Open Access
Add to BibSonomy
Abstract
We report for the first time nonlinear frequency conversion—specifically optical parametric oscillation—in OP-GaP layers grown by hydride vapor-phase epitaxy on 3-inch OP-GaAs templates. Continuous tuning with wavelength coverage from 3.9–12 µm was achieved by using stepped and fan-out gratings having periods from 18.0–35.2 µm, which propagated 150 µm of a 1.2-mm-thick layer before overgrowth. Anti-reflection-coated OP-GaP crystals were pumped at 1040 nm with an ultrafast Yb-fiber laser, yielding idler output powers up to 150 mW (60 mW) in spectra centered at a wavelength of 5.6 µm (10.7 µm).
Published by The Optical Society under the terms of the Creative Commons Attribution 4.0 License. Further distribution of this work must maintain attribution to the author(s) and the published article's title, journal citation, and DOI.
1. Introduction
Orientation-patterned gallium phosphide [1] (OP-GaP) is an exciting new quasi-phasematched (QPM) nonlinear optical crystal, whose wide bandgap avoids two-photon absorption at 1 µm, allowing it to be pumped at this wavelength and so deliver high output powers in the mid-infrared using mature 1040-nm Yb- [2,3] and 1064-nm Nd-based [3,4,5] laser sources. In particular, ultrafast Yb-fiber pumped OP-GaP optical parametric oscillators (OPOs) have been demonstrated as practical sources for coherent spectroscopy in Fourier-transform [2,6,7] and dual-comb [8] modalities across the 5–12-µm molecular fingerprint region. While there have been numerous device demonstrations to date, epitaxial growth and processing of OP-GaP is now severely limited by the quality and availability of GaP substrates. For nearly a decade, researchers have been developing heteroepitaxial growth of GaP on orientation-patterned gallium arsenide (OP-GaAs [9]) templates, successfully demonstrating OP-GaP layers up to 400 µm in thickness with excellent domain fidelity [10]. Despite these successes, no nonlinear frequency conversion has yet been demonstrated using these GaP-on-GaAs QPM layers, suggesting that issues such as stress-birefringence (induced by the lattice and thermal-expansion mismatch between GaP and GaAs) or unwanted absorption may impede device performance. Here we report, for the first time, efficient frequency conversion in OP-GaP layers up to 1.2-mm thick, grown by hydride vapor-phase epitaxy (HVPE) on a 3-inch MBE-grown OP-GaAs template. The OP-GaP crystals were evaluated in an Yb-fiber-laser-pumped optical parametric oscillator (OPO). The wafer design enabled discrete wavelength tuning via stepped adjacent gratings with fixed periods and continuous tuning via fan-out gratings, providing continuous wavelength coverage from 3.9–12.0 µm by simply translating the OP-GaP crystal position in the OPO cavity focus.
2. Crystal fabrication and characterization
2.1 HVPE OP-GaP growth on OP-GaAs templates
Our process for producing thick, quasi-phase-matched GaP structures on OP-GaAs templates is illustrated below in Fig. 1. A 3-inch, epi-ready, semi-insulating (100) GaAs wafer, offcut 4° toward <111B>, is loaded into a Varian Gen II MOD molecular beam epitaxy (MBE) machine (step 1), baked, and de-oxed. A 20-layer GaAs (450Å)/AlAs (50Å) superlattice is grown, followed by a 1000Å GaAs base layer, a 200Å Al0.75Ga0.25As etch-stop layer, and a 100Å GaAs buffer layer, all at 750°C. The wafer is then transferred to an auxiliary chamber (without breaking vacuum) where a 40Å non-polar Ge layer is deposited at ∼ 400°C. The wafer is then returned to the main chamber, and a 1200Å GaAs inverted layer (- polarity) is grown (step 2), followed by a 200Å Al0.75Ga0.25As etch-stop and a 200Å GaAs cap. Initially both + and - domains nucleate on the non-polar Ge layer, but in the right temperature regime (>650°C) the + domains eventually self-annihilate via pyramidal growth at thicknesses > 1000Å, leaving a single-domain surface with inverted (-) polarity relative to the (+) substrate. The wafer is then removed, patterned with a photo-lithographic mask to define desired grating periods, and developed. Exposed domain stripes are then wet-etched back to the substrate surface (step 3) to expose the original + polarity (1:1 HCl:H2O to remove oxides and etch-stop layers, 4:1 citric acid:peroxide to remove GaAs layers). The patterned wafer is then stripped, cleaned, and reloaded into the MBE machine, and a 2-μm-thick orientation-patterned GaAs layer is grown (step 4): domains nucleated on the etched regions have the + substrate polarity, while domains grown on the inverted layer stripes have the opposite – polarity.
The orientation-patterned GaAs template produced by steps 1-4 in Fig. 1 is then loaded into an AIXTRON 103 LP-VPE low-pressure HVPE reactor equipped for growth of GaAs and GaP. HVPE is a near-equilibrium process capable of growth rates over 100 times faster than those achievable by MBE, allowing layer thicknesses ≥ 1mm needed for in-plane propagation of laser radiation through the orientation-patterned QPM structure. The HVPE process is shown in Fig. 2, whereby HCl gas flows over liquid Ga (T ∼ 700°C) to form GaCl, which is then introduced through close-coupled “showerhead” injectors directly above the substrate (rotating via H2 gas flow over grooves on the bottom of the wafer carrier) and reacts with a flowing sheet of hydride – AsH3 or PH3 – to form crystalline GaAs or GaP respectively on the wafer surface according to the reaction GaCl + AsH3 (or PH3) → GaAs (or GaP) + HCl + H2. Hydrogen is used as the carrier gas at growth pressures of 20 mbar and typical substrate temperatures of 710°C for GaAs growth and 793°C for GaP growth. During heating and cooling, AsH3 flows over the GaAs wafer (or PH3 over the GaP wafer) to prevent sublimation loss of the group V element from the substrate wafer. For this experiment we used our standard OP-GaP HVPE growth parameters except that for growth on an OP-GaAs template we flowed arsine instead of phosphine during heat-up, reduced the growth temperature to 774°C, and reduced the total growth time from 10 to 9 hours. The resulting layer thickness was just under 1 mm (990 μm) for an average growth rate of 110 μm/h. Minor parasitic deposits only occurred downstream from the wafer and no changes in growth rate as a function of growth time and layer thickness were observed, which we attribute to the use of a close-coupled showerhead design in our reactor (i.e., the reactant gases are introduced directly at the rotating wafer surface to prevent upstream parasitics). Layer thickness uniformity and growth rates were comparable for growth on GaP and GaAs.
Figure 3 shows the lithographic mask design used with a selection of fixed, stepped and fan grating designs for 1040-nm-pumped mid-infrared generation in OP-GaP (a), along with the corresponding MBE OP-GaAs template before (b) and after (c) growth of the 1-mm-thick OP-GaP by HVPE. Note that for fan gratings the larger grating end of the fan is always closest to the major flat (as indicated by the maximum fan angles in red). The quality of the particular OP-GaAs template in Fig. 3(b) was unusually poor, as indicated by the hazy regions near the center of the wafer which manifested themselves as shiny faceted unpatterned regions on the subsequent HVPE GaP layer grown on the template in Fig. 3(c). We expect this arose from non-uniform growth of the non-polar Ge layer, and noted that the Ge source was nearly empty and needed to be replenished.
Fig. 3. (a) Grating mask design; (b) MBE-grown OP-GaAs template; (c) HVPE OP-GaP (t = 1 mm) on GaAs.
2.2 OP-GaP-on-GaAs grating propagation
Several multi-grating and fan-out crystals with widths of 6.5 mm and 13 mm and of lengths 1.1 mm and 2.9 mm were diced from the wafer shown in Fig. 3. The small side faces (parallel to the major flat) were polished and etched in 1:1 H3PO4:H2O2 under bright light for 25 minutes to reveal the QPM gratings. A typical etched grating cross-section is shown in Fig. 4, which shows parallel domain propagation (grating period Λ = 35.2 μm) for the first 150 μm of growth before the gratings drop out. The reason is unknown for the sudden drop-out of grating propagation during HVPE growth in OP-GaP of short grating periods (< 50 µm), but identical behavior is observed on both GaP and GaAs substrates. Larger periods (> 50 µm) in both cases show better domain fidelity, but since the objective of this work was to produce devices for 1040-nm pumped OPOs, phasematching constraints meant that grating periods in the 20–30-µm region were essential. Despite the limited grating height, these results are promising in that they are very similar to results normally observed for HVPE OP-GaP grown on native OP-GaP templates.
2.3 OP-GaP crystal transmission
After grating etching, the wide input and output faces were polished with a laser-quality finish and a broadband anti-reflection coating was applied to both of the optical faces. Each crystal was mounted with the GaAs substrate side of the crystal fixed onto a metal mount, with the surface of the HVPE-grown GaP exposed on the opposite side. Using an attenuated beam from the 1040-nm ultrafast Yb-fiber laser employed later as the OPO pump source, the transmitted and reflected powers from the crystals were recorded, allowing the average residual losses due to absorption / scattering to be calculated as 2.3% and 6.1% for the 1.1-mm and 2.9-mm crystals respectively, which implied a consistent loss coefficient of 0.21 cm-1. An identical approach applied to an older crystal of OP-GaP-on-GaP, again correcting for the reflectivities of the coatings on the crystal faces, yielded a lower value of 0.12 cm-1. These results, obtained from only two different fabrication batches, do not allow us to state with confidence that OP-GaP grown on GaP has a universally lower loss than OP-GaP grown on GaAs. Indeed, an absorption measurement performed on an independent sample of OP-GaP-on-GaP returned a loss coefficient of 0.21 cm-1, consistent with our OP-GaP-on-GaAs crystal. However, there are indications that the absorption may be batch specific, and may arise from the presence of Fe3+ impurities in the HVPE-grown material, which may vary from one growth run to another. Other researchers observed a sizeable Fe3+ electron paramagnetic resonance (EPR) signal in HVPE OP-GaP grown at BAE Systems [11]. Consistent with this, a deep level in GaP with an absorption peak at 1.8 eV (689 nm) was attributed to the hole activation from Fe3+ to the valence band [12]; this could easily be responsible for the observed absorption loss in HVPE GaP near the band edge. Finally, we note that high purity phosphine gas (PH3) is used as the phosphorus source in HVPE-grown GaP; this can easily form phosphoric acid in the presence of moisture which readily dissolves iron oxide (rust) in steel gas cylinders and stainless steel tubing used in this process, providing a source of such iron impurities.
3. Crystal evaluation in an ultrafast optical parametric oscillator
3.1 OPO systems evaluated
Crystals of both lengths were tested in OPOs synchronously pumped in ring resonators by a 1040-nm Yb-fiber laser (Chromacity Ltd.), similar to the system previously described in detail in [2]. The performance was investigated using two different pump-laser configurations, operating with pulse repetition frequencies of 100 MHz and 200 MHz. At 200 MHz the unfolded ring OPO already had a reduced footprint of 75 cm × 15 cm, comparable to the pump laser size of 43 cm × 25 cm, and therefore making the entire system very compact with a further reduction possible through a folded cavity design.
3.2 Power dependence with pump power
Before describing the specific performance obtained using different crystal lengths, grating designs and pump-laser repetition frequencies, we present in Fig. 5 a direct performance comparison between an OPO configured with a previously reported OP-GaP-on-GaP crystal [2] and the same cavity operated using an OP-GaP-on-GaAs crystal of similar length and grating design. The OP-GaP-on-GaP crystal had a quasi-phase matching period of 21.5 µm and a length of 1.0 mm, while the OP-GaP-on-GaAs crystal quasi-phase matching period was 21.0 µm and its length was 1.1 mm. The crystals were tested in a 100-MHz OPO, with 0.9-ps pump pulses and signal (idler) wavelengths of approximately 1.29 µm (5.3 µm). The idler slope efficiencies are compared in Fig. 5, which shows 6.1% slope efficiency for the OP-GaP-on-GaP OPO and 3.8% for the OP-GaP-on-GaAs OPO. After comparing the coating reflectivities at 1040 nm, we found that the OP-GaP-on-GaAs crystal had an average reflection loss of 7.5% per for each face, compared to 1.5% per face for the OP-GaP-on-GaP crystal. Signal and idler reflection losses were also double for the OP-GaP-on-GaAs sample, at about 1% and 4% respectively. Along with the higher absorption loss detailed in Section 2.3, these factors partially explain the lower slope efficiency of the OPO configured using OP-GaP-on-GaAs crystals. The OP-GaP-on-GaAs crystal also exhibited deformation after post-growth cooling, making it possible that residual stress birefringence due to differential thermal expansion between GaP and GaAs might contribute to the loss, but we have no direct evidence for this at present.
Fig. 5. (a) Idler power at 5.3 μm as a function of Yb-fiber-laser power incident on the pump lens for OP-GaP-on-GaP (red – 1 mm crystal length) and OP-GaP-on-GaAs (blue – 1.1 mm crystal length) OPOs, synchronously pumped with 0.9-ps pump pulse duration at 100 MHz. Slope efficiencies are 6.1% and 3.8% respectively. Inferring the OPO oscillation thresholds from the slope efficiency data provide values of 600 mW for OP-GaP-on-GaP and 721 mW for OP-GaP-on-GaAs. Inset: idler beam profile recorded at a center wavelength of 5.3 µm.
At the highest pump intensities, we observed a spatial self-defocusing effect, as reported in [13]. This effect was reversible and led to no damage, taking the form of a clearly perceptible change in the divergence of non-phasematched SHG light, and may be explained by cascaded χ(2) processes leading to a negative Kerr lens. All data presented in this paper were recorded at pump powers below the values for which this effect became apparent. An example of the idler beam profile recorded in the absence of self-defocusing effects is shown in the inset to Fig. 5 for wavelength 5.3 µm.
3.3 Power dependence with wavelength
The idler output power from the 2.9-mm crystals was characterized for 100-MHz pumping, and representative results are shown in Fig. 6 for idler wavelengths of 5.6 µm, 7.8 µm and 10.7 µm. The lower idler powers at longer wavelengths are partly explained by the decreasing photon energy and lower nonlinear gain when operating further from degeneracy [14,15]. Figure 6(b) represents the data of Fig. 6(a) in terms of the photon flux as a function of input pump power, revealing a lower efficiency for the 32-µm grating used to phasematch idler generation at 10.7 µm. We speculate that this may be caused by the poorer spatial mode overlap of the pump and idler modes when the OPO is operated far from degeneracy, since the confocal parameters differ by a factor of approximately ten.
Fig. 6. (a) Idler output powers from a 100 MHz, 2.5 ps pump pulse duration, OPO configured with a 2.9-mm crystal operated at idler center wavelengths of 5.6 µm (red – slope efficiency 2.87%), 7.8 µm (blue – slope efficiency 1.84%) and 10.7 µm (yellow – slope efficiency 1.32%). (b) Corresponding photon flux.
3.4 Useful crystal aperture
The useable aperture was investigated for two crystals containing step gratings of periods 21–34-µm and of lengths 1.1 mm and 2.9 mm using 2.5-ps pulses with an average power of 2.5 W. Idler power measurements recorded with the 2.9-mm crystal at 5.6 µm and with the 1.1-mm crystal at 7.8 µm and 10.7 µm are shown in Fig. 7 as the crystal was translated laterally across the beam. The edge of the aperture corresponding to the OP-GaP-GaAs boundary was defined as the point where OPO oscillation ceased and there was no second-harmonic of the pump light visible. The calculated focus in the crystal had a 1/e2 waist radius of 18 µm, and the data show a region of only ∼80 µm in width across which the power exceed 90% of the maximum value, corresponding to two beam diameters. As Fig. 7 shows, oscillation was possible over a lateral translation range of 90–140 µm, corresponding approximately to the images of the HVPE growth which show propagation of the OP-GaP domains over a distance of ∼150 µm from the GaAs template before overgrowth occurs. The variability in the grating propagation across the crystals explains the varying width of the useful aperture.
Fig. 7. Idler powers recorded as the crystal was translated laterally across the pump beam, indicating a useful aperture of ∼80 µm and a maximum operating aperture of ∼140 µm. Variable growth quality across the crystals explains the varying width of the useful aperture. On the distance axis, zero is defined as the point where oscillation ceases near the GaP-GaAs interface
3.5 Tunability
Pumping at 1040 nm and resonating the OPO signal wavelength means that a large portion of the mid-infrared from 4–12 µm is accessible for a change in the signal wavelength of only 200 nm, a range that is readily accommodated by the typical bandwidth of a simple dielectric cavity mirror. In practice, realizing an OPO with such gap-free tunability requires using an OP-GaP crystal comprising either many discrete gratings or a single fan-out grating, both covering the period range from 18.0–33.8 µm. Changing the OPO wavelength then becomes a simple matter of translating the crystal across the intracavity focus along the appropriate direction. While such dynamic patterning of the domain inversion period is now routine in QPM crystals made using ferroelectric periodic poling, the situation is dramatically different in the all-epitaxial processing of QPM semiconductors, in which the patterned lithographic grating structures must be preserved throughout the HVPE regrowth process over large distances and high aspect ratios. Examples of stepped and fan-out gratings were reported recently in OP-GaP-on-GaP crystals [16], but until now none has been demonstrated using an OP-GaP-on-GaAs process.
Using a commercial spectrometer (Thorlabs OSA207C) we evaluated tuning for two grating designs included on the same wafer: a 13-mm-wide stepped-grating with periods from 21–34 µm incrementing by 1 µm at 1-mm intervals; and a fan-out grating with periods from 18–35.2 µm across an extent of 13 mm. An example of the tuning achieved by using multiple gratings on the same crystal is shown in Fig. 8. While comparable performance was possible with other laser and OPO configurations, the example shown had the best output powers for 2.5 W pump power and was achieved with a 200-MHz, 1040-nm source operating at 0.75-ps pulse duration and with a step-grating of length 1.1 mm. Despite the pump power loss due to reflection and absorption/scattering, the idler powers achieved in this case are comparable to previous work using OP-GaP-on-GaP [3]. We note that the structure in the spectra from 5.5–7 μm is due to atmospheric water absorptions. This OPO demonstrated discrete-step tunability from 5.6–10.7 µm, a slightly smaller range than the grating design because the crystal mount prevented the gratings at either end of the crystal (Λ = 21 μm and 34 μm) from being accessed. The upper axis indicates the fixed grating period corresponding to each spectrum.
Fig. 8. Idler spectra and associated output powers (diamond symbols, right axis) from a discrete-step grating design of length 1.1 mm. The OPO was pumped by 200-MHz, 2.5-W, 0.75-ps, 1040-nm pulses. Each spectrum corresponds to a discrete grating (Λ = 22–32 μm in 1-μm steps), resulting in idler wavelengths of 5.6–10.7 μm respectively.
For extended gap-free performance, simple tuning of the OPO wavelength was also tested using the fan-out OP-GaP crystal described earlier, this time using a 100-MHz laser of similar performance to that used to characterize the stepped gratings. An example of continuous wavelength tuning on a single 1.1 mm length fan-out device is shown in Fig. 9. Optical spectra were recorded at intervals corresponding to ∼500 nm shifts in the center wavelength by translating the crystal. Only a small adjustment of the cavity length was required when changing wavelength to account for the group-delay dispersion profile of the cavity. No re-alignment of mirror mounts was required when translating between wavelengths. The upper axis in Fig. 9 shows the estimated grating period for each recorded spectrum. This fan-out-grating OPO provided gap-free tunability with spectra centered from 4.0–11.5 µm, and with edge-to-edge wavelength coverage of 3.9–12.0 µm. Based on our experience of pumping other (non-OP-GaP) OPOs constructed using identical crystals but pumped by 100-MHz and 200-MHz oscillators, we expect that the differences in the idler spectral shapes and bandwidths visible in Fig. 8 and Fig. 9 are mainly associated with the different pump pulse spectral characteristics, which partly transfer into the idler spectra.
Fig. 9. Representative spectra from a 1.1 mm fan-out crystal design. Spectra were recorded approximately every 500 nm by translating the crystal. The upper axis shows the estimated grating period at which each spectrum was recorded. The OPO was pumped by 100-MHz, 2.5–W, 0.9-ps, 1040-nm pulses.
4. Summary and conclusions
We have demonstrated the first example of practical nonlinear optical frequency conversion using OP-GaP grown on OP-GaAs templates. The implementation of HVPE heteroepitaxial growth of GaP on GaAs preserved grating integrity to a depth of ∼150 µm, sufficient to enable optical parametric oscillation on grating periods ranging from ∼18–34 µm that yielded continuously tunable idler wavelengths with a wavelength coverage from 3.9–12.0 µm. This achievement is significant because it circumvents issues with problematic GaP growth associated with the limited availability and poor quality of commercial GaP wafers. The approach exploits a robust and well-developed technique for the preparation of OP-GaAs templates, while also gaining the principal advantage of GaP, namely its low two-photon absorption at 1 µm, which allows pumping by mature Yb- and Nd-lasers. We anticipate that this process will now become the method of choice for the preparation of OP-GaP crystals.
Funding
Defence and Security Accelerator (DSTLX1000125804); Engineering and Physical Sciences Research Council (EP/R033013/1).
Acknowledgments
The authors thank Dr. Vladimir Tassev at the U.S. Air Force Research Laboratory for helpful discussions regarding hetereoepitaxy of GaP on GaAs by HVPE.
Disclosures
CF: Chromacity Ltd. (I,E); KJ: Chromacity Ltd. (E); DTR: Chromacity Ltd. (I); PGS: BAE Systems (E); LM, MR and YS declare no conflicts of interest.
References
1. T. Matsushita, T. Yamamoto, and T. Kondo, “Epitaxial growth of spatially inverted GaP for quasi phase matched nonlinear optical devices,” Jpn. J. Appl. Phys. 46(No. 17), L408–L410 (2007). [CrossRef]
2. L. Maidment, P. G. Schunemann, and D. T. Reid, “Molecular fingerprint-region spectroscopy from 5 to 12 μm using an orientation-patterned gallium phosphide optical parametric oscillator,” Opt. Lett. 41(18), 4261 (2016). [CrossRef]
3. L. Maidment, O. Kara, P. G. Schunemann, J. Piper, K. McEwan, and D. T. Reid, “Long-wave infrared generation from femtosecond and picosecond optical parametric oscillators based on orientation-patterned gallium phosphide,” Appl. Phys. B 124(7), 143 (2018). [CrossRef]
4. L. A. Pomeranz, P. G. Schunemann, D. J. Magarrell, J. C. McCarthy, K. T. Zawilski, and D. E. Zelmon, “1-μm-pumped OPO based on orientation-patterned GaP,” Proc. SPIE 9347, 93470K (2015). [CrossRef]
5. H. Ye, S. C. Kumar, J. Wei, P. G. Schunemann, and M. Ebrahim-Zadeh, “1 μm-pumped optical parametric generator and oscillator based on orientation-patterned gallium phosphide,” Proc. SPIE 2018, 71 (2018). [CrossRef]
6. L. Maidment, P. G. Schunemann, and D. T. Reid, “White powder identification using broadband coherent light in the molecular fingerprint region,” Opt. Express 26(19), 25364 (2018). [CrossRef]
7. L. Maidment, P. G. Schunemann, R. J. Clewes, M. D. Bowditch, C. R. Howle, and D. T. Reid, “Systematic spectral shifts in the mid-infrared spectroscopy of aerosols,” Opt. Express 26(15), 18975 (2018). [CrossRef]
8. O. Kara, L. Maidment, T. Gardiner, P. G. Schunemann, and D. T. Reid, “Dual-comb spectroscopy in the spectral fingerprint region using OP-GaP optical parametric oscillators,” Opt. Express 25(26), 32713 (2017). [CrossRef]
9. C. B. Ebert, L. A. Eyres, M. M. Fejer, and J. S. Harris Jr, “MBE growth of antiphase GaAs films using GaAs/Ge/GaAs heteroepitaxy,” J. Cryst. Growth 201-202, 187–193 (1999). [CrossRef]
10. V. L. Tassev, S. R. Vangala, R. D. Peterson, M. M. Kimani, M. Snure, R. W. Stites, S. Guha, J. E. Slagle, T. R. Ensley, A. A. Syed, and I. V. Markov, “Heteroepitaxial growth of OP-GaP on OP-GaAs for frequency conversion in the IR and THz,” Opt. Mater. Express 6(5), 1724 (2016). [CrossRef]
11. Private communication, L. Halliburton and E. Golden, Air Force Institute of Technology (2014).
12. Y. Okuno, K. Suto, and J. Nishizawa, “Deep levels in GaP,” J. Appl. Phys. 44(2), 832–836 (1973). [CrossRef]
13. M. Rutkauskas, A. Srivastava, and D. T. Reid, “Supercontinuum generation in orientation-patterned gallium phosphide,” Optica 7(2), 172 (2020). [CrossRef]
14. J. M. Manley and H. E. Rowe, “General energy in nonlinear reactance,” Proc. IRE 47, 2115 (1959).
15. R. Boyd, Nonlinear Optics (Elsevier, Acad. Press, 2011).
16. P. G. Schunemann, D. J. Magarrell, and L. A. Pomeranz, “HVPE of orientation-patterned gallium phosphide (OP-GaP) with novel quasi-phasematched device structures,” in Conference on Lasers and Electro-Optics, OSA Technical Digest (online) (Optical Society of America, 2018), paper STh4F.1.
