1. Introduction

High power laser diode arrays (LDAs) find extensive applications in solid-state or fibre laser pumping as well as for direct material processing. LDAs clearly surpass other sources in their high efficiency, compactness, and long lifetime. However, poor beam quality and spectral properties restrict the range of possible applications. In recent years there have been many techniques developed to address these problems for both spatial and spectral brightness improvement for LDAs in bars and stacks, including spatial multiplexing, phase-locking, line narrowing and spectral beam combining. Fundamentally, the radiance (commonly referred to as brightness) achievable by any beam-combined diode laser bar or stack is always limited by the beam quality of its single element. Currently, single-lateral-mode operation devices are the most attractive sources for beam combining in high-brightness systems, notably those with narrow-stripe emitters of 3-6μm lateral width or tapered emitters, because of their high intrinsic radiance. However, two major limitations of diode laser bars usually constrain the performance of techniques for brightness improvement. Firstly, fast-axis collimation (FAC) errors caused by smile deformation in bonding of the chip, by lens imperfections and by lens misalignment, produce pointing errors and beam deformation. In the slow-axis direction, light spilling on to the adjacent slow-axis collimation (SAC) lenses can lead to a significant power and beam quality loss, particularly for small emitter pitch.

In previous work [1], we introduced an efficient technique for correction of fast-axis collimation and beam pointing errors for laser diode bars and stacks. Based on an accurate wavefront measurement of the beam at full working current, our approach is to design and laser-cut a corrective phase-plate to cancel errors. At the cost of additional measurements, this method is flexible in that it provides effective correction of both lens errors and arbitrary smile shape. It shows a clear advantage over other techniques for smile correction, such as the use of a magnifying telescope [2], a tilted cylindrical lens [3] or a custom lenslet array [4]. Recently, we reported the use of such custom corrective phase-plates in a 1.8kW 20-bar diode laser system [5]. The technique was also applied to provide fast-axis correction on diffraction-coupling of a 10 element tapered emitter bar with 100μm pitch [6].

In this paper, we report extension of the custom phase-plate technique to provide a single component that incorporates both fast-axis correction and a high accuracy of slow-axis collimation. Such a dual-axis phase-plate is demonstrated with a single mode diode array with 49 emitters on 200μm pitch. Our objective is to provide a laser system where the smile and collimation optics no longer constrain the performance of beam-combining methods, particularly such as phase-locking as in [7,8], but also some of the spectral beam-combining methods [2,9]. Fast-axis correction is particularly important for external cavity laser diode arrays, where the smile-induced pointing errors prevent efficient and uniform feedback along a bar [10]. In the slow-axis, accurate and efficient collimation of small pitch single-mode-emitter bars is a particularly difficult, and the use of an expensive custom focal length lens array is necessary to allow optimization of the beam fill-factor and to avoid drastic power loss in the optical system.

For a collimated bar with separate FAC lens and SAC lens array already attached by the manufacturer, the application of the fast-axis correction technique is less effective because of the increased distance at which the correction plate must be positioned. There is also a difficulty in measuring smile and FAC wave-front errors when viewing them through the distortions of imperfect SAC lenses. Here, we show that by combining the SAC lens array in the fast-axis correction plate, optimized slow-axis collimation and ideal positioning of FAC correction is achieved. This provides a very flexible method for producing an array of high quality, parallel beams with high accuracy collimation in both directions. The technique provides an excellent source for a range of beam-combining experiments.

2. Laser-cut phase-plate technique for fast-axis correction and slow-axis collimation

The laser-written phase-plate technique employs a laser system to cut [11] and smooth [12] a custom refractive surface in silica glass. The smile-induced pointing and collimation errors along the bar are measured accurately using a wave-front sensing procedure developed in-house, based on the Hartmann principle. The compensating refractive optical surface for the fast-axis is micro-machined in silica, typically with a surface depth modulation of 10μm peak-to-peak. Here, the laser cutting technique is used to combine the corrective surface prescription with an array of 900μm focal length cylindrical lenses on 200μm pitch to provide slow-axis collimation for the 49-element single mode emitter bar. The focal length of the lens is determined by the minimum optical distance for the positioning of the correction phase-plate after the FAC lens; thus the effect of the overlapping of the adjacent beams is minimized. On a pitch of 200μm, the single mode beams begin to overlap at a reduced distance of about 700μm from the laser facet. However, the fast-axis collimator and its mounting tab constrain the position of the correction plane to be at a reduced distance of 900μm from the facet (for the physical distances in Fig. 6 ).

 

Fig. 6 Laser diode bar aligned with a slow-axis collimating lens array of bigger pitch.

Download Full Size | PDF

For our experiments, we designed and produced two phase-plates. The first, in Fig. 1(a) , provides only fast-axis correction, similarly to the plate reported in [1]. The second in Fig. 1(b) combines the fast-axis correction with the 200μm pitch lens array. Figure 1(c) shows a section of the lens array scanned with a high accuracy profilometer, with the shape compared with the design profile (dotted line). The major challenge in producing the 200μm pitch micro-lens array is to avoid light scatter at the joints and to minimize the inter-lens dead-space. Typically, the laser cutting/smoothing technique produces rounding of the desired sharp structures at the joint. However, our technique has been developed to minimize this problem, and an active lens width of about 170μm was obtained for each lens with a maximum figure error of about λ/4 (peak-to-valley) and local (AFM measured) roughness below 10nm. The tolerance in the focal length obtained was below 2%. These satisfactory results provide a flexible technique to produce the lens arrays that are not easily available from commercial providers.

 

Fig. 1 Contour map of (a) the corrective phase-plate and (b) of the lens array measured with a profilometer; (c) Cross-section of the lens array taken at the centre of map (b).

Download Full Size | PDF

3. Optical testing with a 49-single mode emitter bar

In our experiments we used a 980-nm array of 49-single-mode 6μm-ridge emitters, manufactured by Bookham (now Oclaro) and similar to the array described in [13]. The bar produces 30W at 40A with a threshold at 1.6A, 55% conversion efficiency and 0.8 W/A of slope efficiency. Manufacturer’s data gives FWHM divergences of 22° and 7° in the fast- and slow-axis, respectively. We performed accurate fast-axis lensing with a 0.60mm focal length FAC from Ingeneric GmbH, and measured far-field divergence of 3.7 mrad (defined as width between 5 and 95% of total power). The far-field measurement setup consisted of a 1m focal length spherical lens capturing the full beam, projecting the far-field pattern on a rotating screen placed in the focal plane of the lens. The choice of field of view of CCD camera and use of Spiricon software’s Ultracal feature ensured the capture of weak tails in the laser beam. For the emitter-resolved far-field pattern of the FAC-lensed bar, shown in Fig. 2(a) and 2(b), a 200mm FL cylindrical lens was placed (temporarily) between the 1m lens and the screen to create the slow-axis image. Smile results in a pointing error of 6 mrad peak-to-valley (as in Fig. 4 ) equivalent to 3.8μm at the bar, widening the fast-axis profile for the full bar. The profile of each of the emitters is also affected by lens errors such as roll, defocus and intrinsic aberrations. Fast-axis wavefront data measured in a plane approximately 0.3mm beyond the apex of the FAC lenses was used to calculate the surface depth required for the correction surface are shown in Fig. 1(a). After the fast-axis correction was applied, both the smile and collimation errors are seen to be successfully corrected in Fig. 2(b). The power in the bucket (PITB) graphs in Fig. 2(c) are calculated from the far-field patterns for the full bar measured in the setup described above and show that correction narrows the beam, transferring power in the tails into the main beam. The far-field divergence is reduced to 2.47 mrad (5-95%). These values correspond to the beam propagation factor M²~1.3, assuming a beam width of 600μm leaving the FAC.

 

Fig. 2 Emitter resolved fast-axis far field pattern of (a) the FAC only bar and (b) the corrected bar. (c) Power in the bucket curves comparing the FAC-only and corrected cases.

Download Full Size | PDF

 

Fig. 4 Fast-axis pointing error along the bar -uncorrected (green points) and corrected (blue points). For comparison the corrected fast-axis divergence is 2.47 mR (5-95%).

Download Full Size | PDF

In Fig. 3 , the slow-axis far-field pattern and power in the bucket graph are presented for the case when the lateral position of the phase-plate is adjusted for minimum width (17mrad, 5-95%) In this case, the slight lens aberration and truncation produces an asymmetric side-lobe in a far-field. Based on the manufacturer’s data, which gives a 208mrad (4σ) un-collimated divergence in the slow-axis, we estimate that 92% of the power is within the 170μm active width of each lens. The fraction of power sent through the lens joints is redistributed into small angles and contributes to the tails that are evident in Fig. 3(a). The absolute power loss from the phase-plate, related to the random scatter outside the measured range, is below 1%.

 

Fig. 3 (a) Far field beam profile of the bar corrected in fast-axis and collimated in both directions; (b) Power in the bucket (PTIB) in the slow-axis direction.

Download Full Size | PDF

3.1 Fast-axis pointing error

In order to evaluate the accuracy of fast-axis correction and slow-axis collimation, we measured the pointing direction for each individual emitter in the bar. To select a single emitter beam at a time, we used a pair of angled mirrors close to the phase-plate to form a ~200μm slit of width, mounted on a micrometer stage, to deflect all other beams into dumps. The selector allowed the pointing direction of each beam to be obtained from the centroid of its position on the rotating screen. Figure 4 presents the pointing direction of each emitter beam for the uncorrected (green triangles) and corrected (blue squares) cases. The smile-affected uncorrected beams exhibit a non-uniform pointing distribution matching Fig. 2(b); by contrast, for the phase-plate-corrected bar all beams are within an RMS error of 3% of the 5-95% far-field divergence of the bar, independently of drive current.

A key result is that the process of smile compensation is unaffected by the addition of the slow-axis collimation functionality to the phase-plate, a very satisfactory outcome showing that the two processes are unambiguously additive.

3.2 Slow-axis pointing error

A similar measurement of the slow-axis beam pointing of selected emitters along the bar was recorded at two drive current levels: 15A and 40A. The two sets of measured values follow the sloping linear fits, as shown in Fig. 5 .

 

Fig. 5 Slow-axis pointing error along the bar at 40A (red) and at 15A (blue). For comparison the slow-axis far-field divergence was 21.4 mrad (5-95%) in this experiment.

Download Full Size | PDF

Such a distribution of pointing direction along the bar indicates that the beams are ‘diverging’: for the vertically mounted bar, the top emitters are pointing upwards and the bottom emitters are pointing downwards, as in Fig. 6. A likely reason for this is a displacement of the beams with respect to the slow-axis collimating lenses. The bottom emitters are above the centre of the corresponding lenses while the top emitters are below the centre of the lens. An explanation may be the occurrence of a slight pitch mismatch between the emitters and the SAC array. The ~2mrad angle of the end emitters corresponds to 1.8μm off-axis displacement of the SAC lens. Such a displacement corresponds to a pitch mismatch of 0.035% of the period, and whilst it is in the correct sense for the chip bonding strain it is perhaps too large to be explained in this way. There may also be a contribution from encoder calibration error in the laser-machining XY table.

The experiment revealed that the slope depends on the temperature of the chip. A temperature change of 18°C was estimated between drive current values of 5A and 40A, based on the thermal tuning of the measured laser emission wavelength. Such a temperature change would give a linear expansion of 0.01% for free GaAs, and somewhat more with the chip bonded to the Cu passive cooler block. The slope change in Fig. 5 corresponds to about 0.015% linear expansion of the chip. Hence the dynamic part of the pointing variation can be explained by thermal expansion of the semiconductor material. These interesting observations are made possible because we use single mode emitters in our experiments. For broad-area emitter bars, the longer focal length of the SAC array, and the larger M² of the individual emitter beams will tend to mask the effects.

Although the change of pointing direction along the bar was detected, the performance of the slow-axis collimating component is still very good. The RMS pointing error along the bar remains very low at a value of 6% of the far field divergence when the pitch mismatch effect is discounted. Otherwise the pointing directions of all the emitters are enclosed within 10% of far-field divergence. It should be noted that any fixed pitch error is correctable by the use of a phase-plate with a suitably modified design.

4. Conclusion

We have demonstrated the design, fabrication and use of a single optical component to perform the dual function of fast-axis correction and slow-axis collimation for a full-length diode laser bar of 49-single mode emitters. We have overcome stringent requirements on the collimating optics imposed by the 200μm emitter pitch. Good pointing accuracy has been achieved in both directions, giving an extremely promising LDA source for further developments in brightness improvement techniques, by spatial multiplexing, external cavity feedback methods, or high resolution wavelength combining.