1. Introduction

Ferroelectric crystals with periodic domain structures can be used to create intracavity elements that modify the emission of a laser. For example, they can be used to change its wavelength, phase or intensity, to switch it on and off, among other functions. In 2002 Zhang et al. [1] reported a green laser that emits pulses with a duration of 15.6 ns and up to 1.96 μJ of energy using a periodically poled lithium niobate (PPLN) crystal intracavity to obtain the second harmonic of a diode pumped Nd:YVO₄ laser and a saturable absorber of Cr⁴⁺:YAG to produce Q-switched output. Jiao et al. [2] reported the fabrication of a Nd:YAG laser with an intracavity optical parametric oscillator (OPO) to simultaneously obtain emission of 7.1 W at 2100 nm and 23.5 W at 1064 nm. Other works take advantage of the sign reversal of the electro-optic coefficient between anti-parallel ferroelectric domains. For example, electro-optic deflectors have been made by applying an electric field on an interface of anti-parallel ferroelectric domains; the abrupt change of the refractive index at the interface causes a deflection of an incident beam [3,4]. The direction of propagation of a beam can also be controlled with a series of prism-shaped domains; this idea has been used to make other devices, such as beam scanners [5–8]. Chiu et al. [9] developed an integrated optical device based on domain structured lithium tantalate that accomplishes second harmonic generation, electro-optic lensing and scanning. Quasi-phase-matching and beam-deflecting structures can be combined in a single device. Chen et al. [10] used a periodically poled Nd:MgO:LiNbO₃ crystal to obtain a diode pumped Q-switched laser that emits at 1085 nm; the Q-switching was produced by inducing birefringence with an electric field applied along the $y$ faces of the crystal. Chang et al. [11] reported the use of a two-dimensional domain structure that acts as both a Bragg reflector and wavelength converter, with which they electro-optically Q-switch a Nd:YVO₄ laser and produce frequency mixing, obtaining emission at 1550 nm with pulse energy of 9.7 µJ.

In this work we present an intracavity Q-switch and periodically poled lithium niobate crystal (QPPLN) that is used to simultaneously Q-switch a laser cavity and convert the laser pulse into another wavelength. The Q-switching is performed by introducing losses via electro-optic deflection and the wavelength conversion is produced by quasi-phase-matched optical parametric generation (OPG). The crystal has two different regions, one consisting of a series of single-domain triangles and another consisting of triangles with an internal periodically poled structure, as shown in Fig. 1. The “c” faces of the crystal have electrodes on them, and when an electric field $E$ is applied to these electrodes the extraordinary refractive index $n_e$ changes by an amount $\Delta n_e$ given by [6]:

 

Fig. 1 QPPLN. The single-domain triangles act as voltage-controlled deflectors and the triangles with periodic domain structures act as wavelength converters. The yellow region is the electrode on the top “c” face; a similar electrode exists on the bottom (not shown).

Download Full Size | PDF

The applied field turns each of the single-domain triangles into prisms capable of deflecting light. The refractive index of the triangles with the internal domain structure also changes; however, since for every triangle there are many alternating 180° domains the index change averaged over each triangle is close to zero (assuming a 50/50 mark-to-space ratio), so the deflection that may be produced by them is negligible. Under the paraxial approximation, a simple calculation shows that the external angle of deflection ${\theta }_{ext}$of an extraordinary polarized beam normally incident on the center of the deflector is given by:

It is straightforward to show that for a beam incident on the center of the input face the conversion efficiency of this crystal corresponds to that which would be obtained with a full periodic structure of half the length $L$, provided that there are many domains in each triangle; this is simply because only half of the crystal is performing the wavelength conversion. However, the wavelength selectivity of the conversion process is the same one obtains with a sample of the same length with a full periodic structure.

A similar device that also is used to Q-switch a laser and convert the output pulse into another wavelength has been reported previously by Huang et al. [12]; the main difference is that in their device the deflection is produced by applying voltage to triangular domains that have a small apex angle; the spacing between these domains determines which quasi-phase-matched process is produced. In principle, for a given applied voltage and crystal length the deflection should be twice the value given in Eq. (1) and the conversion efficiency should be higher than what can be obtained with our device. However, the smaller the periodicity is, the smaller is the apex angle and consequently the harder it is to make these domains with a well-defined shape, which can degrade both the deflection and the conversion efficiency; the device we present does not have this drawback.

2. Experiment

We produced a sample with several QPPLN structures in a 0.5 mm thick congruently grown, LiNbO₃ wafer. The QPPLN structure had a periodicity$\Lambda =30.6$ µm, designed to convert 1064 nm radiation into a ~1600 nm signal (the exact value depends on the temperature). This sample was made using standard electric field poling techniques [13,14]. Two sides parallel to the y-axis were cut and polished; anti-reflection coatings for 1064 nm were deposited on both of these polished faces, and aluminum electrodes were deposited on both c-faces in order to apply an electric field to the sample. The length L of the structures was 30 mm and the width $W$ of each structure was 0.45 mm. The apex angle of each triangle was 120°. This crystal was placed inside a laser cavity comprised of two concave mirrors and a Nd:YAG rod, pumped by a 880 nm fiber-coupled laser diode, as shown in Fig. 2. The output of the fiber (200 µm diameter, NA = 0.22) was collimated by a 25 mm focal length lens and focused by a 20 mm focal length lens onto the Nd:YAG crystal (beam radius ~0.1 mm). The Nd:YAG (1.0% dopant concentration) was 5 mm long, 3 mm diameter with 1064 nm antireflection coatings on each face. The mirror M1 had a radius of curvature of 50 mm, a >99% reflectivity at 1064 nm and an anti-reflection coating for the pump; the output coupler M2 had a radius of curvature of 50 mm, 96% reflectivity at 1064 nm and 32% transmittance at 1600 nm. The separation,$l$, between mirrors M1 and M2 was 45 mm.

 

Fig. 2 Experimental set-up. The signal and idler wavelengths are at $T=60°C$.

Download Full Size | PDF

The cavity was aligned such that without applying voltage across the QPPLN crystal the laser operated in a continuous-wave mode. In order to produce Q-switching, a constant voltage (~1800 V) was applied to inhibit oscillation. According to Eq. (2) and using the value of $r_{33}$ given in [15], this corresponds to an angular deflection of ~1.1° for extraordinary polarized beams. The pulses were obtained by dropping the voltage to zero, thus allowing oscillation. The width of these zero-voltage pulses was 50 µs and the repetition rate was 40 Hz. The QPPLN capacitance (0.38 nF) and the internal impedance of the high voltage supply used in this experiment limited the decay time of the voltage pulses to ~30 ns.

The laser threshold occurs at ~1.2 W of absorbed pump power. Let $x$ be the ratio between the absorbed pump power and the absorbed pump power at threshold. At $x=1$, only 1064 nm radiation is observed. At $x\approx 1.7$ and at room temperature, an OPG signal at 1602 nm appears, as well as the second harmonic of the laser (532 nm), a red (~639 nm) beam due to sum-frequency generation between the 1064 and 1602 nm beams, and a beam at 801 nm due to second harmonic of the 1602 nm signal; these beams are detrimental to the spatial quality of the 1064 and 1602 nm beams since they cause photorefractive damage in LiNbO₃ [16]. To reduce the photorefractive damage the QPPLN crystal was warmed to 60 °C, which shifted the OPG signal to 1617 nm. Figure 3 shows the observable infrared spectrum obtained when the laser is operated in the pulsed mode at $x\approx 1.93$.

 

Fig. 3 Infrared spectrum obtained with Q-switching at $x\approx 1.93$. a) Spectrum of laser and OPG signals at $T=25{}^{\circ }C$; b) OPG signal at $T=25{}^{\circ }C$ and $T=60{}^{\circ }C$.

Download Full Size | PDF

Regardless of the wavelength, all of the pulses are polarized (>125:1) along the $"c"$ axis of the QPPLN crystal. In LiNbO₃, the $d_{33}$coefficient is much larger than all of the other nonlinear coefficients, which explains why the beams generated through nonlinear processes are well polarized along “c.” However, Nd:YAG by itself should not and does not impose a polarization on the beams; this was verified by operating the laser in CW mode with the QPPLN inside the cavity but without applying a field, obtaining an essentially depolarized output. If a low CW voltage (~800 V) is applied, the laser emits an ordinary polarized CW 1064 nm beam. That is because the deflection for this polarization is less since $r_{13}\approx r_{33}/3$, which isn’t sufficient to quench the oscillation. However, for CW voltages above 1200 V no oscillation in either polarization is observed, at least for a pumping factor $x<2$. It seems that the stronger deflection produced by $r_{33}$ - and therefore faster Q-switching when the voltage is turned off - is somehow responsible for the polarization of the 1064 nm pulses.

The temporal profiles of the 1064 nm pulses for different values of $x$ are shown in Fig. 4. Here the time origin is the moment the voltage applied to the QPPLN is reduced to zero; we see that it takes ~1 µs for the pulses to build up and this delay decreases with increasing values of $x$, as expected from standard active Q-switching theory [17]. Also, we observed that for $x\approx 2.10$ and below, the pulses have a Gaussian-like temporal profile; above these values the profile is distorted, as can be seen for the case of $x=2.60$; we ascribe this to depletion of the 1064 nm pulse, due to energy transfer to other wavelengths, mainly to the signal and idler wavelengths of the OPG process.

 

Fig. 4 Pulse shape of the 1064 nm signal at different values of $x$.

Download Full Size | PDF

Figure 5 shows the synchronization of the depletion of the 1064 nm beam with the onset of optical parametric generation. In Fig. 5(a)) the data was taken at $x=2.34$. The appearance of the signal and the idler (not shown) beams depletes the 1064 nm pump. In Fig. 5(b)), taken at $x\approx 2.60$, the same behavior is observed; however, in this case once the pump is depleted significantly the OPG signal disappears, allowing the pump to recover, producing a second peak of the signal. The shape of the pulses varies slightly from pulse to pulse since they originate from noise. This is why the pulses shown in Fig. 4 and Fig. 5(b)) are slightly different, even though they were taken at the same value of $x$.

 

Fig. 5 Synchronization of 1064 nm pump depletion and the onset of OPG. a) $x=2.34$ and b) $x=2.60$. The scales of the intensities are not the same for the two wavelengths; $t=0$ corresponds to the peak of the pump beam. $T=60{}^{\circ }C$.

Download Full Size | PDF

Figure 6 shows the pulsewidths and the output energy per pulse of the 1064 and 1617 nm beams. Figure 6(a)) shows the FWHM pulsewidth; as expected, the duration of the 1064 nm pulses (blue circles) decreases with $x$. The lowest pulsewidth shown in the graph is 9 ns, obtained at $x=2.53$. Data for $x>2.53$ is not presented because the depletion caused by OPG distorts the shape of the pulse so much that this measurement is pointless. The black line is the expected pulsewidth according to active Q-switching theory [17], assuming a 45 mm long cavity (taking into consideration the refractive indices of the components) and total losses of 25%. The reflectivity at 1064 nm of the output coupler is large, 96%, so most of the cavity losses must be due to other sources, mainly scattering inside the QPPLN crystal, quite possibly at the domain walls, and on its surfaces. The pulsewidths of the 1617 nm signal (red circles) are much shorter, ~3 ns at$x=2.53$. The output energy per pulse vs. $x$ at 1064 and 1617 nm is shown in Fig. 6(b)). In order to make sure that we were measuring the energy of pulses and not an integration of a continuous background, these data were measured with a pyroelectric detector, which requires a low repetition rate, <40 Hz. However, the laser was successfully operated at higher (up to 1.5 kHz) and lower (down to 1 Hz) repetition rates; these limits were imposed by the power supply. At 40 Hz we obtained pulses with up to 195 µJ at 1064 nm and 15 µJ at 1617 nm, as can be seen in the figure.

 

Fig. 6 Pulsewidth and energy vs.$x$. a) FWHM pulsewidth; b) Output energy per pulse. Data taken at a repetition rate of 40 Hz. $T=60°C$.

Download Full Size | PDF

3. Concluding remarks

We have presented a device that both Q-switches a laser and converts 1064 nm into 1617 nm radiation by optical parametric generation; the same idea can be used to obtain other wavelengths or even cascaded nonlinear interactions. A drawback of this crystal is that photorefractive damage occurs, mainly due to the parasitic 532 nm beam that is generated, which distorts the wavefront of the beams generated inside the cavity. We reduced this problem by operating the crystal at 60 °C; however, even at this temperature the spatial quality of the beam would distort over time, which also reduced the energy per pulse. A way to circumvent this problem would be to use another material less prone to photorefractive damage, such as a MgO:LiNbO₃ crystal [18], or to operate the crystal at a higher temperature.