Narrow-line fiber-coupled modules for DPAL pumping

Narrow-line fiber-coupled modules for DPAL pumping Tobias Koenning*, Dan McCormick, David Irwin, Dean Stapleton, Tina Guiney, Steve Patterson DILAS Diode Laser Inc., 9070 South Rita Road, Suite 1500, Tucson AZ 85747 ABSTRACT Recent advances in high power diode laser technologies have enabled advanced research on diode pumped alkali metal vapor lasers (DPALs). Due to their low quantum defect, DPALs offer the promise of scalability to very high average power levels while maintaining excellent beam quality. Research is being conducted on a variety of gain media species, requiring different pump wavelengths: near 852nm for cesium, 780nm for rubidium, 766nm for potassium, and 670nm for lithium atoms. The biggest challenge in pumping these materials efficiently is the narrow gain media absorption band of approximately 0.01nm. Typical high power diode lasers achieve spectral widths around 3nm (FWHM) in the near infrared spectrum. Gratings may be used internal or external to the cavity to reduce the spectral width to 0.5nm to 1nm for high power diode laser modules. Recently, experimental results have shown narrower line widths ranging from picometers (pm) at very low power levels to sub-100 picometers for water cooled stacks around 1kW of output power. The focus of this work is a further reduction in the spectral line width of high power diode laser bars emitting at 766nm, with full applicability to other wavelengths of interest. One factor limiting the reduction of the spectral line width is the optical absorption induced thermal gradient inside the volume Bragg grating (VBG). Simulated profiles and demonstrated techniques to minimize thermal gradients will be presented. To enable the next stage of DPAL research, a new series of fiber coupled modules is being introduced featuring greater than 400W from a 600µm core fiber of 0.22NA. The modules achieve a spectral width of <<0.1nm and wavelength tunability of +/- 0.15nm. Keywords: High power diode laser, DPAL, VBG, VHG, narrow line, spectral width, wavelength locked, defense 1. INTRODUCTION In the search for high energy lasers with improved efficiency and scalability, W. F. Krupke et al. in 2003 proposed a type of optically pumped gas laser for directed energy, the diode-pumped alkali metal vapor laser (DPAL) [1] which has since shown very promising results. Development efforts are ongoing, with Shapiro and Teare stating that in order to achieve higher power, the efficiency of coupling between the pump laser energy and the chemical cell must be increased [2]. One of the main reasons for inefficient coupling is the difference in spectrum between the diode laser emission and the absorption of the DPAL medium. According to Perram [3], depending on the DPAL design the diode emission spectrum should be as narrow as 0.02nm (10GHZ) for low-pressure gain cells and as wide as 0.2nm (100GHz) for rubidium gas cells that are operated at about 10 atmospheres of pressure. In 2014, DILAS introduced high power diode laser stacks at 780nm with optical power exceeding 1kW while maintaining a spectral width (FWHM) of <85pm [4]. Due to size constraints and efficiency considerations, resistive heaters were used to temperature tune the Volume Bragg Grating (VBGs) used to lock the diode wavelengths which resulted in a wavelength tuning range of 155pm. The tuning capability was used to first overlap all individual bar spectra within the stack and then to allow fine-tuning of the wavelength of the entire stack with respect to the absorption peak of the gas. Based on results with the 780nm stack, development efforts continue in multiple areas to provide pump modules for DPAL and other applications that require extremely narrow spectral line width. First is the expansion to other emission wavelengths. While chip material at 780nm for rubidium and 852nm for cesium pumping are available at high output power exceeding 100W per bar, DPAL developments based on potassium are limited by the lower pump power of about * t.koenning@dilas-inc.com; phone +1 (520) 282 5986; fax +1 (520) 300 8230; www.dilas.com

50W per bar typically available at 766nm. Data will be shown for new chip material that has been developed at 766nm with higher fill-factor and longer cavity length to decrease both the facet optical intensity and the current density while increasing surface area to improve cooling of the device. The second area of focus is the spectral width of the device. While both the spectral width and the tuning range shown for the 780nm stack represent the state of the art by a good margin, applications demand an even narrower line width and larger tuning range to increase system efficiency and flexibility. At the same time, a large tuning range allows relieving the manufacturing tolerances of the VBGs, one of the cost drivers of the module. Size constraints and efficiency demanded the use of resistive heaters for the 780nm stack; the down-side of using heaters to tune the temperature is the inability to tune to a shorter wavelength than the cold cavity wavelength of the VBGs. Thermo-electric coolers (TECs) can solve this problem and possibly widen the tuning range by allowing tuning in both positive and negative directions. However, due to their typical efficiency of less than 50%, TECs introduce extra heat into the overall system and increase the design complexity both for the mechanics and the electronics. Based on data collected on the 780nm stack, thermal gradients inside the VBGs appear to be a driver for broadening of the spectral width compared to cold cavity measurements of the VBGs. Tailored heat sinks to counteract the thermal gradients are being tested to minimize spectral line width. At the same time, VBGs with increased thickness are used for a further reduction in line width. The third area of further development is the beam delivery. While microchannel cooled, high power diode laser stacks are about the smallest and most efficient type of diode laser platform that exists, fiber coupled modules entail benefits for the user that often justify the increase in complexity, cost, and performance degradation. These benefits include flexibility in placing the laser away from critical areas and easy drop-in module replacements without optical realignment. A new series of fiber coupled modules specifically designed for DPAL applications is under development and includes specialty features like wavelength tuning of individual laser bars, a programmable set-point for center wavelength, and an RS485 bus interface to control multiple laser heads through a common software interface. Similar to the stacks presented in 2014 [4], the capability to tune individual bar spectra serves a dual purpose. First, it is used to minimize the spectral line width by overlapping individual bar spectra; second, it is used to tune the ensemble of laser bars to the desired wavelength, typically the peak absorption of the gain medium. In addition to the new features described above, closed loop temperature control for all VBGs is added to the control electronics in order to increase the overall system stability and minimize wavelength variations induced by changes in optical power or environmental factors. 2. 766nm CHIP DEVELOPMENT Starting with readily available epitaxial designs, long cavity lengths and high fill factor bars have been manufactured and tested to close the gap between 766nm chip material and other DPAL pump wavelengths like 780nm and 852nm, where output power exceeding 100W per bar is already available. Longer cavities reduce thermal resistance by increasing surface area. However, if internal losses dominate the design, the long cavity length may come at the penalty of lower electrical-to-optical efficiency. A high fill factor increases surface area. For the same output power, it also reduces the optical intensity at the front facet which is the limiting factor for many high power diode lasers. At the same time, a larger fill factor increases the threshold current which may impact the efficiency of the device. In order to achieve best possible locking performance, the front facets of the laser bar are coated with a low reflectivity AR coating. Different VBG reflectivities are modeled and tested in order to determine the best trade-off between threshold current, slope efficiency, damage threshold, and locking behavior. no VBG R=5% VBG R=10% VBG R=15% VBG Slope Eff. [W/A] 0.95 0.84 0.79 0.82 Threshold [A] 55.8 44.0 40.0 40.2 Table 1: Measured chip performance with various VBG reflectivities (R) Table 1 shows the measured chip performance at low operating current of up to 80A. Higher operating currents are neglected for the determination of slope efficiency and threshold current in order to exclude thermal roll over effects that

occur at higher operating current. The data shows a reduction in threshold current from about 56A for the low AR coated bar down to 40-44A with different VBG reflectivities. At the same time, the slope efficiency gets reduced with high VBG reflectivities. Figure 1a: Simulated chip performance with various facet reflectivities (4mm cavity length, 50% fill factor) Figure 1b: Measure chip performance with various VBG reflectivities (4mm cavity length, 50% fill factor, low AR facet coating) Figure 1a shows simulated chip performance for various facet reflectivities. The modeled data shown is for a 4mm long cavity device with a 50% fill factor, similar to the chip material used for the experiment shown in Figure 1b. The bar was first tested without VBG, then with three different VBGs with reflectivities of 5%, 10%, and 15%. A maximum power of 120W was achieved at about 200A with minimum thermal roll over at 5% VBG reflectivity. Higher VBG reflectivities show more pronounced thermal roll over with catastrophic failure of the bar at 200A for the 15% VBG reflectivity. The measured threshold current is well aligned with the simulated data; however, the slope efficiency is significantly lower than modeled for all reflectivities, resulting in a maximum output power at 200A about 40W lower than the model predicts. Further efforts in chip design will focus on increasing the slope efficiency by addressing internal losses and further optimizing the cavity length. At the same time, front facets will be passivated to increase the damage threshold and allow operation at even higher output power. The current chip material already pushes the boundary of 766nm chip material past the 100W/bar mark, enabling researchers to conduct experiments with potassium-based DPAL with similar pump power as is available for rubidium. 3.1 Simulation 3. THERMAL GRADIENTS IN VBGS As previously discussed [5], two of the critical parameters that influence the width of the locked spectrum when using VBGs are the thickness of the VBG and thermal gradients inside the VBG. One limiting factor for the thickness of the VBG is the achievable uniformity during manufacture; non-uniformity broadens the line by chirp. Based on previous results and discussions with manufacturers, a VBG thickness of 8mm was chosen compared to 5mm used for previous experiments. In theory, the spectral width of the grating is inversely proportional to the length [6], however, both the non-uniformity of the grating and thermal gradients may broaden the measured spectrum. Transmission measurements of the gratings with a white light source can reveal inhomogeneity inside the gratings and are used as a best-case target for the spectral line width of the laser module. In order to better understand thermal gradients inside the VBG, a closed-form solution of coupled wave equations was used to model the intensity distribution inside the VBG, which was based primarily on work by Bar and Cohen [7]. These equations describe how light is partially reflected from the forward travelling wave at each of the internal VBG index interfaces, creating a backwards travelling wave which couples into the diode laser cavity, locking the laser diode spectrum at the Bragg wavelength. As a reasonable approximation, the sum of these two waves represents the total optical intensity, and heat generation in the presence of absorption at any

position in the VBG. Since the forward wave is slowly donating power to the backwards travelling wave as it moves through the VBG, optical intensity decreases axially through the VBG. This effect is more pronounced with higher diffraction efficiency (DE) and is also most pronounced at the Bragg wavelength. Several assumptions were made to solve the wave-coupled equations and express a closed-form solution solely in terms of power, wavelength input, and intrinsic VBG parameters. These assumptions are very close to expected real-world performance and include a perfect anti-reflective (AR) coating on the VBG facets, low bulk absorption (<<1/cm), perfect collimation of incoming light, and good locking performance of VBG (where most power is within the locked spectrum). The result of the equations is displayed in Figure 2 which shows the axial optical intensity present inside the VBG for different diffraction efficiencies. One implication from the model is that for VBGs of higher DE, a much higher thermal gradient is expected through the VBG due to absorption losses. This suggests that a lower DE grating will produce more manageable thermal effects and less spectral broadening. Figure 2: Modeled axial intensity distribution inside the VBG at various reflectivities Based on the results above, a finite element analysis (FEA) model was created. The model contains the anticipated geometry of the mounted VBG, heat sink, and TEC for temperature control (Figure 3). To model the steady state temperature gradients within the VBG, two thermal sources were used: a fixed TEC temperature and a volumetric heat generation source within the VBG caused by laser absorption. This latter source was modeled as a three-dimensional function which takes into account the Gaussian and Super-Gaussian shape of the collimated emitters passing through the VBG, as well as the axial intensity distribution previously predicted by the wave-coupled equations described above. For simplicity, the laser diode bar was modeled as a single source rather than treating all 23 diode emitters separately. Due to the high fill factor of the beam after collimating optics, the simplification is expected to have a negligible effect on the result. Figure 3a: FEA model of VBG on tailored heat spreader and TEC Figure 3b: Temperature profile through the center plane of the VBG. Lateral profile plots below are shown along the red line.

Since there are multiple factors upon which the model depends which are not well quantified (e.g., total percent absorption of VBG material, air conduction coefficients, etc.), the total power of this VBG heat generation source was tuned to match empirical observations of VBG temperature obtained through previous experiments. The purpose of this model is to explore the interaction between VBG heating and heat sink geometry and minimize thermal gradients. Therefore, an exact accounting of total VBG power is not needed. The result of this analysis is a steady-state model of internal VBG temperature (Figure 3). To explore effects of different heat sink geometries, the center plane of the VBG was considered and the magnitude of axial and lateral gradients tracked. Total maximum internal temperature, flatness of the lateral temperature profile, and peak temperature location within the VBG were also monitored. Figure 4: Lateral temperature through center plane of VBG on unstructured vs. tailored heat sinks (left) and modeled at various TEC temperatures on tailored heat sink (right). Multiple configurations of heat sinks were considered: symmetric axial grooves, addition of a lateral cut, and addition of a copper heat spreader to the top of the VBG. Thus far, the optimal solution found is the addition of two 3mm axial grooves in the heat sinks as shown in Figure 3a. These grooves have the effect of reducing cooling at the lateral edges of the VBG which serves to raise the temperature in these areas and more uniformly match the temperature at the VBG center. The overall lateral temperature profile is flattened and the reduced thermal gradients are expected to significantly improve wavelength broadening due to non-uniform thermal expansion within the grating (Figure 4). These grooves also raise the overall internal temperature, but this effect may be adequately compensated for by specifying the VBG grating layer thickness appropriately and will be calibrated during initial tests. Lastly, the effect of TEC heating on the VBG was investigated for a range of TEC set point temperatures. The modeling showed good control of internal VBG temperature with TEC heating/cooling and little effect on the shape of the internal VBG profiles. This is desired since the locked wavelength selected by the VBG can be tuned by targeting the internal VBG temperature without sacrificing locked bandwidth performance caused by a changing temperature profile shape. 3.2 Experimental results VBGs with three different reflectivities ranging from 5% to 15% have been tested. The VBGs are 8mm long with an aperture of 11mm x 3.1mm. The beam is collimated in both axes to <10mrad divergence prior to the VBG. All VBGs show exceptional locking behavior across the entire power range, from just above threshold to full power, with measurement noise limited side mode suppression ratios (SMSR) of >20dB. Figure 5a shows the spectral performance of one VBG (R=10%) over a power range from 0 to 85W. The VBG was held by a mechanical gripper, but not thermally contacted to a heat sink. The central wavelength shifts by about 1.8pm/W. The measured spectral width FWHM is 45pm at low power and rises to about 50pm at 85W output power. A similar behavior can be seen for the 90% power included spectral line width rising from about 75pm to 92pm.

Figure 5a: Spectral performance of VBG (R=10%) vs. optical power Figure 5b: Spectral performance at 85W optical power for various VBG reflectivities The increase in spectral line width can be explained by the increase in thermal gradient due to higher amounts of optical power. Figure 5b shows a comparison of different VBG reflectivities. Both the central wavelength and the line width are higher for larger reflectivities. This qualitatively matches the calculated intensity distribution shown in Figure 2 which implies larger thermal gradients for larger VBG reflectivities. Tailored heat sinks to further reduce the spectral line width have been tested but experiments have not been finalized yet. Current results show a reduction of up to 10% in FWHM line width compared to the data shown above. However, tests at full operating power have not been conducted yet. 4.1 Design 4. FIBER COUPLED MODULE Optical fibers are a common delivery system for high power diode lasers. They allow the placement of the laser head away from critical areas in the setup and ensure drop-in replaceability superior to most free-space beam delivery systems. For the targeted application of DPAL pumping, brightness is not as critical as for some other applications like fiber laser pumping, while efficiency remains critical. For this reason, a comparatively large fiber diameter of 600µm was selected and a commonly used numerical aperture (NA) of 0.22. As cooling performance is critical to achieving the highest possible output power while maintaining good efficiency, micro channel cooled heat sinks are used to cool the laser bars. Based on beam parameter calculations and power target, the package is populated with 8 laser bars, each about 10mm wide. Commercially available beam shaping optics, in conjunction with DILAS patented prism design for spatial combination, are used to couple the beam efficiently into the fiber. In order to achieve the desired spectral line width of <<0.1nm, VBGs are used to selectively provide feedback into the laser cavity at the desired wavelength, hence locking the laser bar at the desired wavelength by increasing the effective gain at that wavelength per section 3. In order to achieve wavelength tuning capability, TECs are used to control the temperature of each VBG. A closed loop temperature controller has been developed that maintains the desired temperature of eight VBGs individually. Changing individual TEC temperature set points allows overlapping of individual bar spectra, thus minimizing spectral line width. By changing the temperature set point of all channels at the same time, the wavelength of the entire module can be shifted, allowing the user to tune the diode laser and maximize the overlap between the diode laser emission spectrum and absorption of the gas. As DPALs are intended to deliver high output power and state of the art research is reaching into the multi-kilowatt range, scalability is important. This can easily be achieved by adding a multitude of fiber coupled modules to the system while arranging the fiber outputs in close proximity to each other. User interaction required for maintaining a narrow spectrum or keeping the center wavelength stable needs to be minimized. At the same time, the user needs easy access to key parameters like the center wavelength of all laser modules used to pump the gas (e.g., to sweep the laser wavelength through the absorption range of the gas in order to find the absorption peak). Each laser features an industry standard

RS485 bus interface. Via central control software, key laser parameters can be monitored, factory set wavelength tuning coefficients can be modified, and most importantly, the set point for center wavelength of all modules can be changed with a single command. Besides the temperature controllers and communication, the on-board microcontroller monitors parameters critical to the safe operation of the laser and provides an interlock signal to the user. 4.2 Results A prototype has been manufactured based on 780nm chip material similar to the stack presented previously [4]. This particular bar design is limited to about 60W output power per bar to maintain good lifetime. The 5mm long VBGs without the previously described tailored heat sinks have been used. A passively cooled, high power SMA fiber with a core size of 800µm and an NA of 0.22 was used for the first prototype. Future modules will use a water-cooled SMA fiber with a specified 600µm core diameter. RS485 SMA fiber Desiccant Cooling water TEC supply Diode power Figure 6a: LI curve and efficiency for 780nm prototype Figure 6b: DPAL fiber coupled pump module with integrated 8-channel temperature controller The targeted output power of 400W from the fiber was achieved at about 70A. The electrical-to-optical efficiency of the module at full power is 38.5% measured through an uncoated fiber. Anti-reflective coatings on both fiber ends typically increase the power and efficiency by 6-8%. All VBG temperatures were individually tuned to the desired wavelength. The resulting spectral width (FWHM) was 0.061nm at full power where 90% of the power was within 0.113nm. Temperature tuning of individual VBG temperatures was repeated at the low end of the temperature tuning range and again at the high end. The resulting tuning range was 0.232nm (Figure 7a) which is in line with the expected 8pm/ C. Figure 7a: Measured spectra for 780nm prototype at different VBG temperatures across wavelength tuning range Figure 7b: Individual VBG temperature set points for all eight channels at three target wavelengths across the tuning range The temperature tuning range was conservatively limited to 10 C on the low end to avoid condensation and 50 C on the high end to avoid problems with the TECs or epoxy used to mount the VBGs (Figure 7b). A slightly larger tuning range can be achieved by means that avoid condensation or by using TECs and epoxies that can withstand higher temperature.

Another way to increase the tuning range is to select VBGs with better matching center wavelength to minimize temperature variations between VBGs which take away from the tuning range of the module. The module was operated at 60A for one hour to test its stability. Results are shown in Figure 8. Cyclic variations in power can be attributed to changes in water temperature both in the cooling loop for the laser and for the power detector. The center wavelength is stable within <2.5 picometers after an initial stabilization time of about 2 minutes. However, even after a sudden step in power from 0W to 325W, the central wavelength is within 10 picometers of the steady state wavelength within 10 seconds. This behavior is expected to be further improved by implementation of a laser power monitor into the micro controller which increases the temperature set points for all VBGs at lower operating power. During laser-off periods, VBGs will be held at higher temperatures to compensate for the turn-on behavior shown below. Compared to the spectral line width of about 0.06nm and expected line width for the 766nm modules of 0.04nm, none of the variations shown in Figure 8 should impose major problems onto the user. Figure 8: 1-hour stability test at 60A The next step is to build a module based on the 766nm bars. The expected module output power is about 600W with a spectral line width of 0.040-0.045nm and a similar tuning range as presented here for the 780nm module. Firmware upgrades to the microcontroller are planned over the next months to incorporate automatic tuning at different power levels based on wavelength set point. 5. SUMMARY A new fiber coupled module for DPAL pumping or other narrow spectral line width applications has been developed. The module uses eight water-cooled, high power diode laser bars and can deliver between 400W and >600W from a 600µm fiber, depending on wavelength and available chip material. A spectral line width of <0.07nm has been shown, and data presented that suggest line width as low as <0.045nm is achievable in follow-on builds. The module includes a built-in controller that allows tuning the center wavelength over a range of about 0.25nm via an RS485 interface, allowing easy control of multiple modules via the same cable, further easing power scaling by adding more modules to the system. The technology used to achieve the narrow line width is based on volume Bragg gratings (VBGs). Active temperature tuning and tailored heat sinks have been presented that allow overlapping all individual bar spectra of the ensemble as well as tuning the entire module. Compared to previously shown results, the wavelength stability has been improved by almost a factor of ten, keeping the center wavelength within 2.5pm during a one hour test. In a parallel effort, new chip material at 766nm has been developed and tested. Output power of 120W CW has been achieved from a single 10mm laser bar. Efforts continue to improve the efficiency of the material and increase the maximum output power further. At the same time, this chip material enables researchers to conduct experiments on potassium that were previously only possible on other materials like rubidium due to the limited power per laser bar available at 766nm. The technology developed here is transferrable to other diode laser wavelengths and products like water-cooled stacks.

6. ACKNOWLEDGEMETNS The authors would like to gratefully acknowledge the support of the High Energy Laser Joint Technology Office (HEL- JTO) contract#fa9451-12-d-0191 for portions of this work. Opinions, interpretations, conclusions, and recommendations are those of the authors and are not necessarily endorsed by the U.S. government. We also wish to thank Dr. Frank Havermeyer of Ondax for his insightful contributions. REFERENCES [1] W. F. Krupke, R. J. Beach, V. K. Kanz, and S. A. Payne, Resonance transition 795nm rubidium laser, Opt. Lett. 28, p. 2336, 2003. [2] Shapiro, J.; Teare, S. A Modular Control Platform for a Diode Pumped Alkali Laser, Proceedings of the Advanced Maui Optical and Space Surveillance Technologies Conference, held in Wailea, Maui, Hawaii, September 17-19, 2008, Ed.: S. Ryan, The Maui Economic Development Board., p.e70. [3] Perram, Glen. Intensity scaling for diode pumped alkali lasers, SPIE Newsroom (6. December 2012), DOI: 10.1117/2.1201211.004560. [4] Tobias P. Koenning, David A. Irwin, Dean Stapleton, et al., "Narrow line diode laser stacks for DPAL pumping", Proceedings of SPIE Vol. 8962, 89620F (2014). [5] Pandey, R., et al. Narrow-line, tunable, high-power diode laser pump for DPAL applications, Proceedings of the SPIE, Volume 8733, id. 873307 16 pp. (2013), DOI: 10.1117/12.2014301. [6] Kogelnik, H. (1969), Coupled Wave Theory for Thick Hologram Gratings. Bell System Technical Journal, 48: 2909 2947. doi: 10.1002/j.1538-7305.1969.tb01198.x [7] Bar-Cohen, Avram, Bongtae Han, and Kyoung J. Kim. "Thermo-Optic Effects in Polymer Bragg Gratings." Micro- and Opto-electronic Materials and Structures: Physics, Mechanics, Design, Reliability, Packaging. New York: Springer, 2007. A65-110.