April 1, 1992 / Vol. 17, No. 7 / OPTICS LETTERS 529

Squeezed-light generation with a mode-locked Q-switchedlaser and detection by using a matched local oscillator

Orhan Aytiir* and Prem Kumar

Department of Electrical Engineering and Computer Science, Northwestern University, Evanston, Illinois 60208

Received December 13, 1991

We report pulsed squeezed-light generation by means of an optical parametric downconverter that is pumped bythe second harmonic of a mode-locked Q-switched laser. Using the fundamental beam of the laser as a localoscillator, we observe 2-dB squeezing for a parametric gain of 2.0. This local oscillator, however, is nonoptimalbecause of its spatiotemporal mode mismatch with the squeezed mode generated by the downconverter. Wedescribe an experiment in which a matched local oscillator is generated with the use of an optical parametricamplifier that is pumped by the same laser as is the downconverter. In this case, 2-dB squeezing is observedfor a parametric gain of 1.5. The present experimental setup is limited by the uncontrollable phase fluctua-tions that prohibit us from making squeezing measurements at higher parametric gains.

Recently we demonstrated the generation of pulsedsqueezed light by means of an optical parametricdownconverter that is pumped by the second har-monic of a Q-switched multilongitudinal-modelaser.' In that experiment, the degree of observ-able squeezing (=1 dB) was limited by two factors:First, owing to intensity fluctuations underneaththe Q-switched pulse envelopes, which result frombeating of the various longitudinal modes,2 largeparametric gain could not be obtained without dam-aging the downconverting KTiOPO4 (KTP) crystal.Second, a spatiotemporal mismatch existed betweenthe squeezed mode generated by the downconverterand the local-oscillator (LO) mode that was em-ployed in the homodyne detector.3 In this Letterwe describe experiments in which attempts aremade to overcome the above two limitations. Thefirst limitation is removed by pumping the down-converter with a mode-locked Q-switched laser.Mode locking eliminates the intensity fluctuationsunderneath the Q-switched pulse envelopes, andlarge parametric gains can be easily obtained.4 Toovercome the second limitation, a matched LO isgenerated by means of an optical parametric ampli-fier (OPA) that is pumped by the same laser as is thedownconverter.

We first describe the experiment in which the fun-damental beam of the laser is used as an LO. Inthis case a spatiotemporal mode mismatch still ex-ists between the squeezed and the LO modes. Nextwe describe the experiment in which a matched LOis generated. The idler beam from the OPA is em-ployed in the homodyne detection of the squeezinggenerated by the downconverter.

The first experimental setup is shown in Fig. 1. Itis similar to that used in Ref. 1. The second har-monic of a mode-locked Q-switched Nd:YAG laser(Quantronix Model 416) at 532 nm is used to pumpan optical parametric downconverter. Portionsof the fundamental IR beam at 1064 nm are sepa-rated by using beam splitters to generate (i) an LO

for the homodyne detection of the generatedsqueezed light and (ii) a test input signal to thedownconverter (shown as a dashed line) for the char-acterization of its classic response. The Q-switchedenvelopes of the resulting pulses are 270 and 150 nsin duration for the IR and the pump beams, respec-tively.

Parametric downconversion of the pump beamtakes place in a type II phase-matched KTP crystal.The pump beam is linearly polarized along the e axisof the crystal. The downconverted IR beam is com-posed of two orthogonal linear polarization modes,called the signal and the idler modes. The IR testbeam is made collinear and copolarized with thepump to measure the parametric gain and to facili-tate subsequent alignment of the amplified beamsinto the homodyne detection apparatus. The signal(S) and the idler (J) beams emerging from the KTPcrystal are slightly displaced from each other be-cause of walk-off in the birefringent KTP. To makethem collinear again, for the purpose of extractingthe proper squeezed mode, a separation and recom-bination technique is employed with the use ofbeam-splitting polarizers (BSP's).

The recombined beam exhibits squeezing in amode that is polarized at a 450 angle relative to boththe signal and the idler beam polarizations. Thesqueezed mode is superimposed on the properly po-larized LO mode, obtained with the help of variouspolarizers (P's) and half-wave plates (HWP's), byusing a 50-50 beam splitter (BS) as part of the dual-detector balanced homodyne detection configu-ration.5 The output beams from the 50-50 beamsplitter are focused onto InGaAs P-I-N photodetec-tors. Photocurrents from the two detectors aresubtracted from each other, and noise measure-ments on the difference photocurrent are made at28 MHz by using a pulse-noise measurementscheme.6 One mirror in the LO path is mounted ona piezoelectric transducer to sweep the phase of theLO beam with respect to that of the squeezed mode.

0146-9592/92/070529-03$5.00/0 (© 1992 Optical Society of America

530 OPTICS LETTERS / Vol. 17, No. 7 / April 1, 1992

Detection Off BS sclatElectronics Oscillator

Fig. 1. Schematic of the experimental setup forsqueezed-light generation. The fundamental beam fromthe laser is used as a LO in homodyne detection. Thedashed line is a test signal input to the downconverter.

The path lengths of the various beams from theNd:YAG laser to the photodetectors are made equalto each other within a few millimeters to ensurethat the various mode-locked pulses arrive at anygiven point at the same time.

Because squeezed vacuum has no mean field, it isnecessary to have a test signal input to the down-converter for alignment purposes. The signal andidler beams at the output of the downconverter havedifferent spatiotemporal profiles. The idler profileis almost completely determined by that of thepump, whereas the signal profile is influenced byboth the pump and the input signal profiles. Thehomodyne setup is aligned by interfering the LObeam separately with the signal and the idler outputbeams from the downconverter. For this purpose,one of the detectors is blocked while the other isterminated directly into a 50-4 resistor and an os-cilloscope. The interference is observed on theoscilloscope as the phase of the LO is scanned. Themode-matching efficiency (MME) is given byMME = 1pp2/161112, where I, and I2 are the peakphotocurrents when either beam is blocked and Ip isthe peak-to-peak photocurrent when the two beamsare interfering.7 The signal MME, as measured be-tween the signal output beam and the LO beam, isalways found to be larger than the idler MME. Thisis because the spatiotemporal profile of the signalbeam is closer to that of the LO; they are bothderived from the fundamental beam of the laser.This is especially true at low parametric gains. Asthe parametric gain increases, the signal MME de-creases because an increasingly larger portion of thesignal output is due to parametric downconversion.The squeezed mode has a spatiotemporal profilethat is closer to that of the idler. Therefore a mea-surement of the idler MME approximates the MMEfor the squeezed mode.

The boxcar used in the pulse-noise measurementscheme generates an exponential moving averageover 300 samples at the Q-switch repetition rate of1 kHz. Therefore the measurement bandwidth is-1 Hz. This low bandwidth places a stringent re-quirement on the phase stability of the experiment.Unfortunately the phase stability of the presentsetup is not good owing to circumstances beyond

our control. The phase jitter is readily observableon the oscilloscope when the LO phase is notscanned. The random phase fluctuations are typi-cally 0.4 rad peak to peak. This situation hampersour squeezing measurements considerably, espe-cially for parametric gains >2.

Figure 2 shows a measurement of the quadraturenoise as the LO phase is scanned in time. The 0-dBline corresponds to the vacuum-state noise level (thequantum limit). This level is 10 dB higher than thebackground noise level in the detection electronics.The contribution of the background noise levelis subtracted from all the numbers quoted in thisLetter. When the LO is in phase with the squeezedquadrature, the noise goes below the quantum limitby 2.0 dB, and when it is in phase with the de-squeezed quadrature, the noise increases by 5.4 dB.For these data the MME was measured to be 65%.The detectors have quantum efficiencies of 86%, andthe reflection losses at various optical surfaces addup to 2%. These figures give an overall detectionefficiency of 55%. The parametric gain was 2.0.

Squeezing and the observed quantum-noise reduc-tion are given by S = (Vs - 72 St-(V; ±

g2- 1)2, R = -qS + 1 - Aq, and R' = -S' + 1- -.Here g is the parametric gain, S and S' are squeez-ing and desqueezing, and R and R' are the observedquantum-noise reduction and increase, respectively.For 7- = 0.55 and g 2.0, these equations giveR = 2.6 dB and R' = 5.6 dB. The expected noisebehavior as a function of the LO phase for the aboveexperimental parameters is superimposed on thedata in Fig. 2. The measured noise reduction fallsshort of that expected from the Y7 and g values be-cause of the uncontrollable phase fluctuations dur-ing the measurement. These fluctuations have amuch larger effect on R compared with R' becausethe negative peaks are much narrower than thepositive peaks.

Our MME measurements between the LO and thesignal, and between the LO and the idler, point tothe fact that the fundamental beam of the laser isnot a suitable LO for detecting the squeezed modegenerated by the downconverter. A simple modelhelps us gain insight into the problem. Assumethat the fundamental IR beam has a mode-lockedpulse envelope that is a Gaussian with a FWHM of

. The second-harmonic generation is a nonlinear

7.5 1 I l l l I l l t

5

02

.Foz

-50 10 20

Time (s)Fig. 2. Time trace of the squeezed-vacuum noise at28 MHz as the LO phase is scanned. The 0-dB line cor-responds to the vacuum-state noise level. The solid curveis a theoretical fit.

.0 -1 -1 L!111.5 f � � � - /I I' _-Z

.0 i _1� I �, / � �, 'I, J

V Ai/ �r \�f.5

.0 I I

April 1, 1992 / Vol. 17, No. 7 / OPTICS LETTERS 531

BSP

Squeezed VacuumPumps

HWP I Idler A :I- u Ad flu r:a:n

SignalInput

sCi Llatocal P : 1 T U P _OscillatorI P :HWP xP KTP P ~

1 lectronics [ _-_ 1r1c ̂ _ A OPA for MatchedLO Generation

Fig. 3. Schematic of the experimental setup forsqueezed-light detection with a matched LO. The idleroutput from the OPA is used as a LO in homodyning.

7.5

m 5.0

c) 2.5

0

Z o.o

-2.50 10 20 30 40 5'0

Time (s)Fig. 4. Time trace of the squeezed-vacuum noise at28 MHz as the matched LO phase is scanned. The 0-dBline corresponds to the vacuum-state noise level.

process that reduces this pulse width by a factor ofV¶, giving Tsh = Tf/V2 for the pump beam. Be-cause the signal input pulse to the downconverter islonger than the pump pulse, the idler pulse followsthe same profile as that of the pump, at least at lowparametric gains. Therefore the MME between theidler and the LO beams is less than unity owing tothe different pulse widths because our detectionelectronics integrate over the mode-locked pulses.The overlap integral between two Gaussians witha width ratio of V2 is calculated to be 0.92. This isthe mismatch factor that is due to the temporal pro-files only. The two-dimensional transverse profilesof the beams are also Gaussian. The same mis-match effect also takes place in the two space di-mensions because the photodetector diameters arelarger than the beam sizes. Taking the cube of theoverlap integral, we expect a total mismatch factorof 0.78.

To overcome the spatiotemporal mode-mismatcheffect, we use an OPA to mimic the profile of thesqueezed mode.5 Figure 3 shows the modifiedsetup. An OPA, which is pumped by the same laseras the downconverter, is used to generate an idlerbeam having the same spatiotemporal profile as thesqueezed mode. The output signal and pumpbeams of the OPA are blocked with a polarizer. Theidler beam is strongly attenuated with the use of ahalf-wave plate and another polarizer and is used asan LO. Attenuation and balanced homodyne detec-tion5 remove the excess noise on the idler beam,4 and

quantum-limited homodyne detection is easilyachieved.

Using this new LO, we are able to increase theMME from 65% to 85%. The ratio of these twoMME's is 0.76, very close to the 78% mismatch fac-tor predicted by the Gaussian profile model.Figure 4 shows squeezing data taken with thematched LO. The noise reduction and increase are2.0 and 5.6 dB, respectively. The overall detectionefficiency is 72%, and the parametric gain of thedownconverter is 1.5. Once again, the measurednoise reduction falls short of that expected from the-q and g values because of the uncontrollable phasefluctuations during the measurement. Comparingthese data with the mismatched LO case, however,we see that approximately the same level of noisereduction is measured for g = 1.5 with the matchedLO (g = 2.0 with the mismatched LO). In otherwords, using the matched LO enables us to achievethe same R at a lower parametric gain.

Uncontrollable phase fluctuations prohibit usfrom making squeezing measurements at higherparametric gains to increase the quantum-noise re-duction and possibly to observe the limitationpredicted by LaPorta and Slusher.9 The quantum-noise reduction observable with our present setupactually decreases as the parametric gain is in-creased. The matched LO would be more useful inan experiment that is limited only by the detectionefficiency.

A preliminary account of this research was firstreported at the 1990 Optical Society of AmericaAnnual Meeting held in Boston, Massachusetts.This research was supported in part by the NationalScience Foundation and the U.S. Office of NavalResearch.

*Present address, Fibertek, Inc., 510 HerndonParkway, Herndon, Virginia 22070.

References

1. P. Kumar, 0. AytUr, and J. Huang, Phys. Rev. Lett. 64,1015 (1990).

2. A. V Masalov, in Progress in Optics XXII, E. Wolf, ed.(North-Holland, Amsterdam, 1985), pp. 147-196.

3. J. Huang, P. Kumar, and 0. Aytiir, in InternationalConference on Quantum Electronics, Vol. 8 of 1990OSA Technical Digest Series (Optical Society of Amer-ica, Washington, D.C., 1990), p. 44.

4. 0. Aytiir and P. Kumar, Phys. Rev. Lett. 65, 1551(1990).

5. H. P. Yuen and V. W S. Chan, Opt. Lett. 8, 177, 345(erratum) (1983); B. L. Schumaker, Opt. Lett. 9, 189(1984); B. Yurke, P. Grangier, R. E. Slusher, and M. J.Potasek, Phys. Rev. A 35, 3586 (1987).

6. 0. AytUr and P Kumar, Opt. Lett. 15, 390 (1990).7. K. A. Winick and P. Kumar, IEEE J. Lightwave Tech-

nol. 6, 513 (1988).8. 0. AytUr and P. Kumar, in Digest of Optical Society of

America Annual Meeting (Optical Society of America,Washington, D.C., 1990), p. 4.

9. A. LaPorta and R. E. Slusher, Phys. Rev. A 44, 2013(1991).

04% - ju% 71�17 _�_ ( � I

'11-V

11, " ' Bs"I- . . - I