Invited Paper Dense Spatial Multiplexing Enables High Brightness Multi-kW Diode Laser Systems Holger Schlüter a, Christoph Tillkorn b, Ulrich Bonna a, Greg Charache a, John Hostetler a, Ting Li a, Carl Miester a, Robert Roff a, Thilo Vethake a, Claus Schnitzler c a TRUMPF Photonics, 261 US RT 13 S, Cranbury, NJ 8512, US b TRUMPF Laser, ichhalder Straße 39, 78713 Schramberg, Germany c Ingeneric GmbH, Dennewartstraße 25-27, 5268 achen, Germany BSTRCT The materials processing industry has recently mandated the need for more efficient laser systems with higher beam quality and longer life. Current multiplexing techniques, state-of-the-art laser diodes and novel cooling designs are now emerging as possibilities to meet the ever demanding industry needs. This paper describes the design and initial results of a direct diode system that is aimed at delivering 1.5 kw of output power and a beam divergence of 4 mm mrad on a long life macro-channel cooler. The design entails multiplexing 2 wavelength combined beams and 2 polarization combined beams. Each of the four branches of the direct diode system utilizes a novel stacking and cooling design. The results from one of these branches, 1 wavelength and 1 polarization, are presented where the light is coupled into a fiber with a 4 µm core diameter and a N of.22. Each branch consists of 6 diode laser mini-arrays, where each mini-array consists of four 1 µm wide emitters and a lateral fill factor of 5%. n output power of 5W at 1 C water temperature and 42 W at 25 C are demonstrated through the 4 um fiber. Keywords: Multiplexing, direct diode, fiber coupling, high brightness, 9XX nm 1. INTRODUCTION High power laser systems in material processing are mainly characterized by their CW power level, brightness, wavelength distribution, efficiency and cost per Watt. While the high electro-optic efficiency and lower cost per watt make diode laser systems more desirable than solid state and CO 2 lasers, their limited brightness has restricted their use to only a few niche markets such as plastic welding and soldering..1 1 1 1 1 1 Power P [W] Figure 1: Brightness map for different commercially available diode laser systems 2,3,4,6. High-Power Diode Laser Technology and pplications IV, edited by Mark S. Zediker, Proc. of SPIE Vol. 614, 614M, (26) 277-786X/6/$15 doi: 1.1117/12.657267 Proc. of SPIE Vol. 614 614M-1
larger range of applications could be accessible, if diode laser systems could enter the kw class and still be capable of coupling to a fiber with a core diameter of 4 µm and a N of.22. Theoretically, such brightness could be achieved by spatially multiplexing the emission from commercially available 12 Watt arrays 1 without wavelength and polarization combination (see Figure 1). However, issues like smile of the industry standard 1 mm wide arrays and limited efficiency of complex rearrangement optics have prevented such systems from being reported. fl Therefore we are choosing a novel combination of lateral and vertical spatial multiplexing, polarization multiplexing and multiplexing of two different wavelengths (centered around 915 and 95 nm) in order to achieve the high brightness goal of 1.5 kw with 4 mm mrad (see Figure 2). In order to avoid complex rearrangement optics with limited transmission efficiency the width of the individual arrays is chosen so that the slow axis divergence angle matches the fiber entrance diameter and N. While this still requires a step mirror (simple rearrangement optics) due to the small vertical pitch, a transmission efficiency of greater than 92 % can be achieved after the combination of the step mirror and the fast axis collimation lens. Spatial multiplexing Polarization multiplexing Wavelength multiplexing =1- I c] c71 Figure 2: Different types of multiplexing the emission of individual high power diode lasers. 2. EXPERIMENTL SETUP The fiber core diameter, d f, and the fiber numerical aperture, N, define a fiber entrance pupil beam parameter product (BPP). While the fiber aperture has circular symmetry in space as well as in angular space, the emission from a geometrically multiplexed diode laser stack is typically of rectangular shape in space and angular space. Therefore, it is necessary to compromise between beam parameter product and power loss as depicted in Figure 3. P - b * CD Figure 3: Brightness vs. coupling efficiency as a function of the quotient of fiber entrance pupil beam parameter product divided by source beam parameter product (relative beam parameter product BPP). Proc. of SPIE Vol. 614 614M-2
If complex rearrangement optics are to be avoided, the optical properties of the fiber entrance pupil ( d f = 4 µm, N =.22) and the maximum slow axis divergence of the diode laser emitters (< 12, 95% power content) dictate a maximum width of the source of 8µm in the slow axis. mini-array was designed with four 1µm wide emitters and a fill factor of 5%, resulting in a source width of 7 µm (see Figure 4). The smaller width of the mini-array also allows for a macro-channel cooling design. In the fast axis the low M² value allows to stack 6 devices with a vertical pitch of 96µm. The devices have a fast axis divergence of < 5 (95% power content) and almost diffraction limited performance after collimation can be achieved with a fast axis collimation lens ( f = 91µm). The lower limit of the BPP is 25mm*mrad in fast axis and 35mm*mrad in slow axis, which gives enough safety margin to the case of overfilling to account for movements of the fast axis collimation lens during curing and mechanical tolerances between the mounts of the stacked M-Blocks. Geometrical losses of <2% are expected. Fast-axis (6x) Slow-axis (top view) θ S /2 = 6 f=9lojim 2pm 5% Filling Factor Figure 4: Lateral and vertical spacing chosen for the dense spatial multiplexing of the emitters. The vertical pitch of 96µm requires simple rearrangement optics in order to accommodate a feasible thickness for the mount and heat sink. novel design, termed M-Block (see Figure 5) combines the top n-contact, the fast axis collimation micro-lenses and the step mirror. The 12mm thick M-block allows the use of ~2mm cooling channels, thus improving the lifetime of the cooler without increasing the pitch between stack elements. The beams of 6 mini-arrays are stacked vertically, so that the height of a collimated beam from an M-Block is approximately 5.7 mm. Top Contact Hydrocooling M-Block--- Stepmirror InsuIation Wire bonds Mini-array FC Microlens Stacked beams Figure 5: M-Block module including macro-channel heat sink, step mirror, fast axis collimation and current distribution for 6 mini-arrays of 4 emitters each. Proc. of SPIE Vol. 614 614M-3
Ten such M-Blocks are combined into a stack as shown in Figure 6. The distance between two M-Blocks is.15 mm in order to allow serial electrical connections. The M-Blocks are mounted with their precision machined surfaces to the equally precision machined and electrically isolated central mount. The accurate surfaces on the M- Block and on the central mount ensure minimum angular deviation of the stacked beams from the optical axis. The central mount is mounted into a fully integrated diode laser system that also contains: Slow axis collimation doublet for the entire stack Fast axis telescope to adjust the beam size of fast and slow axis Fiber coupling lens pair Fiber coupler The optical setup is shown in Figure 1. Four such branches can be combined in the system utilizing the integrated wavelength and polarization coupling optics. Central mount Kinematic constraints (M-Block) Insulating coating Series connection Stacked modules Kinematic constraints (Stack) Figure 6: Stack of 1 M-Blocks mounted to a Central Mount. 3. RESULTS One stack of ten M-Blocks, which constitutes one branch of the laser system, has been assembled, fiber coupled and tested. The devices were chosen with a central wavelength around 95 nm. In a future stage a second wavelength, centered around 915 nm and a second state of polarization will be added to the system. Figure 7 shows the fast axis and slow axis far field of an individual four element mini-array mounted on a micro-channel cooler operating CW at 2. This is the same current used for each of the six parallel connected diodes on M-Blocks which operate at 12. The slow axis divergence (95% power content) is measured to be 12.. This should allow nearly optimum performance of the optical system that was laid out for a divergence of 12. The fast axis divergence is measured to be 48.7 and therefore also matches the layout of the optical system. The operation on a micro channel cooler is thermally not identical to the operation on an M-Block (which only contains a 2 mm wide bore for cooling close to the chip mounting surface). However, the emission width was monitored during short periods of fully restricted water flow and no noticeable widening of the far field in the slow axis was detected. Therefore it is anticipated that the thermal properties of the M-Block have no severe negative effect on the optical performance. Proc. of SPIE Vol. 614 614M-4
.9 ± ; ± d 9!..2 ngle (degree) -45-4 -3-2 -to 1 2 ZJ 4 45 ne (deqree) Figure 7: Fast axis and slow axis far-field of an individual mini-array with four 1 µm emitters and 5% fill factor at 2 CW. Figure 8 shows the PI curve of an individual M-Block. Six mini-arrays operate in parallel and achieve 115 Watt at 12 total current before transmission through the micro lens and deflection from the step mirror. lso shown is the far field intensity profile of the fast axis of an entire M-Block. Consistently a 95% power content of less than 4µm is measured in the focus of a lens with a focal length with 2 mm. This value equals a divergence of 2 mrad of the entire M-Block. Given the fast axis beam height (95% power content) of 5.7 mm this equals a fast axis beam parameter product of 2.8 for the entire collimated M-Block. The stack of 1 M-Blocks should therefore stay well below 3 mm*mrad which allows for a safety margin given the fiber acceptance of 4 mm*mrad..1-. 1,2.. I I.. F 1, -w1-- $--,8 C,6. t ea,4,2 1 Power 4-- 2 Slope 2 4 6 8 1 12 14 ---- I [] Figure 8: PI curve of an individual M-Block (w/o micro lenses and step mirror) and the fast axis far-field of an M-Block (incl. micro lenses and step mirror) in the focus of a lens with focal length of 2 mm. Figure 9 shows the performance of individual M-Blocks at 11 and 25 C heat sink temperature before and after attachment of micro lenses and step mirror. From these values the loss induced by fast axis collimation and step mirror deflection is calculated and presented. The M-Blocks show a severe performance variability. Near-field measurements of the slow axis suggest the presence of undesired modes spanning more than one emitter. Proper optical isolation between the emitters may solve this problem in the future. In addition, some M-Blocks show higher than expected losses from collimation and step mirror deflection. While a majority of devices show a loss between 6-8%, some devices experience losses between 1 and 16%. On these devices, the step mirror contains a slight ledge (see Proc. of SPIE Vol. 614 614M-5
Figure 9) due to a nonconformance in the manufacturing process. The defect can be compensated by tilting the micro lens about the axis of cylindrical symmetry by 2-4. However, this procedure results in the higher collimation loss. MBlock power at 11 Mirror surface Ledge on mirror surface 11 15 Uncollimated Collimated Loss 3% 25% Step mirror on MBlock 67: 16% total collimation loss Mirror surface Power [W] 1 95 2% 15% 9 1% No ledge 85 8 49 5 57 58 62 64 65 67 68 75 76 78 9 98 116 5% Step mirror on MBlock 49: 6% total collimation loss % MBlock Number Figure 9: Performance of individual M-Blocks before and after attachment of micro lenses and step mirror and partial side views of two different step mirrors. Figure 1 shows the optical setup of one branch and the intensity profile detected with a ceramic detector card at 15 at various positions of the optical setup. The stack of 6 mini-arrays is clearly visible just in front of the slow axis collimation lens. fter the fast axis telescope the beam width and height in slow and fast axis are almost comparable, the entire beam is now collimated. Before the fiber coupling, a water cooled circular aperture is added to limit the numerical aperture and therefore protect the fiber. Slow xis Collimation Circular aperture to limit N Fast xis Telscope Fiber coupling Reflection of unknown origin Figure 1: Optical Setup of one branch containing 1 M-Blocks on a central mount, slow axis collimation, fast axis telescope, circular aperture to limit N and fiber coupling. Proc. of SPIE Vol. 614 614M-6
t a heat sink temperature of 1 C a total output power of 1 Watts was achieved for the stack of 1 M- Blocks at 14. This equals an operation of 5.9 and 4.2 Watt per 1µm wide emitter (see Figure 11). Such a current and power level can be easily sustained by state-of-the-art devices in the 9xx nm wavelength range 1,6,7. Therefore, no issues are anticipated for extended lifetime tests of the system. It is also predicted that the macro-channel cooling design can sustain a long-life time as well. t this current level (14) the power transmitted through the fiber is ~5 Watt equal to a coupling efficiency of approx. 5%. This value is still far below the anticipated 72% (8% coupling efficiency per Figure 3 8% losses at the uncoated fiber entrance and exit) and is the current focus of investigations. Figure 1 shows reflections to the right of the circular aperture, which were also detected on the left side. t 14 these reflections contain several hundred watt of power. Currently their origin is unknown. t 25 C heat sink temperature the maximum stack power achieved was 825 Watt at 12 of which 42 Watt were transmitted through the fiber. Higher currents were not tested, as the system begins to thermally rollover at this temperature and current level. more sophisticated cooling design with smaller bore-diameter macro-channels will enable higher currents without rollover comparable to the current systems operation at 1 C heat sink temperature. The variability of M-Block performance shown in Figure 9 and losses in the optical system contribute to the lower than expected output power of the stack of M-Blocks. However, combination of two wavelengths and two states of polarizations still allow more than 15 Watts of output power from a fiber with 4 µm core diameter and a N of.22. Further improvement of the brightness is possible by utilization of individual slow axis collimation lenses (and thereby improving the lateral fill factor) and the combination of more than two wavelengths. It is important to notice that the power is coupled to a numerical aperture of N=.22. This allows the use of standard fiber couplers, as are commercially available for multi-kw operation. The ability to use such standard components is critical in ensuring long-term operation. Figure 11 shows no degradation in power from the fiber over a period of 2 hours. fter this time the entire system is thermalized and no further degradation is anticipated. 11i 1 9 6 7 6 5 4 3 2 1 5 1 Current (] ISO 1 5 7 6 5 4 C, 3. C) 2 6 5 4 3 2 1 :: :3: 1:: 1:3: 2:: TIne (hj Figure 11: Power of the stack of 1 M-Blocks and power transmitted through the fiber, coupling efficiency and power stability of the fiber coupled power in a 4 µm fiber with a N of.22 at a heat sink temperature of 1 C. Initial power stability measured at 14. Proc. of SPIE Vol. 614 614M-7
4. SUMMRY Performance results from the first branch of four branches of a fiber coupled high power diode laser system were demonstrated. t 1 C heat sink temperature 1 Watts of CW output power were achieved from a stack of 1 M-Blocks. The beams were fiber coupled to a fiber of 4µm core diameter and a N of.22. The power transmitted through the fiber was 5 W. This power was measured behind the fiber stable for more than 2 hours. t 25 C the stack power reached 825 Watts with 42 Watts transmitted through the fiber. This experiment has demonstrated the feasibility of fiber coupled high brightness diode lasers in the multi-kw range. Future wavelength and polarization combination of multiple such branches will allow for a wider range of applications in materials processing. The technology will benefit further from advances in diode laser power levels, which are anticipated for the near future. REFERENCES 1. N. Lichtenstein et al., DPSSL and FL Pumps Based on Telecom Pump Laser Technology: Changing the Industry, Proceedings of SPIE, V 5336, p. 77, 24. 2. Catalog data taken from www.laserline.de, pril 25 3. Catalog data taken from www.dilas.de, pril 25 4. Catalog data taken from www.jold.de, pril 25 5. V. Gapontsev et al., High-efficiency 97 nm multimode pumps, Proceedings of SPIE, V 5711, p. 42, 25. 6. V. Rossin et al., High Power High Efficiency 91-98nm broad rea Laser Diodes, Proceedings of SPIE, V 5336, p. 196, 24. Proc. of SPIE Vol. 614 614M-8