Eye safety implications of high harmonic generation in zinc selenide.

Polycrystalline zinc selenide (ZnSe) has been the subject of many nonlinear optics studies for wavelengths under 4.0 µm including sum/difference frequency generation, harmonic generation, and filamentation. In this report, the conversion efficiency of high harmonic generation (HHG) in ZnSe is quantified for mid-infrared wavelengths ranging from 2.7 µm to 8.0 µm. By increasing the fundamental wavelength, we demonstrate that HHG in thick ZnSe targets is limited by the band gap. The high conversion efficiency of mid-infrared to near-infrared light in ZnSe raises concerns of a nonlinear retinal hazard. We contrast the HHG behavior of ZnSe against the observed harmonic generation of calcium fluoride, BK7, and fused silica over the same wavelengths.

were generated using a 3.25 µm laser [15]. A later study observed even and odd HHG in gallium selenide using a 10.0 µm laser [16]. HHG in semiconductors has been approached theoretically using different methods [17] such as viewing single electron motion with a semiclassical model [15,18,19], solving the time-dependent Schrödinger equation [20,21], and taking the Hartree-Fock approximation while incorporating semiconductor Bloch equations [16,22]. HHG in ZnSe has been considered theoretically, concluding that HHG at 5.0 µm and 6.7 µm is dominated by emission from electron wave packets propagating along the ΓL and ΓK symmetry axes of the Brillouin zones with the ΓL symmetry axis providing stronger harmonic contributions for wavelengths below the bandgap [23]. N-th harmonic generation in ZnSe has been experimentally observed to obey an I 0 N power law dependence for intensities < 0.1 TW/cm 2 (the perturbative regime) and transitions to a clearly nonperturbative regime at higher intensities with the formation of a harmonic plateau [23].
The ease in which HHG can be generated with ZnSe raises questions about the possibility of a HHG driven retinal hazard. Studies of harmonic generation in ZnSe in the 1.5 µm -2.4 µm range observed harmonics of second to fourth order [5], while studies using 3.0 µm, 3.9 µm, and 5.2 µm laser pulses resulted in observations of sixth, seventh and eighth harmonic generation respectively [1,24,25]. HHG up to twelfth harmonic at 5.0 µm and fifteenth harmonic at 6.7 µm have been successfully generated [23]. The harmonics reported in these studies span the visible and near-infrared, wavelengths that can propagate to the retina and cause retinal injury.
Eye safety standards, such as ANSI Z136.1-2014, have set maximum permissible exposure (MPE) limits for laser systems dependent on factors such as the laser's wavelength and pulse duration. The MPE limits for retinal exposure are much higher for mid-IR lasers compared to visible lasers of similar pulse duration because mid-IR radiation is strongly absorbed by the aqueous and vitreous humors of the eye before reaching the retina [26]. Safety standards such as ANSI do not account for nonlinear processes that change pulse wavelengths. HHG poses a risk of retinal injury for high intensity mid-IR laser systems since HHG can convert mid-IR radiation to visible as well as near-IR radiation whose retinal MPE is much lower (0.1 µJ/cm 2 for wavelengths between 400 nm and 1050 nm) [26]. In this letter, we observe and quantify the energy conversion of HHG in polycrystalline zinc selenide (ZnSe). We choose to study the eye safety risks of HHG in ZnSe over other materials because ZnSe based optics are commonly used in mid-IR experiments due to its transparency to mid-IR wavelengths and resistance to optical damage. We then contrast these HHG observations against the low energy conversion observed in CaF 2 , BK7 and UV-Vis grade fused silica under similar conditions. Since our initial presentation at the SPIE Photonics West conference [27], two studies have been conducted that suggest the conversion efficiency of ZnSe is unusually strong. The first study, reported after this letter's submission, showed frequency conversion in ZnSe to be efficient enough that the fundamental pulse can be depleted until the second harmonic intensity exceeds the fundamental pulse intensity [28]. The second study has presented preliminary results at the Frontiers of Optics conference reporting conversion efficiencies of 18-20% for a 5 mm ZnSe target and 26-28% for a 4 mm ZnSe target for midinfrared wavelengths ranging from 3.3 µm to 3.8 µm [29] far in excess of the < 1% conversion efficiencies reported in other materials [30,31]. To our knowledge, this letter details the first systematic study of the HHG conversion efficiency of ZnSe for mid-IR wavelengths.

Methods
Tunable mid-IR, 1 kHz laser pulses were created using difference frequency generation of the signal and idler waves of an optical parametric amplifier (OPA). The OPA was pumped by an 800 nm, 80 fs Ti:sapphire laser (SpitFire by Spectra-Physics). The resulting mid-IR pulses have a nominal pulse duration of 80 fs. The OPA (Light Conversion TOPAS Twins model TT8F12C1) could be tuned to produce excitation wavelengths from 2.7 µm to 12 µm with a maximum pulse energy of 18 µJ ± 2 µJ at 3.5 µm. The mid-IR beam diameter was 3.0 mm. Leakage of the signal, idler and pump frequencies was observed in the beam spectra, so the beam was passed through two dichroic mirrors to minimize these undesired contributions. The knife edge technique was used to find the beam diameters and the pulse energies were measured using a Coherent EM USB J-10MB-LE power meter.
HHG in ZnSe was achieved by focusing the mid-IR laser pulse into a 10-mm thick polycrystalline ZnSe window using a 40-mm CaF 2 lens (see Fig. 1). For each wavelength, the harmonics generated were recorded using two Ocean Optics spectrometers (the USB 2000 + model for visible wavelengths and the NIR-512 model for near-IR wavelengths). HHG spectra were taken for excitation wavelengths between 2.7 µm to 3.0 µm at 0.1-µm increments, and from 3.0 µm to 8.0 µm at 0.5-µm increments. For each wavelength, the ZnSe window was moved into the focal point and the 75 mm collection optic adjusted to maximize the signal recorded by the visible spectrometer. The position of the ZnSe window was adjusted to optimize the signal from the highest harmonic peak observed. After measuring the signals, visible and near-IR spectra without the ZnSe window in the beam path were taken to verify that leakage of the signal, idler or pump frequencies from the OPA were not contributing to the observed spectra. The UV-grade fused silica window cracked when exposed to 2.8 µm -3.0 µm pulses, most likely a result of a rapid rise in temperature due to strong absorption resulting in thermal expansion which cracked the glass. A filter wheel was added to reduce the peak intensity to avoid damaging the fused silica window (see Fig. 2). The laser was committed to other experiments at the time so measurements of the CaF 2 , BK7, and UV-grade fused silica windows (models CF-W-50-3, PW-1037-C and PW-1037-UV with thicknesses of 3.0 mm, 10.0 mm and 10.0 mm respectively) were taken with a beam expander in the beam path (see Fig. 2). Harmonic generation in the three windows was measured with the Ocean Optics USB 2000 + and NIR-512 spectrometers using the same procedure discussed above for ZnSe. New reference measurements without the windows were taken for the modified setup.
The absolute efficiencies of the USB 2000 + and NIR-512 were estimated using an Ocean Optics LS-1 light source. The relative efficiencies were calculated by measuring the spectral distribution of the LS-1 light source propagated through the collection optics, and approximating the actual spectral distribution of the LS-1 as a black body at a temperature of 3100 K. The relative efficiencies were then converted to absolute efficiencies using the LS-1 light source's power as reported by a Coherent PowerMax PM10. By summing over the absolute efficiency curve, the percentage of the initial pulse energy converted to wavelengths under 1200 nm (which are a direct retinal hazard) could be estimated.

Results and discussion
At the maximum beam intensity at 3.0 µm, HHG in ZnSe generated a red beam visible to the unaided eye. A reflected blue spot was observed when the front surface of the ZnSe window was moved through the focal point for excitation wavelengths of 2.9 µm -3.5 µm, consistent with filamentation reported in [6]. However, blue light was not visible in the exiting beam or measured by the visible spectrometer. This is most likely explained by the strong absorption of harmonic wavelengths in ZnSe below 500 nm due to the band gap at 2.71 eV. The collected signal was found to be maximized when the beam was focused into the front surface of the ZnSe window. Nevertheless, HHG could still be detected when focusing into the bulk media.
The intensity of the visible harmonics oversaturated the USB 2000 + spectrometer in the 2.7 µm to 5.5 µm excitation range, so ND filters were placed between the L3 optic (see Fig.  1) and the spectrometer. The ZnSe spectra for excitation wavelengths of 5.0 µm and 7.0 µm are plotted in Figs. 3(a) and 3(b) respectively. The intensities are reported after subtracting background measurements of visible light leakage from the OPA. The intensities recorded by the USB 2000 + in Fig. 3 were initially a factor of six lower than the intensities recorded by the NIR-512 in the overlapping wavelength range, consistent with the difference in detector wavelength bin sizes. Optimization of the alignment of the ZnSe window and collection optic for the 5.5 µm to 8.0 µm excitation measurements using the USB 2000 + were unsuccessful due to the long integration times between adjustments, resulting in a lower count rate. The count rates for the USB 2000 + in Fig. 3 have been raised to correct for the use of ND filters, bin size difference and, in the case of Fig. 3(b) where optimization was not possible, to be consistent with the NIR-512 measurement in the overlapping region. Consistent with literature, we observed harmonic generation for 3.0 µm to 5.0 µm excitation wavelengths for ZnSe with the harmonics terminating near the band gap energy.
The exiting pulses contained a substantial fraction of light in the near-infrared below 1200 nm with values ranging from (8 ± 6) % to (21 ± 14) % of the initial pulse energy (see Table  1).
This corresponds to visible and near-infrared exposures of (1.2 ± 0.8) µJ to (2.3 ± 1.6) µJ of energy, more than sufficient to pose a retinal hazard under the ANSI standard. From Figs. 3(b) and 4 we were able to achieve visible wavelength HHG at excitation wavelengths longer than 5.0 µm consistent with the HHG reported in [23]. For excitation wavelengths above 8.0 µm, the HHG spectra could no longer be observed. This is consistent with expectations because the peak intensity decreases rapidly beyond 8.0 µm due to increasing absorption of the calcium fluoride optics and the increasing beam radius at longer wavelengths (see Table  1). Estimating the visible pulse energy from the USB 2000 + measurements for excitation wavelengths longer than 5.0 µm requires adjusting the spectra to correct for misalignment, as discussed above for Fig. 3(b). Even increasing the fundamental wavelength to 7.5 µm, we were able to convert (13 ± 9) % of the pulse energy into sub-1200 nm radiation more than sufficient to pose an eye hazard under the ANSI standard (assuming a recollimated 3.0 mm beam diameter). Harmonics wavelengths below 460 nm were absent for all measured wavelengths and only weak unconfirmed peaks were observed for wavelengths between 460 nm to 500 nm, which was consistent with the increased absorption of harmonic wavelengths shorter than 500 nm in ZnSe and the thickness of the sample.  These results contrast with observations for CaF 2 , BK7, and fused silica. All three materials required longer integration times than ZnSe and none oversaturated the visible detector. The fraction of pulse energy converted to sub-1.2 nm wavelengths was commonly on the order of 10 nJ (< 1% the input pulse energy), with the highest efficiency conversion of (2.2 ± 1.5) % observed using a 2.9 µm beam for fused silica. The CaF 2 was observed to generate third and fifth harmonics consistent with literature ( Figs. 4 and 5). Unlike ZnSe, CaF 2 did not generate measurable harmonics for excitation wavelengths greater than 3.5 µm despite ZnSe generating third harmonic up to the detector cutoff using a 5.5 µm beam and fifth harmonic using an 8.0 µm beam ( Fig. 4(b)). Placing the front surface of the BK7 window in the focal point resulted in measurable second harmonic, as well as third and fifth harmonic, but at much lower count rates than ZnSe. Higher harmonics were not observed at any wavelength in BK7. Fused silica generated measurable second, third, and fourth harmonics (for excitation wavelengths from 2.7 µm to 3.0 µm). Since a UV-grade fused silica window was used, the front surface of the window had to be placed at the focus to avoid absorption prior to harmonic generation. We observed fifth harmonic generation using a 2.7 µm and a 3.5 µm beam, but were not able to measure the fifth harmonic at 2.8 µm to 3.0 µm. Likewise, we observed weak signals at higher harmonics using a 2.8 µm beam, but did not observe these signals at any other wavelength tested (see Fig. 4). We observe that fourth harmonic generation only occurred in UV-grade fused silica at excitation wavelengths whose bandwidth covered the -OH resonance. We attribute the fourth harmonic generation at those wavelengths to -OH overtone enhanced bulk quadrupole effect. At 3.5 µm, past the -OH resonance, fused silica generates third and fifth harmonics similar to BK7.

Conclusion
We performed a systematic study of HHG in ZnSe using excitation pulses from 2.7 µm to 8.0 µm resulting in HHG at wavelengths not previously studied and the generation of HHG up to twelfth order. We demonstrate that the band gap wavelength acts as an upper limit of efficient harmonic generation in polycrystalline ZnSe. The production of high harmonics in ZnSe corresponds with substantial frequency conversion, with conversions of (8 ± 6) % to (29 ± 20) % in the sub-1.2 µm range. The generation of (2.3 ± 1.6) µJ of sub-1.2 µm radiation using a sub-20 µJ mid-infrared laser suggests that the use of ZnSe optics may pose an unexpected retinal hazard for high intensity mid-infrared laser applications. If the beam had been recollimated to a 3.0 mm diameter, the sub-1.2 µm energy per unit area of the beam would approach (30 ± 20) µJ/cm 2 , well above the ANSI MPE limits (0.1 µJ/cm 2 for sub-1050 nm pulses to 0.8 µJ/cm 2 for 1200 nm pulses) [26]. The efficient conversion of mid-infrared to near-infrared radiation in ZnSe is contrasted by the behavior of other materials whose conversion efficiencies are typically below 1% and is supported by parallel studies reported this year [28,29]. By contrasting ZnSe against fused silica, calcium fluoride, and BK7 this study further emphasizes that harmonic generation in semiconductors is the result of a different process than in other solid optical materials.