statistics of single- and two- even photon numbers. This is a fundamental feature of mode squeezed vacuum states by calculating their inner this state; in fact, one of the earlier names for squeezed product with coherent states. states has been “two-photon coherent states” [7]. This The wavefunction of a coherent state with real ampli- feature follows from the nature of the squeezing operator tude α is (17): in its decomposition into the Taylor series with re-

√ spect to r, creation and annihilation operators occur only 1 −(X−α 2)2/2 ψα(X) = e , (29) in pairs. Pairwise emission of photons is also a part of π1/4 the physical nature of SPDC: due to energy conservation a pump photon can only split into two photons of half its so its inner product with the position squeezed state (3) energy. equals We now turn to finding the photon number decompo- sition of the two-mode squeezed state. We first notice, Z +∞ r 2 2R − R α2 1+R2 by looking at Eq. (26), that |RABi must only contain hα |sqRi = ψα(X)ψR(X)dX = 2 e . −∞ 1 + R terms with equal photon numbers in Alice’s and Bob’s (30) modes. This circumstance allows us to significantly sim- Now we recall that the coherent state is decomposed into plify the algebra. We proceed along the same route as the Fock basis according to outlined above, calculating the overlap of |TMSVRi with the tensor product |ααi of identical coherent states |αi ∞ n X 2 α in Alice’s and Bob’s channels using Eqs. (10) and (29): |αi = e−α /2 √ |ni , (31) n=0 n! hαα|TMSVRi so we have Z +∞ = ψα(Xa)ψα(Xb)ΨR(Xa,Xb)dXadXb −∞ ∞ n r 2 X α 2R 1−R α2 hn |sq i √ = e 2(1+R2) (32) 2R − 2 α2 R 1 + R2 = e 1+R2 . (38) n=0 n! 1 + R2 6

Decomposing the coherent states in the left-hand side into the Fock basis according to Eq. (31) and keeping only the terms with equal photon numbers, we have

∞ 2n 2 X α 2R − 1−R α2 hnn| TMSV i = e 1+R2 (39) R n! 1 + R2 n=0 Now writing the Taylor series for the right-rand side and using Eq. (35), we obtain

∞ X 1 |TMSV i = tanhn r |nni . (40) R cosh r n=0 FIG. 5. Experimentally reconstructed photon number statis- Similarly to the single-mode squeezing, it is easy to tics of the squeezed vacuum state. For low photon numbers, verify that the above result is consistent with state (15) the even terms are greater than the odd terms due to pair- wise production of photons, albeit the odd term contribution for small r. On the other hand, in contrast to the single- is nonzero due to loss. Reproduced from Ref. [10]. mode case, the energy spectrum of TMSV follows Boltz- mann distribution with mean photon number in each position-squeezed 2 vacuum mode hni = sinh r. This is in agreement with our ear- 0 ˆ aˆ b(0) atˆ() btˆ¢() fictitious lier observation that Alice’s and Bob’s portions of TMSV input 2-mode two-mode input vacuum squeezer squeezed considered independently of their counterpart are in the vacuum vacuum atˆ¢() bˆ0 aˆ(0) btˆ() thermal state, i.e. the state whose photon number distri- momentum-squeezed bution obeys Boltzmann statistics with the temperature vacuum given by e−~ω/kT = tanh r. While the present analysis is limited to pure squeezed FIG. 6. Interconversion of the two-mode squeezed vacuum vacuum states, photon number decompositions of and two single-mode squeezed vacuum states. Dashed lines show a fictitious beam splitter transformation of a pair of squeezed coherent states and squeezed states that have vacuum states such that the modesa ˆ0(t), ˆb0(t) are explicitly undergone losses can be found in the literature [8, 9]. In single-mode squeezed with respect to modesa ˆ00, ˆb00. contrast to pure squeezed vacuum states, these decom- positions have nonzero terms associated to non-paired photons. The origin of these terms is easily understood. In accordance with the definition (22) of quadrature If a one- or two-mode squeezed vacuum state experiences observables, Eqs. (41) apply in the same way to the po- a loss, it may happen that one of the photons in a pair sition and momentum of the input and output modes. is lost while the other one remains. If the squeezing op- Applying this to Eqs. (28), we find erator acts on a coherent state, the odd photon number √ terms will appear in the resulting state because they are ˆ 0 ˆ ˆ Xa,b = [Xa(t) ∓ Xb(t)]/ 2 present initially. √ ∓r ˆ ˆ Photon statistics of both classes of squeezed states have = e [Xa(0) ∓ Xb(0)]/ 2 (42) been tested experimentally, as discussed in Section III for the output positions and below. An example is shown in Fig. 5. √ ˆ0 ˆ ˆ Pa,b = [Pa(t) ∓ Pb(t)]/ 2 √ ±r ˆ ˆ C. Interconversion between single- and two-mode = e [Pa(0) ∓ Pb(0)]/ 2 (43) squeezing for the momenta. In order to understand what state this corresponds to, let us assume, for the sake of the If the modes of the TMSV are overlapped on a symmet- argument, that vacuum modesa ˆ and ˆb at the SPDC input ric beam splitter, two unentangled single-mode vacuum have been obtained from another pair of modes by means states will emerge in the output (Fig. 6). To see this, we of another symmetric beam splitter: recall the beam splitter transformation √ aˆ0 = [ˆa(0) − ˆb(0)]/ 2 (44) aˆ0 = τaˆ − ρˆb; (41a) √ ˆb0 = [ˆa(0) + ˆb(0)]/ 2. (45) ˆb0 = τˆb + ρa,ˆ (41b) ˆ where τ and ρ are the beam splitter amplitude trans- Of course, since modesa ˆ(0) and b(0) are in the vacuum state, so area ˆ0 and ˆb0. We then have: missivity and reflectivity, respectively.√ For a symmetric beam splitter, τ = ρ = 1/ 2. In writing Eqs. (41), we Xˆ 0 = e∓rXˆ 0 ; neglected possible phase shifts that may be applied to a,b a,b ˆ0 ±r ˆ0 individual input and output modes [5]. Pa,b = e Pa,b, (46) 7 where superscript 0 associates the quadrature with E. Effect of losses modesa ˆ0 and ˆb0. We see that modesa ˆ0 and ˆb0 are re- 0 ˆ0 lated to vacuum modesa ˆ and b by means of position Squeezed states that occur in practical experiments and momentum squeezing transformations, respectively. necessarily suffer from losses present in sources, trans- Because the beam-splitter transformation is reversible, mission channels and detectors. In order to understand it can also be used to obtain a TMSV from two single- the effect of propagation losses on a single-mode squeezed mode squeezed vacuum states with squeezing in orthogo- vacuum state, we can use the model in which a lossy op- nal quadratures. This technique has been used, for exam- tical element with transmission T is replaced by a beam ple, in the experiment on continuous-variable quantum splitter (Fig. 8). At the other input port of the beam teleportation [11]. splitter there is a vacuum state. The interference of the signal modea ˆ with the vacuum modev ˆ will produce a mode with operatora ˆ0 = τaˆ − ρvˆ (with τ 2 = T and D. Squeezed vacuum and squeezed light ρ2 = 1−T being the beam splitter transmissivity and re- flectivity) in the beam splitter output. Accordingly, we Squeezed vacuum and bright squeezed light are readily have converted between each other by means of the phase- space displacement operator [5], whose action in the Xˆθ,out = τXˆa,θ − ρXˆv,θ. (52) Heisenberg picture can be written as

† † † Because the quadrature observable of the signal and vac- Dˆ (α)ˆa Dˆ(α) =a ˆ + α. (47) 2 uum states are uncorrelated, and since ∆(Xv,θ) = 1/2, it follows that This means, in particular, that the position and momen- tum transform according to ∆X2 = τ 2 ∆(X )2 + ρ2 ∆(X )2 √ θ,out a,θ v,θ Xˆ 7→ Xˆ + Re α 2; (48) = T ∆(X )2 + (1 − T )/2. (53) √ a,θ Pˆ 7→ Pˆ + Im α 2, (49) Analyzing Eqs. (41) we see that the optical loss alone, no so, under the action of Dˆ(α), the entire phase space dis- matter how significant, cannot eliminate the property of places itself, thereby changing the coherent amplitude of squeezing completely. the squeezed state without changing the degree of squeez- ing. ˆ aˆ aout aˆ aˆ -rb signal output

low-reflectivity vˆ vacuum beam splitter b bˆ FIG. 8. The beam splitter model of loss.

FIG. 7. Implementation of phase-space displacement. ρ  1 is the beam splitter’s amplitude reflectivity. Ideal squeezed-vacuum and coherent states have the minimum-uncertainty property: the product of uncer- 2 2 Phase-space displacement can be implemented experi- tainties ∆Xout ∆Pout reaches the theoretical mini- mentally by overlapping the signal state with a strong mum of 1/4. But this is no longer the case in the presence coherent state |βi on a low-reflectivity beam splitter of losses. The deviation of the uncertainty from the min- (Fig. 7). Applying the beam splitter transformation (41), imum can be used to estimate the preparation quality of we find for the signal mode a squeezed state. Suppose a measurement of a squeezed state yielded the minimum and maximum quadrature un- D E 0 2 aˆ = τaˆ − ρˆb (50) certainty values of ∆Xθ,out corresponding to θ = 0 and θ = π/2, respectively. One can assume that the state Given that mode ˆb is in a coherent state (i.e. an eignes- has been obtained from an ideal (minimum-uncertainty) tate of ˆb) and that ρ  1 (i.e. τ ∼ 1), we have squeezed state with squeezing R by means of loss chan- nel with transmissivity T . Using Eq. (5), one can write aˆ0 =a ˆ − ρβ (51) Eq. (53) for θ = 0 and θ = π/2 and solve the obtained system of equations for T and R. These values can then in analogy to Eq. (47). The displacement operation has be compared with those expected from the setup at hand been used to change the amplitude of squeezed light in in order to find out if any unexpected losses are present many experiments, for example, in Ref. [12]. in it [13]. 8

III. DETECTION The detector’s output current is then proportional to the intensity difference A. Balanced homodyne detection ∗ ∗ I−(t) ∝ a1(t) a1(t) − a2(t) a2(t) = a(t)a∗ (t)e−iθ + a∗(t)a (t)eiθ. (55) In order to detect squeezing, we need to perform mul- LO LO tiple measurements of the field quadrature, i.e. the√ ob- Assuming that aLO(t) is real (this is a matter of choosing −iθ iθ † servable Xˆθ = Xˆ cos θ + Pˆ sin θ = [e Aˆ + e Aˆ ]/ 2, the zero phase reference point), the quantity in the right- where Aˆ is the annihilation operator of the mode of in- hand side of Eq. (55) is an instantaneous (or, rather, terest. The task of measuring optical fields in a phase- averaged over the detection electronics’ response time) sensitive fashion may appear daunting, as these fields os- value of the classical quadrature a(t)e−iθ + a∗(t)eiθ. cillate at frequencies on a scale of hundreds of terahertz. Switching to quantum treatment, we replace the clas- Fortunately however, such a measurement can be imple- sical amplitude a(t) by operatora ˆ(t). This operator is mented using a relatively simple interference setup. The defined as technique known as balanced homodyne detection, pro- Z +∞ 1 −i(ω−Ω)t posed in 1983 by Yuen and Chan [14] and subsequently aˆ(t) = √ aˆωe dω, (56) implemented by Abbas et al., [15] to this day remains 2π −∞ the method of choice for quadrature measurements. Ref- wherea ˆω is the annihilation operator of a plane wave erence [16] provides a review of the current state of the mode of optical frequency ω familiar from the electro- art in this area. magnetic field quantization procedure. One can think of Here I start with a brief overview of this technique, in aˆ(t) as the annihilation operator of a photon of frequency the way it is presented in most textbooks. Subsequently, Ω at time moment t. Such a description is of course un- I will discuss a more complex but important question physical because of the time-frequency uncertainty prin- of identifying the temporal mode whose quadrature is ciple; yet sometimes it turns out useful for visualization. being measured. For simplicity, I will start in the classical As to the local oscillator, we recall that it is in a high- language. amplitude coherent state so the relative quantum noise of its amplitude is negligible. Hence we can continue to treat the LO amplitude a (t) as a number, not an dt LO ò operator. Equation (55) simplifies to

−iθ † iθ Iˆ−(t) ∝ αLO(t)[ˆa(t)e +a ˆ (t)e ]. (57) aaˆˆ+ aˆ = LO s 1 aaˆˆ- 2 ˆ LO s There are two primary approaches to the acquisition a2 = signal 2 and analysis of the subtraction photocurrent of the homo- aˆ s symmetric dyne detector. In time-domain analysis, the photocur- local beam splitter rent is measured using a time-resolving device, such as oscillator aˆ piezo an oscilloscope. In frequency-domain measurements, one LO instead looks at the electronic spectrum of the photocur- rent. FIG. 9. The principle of balanced homodyne detection.

Suppose the field to be measured (referred to as sig- B. Time-domain approach nal) is centered at frequency Ω. We write for the field −iΩt ∗ iΩt magnitude E(t) ∝ a(t)e + a (t)e , where a(t) is In the time-domain approach, the goal is to measure slowly varying. For quadrature measurement, this field the quadrature of a limited duration temporal mode de- is overlapped on a symmetric (50:50) beam splitter with fined by annihilation operator a strong laser field at frequency Ω, known as the lo- +∞ cal oscillator (LO), with amplitude aLO(t): ELO(t) ∝ Z −iΩt+iθ ∗ iΩt−iθ Aˆ = ϕ(t)ˆa(t)dt, (58) aLO(t)e + aLO(t)e . The phase θ of the local oscillator is controlled, e.g. by a piezoelectric transducer. −∞ The two beam splitter output fields impinge onto two where ϕ(·) is some normalized real function of bounded photodiodes whose output photocurrents are electroni- support. As follows from Eq. (57), this measurement can cally subtracted (Fig. 9). be realized by multiplying the subtraction photocurrent, In order to see how the subtracted photocurrent in the obtained from the homodyne detector with a constant detector output relates to the signal quadrature, we write LO, by the mode function and integrating it over time: the amplitudes of the beam splitter outputs as Z +∞ √ ˆ ˆ −iθ ˆ† iθ ˆ iθ ϕ(t)I(t)dt ∝ aLO(Ae + A e ) = 2aLOXθ. aLO(t)e ± a(t) −∞ a1,2(t) = √ . (54) 2 (59) 9

This approach works if the temporal resolution of the ac- a) b) quisition electronics (typically on a scale of nanoseconds) is fast compared to the duration of the mode of interest,

[a.u.] such as in Ref. [17]. [a.u.]

spectral power spectral power The opposite extreme that frequently occurs in exper- 1 2 3 4 5 12345 imental practice is that the squeezed state is prepared frequency (MHz) frequency (MHz) using a pico- or femtosecond pulsed laser, and its tempo- ral mode is defined by the laser pulse. In this case, the quadrature measurement can be accomplished in spite of lack of resolution at the electronic level by using the same

photocurrent [a.u.]

photocurrent [a.u.] laser as the local oscillator. We then have aLO(t) ∝ ϕ(t) +∞ 2 4 6 8 10 246810 R ˆ ˆ time [m s] time [m s] and hence −∞ I(t)dt ∝ Xθ. Because of the slow elec- tronics’ response, the integration occurs in this setting c) 10 automatically. The output of the homodyne detector is 8 an electrical pulse whose shape is determined by the re- [dB] sponse function, and magnitude is proportional to the 6 quadrature [18]. 4 Time-domain homodyne detection permits full recon- struction of the state in the acquisition mode. By vary- 2 ing the local oscillator phase θ, one can obtain noise 0 (a) statistics for all quadratures. Probability distributions -2 pr(Xθ) = hXθ| ρˆ| Xθi for all phase angles are sufficient -4 (b) to obtain full information about the density operatorρ ˆ Power Relative to the SNL of the signal state, such as its Wigner function or the -6 photon number representation. This method of measur- -8 (c) ing the quantum state of light is referred to as optical -10 homodyne tomography [5, 16]. 2 3 4 5 10 10 Frequency[Hz]10 10 Homodyne tomography was first proposed in 1989 [19] and implemented experimentally in application to single- FIG. 10. Frequency-domain approach to homodyne mea- mode squeezed vacuum in 1993 [20] and to two-mode surements. a) Simulated output of a homodyne detector ex- squeezed vacuum in 2000 [21]. hibiting noise corresponding to the SQL (bottom). Its spec- trum (top) is flat. b) The same output affected by the drift of the zero point on a time scale of 2µs. Direct variance mea- C. Frequency-domain approach surement of the photocurrent will not give SQL. However, the spectral power remains at the SQL level for frequencies Theoretically, if squeezing is generated in a continuous above 1 MHz and hence allows observation of squeezing. c) Bona fide spectrum of the homodyne detector from Ref. [22] nonlinear process, it could be observed by measuring the showing similar behavior. Curve a corresponds to SQL, b to variance of the homodyne detector output photocurrent squeezed vacuum and c to the detector’s electronic noise (in as a function of the local oscillator phase. In practice, the absence of LO). The noise peaks arise from harmonics of however, this measurement is obscured by various spuri- 50 Hz power line and the phase locking signal (20 kHz). ous noises produced by either the source or the detector. For example, the reflectivity of the homodyne detector’s beam splitter can vary as a function of time due to minute nical noises, by only looking at those sidebands in which perturbations to its orientation. However small such vari- they do not appear [Fig. 10(c)]. A further advantage ation may be, it may affect precise subtraction of the LO of the frequency-domain method is that, by measuring amplitudes. As a result, the mean value of the output the quantum noise at different sidebands, one is able to photocurrent will drift with time, and the drift ampli- analyze the properties of the source and detector; in par- tude can exceed the shot noise level, thereby obscuring ticular, measure the spectral band in which the squeezing the observation of quantum noise (Fig. 10). is present. Fortunately, such technical (classical) noises of the We start our theoretical analysis of frequency-domain photocurrent can be distinguished from the quantum measurements by find the Fourier transform of the pho- noise by analyzing their spectral behavior. Technical tocurrent (57) using Eq. (56): noises often occur within specific frequency bands; for example, the slow drift of the zero point is limited to Z +∞ low frequencies [Fig. 10(b)]. The quantum noise, on the ˆ˜ 1 ˆ iνt −iθ † iθ I(ν) = √ I(t)e dt ∝ aˆΩ+ν e +a ˆΩ−ν e , other hand, is “white”: it is constant for all frequencies 2π −∞ within the detector’s bandwidth [Fig. 10(a)]. One can (60) therefore observe squeezing, even in the presence of tech- ehre ν is the electronic frequency. By some algebra, we 10 can express the right-hand side of this equation as −iθ † iθ aˆΩ+ν e +a ˆΩ−ν e (61)

= Xˆ+ cos θ + Pˆ+ sin θ − iXˆ− sin θ + iPˆ− cos θ ˆ ˆ = X+,θ + iX−,π/2+θ where 1 Xˆ±,θ = √ [Xˆθ(Ω + ν) ± Xˆθ(Ω − ν)]. (62) 2 This means that measuring the real and imaginary parts ˜ of Iˆ(ν) is equivalent to subjecting frequency modes Ω+ν and Ω − ν to a beam splitter operation and performing homodyne measurements of the beam splitter outputs at LO phases θ and π/2 + θ [23]. Suppose that the state entering the homodyne detector is squeezed — that is, the noise of Iˆ(t) is below the stan- dard quantum limit for a certain local oscillator phase θ. But this would also imply that its Fourier transform ˜ — both the real and imaginary parts of Iˆ(ν) — exhibit FIG. 11. Optical homodyne measurements of coherent and fluctuations below SQL. This would in turn mean that squeezed states of light. Top to bottom: coherent state, phase ˆ ˆ (momentum) squeezed light, 45◦ quadrature squeezed light, observables X+,θ and X−,π/2+θ exhibit reduced variance at the same time — that is, modes Ω + ν and Ω − ν amplitude (position) squeezed light, momentum squeezed vac- uum. Left column: statistics of quadrature measurements are in the two-mode squeezed state (cf. Sec. II C). In obtained with a balanced homodyne detector while the local other words, single-mode squeezing in the time domain oscillator phase is varied. Middle column: histograms pr(Xθ) is equivalent to two-mode squeezing in the frequency do- of these measurements associated with specific values θ of main. An explicit experimental demonstration to that the phase. These histograms are integral projections of the effect has been presented by Huntington et al. [24]. Wigner functions onto vertical planes positioned at angle θ Simultaneous measurements of the real and imaginary with respect to the position axis of the phase space. The his- ˜ parts of Iˆ(ν) are possible using lock-in amplifiers. In this tograms are used to reconstruct the Wigner fucntions (right column) of the corresponding states in a procedure similar way, one can perform full quantum-state tomography of † √ to computer tomography scanning in medicine. They were the modes defined by operators (ˆaΩ+ν ± aˆΩ−ν )/ 2. In also used to reconstruct the states’ density matrices in the a classic work of 1997, Breitanbach et al. [10] used this photon number basis (Fig. 5) by means of the quantum state approach for tomography of an extended family of Gaus- sampling method [5]. Reproduced from Ref. [10]. sian states, including coherent, squeezed vacuum, as well as amplitude and phase squeezed light states (Fig. 11). It is common to use the electronic spectrum analyzer IV. PREPARATION rather than lock-in amplifiers for frequency-domain mea- surements. The spectrum analyzer displays the mean squared power of the photocurrent’s sideband: In Sec. II A we had a conceptual discussion of a Hamil-  2 D E tonian that squeezes the phase space. But in fact, al- ˆ˜ ˆ 2 ˆ 2 most any Hamiltonian that is at least quadratic in the I(ν) ∝ X+,θ + X−,π/2+θ . (63) creation and annihilation operators brings about sophis- ˆ ˆ ticated trajectories in the phase space and can result in Because observables X+,θ and X−,π/2+θ are simultane- ously squeezed, the spectrum analyzer will show reduced squeezing. Similarly, a Hamiltonian that is bilinear in the signal at frequency ν. In this way, the spectrum analyzer creation and annihilation operators of two modes is likely can measure squeezing in spite of being unable to resolve to generate two-mode squeezing. Accordingly, there exist the two terms of Eq. (61). This result is consistent with many physical processes that can be employed to prepare the common sense expectation: if the time-dependent single- and two-mode squeezed states of the electromag- photocurrent I(t) exhibits reduced noise, so will its fre- netic field. quency spectrum. An important limitation to the above is the require- An important shortcoming of the spectrum analyzer is ment that the Hamiltonian evolution that leads to that it does not enable quantum state tomography. It squeezing be not compromised by competing non-unitary provides information about the variance of the quadra- processes that increase noise. For example, attempts to ture probability distribution, but not the probability dis- achieve squeezing in atomic systems have for a long time tribution itself. This does not matter, however, if the met with limited success due to incoherent spontaneous variance (i.e. the amount of squeezing) is the only quan- emission into the signal mode, which leads to thermal- tity of interest, and state reconstruction is not the goal. ization of the signal state and loss of squeezing. 11

Most frequently, squeezing is obtained by nonlinear op- with tical wave mixing processes, in which pairs of photons are χeff Ω emitted into degenerate (single-mode squeezing) or non- r = |ap|L. (67) degenerate (two-mode squeezing) modes. An example is nc spontaneous parametric down-conversion, a three-wave We now quantize the signal and idler field according to mixing between the pump field and the two photons of p εs,i → ~Ω/20V aˆΩ±ν (where V is the quantization vol- squeezed vacuum that occurs due to second-order opti- ume), but continue to treat the macroscopic pump field cal nonlinearity. A related method is four-wave mixing, a as classical. This leads to third-order nonlinear process in which two strong waves, interacting with a nonlinear medium, give rise to a pair † aˆΩ±ν (L) =a ˆΩ±ν (0) cosh r +a ˆ (0) sinh r; (68) of photons. Let us discuss these two processes in more Ω∓ν detail. which is identical to Eqs. (27). In other words, if the signal and idler fields of frequencies Ω ± ν before the crystal are in the vacuum state, they will be in a two- A. Via parametric down-conversion mode squeezed state after the crystal. As discussed in the previous section, such a state man- In order to mathematically describe nonlinear-optical ifests itself as single-mode squeezing when a homodyne squeezing, we begin with equations for the propagation measurement with the local oscillator tuned to frequency of classical electromagnetic fields through a nonlinear Ω is performed. To see this, consider a time-domain mode medium. We then quantize the fields and replace their whose annihilation operator is given by Eq. (58). Using amplitudes with corresponding creation and annihila- Eq. (56), we rewrite the mode operator as tion operators, thereby obtaining their evolution in the Z +∞ Heisenberg picture. Aˆ = aˆωϕ˜(ω − Ω)dω Consider a three-wave mixing process in which a strong −∞ pump field of frequency 2Ω interacts with weak signal +∞ and idler fields of frequencies Ω ± ν, respectively, with Z = [ˆaΩ+ν ϕ˜(ν) +a ˆΩ−ν ϕ˜(−ν)]dν, (69) ν  Ω, in a crystal with effective nonlinearity χeff . All 3 0 fields are continuous in time, but the amplitudes εs(z) √ and εs(z) of the signal and idler change with the propa- R +∞ iνt whereϕ ˜(ν) = (1/ 2π) −∞ ϕ(t)e dt is the Fourier im- gation distance z due to the nonlinear interaction. The ∗ pump amplitude ε is assumed to remain constant be- age of ϕ(t). Because ϕ(t) is real, we haveϕ ˜(ν) =ϕ ˜ (−ν). p Using Eq. (68), we find cause εp  εs, εi, so there is no depletion. We further assume that the crystal is perfectly phase matched for +∞ this nonlinear process. Z Aˆ(L) = cosh r[ˆa (0)ϕ ˜∗(ν) +a ˆ (0)ϕ ˜(ν)] In the slowly-varying envelope approximation [25], the Ω+ν Ω−ν equations of motion for the signal and idler fields take 0 † † ∗ the form + sinh r[ˆaΩ+ν (0)ϕ ˜(ν) +a ˆΩ−ν (0)ϕ ˜ (ν)]dν ∂ (Ω ± ν) = Aˆ(0) cosh r + Aˆ†(0) sinh r, (70) εs,i(z) = i PNL(Ω ± ν), (64) ∂z 20nc i.e. single-mode squeezing. The above derivation is valid where n the refractive index and the nonlinear polariza- only if ϕ(t) is sufficiently slowly varying so that its spec- tion amplitude is given by trumϕ ˜(ν) takes on significant values only at such fre- quencies that two-mode squeezing of operatorsa ˆΩ±ν is ∗ present. In practice, this limitation is established by PNL(Ω ± ν) = 2ε0χeff εpεi,s(z), (65) the nonlinear crystal’s phase-matching bandwidth (for χeff being the effective nonlinear succeptibility. With- single-pass squeezing) or the cavity linewidth (for squeez- out loss of generality, we can define the phase of the ing in a cavity). pump such that iap is real and positive. Then, solving One can readily estimate the amount of squeezing one these equations for propagation length L under assump- can obtain. Consider, for example, a L = 5 mm peri- tion ν  Ω, we find odically poled KTP crystal with the signal wavelength of λ = 780 nm and the pump field of power P = 100 ∗ εs(L) = εs(0) cosh r + εi (0) sinh r; (66a) mW focused into a spot of w = 50 µm radius. The rele- ∗ εi(L) = εi(0) cosh r + εs(0) sinh r (66b) vant effective nonlinear coefficient of PPKTP is χeff = 14 pm/V, refractive index n = 1.8. Under these conditions, the pump intensity is Ip = P/πw2 = 1.3 × 107 W/m2 and the field amplitude 3 ikz−iωt p 4 The amplitude is defined according to E(z, t) = ε(z)e + |εp| = Ip/2n0c = 3.6 × 10 V/m. Substituting this c.c., for E(z, t) being the value of the field in space in time. value into Eq. (67), we find r = 1.1 × 10−2. We see that 12 the amount of squeezing obtained by a single pass of a The squeezing is strongest at the threshold point, when continuous-wave pump laser through a nonlinear crystal the amplification in a single pass through the nonlinear of a reasonable size is very small. crystal is equal to the loss occurring in a roundtrip of There are two primary methods of addressing this com- the signal through the cavity, including that at the out- plication. First, one could use an ultrashort pulsed laser, put coupling mirror. The intensity gain factor equals thereby greatly increasing the pump amplitude. The e2r ≈ 1+2r for r  1. In the numerical example studied above theory, developed for continuous-wave pump, has above, 2r = 0.022, so in order to be at the threshold, only limited application for pulsed pump; the amount of the cavity must have the same roundtrip loss. This loss squeezing strongly depends on the shape ϕ(t) of the tem- occurs due to the transmission through the output cou- poral mode chosen for the measurement [13]. Neverthe- pling mirror as well as spurious losses on all other optical less, squeezing has been demonstrated in the single-pass elements inside the cavity. Assuming, for example, that pulsed regime as soon as one year after the first experi- the mirror has a transmissivity of 0.017, and the spu- mental observation of squeezed light [26] and the degree rious losses add up to 0.005, we find for the quantum of squeezing has been increased to several decibels4 in efficiency η = 0.015/(0.017 + 0.005) = 0.77, which means subsequent years [27]. that at the threshold, for ν  γ, we will see a variance The second approach is to place the crystal inside a of V − ≈ 1/2 − η/2, or about 6 dB. Fabry-Perot cavity. The cavity can be resonant to the Let us now estimate the bandwidth within which the pump light, thereby enhancing the effective pump power, squeezing is generated. This bandwidth is the same as or to the signal, effectively allowing multiple passing of the cavity linewidth γ, which, in turn, is the ratio of the the signal through the crystal, or both. The case when cavity’s free spectral range and finesse. Assuming that the cavity is resonant to the signal is most common; this the cavity is of a bow-tie configuration (Fig. 12) with configuration is referred to as the optical parametric os- a full length of Lc = 30 cm, its free spectral range is cillator or amplifier (OPO/OPA). A theory of squeez- c/L = 1 GHz. The finesse is π/T ≈ 160, so γ ≈ 6MHz. ing inside an OPA has been developed by Gardiner and Historically, the first observation of squeezing using an Savage [28] and reviewed, for example, in [8]. Without OPA cavity has been achieved by Wu et al. in 1986 [29]. derivation, we present the result for the quadrature noise The squeezing reached in that experiment was about 3 levels associated with the antisqueezed (θ+ = π/2) and decibels. Since then, many groups made efforts to fur- squeezed (θ− = 0) quadratures: ther develop this approach. One of the most recent re- sults reported a squeezing of 12.7 dB [30]. This remark- p 1 2 P/Pth able achievement required the overall quantum efficiency V ±(ν) = ± η , (71) 2 p 2 2 (ν/γ) + (1 ∓ P/Pth) (including that of the OPA cavity, homodyne detection, mode matching, etc.) to approach 95%. where 2γ is the cavity linewidth, η is the overall quantum OPAs can as well be used successfully to generate two- efficiency, P is the pump power and Pth is the thresh- mode squeezing. The first experiment to that effect was old power, i.e. the pump power at which the nonlinear reported by Ou et al. in 1992 [31]. In that work, the process in the cavity leads to macroscopic optical oscilla- signal and idler fields resonated in the cavity were of the tions. By analyzing this result, we see that the squeezing same frequency, but different polarizations. occurs at sideband frequencies ν less than or on the or- der of the cavity linewidth. This is not surprising: the enhancement effect of the cavity is only present within B. In atomic ensembles its resonance.

nonlinear As mentioned above, high optical nonlinearity is at the crystal pump heart of most squeezing processes. An atom interacting with an optical wave resonant with one of its transitions is an intrinsically nonlinear object. Atoms begin to ex- resonated signal output hibit nonlinear optical properties at intensity levels on a output coupler field squeezed vacuum scale of the saturation intensity, which is many orders of magnitude lower than the intensity levels required for sig- FIG. 12. Squeezing in an OPA cavity. The cavity mirrors nificant nonlinear effects in ferroelectric crystals. There- are reflective to the signal field, but transparent to the pump. fore atomic ensembles have been considered an attractive medium for the preparation of squeezed optical states from early days of quantum optics. A typical mechanism that leads to the generation of 4 Decibel [dB] is a common unit of squeezing in experiment. squeezing is four-wave mixing (Fig. 13). Consider a The degree of squeezing in decibels is calculated according to 2 Λ-shaped atomic energy level configuration with two 10 log10(2 ∆X ). The standard quantum limit corresponds to a squeezing of 0 dB, the reduction of quadrature variance by a ground states coupled to a single excited state by op- factor of 2 to about 3 dB, factor of 4 to about 6 dB, factor of 10 tical transitions of degenerate or nondegenerate frequen- to 10 dB, etc. cies. This configuration is present, in particular, in alkali 13 atoms, where the ground level is split into two hyperfine generation. In recent years, however, atomic systems sublevels. have been revisited and significant squeezing has been demonstrated in experiments involving four-wave mixing [33, 34] and polarization self-rotation [35, 36].

|añ C. In fibers

pump Optical fibers are typically made of glass, an amor- Antistokes phous material with inversion symmetry. Accord- Stokes ingly, they normally possess no second-order nonlinear- ity. However, fibers enable propagation of focused optical pump wavepackets over long distances, so the effects of third- |cñ order (Kerr) nonlinearities on these wavepackets become D significant. One of these effects is squeezing. |bñ Squeezing in optical fiber is best explained in terms of the nonlinear refractive index. In a Kerr medium, the FIG. 13. Quantum four-wave mixing in an atomic Λ sys- refractive index depends on intensity I of the propagating tem leads to emission of the Stokes and anti-Stokes photons light according to akin to signal and idler in parametric down conversion. Two- mode squeezing obtains for non-degenerate and single-mode n = n + n I, (72) squeezing for degenerate ground states. 0 2

where n2 is related to the third-order nonlinear suscep- Suppose the atom is initially in ground state |bi. The tibility χ(3). The phase of light that has propagated pump field of frequency Ω excites the |ai − |bi transition, through such a material will then depend on the intensity, driving the atom into the other ground state |ci through resulting in the transformation of the Wigner function as Raman scattering, which results in emission of a Stokes illustrated in Fig. 14. The parts of the Wigner function photon of frequency Ω−∆. Level |ci is in turn excited by that are associated with higher and lower intensities be- the pump field and the atom goes back into |bi, accom- comes shifted in the phase space with respect to each panied by emission of an anti-Stokes photon of frequency other, resulting in squeezing. Ω + ∆. If the entire process is coherent, the Stokes and anti-Stokes emission modes will find themselves in the two-mode squeezed state. This mechanism was used in the very first observa- tion of optical squeezing by Slusher et al. [32]. In that experiment, atomic vapor of sodium has been used and two Fabry-Perot cavities resonant with the pump and the Stokes/Antistokes fields have been placed around the vapor sample for amplification. A two-mode squeezed state at frequencies Ω ± ∆ was observed using a homo- dyne detector with the local oscillator at frequency Ω. A squeezing of about 0.3 dB has been detected for cor- related quadratures, while the uncorrelated quadratures exhibited extra noise at a level of about 2 dB. The state observed by Slusher and co-workers did not approach the minimum-uncertainty limit. This is largely due to processes in atoms that occur concurrently to FIG. 14. Effect of a Kerr medium on a coherent state. Differ- four-wave mixing and lead to incoherent emission into ent intensities experience different refractive indices, resulting the signal modes, such as Brillouin and Raman scatter- in quadrature squeezing. From Ref. [37] ing. Further hindrance is presented by various dephas- ing phenomena such as time-of-flight decoherence that Homodyne detection of squeezing in this configuration inhibit coherent four-wave mixing. All these processes is complicated by the macroscopic mean amplitude of the contribute to the “thermalization” of the optical state in signal required to take advantage of Eq. (72). The ampli- the signal modes and degrade the squeezing. tude could, in principle, be eliminated by means of phase- This appears to be a common problem in experi- space displacement (see Sec. II D); however, this would ments using atomic ensembles for squeezing. This is the require a powerful laser and excellent phase stabilization. primary reason that ferroelectric crystals, rather than A more common detection method involves causing two atomic systems became the workhorse of squeezed light fiber squeezed fields to interfere with each other so the 14 resulting phase-space displacement and rotation makes states that lie outside the classical domain. Nonclassi- one of the resulting fields amplitude squeezed. Ampli- cal optical states cannot be achieved by linear-optical tude squeezing is then readily observed by measuring the manipulation: interference of coherent states necessar- intensity with a single high-efficiency detector and eval- ily leads to coherent states. Production of nonclassical uating the variance of the photocurrent noise. states therefore requires nonlinear optics. There are a number of ways such interference can be Parametric down-conversion is a nonlinear phe- implemented. For example, in a Sagnac-type interferom- nomenon capable of producing quantum states of light eter the initial laser pulse impinges on a beamsplitter, with high efficiencies and with well-defined spatiotem- after which the transmitted and reflected fields enter the poral properties. This property is unique among exist- fiber from two ends. Upon exiting the fiber, the fields in- ing methods of non-classical light generation (see, e.g. terfere on the same beam splitter, and one of the result- Ref. [45]). However, the only states that SPDC can pro- ing fields is measured [38]. Alternatively, a polarization- duce are the single- and dual-mode squeezed vacua. For maintaining fiber is used, so that the fields in both po- this reason, the past decade has seen extensive efforts to larizations become squeezed at the fiber output. These use these states as “primitives” to produce (“engineer”) fields are then brought into interference using waveplates various other states of light. As we see in this section, positioned at the output end of the fiber [39]. application of tools such as linear-optical manipulation, Squeezing in optical fibers is limited by phase noise interference with coherent states and conditional mea- associated with thermal fluctuations of the refractive in- surements allows one to accomplish this task successfully. dex, in particular guided acoustic wave Brilloun scatter- However small the degree of squeezing may be, even a sin- ing. An additional degrading factor, particularly signifi- gle squeezed resource permits producing a wide variety cant for very short pulses, is Raman scattering [38]. Both of complex optical states [16, 46]. these phenomena allow precise theoretical treatment, and A TMSV with a weak level of squeezing can be used to can be minimized by wise choice of experimental param- generate heralded single photons. To that end, one chan- eters [40, 41]. As a result, squeezing up to about 7 dB nel of that state (idler) is monitored by a single-photon has been obtained [41]. detector. If the detector “clicks”, we know, according to Eq. (40), that a photon must have been emitted into the other (signal) channel as well. If the squeezing parameter V. APPLICATIONS IN QUANTUM r is sufficiently small, the contribution of higher photon INFORMATION numbers in the signal channel can be neglected. In 2001, this technique was used to generate a her- Squeezed light is a primary resource in continuous- alded single photon in a definite spatiotemporal mode, variable5 quantum information processing. In addition characterize it using homodyne tomography and, for the to fundamental interest such as the implementation of first time, observe a negative Wigner function [47]. This the original EPR paradox, it is the basis of many basic method was later extended to generate and measure the applications such as universal quantum computing, dense two- [48] and three-photon [49, 50] states. In these exten- coding and quantum key distribution. The limited vol- sions, the idler channel of SPDC was split into multiple ume of this manuscript does not permit a comprehensive photon detectors, and their coincident “clicks” were re- review of these applications; such a review can be found, quired for a heralding event. for example, in Refs. [43, 44]. Here we will concentrate A modification of this scheme shown in Fig. 15 per- on only two important examples. mits producing arbitrary superpositions of photon num- ber states. Prior to detection, the light in the idler chan- nel is mixed with weak ancillary coherent states on beam A. Quantum-optical state engineering splitters. In this way, a detector registering a photon “does not know” whether it comes from SPDC or from Lasers generate coherent states and their statistical a coherent state. This indistinguishability results in the mixtures — the states of light known as classical. While idler channel of SPDC being effectively projected onto such states are useful for some applications, many emerg- a superposition of Fock states. Thanks to entanglement ing quantum technologies require a supply of optical of the TMSV, this superposition is automatically trans- ferred to the signal channel. The weight of each compo- nent of the superposition can be controlled by the am- plitudes and phases of the ancilla coherent states. This 5 The term ‘continuous-variable’ refers to optical quantum infor- technique has been demonstrated for superpositions of up mation protocols that involve manipulation of a state in phase to the two- [51] and three-photon [50] terms, but can, in space, i.e. displacement, squeezing, quadrature measurements, principle, be extended to higher numbers. One of the pos- etc. It is usually contrasted with ‘discrete-variable’ methods sible applications of this method is the implementation of dealing with manipulating and measuring single photons. This separation is largely of historical and technological nature; in the cubic phase gate for universal quantum computation fact, more and more interesting applications now arise at the in the continuous-variable setting. boundary between the two domains [16, 42]. In the above examples, a low magnitude of the squeez- 15

remote state preparation trigger detector SPCM SPCM trigger photon weak coherent state parametric down-conversion idler down- doubler converter |añ |bñ master laser weak coherent states low-reflectivity homodyne squeezed beam splitter signal detector vacuum state signal local oscillator

homodyne detection local oscillator FIG. 16. (a) Conditional preparation of the “odd Schr¨odinger FIG. 15. Generating arbitrary superpositions of the zero-, kitten” by applying photon annihilation to the squeezed vac- one- and two-photon states. The light in the idler channel of uum state (the “even Schr¨odingerkitten”). The dashed lines parametric down-conversion is brought into interference with shows the additional elements used to generate arbitrary su- two weak coherent states and subsequently detected by single- perpositions of states |αi and |−αi as in Ref. [58]. photon counting modules (SPCMs). A double “click” heralds the generation of the desired state in the signal channel. low-reflectivity beam splitter (Fig. 16). Detection of a photon in the reflected channel indicates that a photon ing parameter does not degrade the fidelity of engineered has been removed from the squeezed vacuum — that is, quantum states. Quite the contrary, it ensures that the a photon annihilation event has occurred [57]. state is not contaminated by higher photon number com- If a weak ancilla coherent state is injected into the ponents. However, low squeezing also reduces the proba- heralding detector using an additional beam splitter bility of the heralding event, which can make the method (Fig. 16), the heralding photon cannot be definitively unpractical. One must choose the degree of squeezing as traced back to the squeezed state or that ancilla. If the a compromise between the fidelity and the state produc- event comes from the squeezed state, the photon sub- tion rate. traction takes place and the signal output is the odd In the next example, in contrast, a non-negligible value Schr¨odingerkitten; if it comes from the coherent state, of squeezing is essential for obtaining the desired state — the signal output is the same as the input, i.e. the even a superposition |αi±|−αi of two coherent states of oppo- Schr¨odingerkitten. Because these two possibilities are site amplitudes. This state is of interest to the quantum indistinguishable, the output state becomes a coherent community because, while being a linear combination of superposition of the even and odd kittens, with the mag- classical states, it is highly nonclassical, and hence remi- nitude and phase of the terms in the superposition de- niscent of the famous “Schr¨odinger cat” Gedankenexper- pendent on the parameters of the ancilla. In this way, iment. arbitrary superpositions of states |αi and |−αi — an op- Remarkably, the squeezed vacuum is quite similar to tical continuous-variable qubit — are generated [58]. the state |αi + |−αi (“even Schr¨odinger kitten”) for α . 1. To see this, recall the Fock decomposition (31) of the coherent state. The sum of two coherent states of op- B. Continuous-variable quantum teleportation posite amplitudes will contain only even photon number terms, Teleportation is a quantum communication protocol in α2 which a quantum state is transferred between two loca- |αi + |−αi ∝ |0i + √ |2i + O(α4), (73) 2 tions without utilizing any direct quantum communica- tion channel. The transfer is enacted by local interference in the same way as the squeezed vacuum state (37). With of the signal state with a portion of the entangled re- a sufficiently small α, only the first two terms of these source shared between the two locations, as well as local decompositions are significant, and setting r = α2 makes measurements, classical communications and local quan- them mutually identical for the two states. tum operations. The teleportation protocol was first pro- Because coherent states are eigenstates of the photon posed for qubits in 1993 by Bennett et al. [59], and for annihilation operatora ˆ, applying that operator to |αi + continuous variables in 1994 by Vaidman [60]. The latter |−αi produces α(|αi − |−αi), i.e. an “odd Schr¨odinger protocol utilizes the two-mode squeezed vacuum as the kitten”. This idea was implemented experimentally by entangled resource; its major advantage is the principal Wenger et al. [52] and later refined in Refs. [53–56]. capability of complete transfer of a quantum state of an For photon annihilation, squeezed vacuum produced by optical mode, independent, in particular, of the number means of degenerate SPDC was transmitted through a of photons therein. 16

Figure 17 shows the scheme of the protocol. The A more rigorous argument can be presented in sender, Alice, has the signal state she wishes to teleport terms of Wigner functions. Let the initial Wigner in modea ˆ. In addition, she and the receiver, Bob, share function of the signal state be Wa(Xa,Pa). The a two-mode squeezed state in modes ˆb andc ˆ. In order Wigner function of the EPR state shared be- to perform teleportation, Alice overlaps modesa ˆ and ˆb tween Alice and Bob is Wbc(Xb,Pb,Xc,Pc) ∝ on a symmetric beam splitter and preforms position and δ(Xb − Xc)δ(Pb + Pc). The three-mode Wigner momentum measurements in its outputs using two ho- function is then Wabc(Xa,Pa,Xb,Pb,Xc,Pc) = modyne detectors. She then communicates the results of Wa(Xa,Pa)Wbc(Xb,Pb,Xc,Pc). After the beam splitter her measurement to Bob via a classical channel. Bob per- in Alice’s channel, it will transform into forms phase-space displacement of modec ˆ in accordance  0 0 0 0  0 0 0 0 0 Xa + Xb Pa + Pb W (X ,P ,X ,P ,Xc,Pc) ∝ Wa √ , √ with that information, after which the state of this mode abc a a b b 2 2 becomes identical to the initial state of modea ˆ.  0 0   0 0  −Xa + Xb −Pa + Pb ×δ √ − Xc δ √ + Pc , (74) Alice 2 2

Bob where the primed indices refer to the quadratures of the P¢ X ¢ 0 b a modes after the beam splitter. A measurement of Xa and 0 Pb will yield, in modec ˆ, ∞ bˆ¢ aˆ¢ ZZ 0 0 0 0 0 0 0 0 Wc(Xc,Pc) = Wabc(Xa,Pa,Xb,Pb,Xc,Pc)dPadXb −∞ ˆ √ √ aˆ b cˆ  0 0  ∝ Wa Xc + Xa 2,Pc + Pb 2 . (75) Again, applying displacement to Bob’s mode, we recover a state with the Wigner function equal to that of the EPR source initial signal — that is, the state identical to the initial. In experimental practice, the teleportation perfor- mance is degraded by a number of factors, of which the FIG. 17. The scheme of quantum teleportation. Operator primary ones are the optical losses, optical phase fluctu- Dˆ(X + iP ) denotes phase-space displacement. m m ations and imperfect squeezing of the TMSV resource. A variety of performance metrics has been proposed [43, 61– In order to visualize the physics of teleportation, let 64]. The most common one is the coherent-state fidelity, us think of the signal state as a point (Xa,Pa) in the which is the average, over all coherent states |αi, of the phase space (neglect the uncertainty principle for a mo- D E fidelity F = α Tˆ(|αihα|) α , where Tˆ(|αihα|) is the ment). Further, we assume the initial two-mode squeez- c ˆ density operator of the teleported state. For a perfect ing of modes b andc ˆ to be infinite: Xb = Xc and teleportation procedure, Fc = 1. On the other hand, Pb = −Pc, with both these quantities being completely uncertain. The beam splitter transformation, in accor- the best fidelity that can be achieved without the use dance with Eqs. (41) , makes the position in modea ˆ equal of entangled resource, simply by Alice’s measuring the √ position and momentum quadratures of the input state to X0 = (X − X )/ 2 while the momentum in mode ˆb a a b √ and Bob’s recreating a coherent state with the same cen- becomes P 0 = (P + P )/ 2. b a b tral position and momentum, is F = 1/2. The value of Suppose now that the position and momentum mea- c Fc reaching a value of 2/3, known as the no-cloning fi- surements of these modes are performed, producing some delity [63], guarantees that nobody else can have a better specific results X0 and P 0, respectively. This means that a b copy of the input state than Bob. For this reason, the ˆ the position of√ mode b prior to the beam splitter has been√ no-cloning fidelity is of relevance to continuous-variable 0 0 Xb = Xa − Xa 2 and its momentum Pb = −Pa + Pb 2. quantum communication. The value of 2/3 is also the ˆ Because of the infinite two-mode squeezing of modes√ b minimum required for obtaining teleported states with 0 andc ˆ this implies,√ in turn, that Xc = Xa − Xa 2 and negative values of the Wigner function. 0 Pc = Pa − Pb 2. The first continuous-variable quantum teleportation We see that, after Alice’s measurement, the position experiment was reported by Furusawa and colleagues and momentum of Bob’s mode become certain and re- in 1998 [11]. The TMSV resource has been ob- lated to those of the initial state. Furthermore, if Alice tained from two single-mode squeezed fields generated 0 0 communicates the observed values of Xa and Pb to Bob as counterpropagating modes in a single OPA cavity. (via a classical channel), Bob will be able to perform a Phase-space displacement was implemented using a low- phase-space displacement operation (see Sec. II D) on his transmissivity beamsplitter, with the amplitude and mode, obtaining the position and momentum equal to Xa phase of the displacement beam regulated by electro- and Pa, respectively, i.e. identical to those of the initial optical modulator. The resulting fidelity, Fc = 0.58, ex- signal state. ceeded the classical benchmark. 17

Thereafter, numerous efforts have been reported to re- a) fine the protocol and teleport increasingly complex quan- tum states. For example, Takei and colleagues [65] in freely suspended end mirrors 2005 demonstrated entanglement swapping (teleporta- tion of one channel of a TMSV state), which is an es- Michelson sential component of quantum repeaters. Yonezawa et interferometer aˆ¢ homodyne al. [66] teleported in 2007 a squeezed vacuum state and optical isolator detector obtained, for the first time, squeezing in the output. The bˆ¢¢ first teleportation of states with a negative Wigner func- tion, such as the single photon and “Schr¨odingerkit- bˆ¢ bˆ ten” [53] was implemented in 2011 [67]. This work was aˆ followed by unconditional high-fidelity teleportation of squeezing dual-rail single-photon qubits [68]. aˆ¢¢ source

VI. APPLICATIONS IN QUANTUM METROLOGY b) P P

Squeezed light can be useful in any task that requires precise evaluation of the optical phase. Such tasks occur, aj aj for example, in optical communications [69] and metrol- X X ogy [70]. Phase evaluation typically involves an interfer- ometer, and its precision is determined by the phase un- certainty of the fields used. The coherent state, which is c) readily obtained from lasers, has a phase uncertainty on a (i) (ii) scale of the inverse of its√ amplitude, or inverse square root Noise of its photon number 1/ N [Fig. 1(b)].However, employ- squeezing ing nonclassical states has a potential to improve the pre- cision up to the fundamental limit ∼ 1/N established by the Heisenberg uncertainty principle. Among the many approaches leading to this goal [71], phase squeezing is perhaps the most straightforward [Fig. 1(f)]. In this sec- tion, we discuss a prominent example: the application of Photo current (rel. units) squeezed light in gravitational wave detection. 0510 Gravitational waves (GWs) are deformations of the Time (ms) space-time continuum caused by accelerating massive ob- jects and propagating at the speed of light. GWs are a FIG. 18. Using squeezed vacuum to enhance the sensitivity primary prediction of Einstein’s general relativity, but of interferometric phase detection. a) Scheme of the setup. The squeezed vacuum is injected through the dark port (mode they have not yet been observed due to their minuscule ˆ magnitude. The strongest GWs reaching the Earth are b) of the interferometer. An optical isolator is used to separate ˆ ˆ00 expected to cause deformations on a scale of 1 part in the input and output modes b and b , and also to prevent scattering of the macroscopic light inside the interferometer 1020, and their detection constitutes one of the most sig- into the squeezed vacuum source. For clarity, the end mirrors nificant challenges faced by modern physics. are sketched as retroreflectors. b) The Wigner function of the Gravitational wave detectors use a Michelson-type state in mode ˆb00 for the vacuum (left) and squeezed (right) laser interferometer to detect small perturbations to po- input in mode ˆb. The momentum quadrature measurement by sitions of massive, freely suspended mirrors in its two the homodyne detector is proportional to the interferometer arms. The action of a GW stretches one of the arms and path length difference; the measurement precision is enhanced compresses the other, thereby affecting the path-length by the initial momentum squeezing of mode ˆb. c) Simulation difference and changing the intensity of the output signal. from Ref. [70] illustrating how squeezing helps revealing a At present, the world’s most sensitive GW detectors small oscillation. Left, no squeezing; right, squeezing present. are TAMA in Japan, GEO 600 in Germany, LIGO in the US and VIRGO in Italy. These detectors utilize a number of techniques in order to enhance their signal. high-finesse Fabry-Perot cavities by means of ad- ditional mirrors placed near the Michelson interfer- • The interferometer arms are constructed up to a ometer beam splitter. few kilometers in length in order to increase the absolute displacement of the mirrors. • Massive (tens of kg) mirrors are used in order to reduce the radiation pressure noise and the mirrors’ • Both arms of the interferometer are turned into Brownian motion. 18

• Laser powers of up to hundreds of watt are em- hancement of sensitivity of up to 2.2 dB for frequencies ployed to minimize the phase uncertainty. down to 150 Hz is reported. Note that this enhance- ment is far below the > 10 dB degree of squeezing pro- Further enhancement of any of these parameters would duced by the source employed. This is because of the be prohibitive in terms of costs and resources. This is losses introduced when injecting the squeezed field into why additional sensitivity improvement associated with the Michelson interferometer, imperfect mode matching squeezing becomes useful. The idea of this improvement with the carrier field, and phase fluctuations. It is ex- was proposed by Caves in 1981 [72] and involves injecting pected that the next generation of LIGO (the so-called squeezed vacuum into the dark port of the interferometer. Advanced LIGO) will address most of these shortcomings Suppose the interferometer input modea ˆ is fed with a [74]. strong laser field in coherent state |αi [Fig. 18(a)]. We assume α to be real. The other input mode, ˆb, is in the vacuum or squeezed vacuum state. The beam splitter 0 VII. CONCLUSION AND OUTLOOK implements√ the mode transformation√ according toa ˆ = (ˆa+ˆb)/ 2, ˆb0 = (ˆa−ˆb)/ 2. Let the interferometer paths be slightly unbalanced in length so that, upon return to Over the past thirty years, the science of squeezed light the beam splitter, mode ˆb0 acquires a small phase shift has experienced enormous progress and made significant ϕ which we wish to evaluate. After interacting, for the influence on the entire field of physics. Its primary effect, second time, on the beam splitter, the modes become in my opinion, was to radically change the physicists’ per- √ √ ception of quantum theory of electromagnetic radiation. aˆ00 = (ˆa0 + ˆb0eiϕ)/ 2, ˆb00 = (ˆa0 − ˆb0eiϕ)/ 2. Using eiϕ ≈ Prior to the observation of squeezing, it was a largely ab- 1 + iϕ, we find stract discipline, having little connection to experimen- ˆb00 = ˆb − iϕa.ˆ (76) tal practice. Observation of squeezing and subsequent development of optical homodyne tomography resulted Because ϕ is small, the second term in Eq. (76) effectively in techniques of creating, manipulating and measuring results in displacement of the (squeezed) vacuum mode ˆb quantum states of light, allowing the postulates of quan- along the momentum axis by ϕα [Fig. 18(b)]. tum theory of light to be directly tested and applied in A momentum quadrature measurement performed on experiment. mode ˆb00 by means of a homodyne detector will yield this The second important contribution of squeezing is that value, with an uncertainly equal to the momentum un- to quantum information science. It provided an en- certainty of the initial state of mode b. If this state is tangled resource for many quantum information proto- momentum squeezed, the measurement sensitivity is en- cols. Additionally, it gave rise to deeper understand- hanced accordingly, as illustrated in Fig. 18(c). ing of parametric down-conversion, allowing preparation The actual measurement procedure that is currently of other important quantum optical resources such as implemented in GEO 600 [73] and LIGO [74] largely fol- polarization-entangled photon pairs. As a result, optics lows the above description. A major challenge is to con- has become, for at least a decade, the main test bed for struct a source capable of generating squeezing in the fre- quantum information science, effectively jump starting quency band compatible with gravitational waves. Typ- this field. ical GWs are produced in the audio range between 150 What developments can be expected in the next years? and 300 Hz, whereas most OPA-based squeezing sources We are currently witnessing the emergence of new means built until recently exhibited significant technical noises of production of squeezing, e.g. by bringing light into at frequencies below 1 MHz. A series of breakthroughs interaction with an optomechanical cavity, i.e. a optical achieved over the past decade helped identifying and cavity with one of its elements suspended so as to form on eliminating the sources of these noises [70]. a high-quality mechanical resonator [77, 78]. The pres- The primary issue turned out to be macroscopic optical sure of light inside the cavity on that resonator results field at the wavelength of the desired squeezing present in optical nonlinearities described by equations similar within the OPA cavity. Mechanical fluctuations of the to (72), thereby leading to the squeezing. The promise cavity length (which occur at low frequencies) randomly of this new method is the possibility to manufacture on- affect the magnitude and phase of that field and subse- chip sources of squeezed light, enabling compact optical quently contaminate the output. The remedy consisted sensors and new fundamental tests of physics. of preventing the ambient laser field from penetrating In terms of applications, major results are awaited in into the cavity. This included using a field of different gravitational wave detection. Although squeezed light has frequency to lock the cavity length [75], using an optical already been integrated into some of the detectors, it has isolator to prevent the reflection of the local oscillator not yet been used in actual data acquisition runs. In from the homodyne detector photodiodes into the OPA Advanced LIGO, the squeezing is expected to enhance cavity [22] and even minimization of scattering from the the sensitivity by up to a factor of ten. Hopefully, such nearby optical elements [76]. a detector will not only be able to prove the existence The most recent result on incorporating squeezed light of GWs, but provide information about their spatial dis- into a GW detector has been reported for LIGO [74]. En- tribution and temporal dynamics. This would result in 19 a fundamentally new method for observing the universe, main are long-term storage of squeezed light [80–82] as which has a potential to revolutionize the entire field of well as methods of distilling the two-mode squeezed state astronomy. that has experienced losses [83–85]. Recently, exciting developments have been reported No less exciting are squeezed light’s contributions to on creation of multimode quadrature-entangled states by quantum information science. Existing techniques of simultaneous pumping of multiple spatial [86], spectral two-mode squeezing and quantum teleportation can be [87, 88], or temporal [89] modes of an SPDC arrange- employed for the development of the continuous-variable ment. In this way, a large-scale, individually-addressable quantum repeater [79], which will dramatically enhance entangled state is created that may be possible to use the quantum communication distance leading to global in measurement-based quantum computation and other “quantum internet”. The unsolved challenges in this do- quantum information applications.

[1] R. Loudon and P. Knight, “Squeezed light,” Journal of 81, 299–332 (2009). Modern Optics 34, 709–759 (1987). [17] W. Wasilewski, T. Fernholz, K. Jensen, L. S. Madsen, [2] A. Einstein, B. Podolsky, and N. Rosen, “Can quantum- H. Krauter, C. Muschik, and E. S. Polzik, “Generation mechanical description of physical reality be considered of two-mode squeezed and entangled light in a single tem- complete?” Phys. Rev. 47, 777–780 (1935). poral and spatial mode,” Opt. Express 17, 14444–14457 [3] K. Schneider, M. Lang, J. Mlynek, and S. Schiller, “Gen- (2009). eration of strongly squeezed continuous-wave light at [18] H. Hansen, T. Aichele, C. Hettich, P. Lodahl, A. I. 1064 nm,” Opt. Express 2, 59–64 (1998). Lvovsky, J. Mlynek, and S. Schiller, “Ultrasensitive [4] V. Buzek and P. L. Knight, Squeezed States: Basic Prin- pulsed, balanced homodyne detector: application to ciples (Springer, Berlin Heidelberg, 2004), pp. 3–32. time-domain quantum measurements,” Opt. Lett. 26, [5] U. Leonhardt, Measuring the Quantum State of Light, 1714–1716 (2001). Cambridge Studies in Modern Optics (Cambridge Uni- [19] K. Vogel and H. Risken, “Determination of quasiproba- versity Press, 1997). bility distributions in terms of probability distributions [6] B. L. Schumaker and C. M. Caves, “New formalism for for the rotated quadrature phase,” Phys. Rev. A 40, two-photon quantum optics. II. Mathematical founda- 2847–2849 (1989). tion and compact notation,” Phys. Rev. A 31, 3093–3111 [20] D. T. Smithey, M. Beck, M. G. Raymer, and A. Faridani, (1985). “Measurement of the Wigner distribution and the density [7] H. P. Yuen, “Two-photon coherent states of the radiation matrix of a light mode using optical homodyne tomog- field,” Phys. Rev. A 13, 2226–2243 (1976). raphy: Application to squeezed states and the vacuum,” [8] M. Scully and S. Zubairy, Quantum Optics (Cambridge Phys. Rev. Lett. 70, 1244–1247 (1993). University Press, 1997). [21] M. Vasilyev, S.-K. Choi, P. Kumar, and G. M. D’Ariano, [9] V. V. Dodonov, O. V. Man’ko, and V. I. Man’ko, “Pho- “Tomographic measurement of joint photon statistics of ton distribution for one-mode mixed light with a generic the twin-beam quantum state,” Phys. Rev. Lett. 84, Gaussian Wigner function,” Phys. Rev. A 49, 2993–3001 2354–2357 (2000). (1994). [22] K. McKenzie, N. Grosse, W. P. Bowen, S. E. Whitcomb, [10] G. Breitenbach, S. Schiller, and J. Mlynek, “Measure- M. B. Gray, D. E. McClelland, and P. K. Lam, “Squeez- ment of the quantum states of squeezed light,” Nature ing in the audio gravitational-wave detection band,” 387, 471–475 (1997). Phys. Rev. Lett. 93, 161105 (2004). [11] A. Furusawa, J. L. S¨orensen,S. L. Braunstein, C. A. [23] F. A. S. Barbosa, A. S. Coelho, K. N. Cassemiro, Fuchs, H. J. Kimble, and E. S. Polzik, “Unconditional P. Nussenzveig, C. Fabre, M. Martinelli, and A. S. Villar, quantum teleportation,” Science 282, 706–709 (1998). “Beyond spectral homodyne detection: Complete quan- [12] R. Schnabel, H. Vahlbruch, A. Franzen, S. Chelkowski, tum measurement of spectral modes of light,” Phys. Rev. N. Grosse, H.-A. Bachor, W. Bowen, P. Lam, and Lett. 111, 200402 (2013). K. Danzmann, “Squeezed light at sideband frequencies [24] E. H. Huntington, G. N. Milford, C. Robilliard, T. C. below 100 kHz from a single OPA,” Optics communica- Ralph, O. Gl¨ockl, U. L. Andersen, S. Lorenz, and tions 240, 185–190 (2004). G. Leuchs, “Demonstration of the spatial separation of [13] W. Wasilewski, A. I. Lvovsky, K. Banaszek, and the entangled quantum sidebands of an optical field,” C. Radzewicz, “Pulsed squeezed light: Simultaneous Phys. Rev. A 71, 041802 (2005). squeezing of multiple modes,” Phys. Rev. A 73, 063819 [25] R. Boyd, Nonlinear Optics (Acad. Press, 2003). (2006). [26] R. E. Slusher, P. Grangier, A. LaPorta, B. Yurke, and [14] H. P. Yuen and V. W. S. Chan, “Noise in homodyne and M. J. Potasek, “Pulsed squeezed light,” Phys. Rev. Lett. heterodyne detection,” Opt. Lett. 8, 177–179 (1983). 59, 2566–2569 (1987). [15] G. L. Abbas, V. W. S. Chan, and T. K. Yee, “Local- [27] C. Kim and P. Kumar, “Quadrature-squeezed light de- oscillator excess-noise suppression for homodyne and het- tection using a self-generated matched local oscillator,” erodyne detection,” Opt. Lett. 8, 419–421 (1983). Phys. Rev. Lett. 73, 1605–1608 (1994). [16] A. I. Lvovsky and M. G. Raymer, “Continuous-variable [28] C. Gardiner and C. Savage, “A multimode quantum the- optical quantum-state tomography,” Rev. Mod. Phys. ory of a degenerate parametric amplifier in a cavity,” Op- 20

tics Communications 50, 173 – 178 (1984). tion of the single-photon Fock state,” Phys. Rev. Lett. [29] L.-A. Wu, H. J. Kimble, J. L. Hall, and H. Wu, “Genera- 87, 050402 (2001). tion of squeezed states by parametric down conversion,” [48] A. Ourjoumtsev, R. Tualle-Brouri, and P. Grangier, Phys. Rev. Lett. 57, 2520–2523 (1986). “Quantum homodyne tomography of a two-photon Fock [30] T. Eberle, S. Steinlechner, J. Bauchrowitz, V. H¨andchen, state,” Phys. Rev. Lett. 96, 213601 (2006). H. Vahlbruch, M. Mehmet, H. M¨uller-Ebhardt, and [49] M. Cooper, L. J. Wright, C. S¨oller,and B. J. Smith, R. Schnabel, “Quantum enhancement of the zero-area “Experimental generation of multi-photon Fock states,” Sagnac interferometer topology for gravitational wave de- Opt. Express 21, 5309–5317 (2013). tection,” Phys. Rev. Lett. 104, 251102 (2010). [50] M. Yukawa, K. Miyata, T. Mizuta, H. Yonezawa, [31] Z. Y. Ou, S. F. Pereira, H. J. Kimble, and K. C. Peng, P. Marek, R. Filip, and A. Furusawa, “Generating su- “Realization of the Einstein-Podolsky-Rosen paradox for perposition of up-to three photons for continuous vari- continuous variables,” Phys. Rev. Lett. 68, 3663–3666 able quantum information processing,” Opt. Express 21, (1992). 5529–5535 (2013). [32] R. E. Slusher, L. W. Hollberg, B. Yurke, J. C. Mertz, and [51] E. Bimbard, N. Jain, A. MacRae, and A. Lvovsky, J. F. Valley, “Observation of squeezed states generated “Quantum-optical state engineering up to the two- by four-wave mixing in an optical cavity,” Phys. Rev. photon level,” Nature Photonics 4, 243–247 (2010). Lett. 55, 2409–2412 (1985). [52] J. Wenger, R. Tualle-Brouri, and P. Grangier, “Non- [33] C. F. McCormick, V. Boyer, E. Arimondo, and P. D. Lett, gaussian statistics from individual pulses of squeezed “Strong relative intensity squeezing by four-wave mixing light,” Phys. Rev. Lett. 92, 153601 (2004). in rubidium vapor,” Opt. Lett. 32, 178–180 (2007). [53] A. Ourjoumtsev, R. Tualle-Brouri, J. Laurat, and [34] V. Boyer, A. M. Marino, R. C. Pooser, and P. D. Lett, P. Grangier, “Generating optical Schr¨odingerkittens for “Entangled images from four-wave mixing,” Science 321, quantum information processing,” Science 312, 83–86 544–547 (2008). (2006). [35] J. Ries, B. Brezger, and A. I. Lvovsky, “Experimental [54] J. S. Neergaard-Nielsen, B. M. Nielsen, C. Hettich, vacuum squeezing in rubidium vapor via self-rotation,” K. Mølmer, and E. S. Polzik, “Generation of a super- Phys. Rev. A 68, 025801 (2003). position of odd photon number states for quantum infor- [36] S. Barreiro, P. Valente, H. Failache, and A. Lezama, “Po- mation networks,” Phys. Rev. Lett. 97, 083604 (2006). larization squeezing of light by single passage through an [55] K. Wakui, H. Takahashi, A. Furusawa, and M. Sasaki, atomic vapor,” Phys. Rev. A 84, 033851 (2011). “Photon subtracted squeezed states generated with peri- [37] I. Rigas, A. B. Klimov, L. L. Sanchez-Soto, and odically poled KTiOPO4,” Opt. Express 15, 3568–3574 G. Leuchs, “Nonlinear cross-kerr quasiclassical dynam- (2007). ics,” New Journal of Physics 15, 043038 (2013). [56] T. Gerrits, S. Glancy, T. S. Clement, B. Calkins, A. E. [38] P. Drummond, R. Shelby, S. Friberg, and Y. Yamamoto, Lita, A. J. Miller, A. L. Migdall, S. W. Nam, R. P. Mirin, “Quantum solitons in optical fibres,” Nature 365, 307– and E. Knill, “Generation of optical coherent-state super- 313 (1993). positions by number-resolved photon subtraction from [39] J. Heersink, V. Josse, G. Leuchs, and U. L. Andersen, the squeezed vacuum,” Phys. Rev. A 82, 031802 (2010). “Efficient polarization squeezing in optical fibers,” Opt. [57] R. Kumar, E. Barrios, C. Kupchak, and A. I. Lvovsky, Lett. 30, 1192–1194 (2005). “Experimental characterization of bosonic creation and [40] J. F. Corney, P. D. Drummond, J. Heersink, V. Josse, annihilation operators,” Phys. Rev. Lett. 110, 130403 G. Leuchs, and U. L. Andersen, “Many-body quantum (2013). dynamics of polarization squeezing in optical fibers,” [58] J. S. Neergaard-Nielsen, M. Takeuchi, K. Wakui, Phys. Rev. Lett. 97, 023606 (2006). H. Takahashi, K. Hayasaka, M. Takeoka, and M. Sasaki, [41] R. Dong, J. Heersink, J. F. Corney, P. D. Drummond, “Optical continuous-variable qubit,” Phys. Rev. Lett. U. L. Andersen, and G. Leuchs, “Experimental evidence 105, 053602 (2010). for Raman-induced limits to efficient squeezing in optical [59] C. H. Bennett, G. Brassard, C. Cr´epeau, R. Jozsa, fibers,” Opt. Lett. 33, 116–118 (2008). A. Peres, and W. K. Wootters, “Teleporting an unknown [42] U. L. Andersen, J. S. Neergaard-Nielsen, P. van Loock, quantum state via dual classical and Einstein-Podolsky- and A. Furusawa, “Hybrid discrete-and continuous- Rosen channels,” Phys. Rev. Lett. 70, 1895–1899 (1993). variable quantum information,” Nature Physics 11, 713– [60] L. Vaidman, “Teleportation of quantum states,” Phys. 719 (2015). Rev. A 49, 1473–1476 (1994). [43] S. L. Braunstein and P. van Loock, “Quantum informa- [61] S. L. Braunstein, C. A. Fuchs, and H. J. Kimble, “Crite- tion with continuous variables,” Rev. Mod. Phys. 77, ria for continuous-variable quantum teleportation,” Jour- 513–577 (2005). nal of Modern Optics 47, 267–278 (2000). [44] U. L. Andersen, G. Leuchs, and C. Silberhorn, [62] T. C. Ralph and P. K. Lam, “Teleportation with bright “Continuous-variable quantum information processing,” squeezed light,” Phys. Rev. Lett. 81, 5668–5671 (1998). Laser & Photonics Reviews 4, 337–354 (2010). [63] F. Grosshans and P. Grangier, “Continuous variable [45] P. Grangier, B. Sanders, and J. Vuckovic, “Focus on quantum cryptography using coherent states,” Phys. single photons on demand,” New Journal of Physics 6 Rev. Lett. 88, 057902 (2002). (2004). [64] K. Hammerer, M. M. Wolf, E. S. Polzik, and J. I. Cirac, [46] F. DellAnno, S. De Siena, and F. Illuminati, “Multi- “Quantum benchmark for storage and transmission of photon quantum optics and quantum state engineering,” coherent states,” Phys. Rev. Lett. 94, 150503 (2005). Physics reports 428, 53–168 (2006). [65] N. Takei, H. Yonezawa, T. Aoki, and A. Furusawa, [47] A. I. Lvovsky, H. Hansen, T. Aichele, O. Benson, “High-fidelity teleportation beyond the no-cloning limit J. Mlynek, and S. Schiller, “Quantum state reconstruc- and entanglement swapping for continuous variables,” 21

Phys. Rev. Lett. 94, 220502 (2005). M. Aspelmeyer, and O. Painter, “Squeezed light from a [66] H. Yonezawa, S. L. Braunstein, and A. Furusawa, “Ex- silicon micromechanical resonator,” Nature 500, 185–189 perimental demonstration of quantum teleportation of (2013). broadband squeezing,” Phys. Rev. Lett. 99, 110503 [78] T. P. Purdy, P.-L. Yu, R. W. Peterson, N. S. Kampel, (2007). and C. A. Regal, “Strong optomechanical squeezing of [67] N. Lee, H. Benichi, Y. Takeno, S. Takeda, J. Webb, light,” Phys. Rev. X 3, 031012 (2013). E. Huntington, and A. Furusawa, “Teleportation of non- [79] H.-J. Briegel, W. D¨ur,J. I. Cirac, and P. Zoller, “Quan- classical wave packets of light,” Science 332, 330–333 tum repeaters: The role of imperfect local operations in (2011). quantum communication,” Phys. Rev. Lett. 81, 5932– [68] S. Takeda, T. Mizuta, M. Fuwa, P. van Loock, and A. Fu- 5935 (1998). rusawa, “Deterministic quantum teleportation of pho- [80] K. Honda, D. Akamatsu, M. Arikawa, Y. Yokoi, K. Ak- tonic quantum bits by a hybrid technique,” Nature 500, iba, S. Nagatsuka, T. Tanimura, A. Furusawa, and 315–318 (2013). M. Kozuma, “Storage and retrieval of a squeezed vac- [69] R. Slav´ık, F. Parmigiani, J. Kakande, C. Lundstr¨om, uum,” Phys. Rev. Lett. 100, 093601 (2008). M. Sj¨odin,P. A. Andrekson, R. Weerasuriya, S. Sygle- [81] J. Appel, E. Figueroa, D. Korystov, M. Lobino, and A. I. tos, A. D. Ellis, L. Gr¨uner-Nielsen et al., “All-optical Lvovsky, “Quantum memory for squeezed light,” Phys. phase and amplitude regenerator for next-generation Rev. Lett. 100, 093602 (2008). telecommunications systems,” Nature Photonics 4, 690– [82] A. I. Lvovsky, B. C. Sanders, and W. Tittel, “Optical 695 (2010). quantum memory,” Nature Photonics 3, 706–714 (2009). [70] R. Schnabel, N. Mavalvala, D. E. McClelland, and P. K. [83] H. Takahashi, J. S. Neergaard-Nielsen, M. Takeuchi, Lam, “Quantum metrology for gravitational wave astron- M. Takeoka, K. Hayasaka, A. Furusawa, and M. Sasaki, omy,” Nature communications 1, 121 (2010). “Entanglement distillation from Gaussian input states,” [71] V. Giovannetti, S. Lloyd, and L. Maccone, “Advances Nature Photonics 4, 178–181 (2010). in quantum metrology,” Nature Photonics 5, 222–229 [84] Y. Kurochkin, A. S. Prasad, and A. I. Lvovsky, “Distil- (2011). lation of the two-mode squeezed state,” Phys. Rev. Lett. [72] C. M. Caves, “Quantum-mechanical noise in an interfer- 112, 070402 (2014). ometer,” Physical Review D 23, 1693 (1981). [85] A. E. Ulanov, I. A. Fedorov, A. A. Pushkina, Y. V. [73] J. Abadie, B. Abbott, R. Abbott, T. Abbott, M. Aber- Kurochkin, T. C. Ralph, and A. Lvovsky, “Undoing the nathy, C. Adams, R. Adhikari, C. Affeldt, B. Allen, effect of loss on quantum entanglement,” Nature Photon- G. Allen et al., “A gravitational wave observatory op- ics 9, 764–768 (2015). erating beyond the quantum shot-noise limit,” Nature [86] S. Armstrong, J.-F. Morizur, J. Janousek, B. Hage, Physics 7, 962–965 (2011). N. Treps, P. K. Lam, and H.-A. Bachor, “Programmable [74] J. Aasi, J. Abadie, B. Abbott, R. Abbott, T. Abbott, multimode quantum networks,” Nature Communications M. Abernathy, C. Adams, T. Adams, P. Addesso, R. Ad- 3, 1026 (2012). hikari et al., “Enhanced sensitivity of the LIGO gravita- [87] N. C. Menicucci, S. T. Flammia, and O. Pfister, “One- tional wave detector by using squeezed states of light,” way quantum computing in the optical frequency comb,” Nature Photonics 7, 613–619 (2013). Phys. Rev. Lett. 101, 130501 (2008). [75] H. Vahlbruch, S. Chelkowski, B. Hage, A. Franzen, [88] M. Pysher, Y. Miwa, R. Shahrokhshahi, R. Bloomer, and K. Danzmann, and R. Schnabel, “Coherent control of O. Pfister, “Parallel generation of quadripartite cluster vacuum squeezing in the gravitational-wave detection entanglement in the optical frequency comb,” Phys. Rev. band,” Phys. Rev. Lett. 97, 011101 (2006). Lett. 107, 030505 (2011). [76] H. Vahlbruch, S. Chelkowski, K. Danzmann, and [89] S. Yokoyama, R. Ukai, S. C. Armstrong, C. Sornphiphat- R. Schnabel, “Quantum engineering of squeezed states phong, T. Kaji, S. Suzuki, J.-i. Yoshikawa, H. Yonezawa, for quantum communication and metrology,” New Jour- N. C. Menicucci, and A. Furusawa, “Ultra-large-scale nal of Physics 9, 371 (2007). continuous-variable cluster states multiplexed in the time [77] A. H. Safavi-Naeini, S. Gr¨oblacher, J. T. Hill, J. Chan, domain,” Nature Photonics 7, 982–986 (2013).