1. Introduction

Coherent beam combining (CBC) is an attractive way to achieve high output power in a single diffraction limited beam. Individual laser emitters are typically limited in output power due to thermal effects which degrade the laser efficiency, distort the wavefront, and limit the beam quality. Limits on individual emitter performance can be lifted by coherently combining several emitters. There are two flavors of CBC corresponding to either (a) tiled aperture or (b) filled aperture CBC. In a tiled aperture configuration, low fill factor results in sidelobes in the far field [1]. In a filled aperture, the individual emitters are combined to form a single output beam ideally.

Here we report on filled aperture CBC that combines 21 individual laser emitters with active phasing. The beam combining system uses an intracavity diffractive optical element (DOE) to combine the laser array into a single output [2]. CBC may be implemented in a master oscillator–power amplifier configuration (MOPA) in which the gain elements are seeded by a master oscillator and thereby act as amplifiers [3–5] or in a laser resonator configuration in which no seeding element is used [2,6–9]. CBC in an amplifier array configuration has been implemented with success in a variety of gain media [3,4,10]. Recently, CBC of 218 amplifier elements has been demonstrated [3].

CBC in a laser resonator configuration on the other hand has achieved limited success with regards to scaling. In particular, as the number of elements have increased, the trend has been for the beam combining efficiency to decrease. Both experimental data and models support this trend [9,11,12]. But in these experiments, no attempt is made to control the relative phases of the elements (passive phasing) [13,14]. Here we demonstrate experimentally that the scaling limits for passive phasing do not apply to an active CBC laser array system.

CBC of individual laser emitters offers the promises of not requiring a master oscillator and improving electrical to optical efficiency. Toward this end, we have developed an active CBC system which is described below.

2. Experimental Configuration

Our beam combining experimental arrangement is illustrated in Fig. 1. A 21-element semiconductor laser array is focused using a $f=500\mathrm{mm}$ focal length lens onto a DOE. The spacing between array elements is 200 $\text{μ}\text{m}$. The lens is chosen to match the angular separation between the diffracted orders of the DOE ($\mathrm{\Delta }\theta$) to the pitch of the array as in $\mathrm{\Delta }\theta =200\mathrm{\mu m}/f$.

 

Fig. 1. Coherent beam combining experimental configuration. Twenty-one element array is incident on a DOE. A diffraction grating provides feedback and wavelength selectivity.

Download Full Size | PDF

The 21 laser elements in the semiconductor array consist of microlensed, slab-coupled waveguide amplifiers (SCOWA). These lasers have a back-facet HR coating of 95% and a front-facet AR coating. The modal field diameter at the microlens is approximately 100 $\text{μ}\text{m}$, resulting in a 50% fill factor. The SCOWAs have been described elsewhere [3,15,16]. Here we highlight some of the key features which make SCOWAs attractive for high-power beam combining.

SCOWAs allow for a large mode ($\sim 6$ $\text{μ}\text{m}$) while maintaining single-mode operation. The higher-order modes in a ridge-guided structure are selectively coupled into a slab by design. Only the fundamental mode does not undergo coupling and therefore experiences the least loss. The large mode increases the catastrophic optical damage threshold by decreasing the intensity in the guide and at the facet. The SCOWA gain elements used in our experiments have a peak gain at approximately $\sim 943\mathrm{nm}$ and are similar to those previously described [15].

The output power of the SCOWA gain elements within an external cavity is largely dependent on the choice of the output coupler used in the experiments. The cavity illustrated in Fig. 1 is designed to provide diagnostic capabilities to characterize the beam combining efficiency and is not optimized for output power. The cavity may be optimized for output power by removing the diagnostic ports and by appropriate choice of the output coupler.

Figure 2 illustrates the total output power from all the ports for a single laser emitter (central element in the array) aligned through the spatial filter with the DOE removed from the cavity. The threshold current of the central emitter is $I_{\text{th}}\sim 143\mathrm{mA}$ with external feedback and is representative of the various emitters within the array. The intrinsic threshold of an individual laser emitter is $I_{\text{th}}\sim 300\mathrm{mA}$ resulting from the finite reflectivity from the front-facet AR coating.

 

Fig. 2. Single gain element output power in external cavity at 960 nm. Output power is the combined total from all ports.

Download Full Size | PDF

The DOE shown in Fig. 1 serves to create 21 diffraction orders with a single-beam input (beam splitting) or the inverse operation of beam combining. The angular separation between orders is 400 μrads. The DOE is used as both a combiner and a splitter in our cavity arrangement in order to achieve a filled aperture output. The DOE is 90% efficient when used as a splitter into 21 beams and so it is critical to prevent the inevitable higher diffracted orders from feeding back into the cavity. The remaining 10% of the power is diffracted into higher DOE orders and is scattered.

To suppress the undesired orders a spatial filter is used at the output of the DOE. The spatial filter consists of a vertical slit inserted into a telescope configuration. The focal lengths of the lenses ($f_1=300\mathrm{mm}$, $f_2=75\mathrm{mm}$) are chosen for compactness and for a practical spot size at the 100-μm-wide slit. For our chosen focal length of $f_1$, the zero-order beam from the DOE has a 60 micron diameter at the slit. The $f_2$ lens results in approximately a 1.25 mm beam diameter at the grating. A diffraction grating is used to create feedback into the cavity and is operated in the Littrow configuration with a 76% diffraction efficiency.

The laser cavity also includes a $50/50$ beam-splitter that serves as an output coupler and provides two output ports for diagnostics. Referring to Fig. 1, the diffracted orders exiting the DOE are incident onto the $50/50$ beam-splitter. The incident light from the laser array is reflected off the $50/50$ beam-splitter and is sent to far-field diagnostics. This reflected port is labeled $P_2$ in Fig. 1. The far-field diagnostics include a camera to image the far field of the DOE. In this arrangement, the camera captures all of the orders exiting the DOE.

The other beam-splitter port (labeled $P_1$ in Fig. 1) consists of a power monitor that measures only the power in the DOE’s zero-order that undergoes feedback from the diffraction grating. All other orders are blocked by the spatial filter.

The combining efficiency is defined as the ratio of the power in the zero-order output of the DOE to the total power. In our cavity, we obtain a measurement proportional to the zero-order output through port $P_1$ and a measurement proportional to the total power through port $P_2$. By measuring the appropriately weighted ratio of the power at the two beam-splitter ports we obtain a measure of the combining efficiency ($\eta =k{P}_1/P_2$), where $k$ is a calibration factor to be determined. Since the combining efficiency of a single beam is 100% with the DOE removed, $k$ may be measured directly by taking the ratio of the powers on the beam-splitter ports $P_2$ and $P_1$ with a single beam turned on ($k=P_2/P_1$). This technique calibrates for the beam-splitter reflectivity, grating diffraction efficiency, and the transmission through the spatial filter.

Two types of experiments were performed: passive and active phasing. For the passive-phasing measurements, the SCOWA array elements are biased equally at 400 mA and the output of the DOE is monitored on the far-field camera.

For our active phasing experiments we have applied SPGD which we have succesfully used in other CBC demonstrations [3,4]. The SPGD algorithm is a hill-climbing algorithm in which phase dithers are used to determine the slope and correction necessary to optimize the cost function (in our case the zero-order output of the DOE) [17]. As shown in Fig. 1 our SPGD detector measures the zero-order output from the DOE. The SPGD signal on the detector is illustrated in Fig. 3 when SPGD is turned on. The convergence time is typically $\sim 4\mathrm{ms}$ for a 6 kHz dither.

 

Fig. 3. Stochastic parallel gradient descent (SPGD) detector signal. SPGD detector measures the zero-order power incident from the DOE. The convergence time is typically 4 ms. The SPGD loop is turned on at $t=0$.

Download Full Size | PDF

In our implementation, the dither phases were applied by adding perturbative currents to the SCOWA gain elements. The dither currents change the phase by $\sim 1/100$th of a wave. The phase change is primarily due to thermal heating in the lasers with applied current. Characterization of the gain material phase change with current has been described elsewhere [15].

The diffraction grating allows for selecting the lasing wavelength. For our passive- and active-phasing measurements described below, the lasing wavelength was set to $\sim 960\mathrm{nm}$ and the average bias current on the gain elements was 400 mA.

3. Results

In the passive CBC case the power distribution among the various DOE diffracted orders is illustrated in Fig. 4 as observed on the far-field camera. The combining efficiency in the passive case is estimated to be $\sim 5\%$ by measuring the ratio of the power in the zero-order to the total power. This result is qualitatively consistent with incoherent combination where the power splitting among the various orders of a 21-order DOE would ideally result in ($P_{\mathrm{DOE},n}=1/21P_i$), where $P_{\mathrm{DOE},n}$ is the power in order $n$ and $P_i$ is the incident power. The power variation among the various orders may be partly attributed to the nonuniform amplitude splitting of the DOE.

 

Fig. 4. Passive phasing: far-field image of DOE. The power in the zero-order is $\sim 5\%$ of the total power ($\eta \sim 5\%$). The current per element is set at 400 mA.

Download Full Size | PDF

Figure 5 illustrates the output of the DOE measured in the far field with active phasing. We observe that only the zero-order output of the DOE has any appreciable power and is constant with time once SPGD has converged. The power in the remaining diffracted orders are negligible within the dynamic range of the camera image (8 bits). This illustrates that the laser elements add constructively in phase to form the zero-order output. The remaining orders undergo destructive interference. The beam quality in the combining direction was measured to be $M_{x,\mathrm{CBC}}^2=1.1$ which is commensurate with that of a single emitter. The beam quality in the the noncombining direction was $M_{y,\mathrm{CBC}}^2=1.6$ which is improved over that of a single emitter $M_{y,\mathrm{SE}}^2=1.9$. The improved beam quality arises at the expense of the combining efficiency and is a result of the cancellation of noncommon wavefronts in the combined beam.

 

Fig. 5. Active phasing: far-field image of the DOE capturing all the diffracted orders when SPGD is turned on. 81% of the power is captured in the zero-order beam when SPGD is activated. The remaining 19% of the power is diffracted into higher DOE orders and is scattered. The average current per element is 400 mA.

Download Full Size | PDF

The combining efficiency was estimated to be 81% by measuring the appropriately scaled power ratio of the power at the two ports of the 50% beam-splitter. The upper bound on the efficiency for our experiment is limited by the 90% efficiency of the DOE. An estimate for the remaining error sources include the SPGD dither (less than 1% ), pointing errors (less than 3%), and microlens aberration (less than 1%). The remaining 4% residual error may be largely attributed to amplitude variations and optical alignment tolerances.

In addition to the far-field diagnostics, a near-field spectrometer was used to characterize the relative spectral output of the array. The near-field spectrometer was constructed by using relay optics and a cylindrical lens to create a vertical line image of each element in the array onto a rotating screen as shown in Fig. 6. Each position in the line creates an array of angles which are spatially selected through the etalon. If the array elements are of equal frequency, their output will reside on a pattern consisting of circular rings (similar to the fringe pattern arising from interference between a plane wave and a spherical wave). The separation between the rings in the vertical dimension corresponds to the free-spectral range of the etalon (20 GHz). Consequently, the vertical axis corresponds to the frequency whereas the horizontal axis corresponds to the array element position.

 

Fig. 6. Near-field spectrometer. Each element in the array is imaged onto a line. An etalon is used in a telescope configuration to map frequencies onto circular rings.

Download Full Size | PDF

The output spectrum of the array is illustrated in Fig. 7. All the array elements are aligned to form ring patterns indicating that they are operating at the same frequency. Dynamically, the modes (the rings) are seen to move in unison. Multimode behavior is also sometimes observed but is of no consequence to the coherently combined output. Optical spectrum analyzer (OSA) measurements indicate that the lasing bandwidth may be as large as 0.4 nm or (130 GHz). The multimode bandwidth is limited by the spot size on the grating.

 

Fig. 7. Near-field spectrometer output. Array position is on the horizontal axis and frequency is on the vertical axis. Array elements with equal frequency are mapped onto circular rings (similar to interference pattern between a plane and spherical wave).

Download Full Size | PDF

The spectrometer and 81% measured combining efficiency results demonstrate that active CBC may achieve large efficiency without requiring a very large gain bandwidth. In addition, the results further indicate that single-mode operation is not required as anticipated.

The total output power from the largest output coupling port (labeled $P_2$ in Fig. 1) is 1.2 W at an average bias current of 400 mA per element of which 81% resides in the zero-order. It is also worth noting that once the phases have been established by SPGD, the active phasing may be turned off without compromising the combining efficiency for extended periods. This observed behavior is similar to the self-sustaining nature of mode-locked lasers. We have observed that when the drive currents are held fixed to the values used in optimizing the phase, the array elements continue to coherently combine at the same nominal efficiency with less than a 1% variation (tested for at least 60 min in the laboratory). This shows that once the path lengths are adjusted, active phasing is not required in order to achieve efficient beam combining for the semiconductor array because of the low phase noise of the semiconductor elements [3] and the optical feedback provided by the diffraction grating.

4. Conclusions

We have demonstrated experimentally that we can achieve high combining efficiency in an optical oscillator configuration using active phase control in a CBC system. The beam combining efficiency of 81% is close to the 90% theoretical limit in our system. Further improvements in the beam combining efficiency may be obtained largely by using a higher efficiency DOE and by minimizing the noncommon path aberrations introduced by the microlens array and imaging optics.

We have tested our CBC laser array with both active and passive phasing. We have found the combining efficiency to be negligible for the passive case, in agreement with scaling models for passive CBC. Comparing our active results with our passive results demonstrates that active phasing is not subject to the same scaling limits imposed on a passive-phasing CBC system.