Fiber-coupling technique for high-power diode laser arrays

Fiber-coupling technique for high-power diode laser arrays H.-G. Treusch, Keming Du, M. Baumann, V. Sturm, B. Ehiers, P. Loosen Fraunhofer Institut für Lasertechnik, SteinbachstraBe 15, 52074 Aachen, Germany ABSTRAT Monolithic linear arrays of diode lasers, also known as diode laser bars, are the basic elements for most high-power laser applications such as solid-state laser pumping or material processing. ylindrical microlenses used as fast-axis collimators for 10-mm diode bars require very high angles of aperture (up to 100 degree FWI/e2, equivalent to a numerical aperture of approx. 0.8) to capture most of the emitted laser power. For the efficient longitudinal pumping of laser rods, or the narrow focusing of the diode laser radiation (fiber coupling, material processing), high-quality microlenses with small lens aberrations are necessary to avoid power losses and beam quality degradation A technique for coupling the output of high-power diode laser bars into one multimode fiber with high efficiency, easy alignment requirements and low manufacturing costs is demonstrated using a single fiber with core diameter down to 400 jim. This technique comprises two micro step-mirrors for beam shaping. The overall efficiency from one diode-laser bar to fiber is 71% with 20 W cw laser power through the fiber. oupling of 12 diode laser bars and power of 200 W out ofa fiber with core diameter of 0.8 mm and NA 0.2 is achievable with this technique. Keywords: Diode laser arrays, laser beam shaping, optical system design, fiber-optic components, micro-optics, aspherical cylinder lenses, 1. INTRODUTION At present, diode laser bars which continuously emit 10-40 W of optical power at near infrared wavelengths are the basic elements for high-power applications of semiconductor diode lasers such as solid-state laser pumping or material processing.'2 The diode lasers operate reliable at these power levels. By increasing the diode current, record values of up to 200 w cw output power of a single diode laser bar have been demonstrated in the laboratory, but only for short operation times.3'4 A single diode bar is a monolithical linear array typically consisting of 20-50 subarrays each containing of 10-20 single emitters (Fig. 1). An alternative configuration is the linear arrangement of 20-50 broad-area emitters active of 50-200 t.tm width. layer The number of emitters and the subarray of 10... 20 spaces between the emitters destripe emitters or pend on the thermal load mansingle broad area emitter. 0.1 agement and can be optimized for a variety of operating requirements (cw or in the case of pulsed operation: peak power, pulse width, duty factor). Although typ. dimensions diode lasers exhibit a high laser of gain-guided efficiency (up to 30-40 % for stripe emitters diode bars available on the mar- = 10 pm 1 pm ket), the removal of the waste =3 5 pm heat through high-performance micro-coolers is a major task Fig. 1: Typical geometry of high power diode laser arrays since it has to be done across the small diode bar area and only 98 SPIE Vol. 3267 0277-786X1981$1o.oo

small junction temperature rises are acceptable.5 The optical cavity of O.6 mm length is formed by cleaving two opposite facets of the semiconductor wafer and using them as laser mirrors. Typical material systems are AlGaAs (e.g. 808-nm pump wavelength of Nd:YAG lasers) and InGaAs (e.g. 940-990 nm pump wavelengths of Er and Yb doped lasers). Due to the tim-sized, non-circular cross-section of the active region, the output radiation of diode lasers quickly diverges in the direction which is perpendicular to the plane of the pn junction ("fast axis") and slowly diverges in the parallel direction ("slow axis'). The beam quality M2 is approx. I in fast-axis direction, and M2 1700 in slow-axis direction. The very high fast-axis divergence 2ê of up to 100 (FW1/e2), equivalent to a numerical aperture NA = sines = 0.8, makes it necessary in nearly all applications (with the exception of some side-pumping configurations for laser slabs or rods) that the first beam shaping has to be the fast-axis collimation. Since lens aberrations increase with the square of the NA, the optical performance of the fast-axis collimator is of crucial importance for the available beam quality of the diode laser assembly, and hence, for example, determines the minimum focal spot diameter (i.e., the maximum laser irradiance) at the workpiece or the minimum diameter of coupled fibers. Different approaches have been proposed to shape the diode laser emission to obtain a circular focus spot. In a straight forward approach the emission from diode laser bars are coupled into a plurality of fibers.6'7 These fibers are arranged in such a way as to generate a circular fiber bundle at the output end. A beam-shaping technique using two diffractive elements is demonstrated.8 The emission of each emitting facets (sub-beam) is incident on a micro prism.9"0 The emerging beam consists of sub-beams which are rotated 90 about their axis of propagation. Two plane parallel mirrors are used to reshape the emission from diode laser bars". However, most of these techniques have drawbacks such as large size, low flexibility, low efficiency or they are difficult to manufacture. Here, we demonstrate a technique using two micro step-minors to rearrange the emission from diode laser bars and to couple it to an optical fiber. 2. BEAM PROPAGATION OF DIODE LASER ARRAYS Detailed information about the beam propagation of diode laser arrays are required to optimize the design of beam shaping devices. The width of the collimated beam in fast-axis is of uppermost importance (discussed in chapter 5) to reach a filling factor close to 1 behind most shaping devices. The width is determined by the divergence angle in fast axis and the focal length of the collimating lens. In Fig. 2a) and b) the intensity distributions in fast axis of two different diode laser arrays are shown. Intensity is measured with a pin diode while rotating the diode laser array. The influence of the driving current could be neglected in both cases, but the low divergence of the Tutcore array offers the advantages of collimating with simple optics and reduced beam width at constant focal length. Siemens G5 1,0 0,8 Tutcore Nr. 2482 l=20a 1= 40 A Gaussian fit 0,6.. U) 0,4 04 Angle (deg] 0,2 3 j HW89: 0,0 ' ' ø/,.. I,,., -100-50 0 50 100 Angle (deg] Fig. 2: Intensity distribution infast axisfor two chfferent diode laser arrays close and above threshold current. a) left: Siemens G5, 20 stripes array, 200 pm wide and 400 pm pitch, 2 = 808 nm, cavity length 0. 6 mm b) right: Tutcore (1996) broad area array, 150 pm wide, 500 pm pitch, 2 =808 nm, cavity length ca. 1 mm The Gaussian fit ofthe data in Fig. 2 claims a beam quality M2 close to 1 for both arrays in the direction ofthe fast axisṫhe emission characteristic in slow axis is dependent on the structure of the emitting area (broad area or multiple stripe) and the length of the cavity. The intensity distribution in the direction of the slow axis, presented in Fig. 3, is detected by it: 23.0 99

the same procedure as in Fig. 2. The distance between diode laser array and pin diode was 2.6 m. The divergence in slow axis increases with the driving current. The beam quality reduces with increasing output power. This effect has to be considered in the design of the beam shaping optics to prevent significant losses in fiber coupling. Again the small angle of divergence of the Tutcore array offers some more advantages in coupling to smaller fiber core diameter (discussed in chapter 5) Siemens G5 Tutcore 1,0 0,8 0,6 U, a) 0.4 0,2 0.0-10 -8-6 -4-2 0 2 4 6 8 10 Angle [deg] Angle [degi Fig. 3: Intensity distribution in slow axis at different driving current, same diode arrays as in Fig. 2. The area enclosed by the curves are kept constant. 3. FAST AXIS KOLLIMATION Low-power diode lasers used in telecommunication and, for example compact disc players, are single emitters or small arrays of emitters. Microlens collimators available for such diode lasers are not suitable for diode bars since the lenses are mostly rotationally symmetric and their NA (typically < 0.5) is not sufficient. ylindrical collimating mirrors, or diffractive elements, as well as combined elements can also be used in specific applications and geometries but in this paper we focus on microlenses. 1.3 mm. t (1.2 mm) diode laser 2A/itd fast axis tx t1 (0.2 (0.88 mm) r (0.68 mm) Fig. 4: Asperical cylinder micro/ens: up) Aspherical contour and some ray paths. Pt and P2 are the paraxial principal planes of the thick lens, f is the paraxial focal length and r is the radius of curvature of the paraxial sphere. left) REM photograph of the macro/ens manufactured by ultraprecision grinding of Schott glas!.ifn21. 100

Fig. 4 illustrates in detail the geometry of the cylindrical microlens, which is used for collimation. The cross-section depicts the piano-concave lens with an aspherical contour in real proportion to the diode bar. The paths of some rays are drawn according to geometrical optics (with an angle cz exaggerated for clarity). In a first approximation, the lens can be described by its paraxial parameters, which are valid for small angles in the vicinity of the optical axis. The distance of the first principal plane to the entrance face is tin, where t is the center thickness of the lens and n the refractive index. The second principle plane is located at the lens exit because of the plane entrance surface. The radius of curvature of the paraxial sphere is r = (n- 1) f =(n-1) (t1 + t/n). Ideal collimation means in geometrical optics that a =0 stands for a non-extended source. In reality, diffraction limits the minimum divergence of the collimated beam. This is depicted in Fig. 4a by the Gaussian beam with waist d and beam divergence 9B In addition to inherent aberrations in a specific lens design, there are aberrations due to manufacturing irregularities such as deviations ofthe actual lens parameters to the rated values. These are, for example, the surface contour, the thickness and the position of the cylinder axis of the lens. Small surface irregularities and lens material inhomogenities cause scattering and an overall decrease ofthe power transmission. Measuring the quality of a microlens is much more complicated than with normal lenses because of the small dimensions and alignment tolerances, the very high numerical aperture, the cylindrical geometry and aspherical surfaces. Frequently, the first suggestion is to measure the focal spot line width when focusing a parallel beam (e.g. HeNe laser) with the microlens. However, the focal spot width is in the order of the diode beam waist of 1 tm, i.e. less than the pixel dimensions of D cameras. To magnify the focal spot for better measurement requires a cylindrical objective with at least the same NA than that of the microlens (up to NA = 0.8). Unfortunately, such an objective is not available. Another approach is to use a diode bar and measure the quality of the collimated beam in fast-axis direction. Since the beam quality M2 of the diode bar in this direction is close to 1,the beam degradation is predominantly caused by the microlens. At least, it is a relative measurement if the same diode bar is used for a test series. 1.0 diode laser divergence 5O x 5 HW1/e2 I I I I I 1.0 T T I I diode laser divergence 33 x 2.4 HW1/e2 i (mrad) (mrad) Fig. 5: Fast-axis profile ofthe collimated diode laser beam measured with a scanning mirror setup at a distance 2. 4 m awayfrom the aspherical micro/ens. The oscilloscope traces ofthe photodiode signal are measuredfor the same lens but with different diode bars. Therefore, we first measured the fast-axis profile of the collimated beam with a scanning mirror setup. The scanner deflects the beam across a slit aperture of a photo detector. Two examples of the photo detector signals are shown in Fig. 5 for the same aspherical microlens measured with the two different diode bars form Fig. 2. From the measured sweep time (FW1/e2), the profile width d(s) is calculated by a calibration factor, and by dividing d(s) by the distance s, the FWI/e2 divergence 202 is determined. In comparison to a D camera this setup has no offset problems to determine the correct ground level. The determination of the beam quality is completed by measuring the beam waist by the knife-edge method close to the microlens. The presented aspherical microlens collimates the fast axis of diode laser arrays up to NA of 0.8 with no significant losses in beam quality. 101

4. BEAM-SHAPING TEHNIQUE Many applications of diode lasers require a beam delivery by fiber optics. Different techniques are available for coupling the multiple arrays of a diode laser bar into a fiber bundle or into a single fiber. Most of them are aiming at maximum beam quality and are based on the physically arrangement of the arrays and so far limited to small filling factors and the size of the arrays. Beam shaping techniques handling the emission of a diode bar as a line source are not depending on this limitations and can use high power diode laser bars with filling factors of up to 90%. One approach and a really simple one is the usage of micro step-mirrors to homogenize the beam quality of one or multiple diode lasers, which is the basic requirement for coupling into a single fiber. The set-up for the transformation of the diode laser radiation consists of two identical micro step-mirrors. This device consists of N highly-reflective surfaces which are arranged in a special manner as depicted in Fig. 6. From which it can be seen that each single mirror surface of the first micro step-mirror is tilted 45 about the slow axis and separated from the neighboring surface by a constant distance d along the axis of propagation. This distance d corresponds to the width of a mirror surface and the width of the collimated beam in the direction of the fast axis. Thus, the centers of the mirror surfaces include an angle of 45 with the axis of propagation. The collimated beam is incident on the first j J micro step mirror Here it is cut into N sub I / f beams along the slow axis and reflected into the fast-axis direction. Each sub-beam which is reflected from the first step-mirror is incident on one surface of the second step-mirror. The surfaces of the second step-mirror are arranged in such a way as to reflect the sub-beams into the slow-axis direction., By means of this deflection, performed by the system of micro step-mirrors, a rearrangement of the diode laser radiation is performed. Before the rearrangement, the sub-beams are aligned next to one another along the slow axis. The beam-quality factor is approximately jt,l2 1300-1700. After the rearrangement by the Fig. 6: Schematic view of the micro step-mirrors first step-mirror. the sub-beams are grouped in a stair-shaped geometry. Finally the outgoing beams are arranged in a line again but now along the fast axis. As described above, each sub-beam exhibits a beam-quality factor in the fast axis of A102 1-3. Neglecting minor effects of diffraction, M of the sum of the sub-beams is M12 N M10 (1) However, this only holds if the fill-factor of the sub-beams after the transformation is 1, which is achieved by adapting the width of the collimated beam in fast axis direction as close as possible to the width of the steps.. Otherwise the resulting beam quality will be further decreased. On the opposite side, the beam quality along the slow axis improves by the factor of N while M correspondingly decreases by N: M 2 M= N (2) Again, diffraction effects caused by the clipping of the beam into sub-beams by the mirror surfaces are neglected. Now, N has to be chosen as a number which will render M2 and M/ similar. This case is the best approximation of a homogeneous distribution of the beam quality with respect to the axis of propagation. The number of steps N included in one micro stepmirror then is given by M02.i I (3) 102

5. OUPLING OF A SINGLE DIODE LASER BAR To demonstrate the beam-shaping technique a pair of step mirrors are manufactured out of copper by diamond cutting. The microlens discussed in chapter 2, which was originally designed for stack applications, is used for collimating the emission in the direction of the fast axis. The throughput of the microlens amounts to 96 /o with a broad-band AR-coating. Due to width of the collimated beam of about 1.1 mm for Siemens G5 bars and 0.8 mm for the Tutcore bars the width of the stepsis choosen to 1 mm. Fig. 7 illustrates the set-up. The Tutcore bar used in this set-up is a laser bar with a total width of 10 mm and an output power of 28 W at an injection current of 40 A. Efficient cooling is achieved through the use of a microchan- nel-watercooling device, de- Micro f = 1 50 mm signed and manufactured by the Step-Mirrors Fraunhofer ILT. The power of the collimated beam, at an in- - - jection current of 40 A, is 27 W. As illustrated above, the shaping 400 JJIT1 of the diode laser emission is Micro Fiber accomplished with two micro Lens step-mirrors. This device is f = 40mm placed directly in front of the Diode Bar microlens. Each micro stepon Heat Sink mirror consists of 13 steps while only 11 steps are used to shape the beam After the transformation, the relation of the beam- Fig. 7: Optical set-up for the fiber-coupling technique quality factors in fast and slow axis is reduced from 600 to 3. To obtain a small spot the 11 stacked beams in fast axis are focused with an achromatic lens doublet off 40 mm. In this case the NA amounts 0.14. In the direction of the slow axis the illuminated steps are projected by a cylindrical lens off 150 rmn into the focal plain of the achromatic lens. Due to the inclination of the step-mirrors the cylindrical lens is also inclined by 450 about the direction of beam propagation to get a clear image of each step in the focal plain. This inclination reduces the focal length of the cylindrical lens close to 120 mm. Therefore the magnification is 0.3 and the NA in slow axis equals 0.21. Due to the rectangular geometry of the beam the maximum NA equals 0.25. The power distribution in the focal plain of the achromatic lens is shown in Fig.8. The spot measures 160 I.Lm x 340 tm encompassing 86% of the total power. According to Eq. (2) the beam-quality factor in the slow axis should evaluate to 120. In the fast axis a fill-factor of 0.8 after the transformation has to be taken into account. The beam-quality factor, stated in Eq. (1), then amounts to Mj' 41 starting with a beam quality of M, 3 in the unshaped line. There is a good conformity of the >. beam-quality factor in the slow and fast axis between the calculated values and the experimental results of M52 138 and A1/ 43. X Position [pm] Fig. 8: Intensity profile of the focused Fiber coupling experiments are conducted with 600 l.tm and 400 im core diameter, AR-coated at 810 rim. The Numerical Aperture of both fibers is 0.22. Fig. 9 shows the power transmitted through the fiber, measured with a thermal absorber. The coupling efficiency of a 600 j.tm fiber is close to that achieved with a 400 i.tm fiber. It can be seen that the coupling efficiency decreases with increasing injection current and decreasing to a higher extend when using the 400 im fiber. Experiments on the 103

radiation characteristics of the diode laser bar (chapter 2) showed that the fast-axis divergence is not altered by variation of the injection current within a range where the slope efficiency of the diode is constant, i.e. the range of operation in practice. However, the slow-axis divergence increases with increasing injection current. Although this effect is not drastic, it leads to an increase of the NA of the focused beam which is at the limit of the fiber. Tab. 1 : Transfer efficiency ofthe optical setup. I diode micro- step-mirror fiber laser lens + I optics optic [pwer(w) 28 26.9 21.3 20.1 trans. effic. (%) 100 96 79 94 The transfer efficiency quoted in Tab. 1 is strongly influenced by the imperfection of the step-mirrors especially the absorption losses of 10% in total. 0 10 20 30 40 It can be assumed that a further optimization of the micro..... Injection urrent [A] step-mirrors according to Eq.3 will yield a square-shaped spot. This means that for the Tutcore bars 22 steps should be included in the micro step-mirrors at a width of the steps of 0.5 Fig. 9:Power through different fiber core diameters mm. Furthermore, the width of the collimated beam in fast axis has to be adapted to 0.5 mm by reducing the focal lenth of the microlens from 0.88 down to 0.6 mm. Using the same diode laser bar, the same throughput of 20 W with a single diode laser bar should be possible with a 200 j.tm fiber. 6. OUPLING OF MULTIPLE DIODE BARS To increase the output power compared to a single diode bar, the emission of multiple diode laser bars has to be cornbined. Using polarization or wavelength coupling the beam quality is not influenced while the power is linear increased with the number of diode bars. The emission of diode lasers is polarized up to 95% normally. The direction of polarization is parallel to the slow axis for the Siemens G5 and parallel to the fast axis for the Tutcore bars and could be rotated by axj2- wave plate. Therefore the emission of two similar diode lasers can be overlapped by a polarization beam splitter (PBS) as depicted in Fig. IOa. The coupling efficiency amounts 95%. If the emitted wavelength of the diode laser is not important for PBS A A/2 [ LD2 LD 3 LD1 LD1 a) b) c) 900 Wavelength. [nm] Fig. 10: a) polarization coupling b) wavelength coupling c) transmission ofwo different edgefiltersfor s-polarization (filter 1) and s+p-polarization (filter 2), the absorbtivity ofbothfilters is neglectable therefore the reflectivity R equals J-T Bothfilters are used to combine the emission of 800 nm, 930 nm and 970 nm diode lasers. 104

the application i.e. for materials processing like welding of plastics, soldering, cutting etc. different wavelength could be combined by dielectric coated edge filters (see Fig lob). Due to the steep change from 93% transmissivity to 99% reflectivity of those filters within 40 nm (see filter 1 in Fig loc) up to 6 different wavelength can be combined within the conimon range for high power diode lasers from 800 to 1000 nm. The total losses for such a setup will not exceed more than 6%. step-rn irrors PBS cylindrical lens m icro lens laser-diode with cooler aspherical lens Fig.]]: 40 Wpower at 808 nm out of a 0.6 mm fiber by two laser diode bars coupled by polarization ompared to Fig. 7 an improved optical setup for fiber coupling with respect to its compactness and output power is shown in Fig. 1 1. Two Siemens G5 diode bars on micro channel coolers are arranged as depicted in Fig. loa in front of the step mirrors (step width 1 mm). After the beam shaping the beam is compressed by two prisms in fast axis to save some of the distance between the optics and their size. The prisms also set upright the steps with respect to the direction of beam propagation. Therefore the beam incidence on the cylindrical lens (f = 40mm) is normal. Within this setup the steps are projected with a magnification of 0.5 using the cylindrical lens and an aspherical lens (f = 20 mm) on to a fiber with a core diameter of 0.6 mm and a NA of 0.22. The same lens is used to focus the beam in fast axis. This results to a rectangular power distribution on the fiber entrance with spot dimension of 0,2 mm in fast axis and 0,5 mm in slow axis. Again more the 70% of coupling efficiency is achieved at a power level of 40 W out of the fiber. fibre To double the output power a second optical setup as shown in Fig 1 1 is arranged to the first one in a way that the beams are stacked up in the direction of the fast axis (see Fig. 12). The focal length of both lenses is also doubled (f= 80 and 40 mm) to keep the magnification and the NA constant. Then the spot size at the fiber is increased in fast axis from 0.2 to 0.4 mm, while the size in slow axis remains constant at 0.5 mm. Due to the dimension of the spot diagonal of 0.64 mm the core diameter of the fiber is also increased to 0.8 mm. laser-diodes 1 and 2 laser-diodes 3 and 4 prisms for beam-corn pression focusing lens Fig. 12: Optical setupfor a 80W diode laser using 4 diode bars. fibre-plug 105

oupling of three diode bars with different wavelength is also proven and leads to an output power of 60 W. The combination of all three coupling methods like wavelength multiplexing, polarization coupling and doubling can lead to a setup with 12 diode bars. First wavelength multiplexing with stacked diode bars is performed. In two of the four stacks half-wave-plates for each wavelength are introduced to rotate the direction of polarization. A polarization beam splitter combines the emission of two stacks. The setup depicted in Fig. 13 is doubled and arranged like the setup in Fig. 12. Maximum power of such a diode laser system will reach values above 200 w out ofa fiber. A/2 plate PBS Fig. 13: Arrangement of six diode laser bars using wavelength multplexing and polarization coupling to increase the output power A beam-shaping technique is demonstrated which yields highly efficient equalization of the beam-quality factor. With this set-up, 71% of the power of the diode laser bar is transmitted through a 400 j.tm fiber. An even higher efficiency should be obtained through further optimization. The whole set-up is compact and comprises only a few optical components. The alignment of the micro step-mirrors is easy to perform. Results in wavelength multiplexing and polarization coupling give rise to the assumption that an increased output power up to 1 kw is possible. This will open the competition between diode and solid state lasers in the field ofmaterials processing. This work was supported by BMBF/VDI 7. ONLUSION 8. AKNOWLEDGMENTS REFERENES 1. W. Koechner, Solid-State Laser Engineering (Springer-Verlag, Berlin 1992). 2. P. Loosen, "Advanced concepts of using diode lasers in material processing," Proc. of SPIE Vol. 3097 (1997) 3. P. Loosen, J. Biesenbach et al., "Design and industrial applications of high-power diode-lasers," in XI. mt. Symp. on Gas Flow and hem. Lasers and High-Power Laser onf, H.J. Baker, Editor, SPIE 3092, 17-20 (1997). 4. M. Sakamoto, J.G. Endriz, D.R. Scifres, Electron. Lett. 28, 197 (1992). 5. T. Ebert, J. Biesenbach, et al., "Optimisation of micro channel heat sinks for high power diode lasers in copper technology," Proc. of SPIE Vol. 3097 (1997) 6. H. Zbinden, and J. E. Balmer, Opt. Lett. 15, 1014 (1990) 7. Th. Graf, and J. E. Balmer, Opt. Lett. 18, 1317 (1993) 8. R. J. Leger, and W.. Goltsos, IEEE J. Quantum Electronics 28, 1088-1100 (1992) 9. P. Albers, H. J. Heimbeck, and E. Langenbach, SPIE Proc. 1700, (1992) 10. S. Yamaguchi, T. Kobayashi, Y. Saito, and K. hiba, Opt. Lea. 20, 898 (1995) 11. W. A. larkson, and D.. Hanna, Opt. Lett. 21, 375 (1996) 106