Design of a solar-pumped semiconductor laser

Since their first demonstration in 1962, solar-powered lasers have attracted interest as a tool for generating and distributing renewable energy. Proposals have ranged from orbiting solar laser power stations beaming energy down to terrestrial photovoltaic receivers, to arrays of solar lasers in desert areas linked to population centers via optical fibers. However, despite several decades of research, solar-powered lasers have yet to reach the levels of efficiency or output beam quality that would make these applications feasible. Here we propose a new solar laser architecture, the solar-pumped vertical external cavity surface emitting laser (SP-VECSEL), as a logical continuation of the proven mode-converting capabilities of diode-pumped VECSELs. In experiments using VECSEL gain samples pumped using sunlight, we demonstrate that no major drop in efficiency is observed beyond that associated with the quantum defect of the pump light, allowing us to predict that SP-VECSEL performance can substantially surpass that of existing solar lasers. © 2015 Optical Society of America


INTRODUCTION
Sunlight can provide a plentiful and inexhaustible supply of clean power, a fact which has motivated research into photovoltaic cells and has spurred on the rapid growth of the photovoltaic market over recent decades.The efficiency with which photovoltaic cells can convert solar power to electrical power has undergone a corresponding increase, reaching 29.1% for single-junction cells [1] and 44.7% for multiple-junction devices [2].However, despite impressive solar cell performance, photovoltaic power as a whole suffers from the more prosaic issues of storage and transmission of electricity.The unpredictability of bright sunshine in many parts of the world makes matching electricity supply with demand difficult, and electrical transmission presents a further problem for photovoltaic sources, as solar power stations are ideally situated in desert or near-desert areas where bright sunshine is guaranteed and competition for land use is low.Transmission from these stations to areas of high demand incurs an unavoidable loss of power in electrical grids.
Recently, researchers at the Tokyo Institute of Technology proposed a solution to these problems in the form of the MAGnesium Injection Cycle (MAGIC) [3,4].Here, solar-powered lasers would be used to thermally reduce magnesium oxide obtained from seawater.The resulting magnesium could then be transported efficiently to the point of use, where it would either be burnt directly or reacted with steam to produce heat and hydrogen gas.It was proposed that this scheme would enable renewable hydrogen power generation by solving the problems of storage and transport of hydrogen: magnesium has an energy density ten times greater than hydrogen and, as a solid powder rather than an explosive gas, it is both more convenient and safer to store and transport.The MAGIC cycle was conceived to offer capabilities that complement those of photovoltaic electricity generation and of the hydrogen economy, and would therefore represent one facet of a robust solar energy generation, transport, and storage scheme.
Despite the promise of the MAGIC scheme, it suffers a major obstacle in the solar lasers themselves.The highestperformance solar lasers to date are based on Nd:YAG or Nd:Cr:YAG.Nd:YAG is a proven laser medium with excellent optical and thermal properties, but when used in solar lasers the resulting efficiency is unavoidably low due to the poor overlap between its pump absorption bands and the solar spectrum [5].Co-doping with chromium can improve this overlap [6], but has only yielded a moderate improvement in solar laser performance.The highest solar laser slope efficiencies to date are 3.9% for Nd:YAG [7] and 9% for Nd:Cr:YAG [8].
In addition to low efficiencies, current solar lasers generally suffer from poor output beam quality as a result of the difficulty of achieving a good spatial overlap between concentrated sunlight and a fundamental Gaussian laser mode.This is not a trivial problem, having given rise to a range of solar concentrator geometries as either theoretical proposals or as experimental demonstrations [9][10][11][12].To date, no solution has been demonstrated which simultaneously achieves high efficiency and beam quality in a scalable manner.
In contrast to the doped crystals or glasses of solid-state lasers, semiconductors would make ideal gain media for solar lasers [13,14], primarily due to their broadband optical absorption but also due to their very short absorption lengths, which simplify the problem of overlapping pump and output beam.However, only with the rapid increases in performance of vertical external cavity surface emitting lasers (VECSELs), also known as semiconductor disk lasers, in recent years have optically pumped semiconductor lasers gained widespread acceptance as a useful laser architecture.Diode-pumped VECSELs have proved capable of producing output powers above 100 W with slope efficiencies in excess of 60% [15] of spectral coverage from the mid-IR [16] to the visible and into the ultraviolet using intracavity frequency conversion [17], and of ultrashort pulses at watt-level output powers when mode-locked either with separate gain media and saturable absorbers [18] or with these components integrated into a single chip [19].Their proven capability for high power and efficiency, and good beam quality when pumped with divergent, low-beam-quality optical pumps make VECSELs highly attractive candidates for solar-pumped semiconductor lasers.
In this work, we describe experiments to investigate the performance of a VECSEL gain medium under solar pumping.The first experimental section will focus on a comparison between the performances of VECSELs pumped by 808 nm diode lasers and 532 nm solid-state lasers.In the second section, measurements of the gain medium quantum efficiency are used to explain the differences in laser performance.Subsequently, quantum efficiency measurements taken with the sun as a pump source demonstrate constant quantum efficiency across the solar spectrum, and thus that the reduction in gain medium efficiency when switching from laser diode to solar pumping is dominated by that associated with the quantum defect.The final section considers the design and performance of a future solar-pumped VECSEL.

532 AND 808 NM PUMPED VECSEL PERFORMANCE
The VECSEL gain chip used throughout this work was designed for emission at 1010 nm and is produced by NAsP GmbH of Marburg, Germany, to a similar design to the sample used to obtain an output power of 100 W [15], though with a shorter emission wavelength and a Bragg mirror with 22 repeats rather than 17.The sample contained 10 InGaAs quantum wells, spaced one per E-field antinode, in a strainbalanced GaAsP active region of thickness 1.7 μm.It was grown upside down and bonded using indium to a 300 μm thick diamond before having its substrate removed by chemical etching to leave a GaAsP cap layer, whose thickness made the sample resonant at the laser wavelength.When used in lasing experiments, the sample was placed in a near-hemispherical cavity completed by a 0.7% transmission output coupler with 100 mm radius of curvature.
Two pump sources were used for photoluminescence (PL) and lasing measurements: a Dilas 808 nm fiber-coupled diode module capable of emitting up to 200 W in a fiber with 200 μm core diameter, and a Laser Quantum Finesse, emitting up to 10 W at 532 nm in a TEM 00 beam.While the sample and diamond heatsink were identical for pumping at both 808 and 532 nm, the mounting of the diamond was different.When pumping at 532 nm, the diamond was mounted on a water-cooled copper heatsink using thermal paste, in contrast to 808 nm when the sample was placed in a mount with a cooling water flow impinging directly on the back of the diamond, leading to improved heat removal.Figure 1 shows a schematic of the gain sample structure along with the calculated penetration depths of the two pump sources using absorption coefficient values from [20].
Figure 2 shows output powers of VECSELs with 0.7% output coupling pumped with 808 and 532 nm as functions of pump power.The 808 nm pump had a top-hat profile with 300 μm diameter on the sample surface, and the 532 nm pump has a Gaussian profile with a 280 μm beam width.The 532 nm pumped VECSEL can be seen to have a similar threshold but lower slope efficiency than the 808 nm pumped laser.That the threshold and slope efficiency are at all comparable demonstrates immediately that the difference in the pump absorption lengths shown in Fig. 1 is not an issue.The similarity between the values indicates that carrier diffusion is sufficient to redistribute the carriers across the quantum wells.
VECSEL slope efficiency can be expressed as the product of the output coupling efficiency η out , the quantum defect η QD , and the quantum efficiency η QE , defined as the number of output photons emitted per incident pump photon [21].The output coupling efficiency depends on the cavity loss, and so cannot be known with a high degree of accuracy without additional measurements, preventing a direct calculation of the quantum efficiency.However, if the cavity loss is assumed to be similar for the two lasers in question, it is possible to calculate the ratio of the quantum efficiencies when pumping at the two different wavelengths, and to use this ratio as a measure of the reduction in performance caused by pumping at 532 rather than 808 nm.This ratio is shown as the blue unbroken line in Fig. 2. That the ratio is close to one is encouraging, as it implies that the loss of efficiency when pumping at 532 rather than 808 nm is dominated by the quantum defect.However, the ratio can be seen to decrease with pump power.This can be attributed to the increased sensitivity to thermal effects when pumping with a large quantum defect.Thermal rollover is evident at a lower power in the case of the 532 nm pumped laser, and had pumping with more than 10 W at 532 nm been possible, we expect that the ratio of quantum efficiencies would have been seen to decrease even more rapidly at higher powers.It currently not clear whether this results primarily from the quantum defect or from the less efficient cooling of the diamond mount used in this case.This is not a point of immediate concern for an SP-VECSEL as it would be very difficult to concentrate a solar pump to intensities high enough to reach thermal rollover.

532 AND 808 NM PUMPED PHOTOLUMINESCENCE MEASUREMENTS
While providing a useful guide as to the effects of large quantum defect pumping, the method for estimating the ratio of quantum efficiencies above is not ideal in several respects.In particular, without accurate values of the cavity loss it can only provide an estimate of the values of the quantum efficiency.Direct measurements of the quantum efficiency can be achieved by integrated PL measurements, as described in [22].The experimental setup used for such a measurement is simple, consisting of a photodiode with known detector area placed a distance from a pumped gain sample.Longpass filters are used before the photodiode to block scattered pump light, ensuring that the signal detected results solely from PL. Figure 3(a) shows the quantum efficiency of the gain sample in question when pumped by both 808 and 532 nm lasers with the same pump spot profiles as those used in the lasing measurements described in the previous section.In both cases the quantum efficiency starts from a value close to 0.5 at low pump powers and decreases at higher powers due to thermal effects, with the quantum efficiency at 532 nm decreasing more rapidly, though as before to what extent this is due to the thermal impedance of the diamond mount is unknown.
The ratio of the two curves shown in Fig. 3(a) can be used to provide a direct comparison between quantum efficiency data from lasing measurements and from integrated PL measurements.Shown in Fig. 3(b) are the ratios of 532 and 808 nm quantum efficiencies measured by both techniques.Both have similar values and follow similar trends, but the values derived from lasing measurements are higher for a given pump power.This is attributed to a reduction in thermal effects caused by the cooling of the sample by lasing, and demonstrates that the values of quantum efficiencies derived Fig. 2. VECSEL output powers when pumped at 808 nm (black squares) and 532 nm (red circles).The unbroken blue line shows the ratio of the internal efficiencies at the two pump wavelengths.from PL measurements are in general likely to be lower than the values achieved in lasers.
It should be noted that the quantum efficiencies are measured with respect to incident pump power rather than absorbed pump power.Pump absorption was 67% and 63% for the 808 and 532 nm pumps, respectively.Interestingly, the absorption length at 532 nm is short compared to the thickness of the sample active region, as shown in Fig. 1.On this basis, one might assume that the quantum wells at the rear of the sample would be unpumped, leading to absorption and to a drop in efficiency in both lasing and integrated PL measurements.That this is not observed implies that diffusion is sufficient to redistribute the carriers over the active region and somewhat relaxes the VECSEL design constraint that the active region not be thicker than the pump absorption length.

SOLAR-PUMPED PHOTOLUMINESCENCE MEASUREMENTS
Integrated PL measurements were also taken using sunlight as a pump source.The arrangement of the gain sample and photodiode were the same as in the laser-pumped PL measurements.A large aluminum mirror was used to direct sunlight onto an uncoated BK7 lens with 62 mm focal length and 25.4 mm diameter, leading to a pump spot of diameter ∼600 μm on the sample.The loss of the optics directing the sunlight onto the sample was measured to be 48%.Neutral density filters were used to further reduce the power incident on the sample, and five different longpass colored glass filters were used to block portions of the solar spectrum so as to allow a characterization of the response of the sample to different wavelength ranges.Measurements were taken over 20 min on a cloudless day, with the average insolation over the course of the measurements being 720 W∕m 2 as measured by the Dundee Satellite Receiving Station [23], located on the same roof as that where the experiment was performed.
Figure 4(a) shows the number of laser-wavelength photons emitted by the gain sample calculated from the photodiode signal as a function of the neutral density thickness blocking the pump light.Calculated spectra of the light for the different longpass filters are shown in Fig. 4(b) up to the bandgap of Gallium Arsenide and based in an AM1.5 standard solar spectrum.Integrating these pump light spectra allows the total number of pump photons incident on the sample, and thus the quantum efficiency, to be calculated.
The quantum efficiencies are shown as a function of the longpass filter cutoff in Fig. 5.These values are constant to within measurement error across the spectrum, which is consistent with the 808 and 532 nm data at low pump powers in Fig. 3(a).On the basis of the measurements described in Section 3, the quantum efficiencies at short wavelengths can be expected to decrease more quickly at higher powers, but nonetheless the large solar-pumped quantum efficiencies measured here are encouraging for the prospects of solarpumped VECSELs.

PREDICTIONS OF SOLAR-PUMPED VECSEL PERFORMANCE
The pump intensity needed to reach lasing threshold in highpower VECSELs is typically in excess of ∼20 MW∕m 2 [15,24], depending on the specifics of the gain sample and cavity loss.The irradiance at the earth's surface has an intensity of ∼1000 W∕m 2 , of which approximately 65% is at wavelengths which can be used to generate carriers in a GaAs-based VECSEL's active region.It is therefore clear that sunlight must be concentrated by a factor of at least 31,000 in order to even reach lasing threshold, and by a larger factor to achieve useful power output.Thermodynamic arguments limit the maximum concentration of sunlight, C, in a medium with refractive index n to C 1∕n 2 sin 2 Θ S , where Θ S is the angular subtense of the sun.In air this factor is 48,600, meaning that an intensity of 1.6 times the typical VECSEL threshold can be reached at useful pump wavelengths in air.
To reach this intensity requires nonimaging optics rather than just imaging the sun onto the sample surface.We expect the most appropriate types of nonimaging concentrators for SP-VECSEL pumping to be either compound parabolic concentrators (CPCs) [25] or their close relatives, dielectric totally internally reflecting concentrators (DTIRCs) [26].Both have been used to achieve solar concentrations above that predicted for SP-VECSEL thresholds with, to the best of our knowledge, the current record being concentration by a factor of 84,000 [27] achieved in a system using a sapphire DTIRC.Furthermore, both CPCs and DTIRCs are geometrically suitable as VECSEL pumps, producing approximately constant illumination across a circular area with a typical diameter of the order of one millimeter, and being capable of accommodating a laser mode perpendicular to the output aperture.
To use CPCs or DTIRCs as single-stage solar concentrators is unfeasible, as the concentrator lengths would be excessive for any sensible concentration ratio and pump spot size.It is therefore envisaged that a CPC-or DTIRC-pumped VECSEL would use a two-stage concentration system similar to that described in [27], with a circular or parabolic primary mirror producing an image of the sun on the input aperture of the nonimaging optic, which would then concentrate further to the necessary pump intensity.A schematic of an example of a SP-VECSEL pumped by a two-stage solar concentrator is shown in Fig. 6 as an illustration of how the two devices might be integrated.
By referring back to Fig. 2, where the output powers of 808 and 532 nm pumped VECSELs are shown, it is possible to make a basic estimate of the probable output power of a solar-pumped VECSEL.The useful solar power that can be concentrated by a glass (n 1.5) nonimaging concentrator into a spot of equivalent size to those shown in Fig. 2 is ∼5.7 W. Taking a weighted average of the quantum defect across the solar spectrum, and taking the quantum efficiency to be constant with wavelength which, as demonstrated in Fig. 3, is justified at this power well below thermal rollover, an estimated output power and slope efficiency of 0.38 W and 11% can be calculated.This is based on data taken with a pump spot of diameter 300 μm, which is relatively small for a high-power VECSEL and was limited by the available 532 nm pump power.The highest-power VECSEL to date featured a pump spot area of 7.8 × 10 5 μm 2 .If the calculation above is scaled to this pump spot size, then a power of 4.2 W would be achieved.
Two figures of merit are used to characterize the performance of solar-powered lasers, the output power emitted per unit primary collection area which characterizes the laser efficiency, and the output power divided by the square of the M 2 beam quality factor, taken as a metric of the potential usefulness of the output.Current records for these factors are 20 W∕m 2 [28] and 1.1 W [29], achieved in different systems.On the basis of the conservative estimates above, an SP-VECSEL with 1 mm pump spot diameter would reach 49 W∕m 2 and 0.47 W (assuming a typical value of M 2 3 for a VECSEL with a 1 mm diameter pump spot).Furthermore, these values are derived from the behavior of a gain sample which was not designed for solar pumping and a laser which was optimized for operation over a broad range rather than for maximum power.The well-known theoretical limit for a single-junction photovoltaic cell, the Shockley-Queisser limit [30] does not apply exactly to a solar-pumped semiconductor laser, as the stimulated emission will suppress recombination loss and the energy gap between the bandgap of the absorber and the laser wavelength is not taken into account.However, the dominant energy loss mechanisms are the same in both cases, and so it is instructive to note that the value of the theoretical maximum efficiency of a solar-pumped semiconductor laser is likely to be close to the 34% value for a single-junction solar cell.

SUMMARY
We have identified the solar-pumped VECSEL as a candidate to overcome the problems which currently limit the performance of solar-powered lasers.Measurements of solar-pumped PL and of lasing performance when pumped at two different wavelengths allow us to consider key aspects of SP-VECSEL design and to make estimates of key metrics of laser performance.Even these conservative estimates are predicted to surpass existing solar laser performance by large margins.

FUNDING INFORMATION
Engineering and Physical Sciences Research Council (EPSRC) (EP/J017043/2).Fig. 6.Example of a system where a two-stage solar concentrator is used to pump a VECSEL.A primary mirror forms an image of the sun on the input face of a nonimaging concentrator, which further concentrates sunlight to the necessary pump intensity.The VECSEL cavity mode is normal to the gain sample surface and to the nonimaging concentrator input face.

Fig. 1 .
Fig. 1.Layer structure of the VECSEL gain sample along with curves demonstrating the absorption lengths of the 808 and 532 nm pumps.

Fig. 3 .
Fig. 3. (a) Quantum efficiencies of the gain sample when pumped at 808 and 532 nm measured by integrated PL.(b) Ratios of 532 to 808 nm quantum efficiencies derived from PL measurements (unbroken black line) and laser power curves (broken red line).

Fig. 4 .
Fig. 4. (a) Integrated PL measurements pumped by sunlight through combinations of longpass and neutral density filters.(b) Calculated pump spectra for the different longpass filters.

Fig. 5 .
Fig. 5.Quantum efficiencies across the solar spectrum derived from the data shown in Fig. 4.