Macro-channel cooled, high power, fiber coupled diode lasers exceeding 1.2kW of output power Tobias Koenning*a, Kim Alegriaa, Zoulan Wanga, Armin Segrefa, Dean Stapletona, Wilhelm Faßbenderb, Marco Flamentb, Karsten Rotterb, Axel Noeskeb, Jens Biesenbachb a DILAS Diode Laser Inc., 9070 South Rita Road, Suite 1500, Tucson AZ 85747; b DILAS Diodenlaser GmbH, Galileo-Galilei-Strasse 10, 55129 Mainz, Germany ABSTRACT We report on a new series of fiber coupled diode laser modules exceeding 1.2kW of single wavelength optical power from a 400um / 0.2NA fiber. The units are constructed from passively cooled laser bars as opposed to other comparably powered, commercially available modules that use micro-channel heat-sinks. Micro-channel heat sinks require cooling water to meet demanding specifications and are therefore prone to failures due to contamination and increase the overall cost to operate and maintain the laser. Dilas’ new series of high power fiber coupled diode lasers are designed to eliminate micro channel coolers and their associated failure mechanisms. Low-smile soldering processes were developed to maximize the brightness available from each diode laser bar. The diode laser brightness is optimally conserved using Dilas’ recently developed propriety laser bar stacking geometry and optics. A total of 24 bars are coupled into a single fiber core using a polarization multiplexing scheme. The modular design permits further power scaling through wavelength multiplexing. Other customer critical features such as industrial grade fibers, pilot beams, fiber interlocks and power monitoring are standard features on these modules. The optical design and the beam parameter calculations will be presented to explain the inherit design trade offs. Results for single and dual wavelengths modules will be presented. Keywords: High power diode laser, fiber coupling, materials processing
1. INTRODUCTION Two applications drive the development of high power fiber coupled diode lasers. The materials processing industry has mandated the need for more efficient laser systems with higher beam quality and longer life [1]. While lower powered diode lasers are well established in soldering and plastics welding applications, solid state, fiber, and CO2 lasers dominate the industrial welding market. Diode lasers are physically smaller than other lasers, and their initial capital cost is not as large as it might be for traditional welding lasers because diode lasers have fewer system components [2]. While the beam quality achieved today does not allow the use of diode lasers for key-hole mode welding, the higher absorption of Aluminum at 800nm compared to CO2 and Nd:YAG lasers combined with the generally higher system efficiency makes the diode laser an attractive tool [3]. Another advantage of the diode laser is the capability of burr-free welding of stainless steel. With the beam quality achieved here, the lasers can be used in combination with scanners in applications like contour and quasi-simultaneous welding. Besides materials processing applications, the demand for stronger pump sources for both solid state and fiber lasers with ever better beam quality rises quickly. The single wavelength version presented here is well suited for this task and multiple systems can be combined by use of fiber pump combiners or different free-space configurations depending on the pumping application. While dual wavelength pumping is not common today, research is being conducted in this area. When fiber-coupling diode lasers, two key factors determine the outcome of the final product. The first is the source design. The brightness of the source sets the boundary conditions and the final result can never be better than the source. The second major challenge is to maintain the brightness of the source through all steps of the optical design. Other design requirements include lifetime, efficiency, and extra features like safety interlocks, power monitors, and pilot
*
[email protected]; phone +1 (520) 282 5986; fax +1 (520) 232 3499; www.dilas.com
beams. While the laser shown here is based on standard chip material, all other aspects described above have been addressed during the design of the laser and will be presented here.
2. DESIGN 2.1 Heat sink Multiple design requirements for the new Dilas kilowatt-class fiber coupled diode lasers made it necessary to come up with a heat sink solution that has not been used for fiber coupled modules before. While most existing fiber coupled packages at Dilas are based on CS-type heat sinks, these heat sinks require about an inch-by-inch area per bar. For modules exceeding 1KW of optical power this leads to package dimensions that are not desirable. Besides that, the thermal resistance of CS-type heat sinks has its limitations with ever increasing power capabilities of modern laser bars. Micro-channel coolers meet the size goal and have a low thermal resistance compared to CS-type heat sinks but adhere to other disadvantages. For one, the cooling water has to be tightly controlled and small particle filters are required to maintain industrial grade life times. Even with perfect water conditions the flow inside the micro channels causes wear and therefore the lifetime of a laser diode is in most cases limited by the lifetime of the heat sink [4]. To achieve long term stability in copper, a flow rate of 2 m/s should not be exceeded. Otherwise, erosion will increase the thermal resistance of the cooler over time and limit the life time of the device. Another issue with micro-channel coolers is the amount of ‘smile’ that becomes inherent when mounting the semiconductor bar to the heat sink. In general it is easier to achieve low ‘smile’ on a solid heat sink compared to a micro-channel cooled heat sink where micro channels are located only a few hundred micron below the laser bar. In order to combine the advantages of both technologies, Dilas uses a macro-channel cooled heat sink. By flowing the cooling water through the heat sink of each laser bar, a low thermal resistance of about 0.68 K/W is achieved. At the same time material thicknesses and water channels are designed large enough to guarantee stable mounting of the laser bar with low ‘smile’ and prevent any problems caused by particles in the cooling water. In addition the heat sink is designed in a way that protrudes forward in the propagation direction of the laser. This design adds one axis for the heat flow and thus reduces the overall thermal resistance of the heat sink. As shown in Figure 1 (right) the heat sink is open to the bottom, which allows an efficient application of protective coatings to prevent corrosion. Thermal Resistance [K/W] 1.3 Passively cooled
Passively cooled - fit
Macro-channel
Macro-channel - fit
"Micro-channel"
"Micro-channel - fit"
1.2 1.1 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0
5
10
15
20
25
30
35
40
Flow Rate [l/h]
Figure 1: Thermal resistance measurement of various heat sink types for one type of 976nm chip material (left); macro-channel cooled heat sink (right) 2.2 Spatial beam combining by ‘rotating prism’ Currently, two general methods of spatial beam combining for high power diode lasers are well established. The first method is to simply stack heat sinks on top of each other. While this method works well for micro channel cooled heat sinks, the stacking pitch is determined by the thickness of the heat sink and typically not suitable for passively cooled
heat sinks. Here, the typical method of spatial beam combining is the use of mechanical steps in the base plate of the laser in combination with a 90 degree redirection of all beams by means of mirrors. While this method is well established, it has two disadvantages. The first disadvantage is the fact that steps have to be machined into the base plate of the laser. For passively cooled laser bars this base plate acts as a cooling interface and surface roughness is critical in order to achieve good thermal contact between each laser bar and the base plate. Steps add weight and make it more difficult to achieve a low surface roughness compared to an even surface and increase overall machining cost. If water is used to cool the base plate, channel structures become more complex or the thermal resistance may differ between bars due to different material thicknesses caused by these steps. This may result in a broader spectrum of the combined beam due to a varying drift in center wavelength between bars depending on the position of each bar on the step structure.
Figure 2: Common methods for spatial beam combining: “step-and-mirror” (left); vertical stacks (right) A second disadvantage of the step structure is the mechanical tolerance of the step and the inflexibility of the design. When trying to maintain the brightness of all laser bars through the optical train, it is crucial that beams are stacked as dense as possible in order to achieve the lowest beam parameter product possible. Mechanical steps inherit tolerances that make it very difficult to achieve this goal. Typically, a step size is chosen that minimizes clipping but as a compromise leaves some gaps between stacked beams to compensate for tolerances which leads to an increased BPP of the combined beam. Also, mechanical steps cannot be adjusted easily to vary the pitch. In praxis a variable pitch is very much desired as it not only allows trying different optical designs in a prototyping phase, but it also allows an active adjustment of the pitch depending on the actual beam. Especially when stacking beams in fast axis, the beam size per bar in some distance of the laser bar is highly dependent on the ‘smile’ of each laser bar. Dilas has developed a new method of spatial beam combining that overcomes both of these disadvantages. By adding a second reflection to the beam path, one degree of freedom is gained. The previously described method with one mirror and mechanical steps allows controlling the far field pointing in the horizontal and vertical axis and the lateral position of the beam in the horizontal axis. The lateral position of the beam in vertical axis and thus the pitch is fixed by the mechanical step size. Dilas’ new ‘rotating prism’ technology redirects the beam into the vertical axis prior to reflecting it by 90 degrees in the horizontal axis. In addition to a redirection of 90 degrees, this leads to a rotation of each beam by 90 degrees and beams are no longer stacked on top of each other vertically but arranged next to each other horizontally (Figure 3). As a result, all laser bars can be arranged on a flat mechanical plate which allows for even cooling, simplified cooling channels and easy polishing of the plate for low surface roughness if required. In addition to these mechanical benefits, the stacking pitch can now be adjusted by a lateral movement of the ‘rotating prism’ in the stacking direction (propagation direction of the incident beam). At the same time the lateral position of the beam in the vertical axis can be controlled by a left and right movement with respect to the incident beam. Pivoting the prism around two axes allows for a pointing correction of the beam both in the horizontal and vertical direction. These four degrees of freedom, compared to only three for the ‘steps and mirror’ method, allow stacking all beams accurately with full control over the lateral position and pointing of
each beam. As a result, beams can be stacked more densely with the same power throughput compared to previously used methods and thus the beam quality of the laser bar can be better maintained through the process of spatial beam combining.
‘Rotated’ beam Incident beam
Figure 3: New method for spatial beam combining by “rotating prism”. Single prism and beam path (left); three bars spatially combined by rotating prisms (right) 2.3 Optical design Any fiber coupled diode laser module depends on the brightness of the laser bars that are used. While this brightness defines the limitations of what can be achieved, the goal for the optical concept is to maintain this brightness in the best possible way through the optical train and all the way into the fiber. The laser bar used is a 19 emitter bar with a fill factor of 20%. Each emitter is 100um wide and the center to center spacing between emitters is 500um. The fast axis divergence was measured as 46 degrees (95% power included) and the slow axis divergence is 7 degrees. Both, fast and slow axes are collimated using micro lenses with focal lengths of 600um and 4.2mm respectively. After slow axis collimation, a monolithic polarization beam combiner is used to reduce the beam width for each bar from 10mm to 5mm. Each twelve bars are then optically stacked in fast axis direction by use of the ‘rotating prisms’ described above. Two modules of twelve laser bars each are arranged next to each other. The beams from both 12-bar modules are expanded in slow axis to achieve a more symmetric beam and are spatially combined in fast axis direction. This results in a combined beam of 24 bars in fast axis direction, and half a bar (or 9.5 emitters) in slow axis direction. This combined beam is then focused into the fiber. When fiber coupling high power diode laser bars, the beam quality in slow axis is typically the limiting factor. For smaller fiber core sizes (i.e. 100um and 200um) various beam shaping methods exist to equalize the beam parameter product (BPP) in both orthogonal axes. However, when coupling a laser bar with the above described properties into a 400um core fiber with 0.2NA, it is sufficient to use polarization beam combining to effectively reduce the near field beam width of the bar from 10mm to 5mm, without affecting the far field divergence. In order to maximize the power that can be coupled into such fiber, multiple bars are stacked in fast axis direction until the BPP of the combined beam matches the BPP of fiber.
θ slow =
d emitter , slow f SAC
=
50µm ≈ 12mrad 4.2mm
(1)
The beam width in slow axis after micro optics and polarization beam combiner is given by the pitch of the slow axis lens array, times the number of elements, divided by 2 when taking the polarization beam combiner into account.
d slow ≈ 2.5mm
(2)
This leads to a beam parameter product in slow axis of
BPPslow = θ slow ⋅ d slow = 30mm ⋅ mrad .
(3)
In fast axis similar calculations cannot be performed as easily as ‘smile’ has a big influence on the beam quality. The FAC lenses are specified to delivery >85% of the optical power within 3mrad (full angle). Results have shown that with some selection of lenses and well controlled smile 4mrad full width 95% percent power included can be achieved consistently. While the beam height at its waist in fast axis is only about 650um (full width), beams are optically stacked some distance away from the FAC lens and in order to keep the efficiency high an average pitch of 0.9mm seems adequate. These numbers yield a beam parameter product in fast axis for n stacked bars of
BPPfast = n ⋅ 2mrad ⋅ 0.45mm = n ⋅ 0.9mm ⋅ mrad .
(4)
A ‘diagonal’ beam parameter product to combine fast and slow axis and to compare with the rotationally symmetric BPP of the fiber is defined by [5] as BPPtotal = BPPfast + BPPslow. With a BPP for the fiber of
BPPfiber = 0.2 NA ⋅ 0.2mm ≈ 40mm ⋅ mrad the maximum number of bars that can be
coupled into the fiber can be estimated to
BPPfiber = BPPfast + BPPslow = n ⋅ 0.9mm ⋅ mrad + 30mrad ⇒ n ≈ 11
(5)
However, this calculation assumes an even intensity distribution within a rectangular shape for both the near field and the far field distribution of each laser bar. This is not true and corners contain less power compared to the center. Based on previous experience and simulation results a total number of bars of 24 is selected. While this is more than twice the amount than estimated above, this does not mean that the expected power loss is >50%. The combined beam parameter product of fast and slow axis for 24 bars is
BPPtotal , 24 = 24 ⋅ 0.9mm ⋅ mrad + 30mm ⋅ mrad = 51.6mm ⋅ mrad .
(6)
This is about 25% higher than the acceptance of the fiber when assuming even intensity distributions. Optical simulations show that the power loss for a ‘real’ beam with its underlying intensity distribution is still within acceptable bounds. However, the system design is not ‘loss-less’ at this point and losses are accepted to achieve a higher brightness at the penalty of efficiency.
In order to allow the use of spherical lenses for focusing, a telescope is used in slow axis. As shown by [5] there is an optimum magnification for this telescope in situations like this that minimizes the total BPP. In order to minimize the total BPP the ratios of beam diameter and divergence in fast and slow axis respectively have to be equal.
d fast d slow With d slow,magnified = M ⋅ d slow and magnification.
Θ slow,magnified =
=
Θ fast Θ slow
(7)
Θ slow equation 7 can be used to calculate the optimum M
d fast ⋅ Θ slow
M=
=
d slow ⋅ Θ fast
10.8mm ⋅ 12mrad = 5.09 2.5mm ⋅ 2mrad
(8)
A computer optimized pair of spherical lenses is used to couple the now almost symmetric beam into the fiber. A water cooled aperture in front of the focusing optics ensures that no light with an NA greater than 0.2 can reach the fiber.
3. RESULTS 3.1 1KW unit based on single diode wavelength Lasers with different configurations have been built. While the first prototype was based on a 400um 0.2NA fiber, modules with 800um core fibers and 0.12 NA optics have been built for use with industrial laser processing heads. Either an Optoskand QBH or Q5 (compatible to LLK-B) fiber connector may be chosen. Both configurations have been tested. Figure 4 shows a maximum power of about 1,250W and an electro optical efficiency of about 40% at the highest power level tested. The peak wavelength was measured at 976.0nm with a spectral width of 2.8nm FWHM (4.1nm 90% power included). This result makes this laser not only useful in materials processing application but qualifies as a pump source for solid state and fiber lasers. With dimensions of 250 x 230 x 86 mm3 the laser head can easily be integrated into a 19” rack-mountable box like the Dilas Compact Series. So far the weight has not been optimized. While the laser is not particularly heavy for a laser of this power level, large areas of material can be eliminated without affecting stability. Alternative material choices could also be used to reduce weight.
Efficiency [%]
Power [W] 1400 Power 1200
Intensity 45
1
40
0.9
35
0.8
Efficiency
1000
30
800
25
600
20
0.7 0.6 0.5
15
400
10 200
5
0
0 0
10
20
30
40
50
60
70
Current [A]
80
0.4 0.3 0.2 0.1 0 960
965
970
975
980
985
990
Wavelength [nm]
Figure 4: Results of single wavelength module. Power and efficiency (left); wavelength (right)
3.2 2KW unit based on wavelength multiplexing A prototype unit has been built that uses two modified 1KW modules that operate at different wavelengths. For the first unit, diode wavelengths of 915nm and 976nm have been selected. A water cooled QBH fiber was used, which is ideal for high power applications due to its internal water cooling. A maximum of about 2200W of optical power was achieved out of the fiber. The electro-optical efficiency at the highest operating power was above 40%. With two wavelength peaks which are separated by about 60nm, this unit is not designed as a pump source but as an industrial tool for materials processing. With the ability to add more wavelengths, this laser type can be scaled up in power to achieve multiple kilo-watts of optical power within a beam parameter product of 40 mm·mrad.
2500
Efficiency [%]
Power [W]
45
Power
40
Efficiency
2000
35 30
1500
25 20
1000
15 10
500
5 0
0 0
10
20
30
40
50
60
70
Current [A]
Figure 5: Results of dual-wavelength module. Power and efficiency (left); picture of first prototype (right)
4. SUMMARY Challenges in designing a high power, high brightness fiber coupled laser have been discussed. A new semi-active heatsink that does not rely on micro channels but delivers efficient cooling at a small footprint has been shown. A big challenge in designing a high brightness system is to maintain the brightness of the laser bars through the optical train. Often beam quality gets lost when optically stacking multiple laser bars. A new technology has been presented that allows for an active alignment of the stacking pitch for each element. This not only makes the system extremely flexible in a prototype phase, but also allows maximizing brightness while monitoring efficiency. The optical design has been described and trade-offs between highest efficiency and highest brightness have been shown. Results have been shown for different laser heads that have been built. While all lasers of this series include temperature sensors, fiber interlocks, and pilot beams, different options are available for the fiber connector type and NA. Options with Optskand QBH and Q5 (compatible to Trumpf LLK-B) fiber connectors are available. Fibers with either 400um core and an NA of 0.2 or 800um core and incoupling optics for 0.12NA can be used. For a single diode wavelength, about 1,250W have been measured at the fiber output. The wavelength was measured at 976.0nm with a spectral width of 2.8nm FWHM (4.1nm 90% power included) and electro optical efficiency of >40% which makes this unit an ideal pump source for solid state or fiber lasers. Other wavelengths are available upon request. A prototype unit with two diode wavelengths at 915nm and 976nm has been built based on modified versions of the 24bar module described above. 2200W at the fiber output have been achieved at 40% electro optical efficiency. Moving forward it is possible to combine more than 2 wavelengths to increase the total output power to multiple kW while maintaining a beam parameter product of 40 mm·mrad.
REFERENCES [1] H. Schlüter, C. Tillkorn, U. Bonna, G. Charache, J. Hostetler, T. Li, C. Miester, R. Roff, T. Vethake, C. Schnitzler; “Dense Spatial Multiplexing Enables High Brightness Multi-kW Diode Laser Systems”; Proc. SPIE Vol. 6104,(2006). [2] S. Venkat; “Welding with diode lasers”; www.the fabricator.com,(October 14, 2008). [3] K. Howard, S. Lawson, Y. Zhou; “Welding Aluminum Sheet Using a High-Power Diode Laser”; Welding Journal, (May 2006); pp 101-106.
[4] W. Horn; “High Power Diode Lasers for Industrial Applications”; ICALEO,(2007). [5] Z. Wang, A. Segref, T. Koenning, R. Pandey; “Fiber coupled Diode Laser Beam Parameter Product Calculation and Rules for optimized Design”; SPIE LASE 7918-13,(2011).
Copyright 2011 Society of Photo-Optical Instrumentation Engineers. Speaker: Tobias P. Koenning, DILAS Diode Laser, Inc. Paper Title: Macro-Channel Cooled, High Power, Fiber Coupled Diode Lasers Exceeding 1.2kW of Output Power Session 3: Mounting and Cooling High Power Laser Diodes Paper: 7918-13 of Conference 7918 Date: Sunday, 23 January 2011 Time: 3:30 PM - 5:10 PM Copyright 2011, Society of Photo-Optical Instrumentation Engineers. One print or electronic copy may be made for personal use only. Systematic reproduction and distribution, duplication of any material in this paper for a fee or for commercial purposes, or modification of the content of the paper are prohibited.
