1. Introduction

Diode lasers have many distinctive advantages such as high electrical-optical conversion efficiency, low cost, and wide commercial availability. Consequently, it has wide range of applications including optical pumping of solid state lasers, optical communication, medical, and military. To achieve Watt-level diode radiation with good brightness, various designs incorporating elements such as ridge diode laser and tapered amplifier have been proposed [1–4]. However, these configurations generally require specialized designs and fabrication techniques. For more common and low cost diode solution, one may choose to work with broad area laser diode which are widely available in many different wavelengths. For example in high power solid state laser, one or several broad area diodes are used to pump Cr: LiSAF, Cr: LiCAF, Yb: KYW and Yb: KGW [5–9], producing several watts of femtosecond pulses. A typical single broad area emitter of 1 by 100 µm active area can produce 2 W of diode radiation with M² of 1 by 15. To achieve tens of Watts-level, such emitters are arranged side by side in a bar to give an array configuration. To manage the thermal load in the single bar, typical pitch spacing is around 500 µm, which can easily result in beam quality degradation of M² > 1000 in the slow axis. To retain the beam quality while increasing the output power, the diode array can be coherently combined in an external cavity.

Coherent beam combining of diode array has been studied for many decades. The phase locking methods can be in active [10–12] or passive mechanisms (for example, evanescent/leak wave coupling [13], Talbot cavity [14–19], and coherent polarization locking [20]). Active coherent combining requires external control to maintain precise optical path difference (better than λ/10) of the diode elements. The implementation often involves complex electronic controls. On the other hand, passive locking is relatively simpler and requires much less control over individual emitters. Recently, high power multi-watt combining of diode array have been demonstrated by using Talbot cavity [15,16]. Talbot cavity is based on Talbot imaging condition in which the output coupler provides diffraction feedback that couples the emitters’ phases in the array. Although the emitters are all phase-related, the beams are not spatially overlapped at the output coupler which resulted in multi-lobes spatial profile in the far field. Such multi-lobes profile may not be ideal for many diode applications where a single tight focusing spot is required.

An alternative method to achieve single-lobed output is by polarization superposition. In diode pumped solid state laser, the phase locking mechanisms have been demonstrated by active [21] and passive [22] controls, with both achieved near-perfect combining efficiency. The basic principle of passive coherent polarization locking is coherent superposition of several beams to achieve polarization with minimum loss in the cavity. Due to the competitive nature in light amplification, the most favorable mode will dominate and suppress other modes in the cavity.

Recently, we extended the idea of polarization locking to the broad area laser diode array by using specially cut birefringent wedges and cascades of walk-off crystals as the compact solution for combining diode emitters [20]. The external cavity is designed for Gaussian mode with maximum achievable power limited to 1 W. In this paper, in order to power scale the coherent locking output to multi-Watts laser, we propose two key changes in the external cavity; an imaging cavity configuration that utilizes retro-reflection feedback in the fast axis and novel design of the spatial combiner that consists only of walk-off crystals and waveplates. These changes improve the efficiency and stability of coherent locking.

2. Power scaling issues

There are several conditions which need to be fulfilled for power scaling of coherent polarization locking output. First, mode matching at the diode facets is very important for optimum gain extraction. Broad area laser emits multimode beam in the slow axis due to the coupling factor and thermal variation across the emitter [23,24]. We use an imaging cavity that provides a resonator for the multimode beam profiles [24]. The cavity consists of lenses that form an image plane at the output coupler (OC).

In fast axis direction, although the beam is nearly diffraction limited, proper mode matching is often impeded by the smile effect. Smile effect is the mechanical induced “smile” pattern that causes the emitters to be located at different heights. It can lead to significant optical loss at the diode entrant surface. The maximum smile displacement in the experiment is 1 µm. To mitigate the issue, we implement a retro-reflection feedback in the fast axis. As illustrated in Fig. 1 , the retro-feedback together with the beam collimation setup is able to reflect incoming lights back to its sources regardless of the heights from which the beams are emitted. The retro-reflector configuration consists of a focusing lens and a plane mirror located at a focal distance. Similar design has also been demonstrated in spectral combining experiment [25].

 

Fig. 1 Illustration of smile effect compensation by using retro-reflection feedback.

Download Full Size | PDF

Lastly, another important consideration is to ensure a stable coherent locking. Apart from the intrinsic stability of individual emitters, the locking stability is determined by the cavity’s capability in finding the common longitudinal modes for coherent polarization locking, similar to the Vernier effect [26]. The number of common modes can be increased by introducing relative path differences among the emitters’ sub cavity. In our experiment, we inserted the path differences by configuration of walk-off crystals and half waveplates (HWP) whilst combining the emissions into a single beam. The schematic design of the walk-off crystals and waveplates are shown in Fig. 2 . We use 4.95 mm YVO₄ crystals with optic axis cut at 45° to obtain 500 µm walk-off displacements. It can be traced from the figure the relative path lengths of L_i (i = 1, 2, 4) with respect to L₃ (the emitter with shortest cavity length) are given by Eq. (1)-(3).

 

Fig. 2 Schematic of the spatial combination of emissions from four emitters by cascades of walk-off crystals and half waveplates. The beams are shown in their state of polarizations with ray tracings from individual emitters.

Download Full Size | PDF

n is the HWP refractive index (1.54), t is the thickness of HWP (1 mm), n_e is the extraordinary refractive index at 45° (2.06), n_o is the ordinary refractive index (1.96), and d is the thickness of the crystal (4.95 mm).

3. Experiment

A 980 nm diode bar from Jenoptik GmbH is used for the experiment, with 19 emitters per bar at a 500 µm pitch. Each broad area emitter is 1 µm by 100 µm in surface area, with average M² of 1.5 by 13. The front surface of the emitters are anti-reflection (AR) coated with residual reflectivity ~0.5%. The emitters are collimated by fast axis collimator (FAC) with focal length of f_FAC = 0.9 mm and slow axis collimator (SAC) with focal length of f_SAC = 1.81 mm.

The experimental setup is shown in Fig. 3 . The cavity consists of the diode bar, series of polarization optics, two cylindrical lenses in the slow axis, one cylindrical lens in the fast axis, and the OC with transmission of 80%. The polarization components first spatially combine the beams from four emitters by using cascades of YVO₄ crystals and HWPs. The beam is then focused by a 100-mm cylindrical lens (slow axis) to pass through a slit that ensures single beam output. The polarization of the combined beam will be in 45° plane-polarized if all the emitters were coherently locked. Then, a HWP with optic axis at 22.5° is placed, so that the beam will transmit through polarizing beam splitter (PBS) without any polarization loss. Finally, the beam is focused by 100-mm lens (slow) and 50-mm lens (fast), to form the image plane at the OC and the aforementioned retro-reflector configuration. Note that the dimension of the external cavity largely depends on the selection of the cylindrical lenses. Therefore, using lenses with shorter focal lengths will help to minimize the footprint of the system.

 

Fig. 3 Diagram of the experimental setup as viewed from the top. The lines represent the ray tracing of the laser.

Download Full Size | PDF

4. Results and discussion

Prior to the external cavity, the bare laser array pumped at 56 A produced power of 4.3 W. By using coherent polarization locking in the external cavity configuration, we achieved maximum power of 7.2 W with M²_fast = 1.5 and M²_slow = 11.5. The beam quality at the fast axis direction is slightly worse than diffraction limited beam due to the misalignment of the collimators. In the slow axis, the M² is close to that of a single broad area emitter. Fig. 4 (a) and Fig. 4 (b) show the caustic measurement and the far field beam profile in the slow axis. Without spatial combining, emission from four emitters has M²_slow = 224 and powers obtained independently from each emitters are 2.8 W, 2.5 W, 2.2 W, and 2.8 W. Thus, coherent polarization locking has improved the beam quality by more than 20 times and brightness scaling over an order of magnitude. Perfect power efficiency is inhibited mainly by the imbalance ratio of emitter’s gain and wavefront mismatch among the sub-cavity.

 

Fig. 4 Beam measurement at maximum output power 7.2 W. (a) Caustic measurement in the slow axis with ω_o = 227 μm and M² = 11.5. (b) Far field pattern of the output beam in the slow axis at the minimum waist. (c) Output spectrum. The dashed line denotes the modulation envelopes.

Download Full Size | PDF

The spectrum of the coherent output is shown in Fig. 4 (c). It consists of two big envelopes that contain 1 nm periodic spectral lines. The two envelopes are mostly due to the inherent profile of the diode laser, which exhibit similar two local spectral peaks. The 1 nm periodic lines are caused by the relative path differences among the emitters when traversing through the walk-off crystals and HWPs. As discussed in the previous section, these path differences are inserted to obtain stable coherent locking. We observe fluctuation of less than 1.5% at the maximum power.

One of the interesting physical observations from the experiment is the feasibility in coherent locking of multimode beams. Efficient combining of multimode spatial profiles have been demonstrated previously for diode pumped solid state laser system [27]. Our result shows that the same phenomenon can be extended for diode lasers. We explain the combining mechanism by simultaneous superposition of several higher transverse modes of the same order to achieve minimum polarization loss in the cavity.

To further increase the output power, coherent polarization locking scheme can be extended to combine more than four emitters. From the design of the spatial combiner in Fig. 2, it can be inferred that by adding more walk-off crystals and waveplates in the cavity, the number of emitters can be scaled to 2^N. However, combining large number of emitters would require better uniformity in each diode lasers. The locking scheme requires the gain of all the emitters to be the same in both the magnitude and the transverse distribution. Imbalance gain ratio or different sets of transverse modes will induce polarization mismatch that reduces the combining efficiency. Thus, for larger number of emitters, the condition will be more stringent. Alternatively, one can increase the total power by improving the gain of individual emitter. Recent product from Lumics GmbH [28] demonstrates power loading of up to 8.5 W in a single emitter by expanding the gain volume of a broad area laser. By locking several of these emitters, tens of Watts of coherent power would be achievable in near future.

Lastly, the beam quality of the combined beam can be further improved in the slow axis. Mode selective technique such as off-axis alignment [29] in principle can be implemented in the coherent locking cavity, for example, after the beams are spatially combined. This technique might increase the brightness significantly.

5. Conclusion

In summary, we demonstrated multi-Watts output from coherent polarization locking of diode lasers in a single-lobed beam. Power scaling is achieved by designing a resonator with good mode matching and high stability. We obtained maximum power of 7.2 W from a 980 nm broad area laser diode array, with beam quality M² of 1.5 by 11.5. Further power scaling would be possible by increasing the number of emitters and power scaling of individual emitter.