# PhysRevA.84.023825 .pdf

Nom original:

**PhysRevA.84.023825.pdf**

Ce document au format PDF 1.3 a été généré par LaTeX with hyperref package / Acrobat Distiller 9.4.0 (Windows), et a été envoyé sur fichier-pdf.fr le 31/08/2011 à 22:12, depuis l'adresse IP 90.60.x.x.
La présente page de téléchargement du fichier a été vue 1423 fois.

Taille du document: 595 Ko (4 pages).

Confidentialité: fichier public

### Aperçu du document

PHYSICAL REVIEW A 84, 023825 (2011)

Coherent phase-modulation transfer in counterpropagating parametric down-conversion

G. Str¨omqvist,1 V. Pasiskevicius,1 C. Canalias,1 and C. Montes2

1

Department of Applied Physics, KTH, Roslagstullbacken 21, SE-10691 Stockholm, Sweden

LPMC-CNRS, Universit´e de Nice–Sophia Antipolis, Parc Valrose, FR-06108 Nice, France

(Received 5 November 2010; published 16 August 2011)

2

Distributed positive feedback established by a counterpropagating three-wave mixing process in a submicrometer structured second-order nonlinear medium leads to mirrorless parametric oscillation with unusual

spectral properties. In this work, we demonstrate experimentally and theoretically that the phase modulation of

the pump is coherently transferred to the co-propagating parametric wave, while the counterpropagating wave

retains a narrow bandwidth and high coherence. The main mechanisms responsible for these properties are

the maximized convective separation between the counterpropagating waves and the associated phase-locking

between the pump and the co-propagating parametric wave. This suggests that mirrorless optical parametric

oscillators can be pumped with incoherent light and still generate highly coherent backward-propagating

radiation.

DOI: 10.1103/PhysRevA.84.023825

PACS number(s): 42.65.Lm, 42.65.Yj, 42.70.Mp, 78.67.Pt

I. INTRODUCTION

The lowest-order nonlinear interactions, collectively called

three-wave mixing (TWM), are important in many different

physical contexts, e.g., in plasmas [1,2], hydrodynamics, and

internal gravity waves [3], in planetary Rossby waves, which

are related to the global climate [4], in bulk and surface

acoustic waves [5,6], in matter waves [7], and, indeed, in

nonlinear optics [8]. Optical parametric amplification (OPA)

of lower frequency waves in a TWM process where photons

h

¯ ωp of the wave of the highest frequency, the pump, are split

into two photons of lower energy, the signal h

¯ ωs and the

idler h

¯ ωi , such that the energy conservation, ωp = ωs + ωi ,

and the momentum conservation, kp = ks + ki , conditions are

satisfied, has large practical importance for the generation of

high-energy ultrashort pulses in attosecond systems [9], as

well as in quantum optics experiments for entangled photonpair generation [10] and generation of squeezed vacuum

states [11].

The coherence properties of the waves generated in the

TWM down-conversion process have been the subject of investigation from the very beginning of nonlinear optics [12,13].

For instance, from simple quantum optical considerations it

was shown [12] that OPA behaves like a perfect phase-sensitive

amplifier, where the output photon number and phase variances

of the amplified signal wave faithfully represent the input

variances within the uncertainty principle. It is a general

property of TWM that the total interaction phase is fixed. In

the case of a down-conversion process, the phases φj (where

j = p,s,i) satisfy the relation

φp − φs − φi = π/2.

(1)

This feature has been utilized in passive all-optical carrierenvelope phase stabilization schemes [14]. In an optical

parametric generation (OPG) process, which, by definition,

is seeded by zero-point fluctuations and thermal background

radiation, the phases of the individual signal or idler waves

remain to a large extent random, although the pairwise

correlation between the signal and idler waves does exist

[15]. As a consequence, the signal and the idler waves

generated in background-noise-seeded singly-resonant optical

1050-2947/2011/84(2)/023825(4)

parametric oscillators (OPOs) can, in general, be considered

incoherent and their spectral widths are much larger than

that of the pump wave, unless phase-sensitive filtering is

artificially imposed [16]. The spectral width of the parametric

gain is determined primarily by the dispersion of the nonlinear medium and it is maximized when the group velocity

dispersion is zero, ∂ 2 k/∂ω2 = 0, at the degeneracy point

(ωs ωi ). Away from degeneracy, the convection operators

in the coupled wave equations, vg,j · ∇, where vg,j are the

respective group velocities, determine the spectral widths of

the signal and the idler waves. Several scenarios leading to

increased coherence of the generated parametric waves by

exploiting convective wave separation have been theoretically

investigated [17,18].

In this work, we demonstrate experimentally and numerically that in a mirrorless optical parametric oscillator

(MOPO), the device which relies on counterpropagating

TWM for establishing a distributed positive feedback for

sustained oscillation above threshold [19], the convective

wave separation is maximized and responsible for the unusual

spectral and coherence properties of the generated signal and

idler waves. Specifically, this physical mechanism ensures that

the frequency bandwidth of the backward-propagating idler

will be narrow compared to that of the signal when a broadband

pump is used. Moreover, as a consequence of the phase relation

in Eq. (1), the phase modulation of the pump is then effectively

transferred to the co-propagating parametric wave.

II. PHENOMENOLOGICAL DESCRIPTION

Proposed in 1966 [20], the first experimental demonstration of a MOPO was reported in 2007 [21]. The major

difficulty in achieving mirrorless parametric oscillation is

associated with satisfying the momentum conservation condition, which in scalar form reads kp = ks − ki . In homogeneous dielectrics, the phase-matching condition can possibly

be satisfied only for large signal-idler detunings, with the

signal frequency close to that of the pump and the idler

frequency in the terahertz region. However, due to the very

different diffraction properties of the near-infrared signal

023825-1

©2011 American Physical Society

¨

STROMQVIST,

PASISKEVICIUS, CANALIAS, AND MONTES

and the far-infrared idler, achieving oscillation in homogeneous dielectrics is problematic. By employing ferroelectric

nonlinear crystals structured with submicrometer periodicity,

, quasi-phase-matching (QPM) [8],

k = kp − ks + ki = 2π/ ,

vg,i (vg,p − vg,s )

.

vg,p (vg,i + vg,s )

details of the TWM dynamics. A numerical solution of the

coupled wave equations will show that the above reasoning

regarding the phases of the waves is qualitatively correct.

(2)

can be utilized to achieve MOPO with counterpropagating

idler ki in the mid-infrared spectral range. In the experimental

investigation of the spectral properties of a MOPO, we here

employ a 6.5 mm-long periodically poled KTiOPO4 (PPKTP)

crystal with a QPM period of = 800 nm to convert the pump

around 813.5 nm into a co-propagating signal at 1123.0 nm

and a counterpropagating idler at 2952.2 nm. The crystal

fabrication details have been reported elsewhere [21,22].

From the momentum conservation condition in the TWM

with counterpropagating idler [Eq. (2)], one can derive an

estimate of the frequency variation in the counterpropagating

wave, ωi , as function of the frequency variation in the pump,

ωp :

ωi = ωp

PHYSICAL REVIEW A 84, 023825 (2011)

(3)

The sum of the group velocities in the denominator of Eq. (3),

which is the result of the counterpropagating geometry, ensures

that the frequency of the back-propagating wave is not sensitive

to changes in the pump frequency. This also means that the

bandwidth of the counterpropagating wave is much smaller

than that of the pump and it should reach minimum in a

special case when the group velocities of the pump and the

co-propagating signal wave are equal. In this limit when

ωi / ωp → 0, the temporal phase of the counterpropagating

idler wave is φi const. Then from Eq. (1), it follows

that the partial derivatives of the co-propagating signal and

pump phases with respect to time are approximately equal,

∂t φs (t) ∂t φp (t). These are indeed very unusual properties

for parametric oscillation. In a process which is seeded by

random noise, the interaction establishes a counterpropagating

wave with a coherence time much longer than both the seeding

noise and the coherence time of the pump. At the same time, the

phase of this counterpropagating wave acts as a stable phase

reference, thereby forcing the phases of the pump and the

co-propagating signal wave to lock and vary in synchronism.

It should be stressed that Eq. (3) is only an estimate, because

the actual spectra of the generated waves also depend on the

III. EXPERIMENTS

For the experimental verification of the convective phase

locking in MOPOs, the above-mentioned PPKTP crystal was

pumped with pulses centered at 814.5 nm with a frequency

bandwidth of 1.21 THz. The pulse was stretched in a normaldispersion stretcher and amplified in a Ti:sapphire regenerative

amplifier. The full width at half maximum (FWHM) length

of the amplified pulse was 480 ps and it was used for MOPO

pumping. A small portion of the pump pulse was recompressed

to close to a transform-limited duration of 1 ps and was

used as a reference for cross-correlation measurements. The

MOPO reached threshold at a pump intensity of 0.8 GW/cm2 .

The measured pump spectra (spectrometer Ando AQ-6315A,

with a resolution of 0.05 nm) at 1.3 times the threshold

intensity are shown in Fig. 1(a), where the undepleted pump

spectrum was measured when the pump beam propagated

outside the structured area of the PPKTP and the MOPO

was not operational. The linear frequency chirp of the pump,

measured using cross-correlation with the reference pulse,

was dωp /dt = 15.7 mrad/ps2 . This linear chirp provides a

direct frequency-to-time mapping, implying that the spectral

measurements also contain temporal information about the

MOPO dynamics. By comparing the spectra in Fig. 1(a),

we can see that, at this pump intensity, the MOPO starts

oscillating after a delay of about 260 ps, corresponding

to the difference in time between the beginning of the

pulse (816.1 nm) and where the pump starts being depleted

(814.7 nm). After the start of the MOPO oscillation, the

pump is rapidly depleted and the depletion level of about

60% remains approximately constant until the end of the

pump pulse. This shows that the counterpropagating TWM

process indeed is very efficient. The measured spectrum of the

MOPO signal had a central wavelength of 1123.0 nm and a

FWMH spectral width of 410 GHz, as illustrated in Fig. 1(b).

Cross-correlation measurements revealed that the pulse had a

FWHM length of 160 ps, containing a positive linear frequency

chirp of 16.0 mrad/ps2 . This value is close to the chirp of the

pump, as expected if the picture of convective phase-locking

and phase modulation transfer between pump and signal is

FIG. 1. (Color online) Experimental spectra: (a) undepleted (solid line) and depleted (dashed line) pump, (b) co-propagating signal, and

(c) counterpropagating idler. The graphs show that the FWHM widths of the pump and signal are 1.21 THz and 410 GHz, respectively. The

deconvoluted idler FWHM is 13 GHz.

023825-2

COHERENT PHASE-MODULATION TRANSFER IN . . .

PHYSICAL REVIEW A 84, 023825 (2011)

feature of the MOPO, enabled by the counterpropagating

TWM process.

IV. NUMERICAL SIMULATIONS

FIG. 2. (Color online) Cross-correlation traces of the MOPO

signal: uncompressed pulse (squares), with a compressor length of

50 cm (circles), and with a compressor length of 86 cm (triangles).

correct. The measured idler spectrum (spectrometer Jobin

Yvon iHR550, with a 0.55 nm resolution) shown in Fig. 1(c)

reveals that the FWHM bandwidth is 23 GHz. By taking

into account that the resolution of the spectrometer is limited

to 19 GHz at 2.952 μm and assuming a Gaussian spectral

shape, the deconvoluted idler frequency bandwidth is then

about 13 GHz. Within the resolution of our measurements, the

idler bandwidth did not depend on the amount or sign of chirp

imposed on the pump wave. This was checked by passing the

pump pulse through a delay line with anomalous group-delay

dispersion so that the pump pulse with the same spectrum as in

Fig. 1(a) had an opposite frequency chirp and a different pulse

length.

As a final experiment, we investigated the compressibility

of the signal pulse. If the linear frequency chirp is transferred

from the pump to the co-propagating signal, the MOPO

signal pulse should be compressible. To test this, we built a

compressor with anomalous group-delay dispersion involving

four bounces off a 1200 lines/mm diffraction grating. The

MOPO was pumped with positively chirped pulses of similar

spectral and temporal widths as in the previous experiment.

The signal pulse before and after the compressor was characterized using cross-correlation with the 1-ps-long reference

pulse at the pump wavelength. The cross-correlation traces

measured by employing noncollinear sum-frequency mixing

in a 200-μm-long BBO crystal are shown in Fig. 2 for different

compressor lengths. The compressibility of the MOPO signal

proves that the phase modulation which is imposed on the

pump pulse is transferred to the signal. This is a unique

The transfer of phase modulation from the pump to the

signal can also be shown theoretically. In order to elucidate

the unusual coherence properties in the MOPO and also to

model the experimental situation, we numerically solved the

coupled wave equations using the slowly varying envelope and

plane-wave approximations for parametric down-conversion

in counterpropagating geometry:

(∂t + vg,p ∂z + γp + iβp ∂tt )Ap = −σp As Ai ,

(∂t + vg,s ∂z + γs + iβs ∂tt )As = σs Ap A∗i ,

(∂t − vg,i ∂z + γi + iβi ∂tt )Ai = σi Ap A∗s .

(4a)

(4b)

(4c)

Here Aj , βj , γj , and σj = 2π deff vg,j /(λj nj ) are the

field amplitude, the group-velocity dispersion, the loss,

and the nonlinear coupling coefficients for the j =

p,s,i wave, respectively. A model PPKTP crystal with

a length of 6.5 mm and an effective nonlinearity of

deff = 9 pm/V was excited with an asymmetric superGaussian pump pulse with amplitude Ap (t) = Ap0 [1 −

t/(2 t0 )] exp{iφp (t) − [(t − t0 )/ t1 ]8 /2}. The pulse was centered at t0 = 441 ps and the temporal widths t0 = 1766 ps

and t1 = 220 ps were chosen in order to fit the experimental

intensity pump shape and to get a FWHM pulse length of

tp = 400 ps. The theoretical pump pulse contained a linear

frequency chirp, given by the quadratic phase modulation

φp (t) = α2 [(t − t0 )/τ0 ]2 , where τ0 = 2/(σp Ap0 ) is the characteristic nonlinear interaction time, defining the characteristic nonlinear interaction length, Lnl = vg,p τ0 . For a pump

intensity of 1.1 GW/cm2 , which is a pump intensity close to

the experimental input, the characteristic parameter values are

τ0 = 6.9 ps and Lnl = 1 mm. The pump chirp parameter α2 =

0.46 was chosen to give a positive chirp and a spectral width

of 1.27 THz. The dispersion parameters for the KTP crystal

were calculated using the Sellmeier expansion from Ref. [23]

and the quasi-phase-matched interaction generated signal and

idler waves around 1125 and 2952 nm, respectively, for a

pump wavelength of 814.5 nm. For these central wavelengths,

the spectral compression in the idler, as estimated from Eq. (3),

is ωi / ωp 1/92.

FIG. 3. Calculated spectra: (a) undepleted (solid line) and depleted (dotted line) pump, (b) co-propagating signal, and (c) counterpropagating

idler. The zeros on the frequency scales correspond to 814.5 nm and approximately 1125 and 2952 nm for the pump, signal, and idler, respectively.

The corresponding FWHM widths are 1.27 THz, 440 GHz, and 6.5 GHz.

023825-3

¨

STROMQVIST,

PASISKEVICIUS, CANALIAS, AND MONTES

PHYSICAL REVIEW A 84, 023825 (2011)

understand this by realizing that the initial pulse front of the

backward-propagating idler wave is rapidly separated from

the signal and the pump, akin to fast convection, and seeds the

parametric amplification process upstream, while maintaining

its initial phase and enforcing the phase of the signal, owing

to the phase relation in Eq. (1). Simulations of the MOPO

pumped with a wave containing random phase modulation give

qualitatively similar results, i.e., the convective phase-locking

mechanism ensures the generation of a narrowband coherent

counterpropagating idler wave, while the signal inherits the

random phase modulation from the pump.

V. CONCLUSIONS

FIG. 4. Phase distribution inside the crystal for the pump, the

signal (locked to the pump), and the backward idler (invariant). The

x axis is in units of Lnl = vg,p τ0 = 1 mm.

Simulations of the MOPO at a pump intensity of

1.1 GW/cm2 reveal that the oscillation develops after an initial

delay of 300 ps and TWM starts depleting the trailing end of the

pump pulse where the high-frequency components are located

due to the chirp. This can be seen in the comparison between

the input and output pump spectra in Fig. 3(a). Figures 3(b)

and 3(c) show that the generated signal obtains a FWHM

spectral width of 440 GHz, whereas that of the idler is only

6.5 GHz, i.e., about 200 times narrower than the initial pump

spectrum and 70 times narrower than the signal. The strong

asymmetry in the spectral widths of the signal and idler is

related to the amount of phase modulation in the waves. This

is illustrated in Fig. 4, where the phase of the idler wave

shows very little variation along the crystal, while the phases

of the signal and pump remain locked. Qualitatively one can

[1] A. Bers, D. J. Kaup, and H. Reiman, Phys. Rev. Lett. 37, 182

(1976).

[2] M. N. Rosenbluth, Phys. Rev. Lett. 29, 565 (1972).

[3] F. P. Bretherton, J. Fluid Mech. 20, 457 (1964).

[4] M. S. Longuet-Higgins and A. E. Gill, Proc. R. Soc. London

Ser. A 299, 120 (1967).

[5] N. S. Shiren, Proc. IEEE 53, 1540 (1965).

[6] L. O. Svaasand, Appl. Phys. Lett. 15, 300 (1969).

[7] H. Y. Ling, in Coherence and Quantum Optics VIII, edited by

N. P. Bigelow, J. H. Eberly, C. R. Stroud, and A. Walmsley

(Kluwer Academic Plenum, Dordrecht, 2003), pp. 573–574.

[8] J. A. Armstrong, N. Bloembergen, J. Ducuing, and P. S. Pershan,

Phys. Rev. 127, 1918 (1962).

[9] Z. Major et al., Rev. Laser Eng. 37, 431 (2009).

¨

[10] T. Yamamoto, M. Koashi, S. K. Ozdemir,

and N. Imoto, Nature

(London) 421, 343 (2003).

[11] L. A. Wu, H. J. Kimble, J. L. Hall, and H. Wu, Phys. Rev. Lett.

57, 2520 (1986).

[12] W. H. Louisell, A. Yariv, and A. E. Siegman, Phys. Rev. 124,

1646 (1961).

In conclusion, we have employed a mirrorless OPO

realized in a submicrometer periodically poled KTP crystal

for experimental demonstration that the phase of the pump

wave is coherently transferred to the co-propagating OPO

signal, as a result of the maximized convective separation of

the pump and idler pulse fronts. This convective mechanism,

as was elucidated by the numerical modeling, gives rise to a

narrowband counterpropagating idler wave, regardless of the

details of the pump phase.

ACKNOWLEDGMENTS

This work has been supported by grants from the Swedish

Research Council (VR), the Knut and Alice Wallenberg Foundation, and the G¨oran Gustafsson Foundation. The authors

acknowledge the GDR PhoNoMi2 No. 3073 of the CNRS

(Centre National de la Recherche Scientifique) devoted to

Nonlinear Photonics in Microstructured Materials. The authors

are thankful for useful discussions with Fredrik Laurell, Pierre

Aschieri, and Antonio Picozzi.

[13] R. J. Glauber, Quantum Theory of Optical Coherence (WileyVCH, New York, 2007).

[14] A. Baltuska, T. Fuji, and T. Kobayashi, Phys. Rev. Lett. 88,

133901 (2002).

[15] B. Dayan, A. Peer, A. A. Friesem and Y. Silberberg, Phys. Rev.

Lett. 93, 023005 (2004).

[16] M. Henriksson, M. Tiihonen, V. Pasiskevicius, and F. Laurell,

Opt. Lett. 31, 1878 (2006).

[17] A. Picozzi and M. Hælterman, Phys. Rev. Lett. 86, 2010

(2001).

[18] A. Picozzi, C. Montes, and M. Hælterman, Phys. Rev. E 66,

056605 (2002).

[19] Y. J. Ding and J. B. Khurgin, IEEE J. Quantum Electron. 32,

1574 (1996).

[20] S. E. Harris, Appl. Phys. Lett. 9, 114 (1966).

[21] C. Canalias and V. Pasiskevicius, Nature Photonics 1, 459

(2007).

[22] C. Canalias, V. Pasiskevicius, R. Clemens, and F. Laurell, Appl.

Phys. Lett. 82, 4233 (2003).

[23] K. Kato and E. Takaoka, Appl. Opt. 41, 5040 (2002).

023825-4