1046
OPTICS LETTERS / Vol. 19, No. 14 / July 15, 1994
Continuous-wave total-internal-reflection optical
parametric oscillator pumped at 1064 nm
D. K. Serkland, R. C. Eckardt, and R. L. Byer
Department of Applied Physics, Stanford University, Stanford, California 94305
Received February 7, 1994
We report what is to our knowledge the first cw 1064-nm-pumped optical parametric oscillator (OPO), using a
critically phase-matched (0pm = 45.5), monolithic total-internal-reflection, doubly resonant OPO fabricated from
congruent LiNbO3 . Frustrated total-internal-reflection provides variable output coupling for the signal and the
idler.
The measured finesse of 6000 at 2014 nm implies a round-trip power loss of 0.1%.
The low round-trip
power losses compensate the 20 Poynting vector walk-off and give a threshold of 130 mW. A passive thermal
feedback mechanism causes the OPO to oscillate stably in a single axial-mode pair for more than 30 min.
The
OPO output tunes in 0.1-nm steps from 2040 to 2225 nm by tuning the pump frequency.
Until quite recently, only one type of cw optical
parametric oscillator (OPO) existed: doubly resonant OPO's (DRO's) that phase matched with little
or no Poynting vector walk-off.' Critically phasematched OPO's have generally required high-peakpower Q-switched pump lasers to exceed threshold.
In this Letter we report operation of a cw DRO that
has significant Poynting vector walk-off.
The single frequency, narrow linewidth, and
high output power of cw 1064-nm Nd:YAG lasers
makes them ideal OPO pump sources. However, the properties of available low-loss nonlinear materials have forced researchers first
to frequency double the 1064-nm laser output and then pump their OPO's with 532-nm
radiation.
Pumping available nonlinear crystals
at 1064 nm requires critical phase matching, which
reduces parametric gain because of walk-off. The
low losses of our total-internal-reflection
(TIR) cav-
ity reduced the threshold to a level attainable by
modest cw pump lasers, enabling us to operate a
cw 1064-nm-pumped OPO, with oscillation near the
2128-nm degenerate wavelength.
Examination of the OPO threshold equation reveals the difficulty of direct 1064-nm pumping. The
threshold pump power required for oscillation is
given by
Pth =
K d 2a~a
lhm
(1)
where AP is the pump wavelength, a, and ai are
the round-trip power losses at the signal and idler
wavelengths, d is the effective nonlinearity of the
crystal, I is the phase-matched interaction length,
hm is a generalized Boyd-Kleinman gain reduction
factor, and
K
=
8,2(1
-
,2)2
(2)
is a slowly varying function of frequency near degeneracy since y = (ws - &w,)/1jP.The Boyd-Kleinman
0146-9592/94/141046-03$6.00/0
hm factor represents a reduction in effective parametric gain because of poor pump-signal-idler mode
overlap, which results from Poynting vector walk-off
(for non-90' phase matching) and/or nonideal pump
focusing.
Researchers
typically achieve hm
1 for
noncritical (900) phase matching. Clearly Poynting
vector walk-off and long pump wavelengths increase
threshold.
We selected LiNbO3 for our nearly degenerate,
1064-nm-pumped cw DRO since it has low losses and
a relatively large nonlinearity. However, noncritical
phase matching would require heating the crystal
to near 550'C. We chose to operate at a more
convenient temperature of 40'C, where the phasematching angle is 45.5°, which results in a 20 Poynting vector walk-off and hm - 0.05, increasing the
threshold by a factor of 20. The pump wavelength
scaling in Eq. (1) leads to an additional factor-of-8
increase in threshold compared with a green-pumped
OPO. Combining these factors implies that the
threshold power for a 1064-nm-pumped LiNbO3 OPO
is inherently 160 times higher than for a comparable green-pumped OPO. Since typical thresholds for
green-pumped LiNbO 3 OPO's are 20 mW, we expect a
3.2-Wthreshold for a 1064-nm-pumped LiNbG3 OPO.
We strove to reduce threshold by lowering the
losses, selecting a monolithic TIR design to eliminate unnecessary interfaces and dielectric coatings.
Figure 1 shows a schematic of our TIR cavity. The
thick lines in Fig. 1 indicate the pump beam Poynting
vector direction. The 1064-nm pump beam enters
through the spherical surface with 85% transmission. Because of bireflection and walk-off the pump
does not form a closed path; it makes two round
trips before leaving the crystal with -85% transmission. The cavity geometry defines a unique, 16mm triangular ring path for the ordinary-polarized
signal and idler waves, which experience TIR at
all three surfaces. The two flat TIR surfaces are
polished with a 117.50 angle between their surfacenormal vectors. The spherical surface has a 51mm radius of curvature, which forms a stable cavity. We use a SF-6 glass prism (index 1.756 at
© 1994 Optical Society of America
July 15, 1994 / Vol. 19, No. 14 / OPTICS LETTERS
H
PbS
To oscilloscope
Fig. 1. Schematic of our LiNbO 3 TIR OPO. All three
surfaces give TIR at 2.1 ,um. We use a SF-6 glass prism
for frustrated TIR output coupling. Phase matching occurs only along the top segment of the triangular path.
The pump enters through the curved surface with 85%
transmission.
2128 nm) for frustrated TIR output coupling the signal and idler waves.5 A prism-crystal separation
of one wavelength gives 0.16% transmission,
which
decreases exponentially to zero with increasing
separation.
Input coupling of the pump radiation poses a challenge for TIR oscillators since TIR occurs at all surfaces. In this Letter we introduce a novel method
for input coupling of pump radiation that works for
critically (non-90') phase-matched interactions. In
Fig. 1, phase matching occurs only along the upper
segment of the triangular path, where n = 2.1928.
Along the other two segments, the extraordinary
pump wave sees a different index. Along the righthand segment, the pump index is only 2.164 (close
to the pure extraordinary value). Because the pump
index is lower than the signal and the idler indices, the critical angle for pump TIR is larger. We
have chosen the internal angle at the spherical surface to be slightly greater than the 2128-nm critical angle but less than the 1064-nm critical angle.
Figure 2 illustrates that, at our chosen angle of
incidence, the 1064-nm pump passes through the
spherical face near Brewster's angle, with 85% transmission, while we retain TIR at 2128 nm. Although
all k vectors are collinear along the phase-matched
segment, bireflection and walk-off cause a shift in
the 1064-nm internal angle at the spherical surface,
compared with the signal/idler internal angle. Another method of introducing pump radiation has been
demonstrated by Schiller and Byer,6 in which external resonant second-harmonic generation produces
pump radiation inside the crystal. A third method,
proposed by Fejer,7 employs an additional, birefringent prism. One orients the birefringent prism so
that the pump polarization sees the larger index and
passes through the interface, while the signal and
the idler polarizations see the smaller index and suf-
1047
fer TIR. Fiedler et al. recently employed birefringent
prisms for second-harmonic-generation coupling."
Before pumping the OPO, we probed the cavity
with a 2014-nm single-frequency Tm:YAG laser to
measure the loss at 2 1ttm and to aid in alignment
of the OPO. We resonantly frequency doubled the
2014-nm radiation in our TIR cavity to determine
the phase-matching temperature. 9 The SHG experiment revealed a 1800 misorientation of an earlier
LiNbO3 crystal, which reduced the effective nonlinearity by a factor of 4 and prohibited a low-threshold
OPO demonstration. We subsequently fabricated
a second LiNbO3 TIR resonator and measured an
asymptotic finesse (as the coupling approaches zero)
of 6000. This finesse corresponds to a 0.1% roundtrip power loss inside the LiNbO3 crystal, to our
knowledge the lowest loss reported to date for a
nonlinear cavity. Frequency doubling of 2014-nm
radiation phase matched at 90'C, probably because
of a slight rotation of the crystal during fabrication.
Irises placed along the 1007-nm second-harmonic
output gave us alignment marks for the 1064-nm
pump beam.
We reduced the threshold to a minimum by separating the prism from the crystal to produce zero
output coupling. We scanned the pump frequency
repetitively over a 12-GHz range and looked for para-
metrically generated 2.1-,ttm radiation scattered onto
a PbS detector placed near the crystal (see Fig. 1).
We observed several spikes as successive signal/idler
mode pairs satisfied phase-matching and cavity
resonance conditions. By rotating the pump polarization until oscillation ceased we inferred a threshold of 130 mW.
We restored
the optimum pump
polarization and increased the output coupling, by
moving the prism toward the LiNbO3 crystal, so as
to maximize the output power. We observed 5 mW
of total output power with a 300-mW pump source.
We also measured a pump depletion of 12%.
A theoretical prediction of threshold requires extension of the Boyd-Kleinman analysis to the case
of arbitrary focusing conditions. Our resonator has
elliptical signal and idler modes, with degenerate
(tan) = 73 jm.
A
waist sizes wo(sag) = 78 /p&mand wo
1.2
.>
0.8
Ut
8)
2!
8)
8)
0.6
0.4
0.2
0L'
25
25.5
26
26.5
27
27.5
28
Internal angle (degrees)
Fig. 2. Fresnel power reflectivities for s-polarized
2128-nm waves and p-polarized 1064-nm waves versus
internal angle of incidence. The internal angle for
2128 nm is 27.5°. The internal angle for 1064 nm is
shifted because of bireflection and walk-off.
1048
OPTICS LETTERS / Vol. 19, No. 14 / July 15, 1994
100
500 MHz
75
0
signal
/2067 nm
50
idler
2193 nm
U,
E.0
25
0
-100
I
l
l
-50
'I U
|
0
l~-
50
100
PZT voltage (V)
Fig. 3. Power transmission through a scanning confocal
Fabry-Perot interferometer as the voltage applied to a
piezoelectric transducer (PZT) is linearly ramped. The
free spectral range is 500 MHz.
single mode-matching lens focuses the pump to an
elliptical waist at the entrance face, with waist sizes
(sag)
wo
=
(tan)
96 /im and wo
= 74 tzm.
A paraxial ray
analysis shows that a tight tangential focus forms
far after the phase-matched region. To calculate
km we evaluate a general mode-overlap integral,
analytically in the transverse dimensions and numerically in the longitudinal dimension. We find
that hm = 0.028. We then calculate the threshold
assuming 85% pump transmission into the crystal and d = d15 sin E) + d22 cos 0 = 5.95 pm/V
(taking d15 = 5.02 pm/V and d2 2 = 3.31 pm/V). Assuming the round-trip
power loss to be 0.1%, we
calculate a 45-mW threshold. However, we expect
more loss at 2128 nm than at 2014 nm since index
dispersion causes longer wavelength resonator modes
to have higher losses, because of their larger TIR critical angle. The measured 130-mW threshold implies
a round-trip power loss of 0.17% at 2128 nm.
Figure 3 shows the OPO output, observed with a
scanning confocalFabry-Perot interferometer, which
clearly indicates that the DRO oscillates stably in a
single axial-mode pair without active feedback. We
inferred signal and idler wavelengths from the spacing of the transmission peaks in Fig. 3. We readily
adjusted the signal/idler wavelengths by tuning the
pump frequency over a 1-GHz range. Excluding instances in which cluster hops occur, tuning the pump
frequency by -50 kHz causes nearly degenerate signal and idler frequencies to hop to adjacent axial
modes, with a mode spacing of 8.5 GHz, or 0.13 nm.
We measured the DRO output wavelengths with a
1-m grating spectrometer and found a total range of
2040 to 2225 nm, for various crystal temperatures
between 80 and 85°C. Oscillation ceased for signal wavelengths longer than 2225 nm because higher
resonator losses raised threshold above the available
300 mW of pump power.
The DRO exhibits remarkable power stability (a
few percent fluctuations), which we credit to an
internal thermal feedback mechanism, caused by
absorption at the signal and idler frequencies. We
hypothesize that the DRO oscillates stably with the
signal and the idler on the high-frequency side of two
cavity resonances, where thermal feedback opposes
perturbations. For example, suppose that the environment heats the LiNbO3 crystal, thereby shifting
all resonances to lower frequencies. The signal and
idler experience more loss since the resonances are
farther away. The circulating signal and idler powers decrease, which results in less absorptive heating.
Thus the crystal cools, counteracting the original perturbation. This passive thermal feedback appears to
have a time constant shorter than 1 ms.
In summary, we have constructed an extremely
low-loss nonlinear cavity that enabled us to demonstrate a 1064-nm-pumped cw DRO. Variable coupling by frustrated TIR has proved advantageous,
allowing us to achieve a 130-mW threshold with 0%
output coupling. Adjusting the coupling to maximize
output produced 5 mW of total output power with a
300-mW pump. We have proposed an intrinsic thermal feedback mechanism that explains the observed
stable oscillation in a single axial-mode pair for more
than 30 min. We will continue to investigate the
DRO tuning properties and plan to measure a sig-
nal-idler beat note.
We gratefully acknowledge the generosity of
T. Kubo, of Lightwave Electronics, for loaning us
a Tm:YAG laser and T. Day, of New Focus, for loaning us an electro-optic modulator. Joe Vrhel and
Greg Mizell deserve credit for carefully fabricating
our monolithic cavities, using congruent LiNbO3 from
Crystal Technology.
We thank A. D. Farinas, E. K.
Gustafson, and M. M. Fejer for helpful discussions.
D. K. Serkland thanks IBM for providing fellowship
support during part of this research. We also acknowledge the support of the U.S. Army Research Office, through grant DAAL03-90-C-0026,and the U.S.
Office of Naval Research, through grant N00014-92J-1903.
References
1. For reviews of OPO's, see R. L. Byer, in Treatise in
Quantum
Electronics,
H. Rabin and C. L. Tang, eds.
(Academic, New York, 1973), p. 587; R. G. Smith, in
Advances in Lasers, A. K. Levine and A. J. DeMaria,
eds. (Dekker, New York, 1976), Vol. 4, p. 189.
2. C. D. Nabors, R. C. Echardt, W. J. Kozlovsky, and
R. L. Byer, Opt. Lett. 14, 1134 (1989).
3. S. T. Yang, R. C. Eckardt, and R. L. Byer, Opt. Lett.
18, 971 (1993).
4. G. D. Boyd and D. A. Kleinman, J. Appl. Phys. 39, 3597
(1968).
5. For a review of frustrated TIR, see S. Zhu, A. W. Yu,
D. Hawley, and R. Roy, Am. J. Phys. 54, 601 (1986).
6. S. Schiller and R. L. Byer, J. Opt. Soc. Am. B 10, 1696
(1993).
7. M. M. Fejer, E. L. Ginzton Laboratory, Stanford University, Stanford, Calif. 94305 (personal communication, March, 1991).
8. K. Fiedler, S. Schiller, R. Paschotta, P. Kfirz, and
J. Mlynek, Opt. Lett. 18, 1786 (1993).
9. D. Serkland, R. C. Eckardt, and R. L. Byer, in Conference on Lasers and Electro-Optics, Vol. 12 of 1992
OSA Technical Digest Series (Optical Society of America, Washington, D.C., 1992), p. 530.