1. Introduction

In most applications of surface emitting lasers such as disk lasers, microchip lasers, VCSELs and polymer dye lasers, there is a need for an integrated element which controls the emitted polarization. The planar structure of such lasers offers strong configurational and manufacturability advantages, which forbids for instance the use of a bulky Brewster element. Multilayer mirrors are usually not polarization selective. Grating coupling of one of the intra-cavity polarizations to a guided mode of a multilayer mirror was proposed [1] as a means of inducing selective reflection losses in the coupled polarization, since the guided TE or TM guided slab modes of a multilayer stack are most often non-degenerate. Since the discovery [2] and analysis [3] of the effect of zero^th order resonant reflection from a grating slab waveguide, it is possible to exert the polarization control of laser emission by highly polarization selective reflection as first demonstrated by Sychugov et al. [4] in the case of a semiconductor laser with external cavity mirror. Zero^th-order resonant reflection can theoretically reach 100% for one polarization [5]. This means that a single corrugated high index layer could be at once both the needed high reflectivity mirror of the laser and the monolithically integrated linearly polarizing element. However, there are two reasons for rejecting such a single layer mirror scheme. The first reason is related to the strong field accumulation at resonance which would lead to a dramatic reduction of the laser damage threshold. The second reason is not of concern in the present structure where the resonant grating is associated with the laser output coupler, but it is worth mentioning: it is difficult to obtain more than about 98% reflection in practice. The reasons are scattering losses due to an imperfect corrugation, absorption losses of the waveguide and usually imperfect matching between the incident beam size and the propagation length in the corrugated waveguide.

The present paper demonstrates a polarizing mirror solution which takes the best of the two mirror technologies: the well established non-selective multilayer technology which ensures most of the needed reflection, and the corrugated slab technology which does not reflect much light but essentially produces the polarization selectivity. Stimulated emission does the rest by amplifying the sole polarization which experiences the larger reflection coefficient. The specific feature in the present combination of the multilayer and resonant submirrors is the destructive interference condition between fields which each submirror reflects which implies that it is the polarization which experiences resonant reflection which is filtered out.

This is not the only approach for monolithically polarizing laser emission by means of a periodic structure: first, referring to the above, the combination between the reflection product of each submirror can have a constructive character by adjusting the spacing between submirrors and lead to the lasing of the polarization experiencing waveguide coupling and resonant reflection [6] ; there are important differences in terms of bandwidth and power flux resistance. There are other polarizing mechanisms as well: for instance, a deep binary grating etched into a thick high index layer such as a semiconductor layer in a VCSEL can exhibit 100% reflectivity for the TM polarization according to the GIRO mirror scheme [7]. A shallow grating of relatively large period can also induce a reflection differential as demonstrated in VCSELs [8]; such an approach is however risky as a large period gives rise to propagating diffraction orders or couples the incident wave to unforeseeable guided modes of the multilayer whose density may be large. Another, more predictive approach, uses deep trenches into the multilayer with the objective of giving rise to photonic crystal effects [9]; the risk here is the scattering losses associated with an imperfect trench pattern of very high index contrast. The present approach has none of the above mentioned handicaps: it is predictive in that it excites a well identified mode of the multilayer without losing power in radiated diffraction orders and it is much less traumatic than a photonic crystal to the mirror and photon generation process, since it is a soft post-process on top of an essentially standard multilayer. As compared with the alternative scheme using constructive interference between reflection products [6], the present destructive scheme can sustain larger power flux before damage occurs [10], and can be more robust regarding photothermal drift of the operation point.

2. Associating multilayer and resonant submirrors

Both low mirror loss and polarization selectivity requirements can be satisfied by resorting to a proper combination of a wide band non-selective multilayer submirror and a resonant grating placed on top of the latter [11]. The multilayer submirror ensures a low-loss, high-reflection pedestal, whereas the resonant mirror adds its polarization selectivity to the smaller part of the reflection it provides; the key point in the present design is that this polarization selective reflection is in phase opposition relative to the multilayer reflected field and leads to a polarization selective damping of the overall reflection. The relative part of the reflection coefficient requested from each submirror depends on the reflection differential needed in a given laser cavity scheme to suppress the undesired polarization. The reflection pedestal provided by the multilayer submirror is set at the reflection level needed for laser operation of the desired polarization whereas the reflection dip due to waveguide coupling and resonant reflection must have a depth below threshold to prevent the amplification of the undesired polarization. The sketch of the superposition of the two submirrors is shown in Fig. 1.

 

Fig. 1. Destructive interference polarizing mechanism resulting from the association of a multilayer submirror and a polarization selective resonant submirror.

Download Full Size | PDF

The resonant submirror is here represented by a single high-index slab; however, it can itself be a bilayer or even a multilayer having desired dispersion properties, for instance to broaden the resonance, i.e., to filter out the undesired polarization over the whole gain bandwidth of the active material. The two submirrors are separated by a low-index buffer, whose thickness w_b is adjusted on the basis of the following criteria: first, the buffer layer thickness dictates the phase shiftΔϕ between the field reflected by the multilayer and that reflected by the resonant grating. If the grating is made at the cover side of the mirror (usually at the air side), Δϕ is given by

where k ₀ = 2π/λ. is the wave number in the cover (here considered as air), n _b and n _w are the refractive index of the buffer and waveguide slab layers of thickness w _b and w _w respectively. The refractive index n _b and n _w can be the same as those of the low- and high-index multilayer, n _l and n _h, if the technology of the submirrors is the same. They can be different in case the multilayer submirror technology is epitaxial growth and the resonant mirror technology is a metal oxide post-process; this would be the preferred option for the control of polarization of VCSELs. The last term of expression (1) is the phaseshift upon resonant reflection known to be π [4]. In the present case Δϕ must be close to π modulo 2π for destructive interference between the multilayer and resonant submirror reflected fields as illustrated in Fig. 1: the polarization which is coupled to the slab waveguide does not lase since its reflection coefficient is decreased by destructive interference, whereas the non-coupled polarization “sees” the multilayer as if there were no grating (except for some scattering which will be discussed later) and will lase since the reflection pedestal level is set accordingly.

3. Design of the polarization selective mirror

As a practical example for the implementation of such a device, the polarizing function will be monolithically integrated into the multilayer mirror of a passively Q-switched microchip laser (Nd:YAG as the gain medium, bonded to a Cr:YAG saturable absorber). This example was chosen because of its many industrial applications, most of which require polarized output beam. Current laser systems are based on creating uniaxial optical properties in the otherwise isotropic microchip, either by applying a mechanical stress, or by creating a unidirectional thermal gradient perpendicular to the optical axis of the cavity. One alternative would be to use anisotropic laser material, such as Nd:YVO₄ ; however, the intrinsic material properties (e.g. gain and absorption cross-sections) restrains the achievable laser performances (repetition rate, pulse duration and energy), so that YAG remains the material of choice for most of the industrial applications. As a consequence, an independent, reliable, compact and cost-effective way for controlling the polarization of the output beam remains an open question.

The incidence is from the substrate side as suggested in Fig. 1. The substrate index at the emission wavelength of 1064 nm is taken as n_s = 1.82. The multilayer mirror technology is HfO₂/SiO₂ sputtering with assumed layer refractive indices of 2.10/1.48 as obtained by ion plating. The targeted reflection coefficient of the laser output coupler for the lasing polarization, i.e. the multilayer pedestal, is set at about 85%, whereas the polarization selective resonant dip is set at 60% reflection at most to prevent the lasing of the undesired polarization.

A first criterion for the choice of a mode to give rise to resonant reflection is its resonant width. A wide resonant width implies a better tolerance on the multilayer and grating parameters. As a matter of fact, the well established multilayer technology can not presently guarantee layer index within better than 0.01 and layer thicknesses within better than a few nanometers at the scale of the wafer diameter. This is not a problem for most applications of multilayer coatings, since what mainly matters there is the optical path, and therefore the product of the index by the thickness of the layers. On the contrary, the spectral position of resonant phenomena does rely upon the layer index and thickness separately since the dispersion equation of a guided mode contains the Goos-Haenchen shifts at the waveguide boundaries, and requires a better control of these quantities than presently ensured by today’s technologies. For instance, setting the resonant reflection under normal incidence within the gain bandwidth of Nd:YAG (0.6nm FWHM) requires a control of about 0.002 on the layer index, and of better than one nanometer on the layer thickness. It is therefore an important design task to provide enough tolerance for the resonance spectral dip to amply overlap the gain spectrum under normal incidence conditions without technology trimming. In the present polarizing structure it was decided to cope with the present status of multilayer deposition technology, to choose the deposition technology showing the best reproducibility, and to set up the conditions for a reproducible grating fabrication process. The chosen deposition technology is therefore ion plating. HfO₂ was chosen for its high optical damage threshold. The most critical point as far as reproducibility and uniformity are concerned is the corrugation fabrication: the effective index of the grating waveguide mode used in the resonant reflection effect is highly dependent on the corrugation depth as well as on the groove line/space ratio. This still is the most limiting factor for a general use of resonant reflection, the more so as the etching of a material like HfO₂ is by far not as easy and developed as that of silica for instance.

The search for a suitable operation point of the polarization filtering laser coupler was undertaken as follows. A stack of 10 SiO₂/HfO₂ quarter wave layers at 1064 nm wavelength under normal incidence gives first the ca 80% reflection pedestal. In a second step, the additional pair of buffer and waveguide layers is designed: for a modal field concentration to take place in the high index waveguide layer; the latter is made significantly thicker than a quarter wave layer of 127 nm. 200 nm is a first guess (there is no grating in this layer at this stage); less would mean that it wouldn’t be possible to concentrate the modal field in the last layer where the grating is etched; more would mean that the normalized modal field in the grating region would be too small. Besides, for increasing the isolation of the modal field of the waveguide layer from the rest of the multilayer, the low index buffer layer is also made thicker than a quarter wave layer of 180 nm; giving it a half wave thickness implies that the previously designed reflection pedestal is little influenced by the additional pair of layers. Furthermore, the experience shows that a half wave thickness provides sufficient isolation. The balance between w_w and w_b is roughly adjusted so that expression (1) is satisfied. There is one more interesting property which a thicker buffer layer brings into this preliminary mental design: it gives the possibility of altering the set of propagation constants of the multilayer modes and to bring two modes close to degeneracy; this is a very important feature giving an opportunity to broaden the fabrication tolerances. These approximately, but intelligibly defined data are then fed into a grating optimization code [12] based on the modal method which searches for the desired dichroic properties of the polarization filter. These are: broad band TM reflection at about 85 %, and a broad dip in the TE reflection spectrum centered at 1064 nm wavelength.

 

Fig.2. Theoretical 0^th order TM and TE reflection spectra and reflected ± 1^st order TE diffraction efficiency with excitation from the substrate side.

Download Full Size | PDF

Figure 2 gives the TE and TM reflection spectra of the resonant grating multilayer optimised by means of Lyndin’s design code [12]. The multilayer submirror consists of 10 alternate quarter wave SiO₂ and HfO₂ layers on the YAG substrate, the first layer being silica, the last HfO₂. The grating is a binary corrugation of 605 nm period, 145 nm depth and 1/1 line/space ratio. The unetched HfO₂ thickness is 185 nm; the total thickness of the waveguiding layer before etching is thus 330 nm. The optimised thickness of the low index SiO₂ buffer is 400 nm. Figure 2 gives the theoretically expected TM and TE reflection spectra as well as the ±1^st order TE reflection into the active substrate with incidence from the active substrate side. As expected, it has been possible to bring two neighbouring TE modes close to 1064 nm which permits to considerably broaden the tolerances. Their effective index is n_e = λ/Λ = 1.776 and 1.767. Plotting the electric field of these two modes reveals that the dominant modal field (n_e = 1.776) has no zero crossing and is the fundamental mode of the whole structure. The closely neighbouring mode with n_e = 1.767 with its single zero crossing is the TE₁ mode.

 

Fig. 3. Calculated electric field distribution of the two neighboring lowest order TE modes of the optimized structure giving rise to the spectra of Fig. 2.

Download Full Size | PDF

Figure 3 illustrates the electric field profile of these two modes at their resonance wavelength. In order to distinctly identify them and analyse their relative pattern and interplay, the influence of the rather strong perturbation of the field in the corrugated high index layer is made smaller by gradually decreasing the grating depth and increasing the non-etched part of the high index layer so as to keep the spectral position of the two modes identical. The structure one ends up with is identical except for a non-etched layer thickness of 228 nm with a grating of 15 nm depth which has just become the weak antenna capable of exciting the two interesting eigenmodes of the structure under normal incidence from the air side. As intended as from the start, both fields exhibit a strong maximum in the high index waveguiding layer and a slow decrease at the substrate side. The major difference is their parity: the dominant mode has no zero crossing whereas the second modal field has its single zero in the buffer layer. One understands now why it was possible to bring the effective index of these two modes close to each other: thanks to the wide low index buffer layer the structure is equivalent for the two first order modes to a directional coupler with its two supermodes of neighbouring effective index, the first waveguide being the last high index waveguiding layer, the second waveguide being the rest of the multilayer. This is why it was possible to obtain both high field confinements in the grating region and spectrally wide polarizing effect. These TE₀ and TE₁ modes are not fully guided modes since their effective index is smaller than 1.82. However their field is mostly concentrated in the upper layers of the multilayer and is very small at the substrate boundary. The phase difference given by expression (1) is approximately 16 radians which is close to the condition of destructive interference of 5π. The high index slab layer thickness w_w is considered as an equivalent thickness, i.e., the waveguide layer thickness prior to the etching minus half the corrugation depth of 145 nm. The resonance condition is also characterized by low efficiency resonant coupling of the guided mode into the ±1^st TE reflected diffraction orders as seen in Fig. 2 with a peak efficiency of 10%; this is a result of the guided mode having a slight leaky mode character, but the main polarizing mechanism is the destructive interference along the 0^th reflected order. The fact that the guided mode giving rise to resonant reflection is slightly leaky may be dangerous because the grating might well exert a non-resonant coupling of the lasing TM polarization to the ±1^st TM orders propagating in the substrate and hence represent a loss penalty for the lasing polarization. This is however not the case because of an additional merit of the thick low index buffer layer: the ±1^st TM reflected orders generated at the waveguide-air interface by the grating are evanescent in the buffer layer and can therefore not tunnel through it. The ±1^st TM diffraction orders are not represented in Fig. 2 because their efficiency is below 10^-8.

4. Fabrication of the grating multilayer.

The prescribed SiO₂/HfO₂ multilayer was deposited onto 12.5 mm diameter, 1.23 mm thick Nd:YAG wafers (dopant concentration 1.3 at. %) bonded to a 0.33 mm thick Cr:YAG saturable absorber (absorption 6 cm^-1). The resist-coated wafers were exposed according to the design to a 605 nm period interferogram at 442 nm wavelength. Prior to etching into the last layer of the multilayer of the doped substrates, the gratings were first etched into a number of single layers of ion plated HfO₂ on blank fused quartz substrates in order to calibrate the RIBE process. Two different etching processes have been tested. The first one is a chemically neutral beam of argon ions. Different etching times lead to different groove depths as shown in Fig. 4(a) exhibiting a clear linear dependence and an etching rate of about 29 nm/min. The straight line is however off the origin which is attributed to a specific difficulty of RIBE with neutral ions: the ion kinetic energy is to a large extent dissipated in the 300 nm thick resist layer on its low thermal conductivity quartz substrate which leads to an overheating of the latter and partial carbonization. These resist remnants are extremely difficult to remove, making the AFM measurement difficult to interpret. The second process uses a 60% CF₄ and 40% argon ion beam under the same conditions of flow and acceleration. The required etching depth of 145 nm was targeted and obtained in a total time of 5 minutes under fixed conditions of flow and acceleration. The compact cluster of points close to 145 nm depth in Fig. 4(b) and a point corresponding to a shorter etching time permit us to draw a slightly steeper straight line closer to the origin. Under these reactive etching conditions the resist rests can easily be removed by means of hot acetone. This is the process which was used to etch down the grating of the YAG substrate mirror.

 

Fig. 4. RIBE etching rate calibration tests. (a) Chemically neutral argon ion beam, (b) 60% CF₄ - 40% Ar ion beam.

Download Full Size | PDF

The grating depth measurement was made by AFM. Fig. 5 shows a typical AFM calibration scan. The calibration curve of Fig. 4 was used to set the etching conditions of the last HfO₂ layer of the multilayer mirror deposited onto the doped YAG substrates. The thermal conditions are not the same since YAG has a much larger thermal conductivity than fused quartz and the fused quartz substrates are 1 mm thick whereas the YAG substrates are 1.56 mm thick. However, the fragmentation of the total exposure time into a number of short time exposures separated by long rest periods prevents the thermal effects from playing an important role; this was confirmed by the fact that standard wet removal of resist rests was sufficient to clean all etched samples; there was no trace of overheated or carbonised resist. Figure 5 shows 3D AFM scan representation of the final sample. The grating depth obtained in the microchip laser sample is 143 nm.

 

Fig. 5. 3D representation of AFM topographic data. The measurement was performed with the intermittent contact mode using a high aspect ration silicon probe.

Download Full Size | PDF

The corrugated multilayer was submitted to the incidence of a polarized microchip Nd:YAG laser beam at 1064 nm wavelength in the neighbourhood of normal incidence with incidence from the air side. This does not correspond to the operational conditions of the microchip laser, but this is the only way of checking whether the resonant mechanism is present and what its spectral position is because the presence of the saturable absorber prevents a transmission measurement as well as a reflection measurement with incidence from the active substrate side. The laser experiment only will tell whether the TE/TM reflection difference is sufficient to exert the desired polarization filtering. Figure 6 gives the reflected power measured for both polarizations upon a variation of the incidence angle. As expected the reflected TM polarization exhibits a monotonic angular dependence since no TM resonance is expected in this region. The measurements were made without power reference. The laser stability was measured to be about ±1 % which partially accounts for the fact that some TM points are slightly off a monotonic curve. The TE polarization shows a pronounced reflection fall in the form of one wide dip at either side of the normal. There is also a distinct but incomplete reflection fall at normal incidence. The absence of symmetry relative to the normal in Fig. 6 is an experimental artefact: the reflection measurement in the neighbourhood of normal incidence imposed a slightly conical mount to enable the detection of the reflected beam and to prevent feedback into the measurement laser. This led to some alignment difficulties and variable partial overlap between reflected beam and detector which gave rise to the slope of the TE and TM reflection spectra. The main features of the resonance effect can nevertheless be clearly identified and quantified. As a whole there is a polarization eye between the TE and TM polarizations exhibiting a TE reflection coefficient at least 20% smaller than the TM’s over a full angular range Δϕ of about 3.4 degrees. Under close to normal incidence, and neglecting the wavelength dependence of the effective index, this corresponds to a wavelength range Δλ. = ΔΔϕ = 18 nm which covers well over the Nd:YAG gain bandwidth which is about 0.6 nm. This also has to be compared with the divergence of the laser beam which is about 10 to 15 mrad (full angle at 1/e²) for typical microlaser cavities like the ones used in this work. The angular width of resonant reflection is about four times larger than the angular width of the lasing mode. As a consequence, the variation of the reflection throughout the beam is low enough and does not induce a noticeable change in the laser parameters. The angular offset of the TE reflection dip is clearly the sign of a discrepancy between the specified and actual layer index and thickness. Assuming that the discrepancy is mainly caused by the error on the high index of the last HfO₂ layer, the error on the index is equal to δϕ = 0.02 in refractive index unit where δϕ is the angular position of the TE dip. This would imply that the index of ion plated hafnia is about 2.08 instead of 2.10. This error has the effect of broadening the polarizing bandwidth; however, assuming that the TE reflection dip would be centered at normal incidence, its ca 0.8 degree angular width still provides a polarizing bandwidth sufficiently broad (about ΔΔϕ = 8 nm) to cover the whole gain bandwidth of Nd:YAG lasers. This polarizing bandwidth of 8 nm approximately corresponds to the spectral width of the theoretical TE double dip of Fig. 2 which is approximately 10 nm.

 

Fig. 6. TE and TM reflected power at 1064nm versus incidence angle measured upon incidence from the air side.

Download Full Size | PDF

The wafer scale uniformity and reproducibility of the layer index and thicknesses represent the fundamental limitation to overcome for such monolithic polarizing solution to be used industrially. Ion plating appears to be a good candidate in this regard. The limitation in the present early prototype is however mainly represented by the non-uniformity of the grating corrugation profile. In order to evaluate this feature, several AFM scans were performed over the microchip laser sample. These reveal a rather good uniformity: the grating depth is 143 nm within 5 nm, and the line/space ratio is 1.0 ± 0.1. The line/space ratio is particularly difficult to estimate from an AFM scan in a submicron grating. However, the uniformity can be assessed with a high degree of confidence. The scans were made by using a high aspect ratio AFM tip. The reproducibility of the resist grating printing (essentially that of the line/space ratio) and of the RIBE steps remain the main hurdle towards manufacturability although the calibration results of Fig. 4(b) are quite encouraging.

6. Damage threshold and scattering level

It is known that the presence of a grating at the surface of a multilayer may significantly lower the laser damage threshold, the more so if the grating is a resonant grating [10]. The polarizing mechanism used here does involve a resonant grating. However, the specific embodiment of resonant diffraction is here particularly astute: the resonant waveguide coupling concerns the polarization which doesn’t lase. Consequently, if the laser cavity is so designed that there is no intra-cavity coupling to the undesired polarization, the laser damage threshold (LDT) should hardly be lowered.

This is what the LDT measurement of such type of structure fabricated by means of another layer deposition technology have confirmed indeed [14]: the LDT of the TM polarization at the resonance of the TE polarization was measured to be 4.8 J/cm² instead of 6 J/cm² for the same multilayer without grating, whereas the LDT of the TE polarization at resonance exhibited a dramatic reduction of the LDT to 0.12 J/cm². The measurements were made at 1064 nm wavelength, 20 Hz repetition rate, 10 ns pulse duration.

Systematic but low sensitivity scattering measurements have been performed on such a device by directing the collimated beam of a white light source onto the grating region from the substrate side [14]. Although the substrate in this measurement was fused quartz, it can be considered that the experimental conditions are close to the operational conditions of a YAG substrate. The scattering level is evaluated on the reflected beam by repeatedly moving the grating area into and outside the beam cross section. The results reveal that the scattering level on the reflected TE polarization (that which couples the guided mode) is at most 0.6%, the scattering measurement being made outside the reflection dip at 1056 nm wavelength, and that the scattering of the reflected TM polarization (the lasing polarization) is at most 0.14%, the scattering measurement being made at the wavelength of the TE reflection dip. Although finer scattering measurements are needed, this first experimental evaluation of the scattering penalty demonstrates that the lasing polarization is much less affected than the polarization which is filtered out, and that a polarization filtering scheme relying upon destructive interference is a promising option.

7. Laser experiments

In order to validate the polarizing effect, we measured the laser performances of the above-described Nd:YAG/Cr:YAG laser, before and after processing of the grating multilayer. The wafer was attached in a circular metallic holder; when mounting the wafer in the holder, we paid particular attention not to create anisotropic mechanical or thermal stress that would bias the measurements.

Pumping was achieved with a multimode diode at 808nm (1*100μm emitter, linear horizontal polarization). The pump power incident on the laser cavity was 0.8 W, focused by a gradient-index lens into a beam size of 60μm. The distances between the pump diode, the lens and the microlaser were adjusted in order to maximize the repetition rate. Before etching the grating, the repetition rate was 10.4 kHz, the pulse energy 5.9 μJ, and the pulse duration 0.59 ns. After processing the grating, the frequency was 11.16 kHz and the pulse energy 6.1μJ. This shows that the process did not degrade the multilayer coating. This is also highly consistent with the above scattering measurements: in the considered Nd:YAG microlasers, the loss penalty due to the grating itself is at least two orders of magnitude lower compared with the other passive losses of the laser cavity (output coupling and non-saturable absorption of the passive Q-switch). As a consequence, this additional loss does not play any noticeable role on the laser performance. It should however be noted that the loss penalty would be more limiting in microlasers with a higher-Q cavity (e.g. using a laser medium with lower gain); in cases where the additional loss is comparable with other cavity losses, the design of the microlaser and of the output coupling coefficient have to be optimised with respect to this limitation in order to achieve the targeted laser performances.

We measured the polarization of the 1064nm output beam before etching the grating. The output polarization was partially linear, and we rotated the wafer to maximize the polarization extinction ratio (PER), which corresponds to the preferred orientation always horizontal, similar to the pump. We marked this preferred axis on the wafer as a guide for the alignment of the grating. Without grating, the PER was very inhomogeneous throughout the surface of the wafer, ranging from 4 dB to 34 dB; at the center of the wafer, measurements on four wafers varied from 11 dB to 33 dB. These huge variations confirm that an external mean is needed to obtain reproducible controlled output polarized beam from a microlaser.

Figure 7 shows the polarization data measured at the center of the wafer after etching the grating. First, the PER is larger than 25 dB whatever the angle α between the grating and the horizontal axis (i.e. orientation of grating lines with respect to the polarization of the pump diode). Second, the polarization of the output beam is at 90° from the lines of the grating for every orientation of the wafer. These data clearly demonstrate that the grating ensures an efficient control of the polarization of the laser beam.

 

Fig. 7. Bottom curve: direction of polarization (measured from horizontal) versus angle α between the grating lines and the horizontal. The top curve represents the corresponding variation of the PER.

Download Full Size | PDF

8. Conclusion

This paper reports on the first demonstration of the monolithic control of the polarization emitted by a microchip laser by means of a resonant grating effect suppressing the emission of the undesired polarization. An appropriate division of functions between a standard multilayer submirror and a post-processed two-layer resonant submirror combines the low-loss feature of laser mirror technology and the polarization selectivity of resonant reflection, the latter causing the destructive interference of the two submirror reflected fields. The power flux resistance of the microchip laser is hardly affected by the action of the resonant grating since the latter only acts on the non-lasing polarization. For the same reason, the loss penalty for using a grating is hardly noticeable.

The experimental demonstration with a passively Q-switched laser microchip at 1064 nm shows potential interest for industrial application. Subsequent process development for solving of the above-mentioned uniformity and reproducibility issues should allow us to take full advantage of the wafer scale processing capability of this technology. Further investigation should also check the reproducibility of the polarization control after dicing of the wafers into microchip, since such process can create stress and induce birefringence in the microlaser. After these process optimization steps, the grating multilayer mirror could represent a cost-effective solution for independent control of the polarization of microlasers. Compared to current solutions discussed above, the polarizing mirror offers interesting packaging alternatives for the realization of compact and inexpensive lasers.