Imaging of transient phenomena with low coherence lasers comprising arrays of independent microbeams: A laser version of Harold Edgerton’s stroboscope

Full-field imaging of transient phenomena, such as the motion of a mosquito or laser ablation of material surfaces, requires bright illumination sources having pulse durations sufficiently short to capture the event of interest without motion blur. Harold Edgerton pioneered short-pulse optical imaging by introducing stroboscopic photography, and his images of hummingbirds (1936) or a bullet exiting an apple (1964), for example, are renown and vividly demonstrate the value of defining the temporal resolution of an image through the pulse width of an incoherent optical source. Since the discovery of the laser in 1960 and the advent of short-pulse generation techniques such as Q-switching and mode-locking, in particular, numerous groups have attempted to extend Edgerton’s imaging concept to shorter temporal scales, while improving the spatial resolution and source brightness, with laser versions of the stroboscope. Virtually all such efforts have been hindered by laser speckle, the granular intensity pattern produced by a laser beam illuminating a surface having roughness on the order of the optical wavelength.^1–3 The interference between mutually coherent wavefronts leaving a reflective or scattering surface produces an intensity pattern that often obscures or completely masks the object of interest. Consequently, although the spatial coherence of conventional lasers is an asset for interferometry and holography, for example, it is often a liability for laser imaging in situations for which spatial resolution is a priority.

Various approaches to suppressing or circumventing speckle for full-field imaging applications have been reported but most have sacrificed output power in order to satisfy the speckle reduction requirement of “diversity in polarization, space, frequency, and time.”² With respect to enhancing time diversity, the insertion of a rotating diffuser,⁴ vibrating mirror,⁵ or other optomechanical device into the laser beam, for example, produces a time-varying speckle pattern.^2–6 Although time-integrating such intensity maps yields the desired reduction in the speckle amplitude, the temporal resolution of the composite image is degraded, but this compromise is often acceptable for display applications, for example. Spectral diversity is readily achieved with ultrashort pulse (i.e., ps/fs) lasers, supercontinuum sources, or through the generation of broadband amplified spontaneous emission (ASE),^7,8 but the desired reduction in temporal coherence often comes at the expense of the spectral power density. Although the latter may be acceptable for select applications, the inherent decline in the spectral power density is ill-suited to those situations in spectroscopy and imaging, for example, in which a spectrally narrow optical field probe is required. The cost and complexity of sub-ps and supercontinuum sources are also barriers to their widespread adoption for full-field imaging. In short, addressing speckle reduction by suppressing temporal coherence has several drawbacks that argue for an alternative approach.

Other efforts to reduce speckle have focused on optical control of the laser’s spatial coherence. Methods involving propagating laser light through multimode fibers or fiber bundles have been demonstrated successfully,^9,10 but more recent efforts have sought a reduction in spatial coherence directly from the laser cavity itself by employing random lasers,^11–13 uncoupled vertical-cavity surface-emitting laser (VCSEL) arrays,¹⁴ semiconductor lasers in chaotic or near-concentric resonator configurations,^15–17 and degenerate optical cavity lasers.^18,19 The principle underlying such lasers is the emission of beams comprising >10³ transverse modes, which, with sufficient angular separation, generate speckle patterns that are uncorrelated on timescales much longer than the inverse of the laser bandwidth, thus reducing speckle.^20,21 However, the output powers currently available with intracavity approaches to speckle suppression are quite limited, generally precluding the possibility of imaging on the nanosecond timescale.

We describe here an optical resonator capable of generating thousands of parallel, mutually incoherent microlaser beams that coalesce in the far field into a single beam having a near-Gaussian transverse intensity distribution. The speckle contrast ratio C observed with this composite laser, defined by the ratio $σ/Ī$⁠, where σ and $Ī$ represent the standard deviation and mean value of the speckle pattern intensity, respectively, is found to scale inversely with the square root of the number (N) of individual microlaser beams in the ensemble.² By stabilizing a Fabry–Pérot optical cavity with an internal array of microlenses and incorporating one of three gain media, a microlaser associated with each lens is produced, thus allowing C to be tuned in a manner independent of the transverse mode of the individual microlasers. Virtually all of the results presented here were obtained with N > 1000 microlasers, most of which are operating in the fundamental transverse mode (TEM₀₀). This, in turn, results in a composite beam with a radial intensity distribution that is approximately Gaussian and a far-field divergence matching that of the individual, nearly diffraction-limited microbeams (as small as ∼5 mrad). Although the resonator design presented here allows for composite laser beams built with microlaser arrays operating on higher transverse Gaussian modes (m, n > 2), the experiments described in the following sections typically employed transverse modes no higher than TEM₁₀ so as to optimize the composite beam brightness for a given value of N and to stabilize beam characteristics in the far field. Because the resonator architecture is compatible with a range of gain media and accommodates output emission spectra ranging from ∼60 GHz to tens of nm in bandwidth, this laser imaging system is capable of probing objects or monitoring transient phenomena with sub-10 ns resolution in several wavelength regions simultaneously for a variety of applications, including microscopy, spectroscopy, or imaging at distances up to 5 m (at present). Employing either solid-state (Ti:sapphire) or liquid (colloidal quantum dots or a dye) gain media, compact oscillators producing single pulse energies of 8 or 11 mJ (respectively) from as many as 4000 microbeams have been designed and tested. This laser architecture simultaneously produces N independent speckle patterns of comparable intensity, thereby suppressing C in an optical cavity of modest dimensions (<2 cm in length and 5–8 mm in diameter) and preserving the capability of the gain medium to deliver pulsed (or CW) powers at least two orders of magnitude higher than those available with previous lasers of low spatial coherence. The unique characteristics of this oscillator have permitted the imaging of, for example, a turbomolecular pump rotor operating at 56 000 rpm and the movement and internal structure of Chlamydomonas reinhardtii (in vivo) with spatial resolution as small as 100 and 1.5 μm, respectively. Each image was recorded with a single laser pulse, and none exhibit motion blur or speckle. Imaging of a target through fog is also demonstrated for distances up to 5 m.

Figure 1(a) illustrates the resonator design, which is based on the introduction of an array of microlenses into the optical cavity. Recalling that a Fabry–Pérot cavity is critically stable, one is able to stabilize the resonator by placing a lens of the proper focal length at an appropriate location within the resonator.²² Consequently, the addition of a gain medium and pump radiation results in lasing along a line parallel to the axis of the optical cavity and passing through the lens. Inserting a planar array of such microlenses in the cavity, oriented so as to lie transverse to the resonator’s optical axis, yields a microlaser array of arbitrary size for which each microlaser is associated with a specific microlens. Elsewhere in this transverse plane, diffraction losses are prohibitive and lasing does not occur. In effect, the laser array pixelates the resonator cross section with a spatial resolution given by the pitch (center-to-center separation) of the lenses in the array. Each of the microlasers is uncorrelated with its neighbors with respect to phase. In the far field, therefore, these mutually incoherent beams merge to form a composite beam of low spatial coherence. For the experiments to be described in the discussion to follow, the monolithic microlens array was fabricated from fused silica and the diameter, pitch, and nominal focal length of the plano-convex lenses are 146 μm, 150 μm, and 5.2 mm, respectively, unless otherwise noted. With this architecture, composite beams comprising N > 4000 microbeams are generated simultaneously with microlens arrays of modest dimensions. The diameter of the largest composite beam exiting the resonator was 8.6 mm, and the measured mean far-field divergence of the beams characterized to date is ∼5 mrad, matching the theoretical value for the individual microbeams to within the experimental error.

Several gain media have been tested extensively, including Ti:sapphire (Ti:Al₂O₃), colloidal quantum dots, and the near-infrared dye LDS 798 (Styryl 11). Optical pumping was provided by ∼8 ns, 532 nm pulses from a frequency-doubled Nd:YAG laser, but all experiments and calculations indicate that this resonator is also suitable for CW operation. Figure 1(b) is a schematic diagram of the optical testing arrangement adopted for imaging experiments. Camera placement was designed to allow for imaging objects with transillumination and for those situations in which backscattered light is of interest. Provision was also made for viewing the interior of the laser resonator and imaging biological media with a custom microscope. Further details can be found in the section titled “Materials and Methods”.

A series of near-field optical micrographs of the individual microbeams produced with two gain media is presented in Fig. 2. Figure 2(a) is a false-color image of a section of an array of Ti:Al₂O₃ microlasers, acquired in plan view with a 4× objective and a CMOS camera. In recording this image, the optical length of the cavity (mirror separation L) was set at 5.2 mm so as to display a few of the higher-order transverse modes that are readily produced, despite the relatively low stimulated emission cross section for Ti:Al₂O₃ and the thickness of the crystalline disk (500 μm). Most of the modes are immediately recognizable as Hermite–Gaussian, Laguerre–Gaussian, or Ince–Gaussian patterns with mode indices m,n < 3. Others have not been identified but appear to be coherent superpositions of two or more eigenmodes. If one now increases the resonator length to 8.0 mm, only the TEM₀₀ mode is observed [Fig. 2(b)] because the diffraction losses suffered by modes of higher order are prohibitive and the microlenses act as spatial filters. Differences in the output of individual microlasers in Fig. 2(b) are the result of a nonuniform pump intensity distribution in a plane transverse to the optical axis of the oscillator, thereby preventing a number of oscillators from reaching threshold. We also wish to point out that the observed beam spot sizes are in accord with the ABCD matrix analysis of the resonator that has been described previously.²² Calculations of the complex beam parameter q(z), where z is the coordinate associated with the optical axis of the cavity, find that the theoretical spot size for the resonator configuration of Fig. 2(b) (L = 8.0 mm, distance from lenses to output mirror = 3.7 mm), measured at the output mirror, is 42 μm whereas that for Fig. 2(b) is ∼40 μm.

A panoramic plan view of a 38 × 24 segment of an N ∼ 2000, Ti:Al₂O₃ microlaser array, given in panel (c) of Fig. 2, shows clearly the Fresnel diffraction rings of the injected 532 nm pump radiation. Note that only TEM₀₀ and TEM₁₀ modes are generated because L was reduced to ∼6.5 mm. Multiple variations of the resonator structure have been tested, and as one example, Fig. 2(d) is another false-color image of the interior of the resonator when Ti:sapphire again serves as the gain medium, but in this case, the microlenses are 300 μm squares and the pitch of this array was also increased to 300 μm. Similar images were acquired for other gain media in the resonator, and Fig. 2(e) is a color photograph of a portion of an array of red lasers (λ_max ∼ 657 nm) observed with CdSe/ZnS colloidal quantum dots and L ∼ 4 mm. The short cavity length and larger gain coefficient for the quantum dots enable transverse modes with m, n up to 3 to be supported. Most of the microlaser beam cross sections are distinctly square, owing to the gain provided by the quantum dots but also the use of a different microlens array (100 μm lens pitch) comprising square lenses.

For optical imaging applications, operating the microlasers in the fundamental mode is generally preferred because divergence is minimal and the far-field intensity distribution is stable because transverse mode-hopping is readily eliminated. Figure 3(a) is a far-field image in false color of the composite beam produced by the diffraction of ∼1000 Gaussian microlasers emitted from a Ti:Al₂O₃ disk. The transverse intensity distribution of the beam is found to be almost precisely Gaussian as confirmed by the inset to Fig. 3(a), which compares an arbitrary, experimental intensity lineout (blue) with a fitted Gaussian (red). A slight flattening of one side of the far-field intensity map of Fig. 3(a) (owing to pump beam inhomogeneities and the spatial extent of the microlens array) is responsible for the only significant departure of experiment from theory. The observation of a Gaussian profile for the composite beam cross section is consistent with the absence of sidebands or oscillatory structure in the experimental intensity map and the presumption that the individual microlaser beams are mutually incoherent. It should be noted that sidebands do appear with higher-gain media and strong pumping (i.e., approaching the saturation fluence), which suggests the onset of coherent coupling between the microlasers. In such cases, however, the central lobe of the intensity distribution exhibits low coherence, regardless of the pump fluence, which indicates that the degree of inter-beam coupling is small. Sidebands of significant magnitude were observed only when the output coupling mirror was replaced by a volume Bragg grating (VBG) having an angle of <1° between the plane of the grating and the substrate surface. Doing so resulted in inter-beam coupling primarily along one axis of the array, which is attributed to Fresnel reflections from the tilted substrate. The results to follow were acquired with a standard output coupling mirror as opposed to the VBG, unless otherwise noted.

Laser spectra for two of the gain media explored in these experiments are shown in Fig. 3(b). The bandwidths (FWHM) of the normalized spectra are 18 and 27 nm for LDS 798 dye (in glycerol) and Ti:Al₂O₃, respectively, and the laser pulse waveforms [Fig. 3(c)] for these gain media resemble that for the pump with pulse widths of 8–9 ns. Panel (d) of Fig. 3 illustrates the dependence of the composite Ti:Al₂O₃ laser pulse energy (E_o) on the 532 nm pump pulse energy (E_p) for E_p < 100 mJ and a 3 mm thick crystal. Threshold occurs for E_p ∼ 25 mJ for N ∼ 1000, and the optical-to-optical conversion and slope efficiencies for this gain medium/resonator combination are measured to be 8.0% and 12.3%, respectively, yielding single pulse energies up to 8 mJ for the highest 532 nm pumping fluences available (∼300 mJ/cm²). No indication of saturation was observed for the Ti:sapphire experiments because the saturation fluence for this doped crystal is known to be of the order of 600 mJ/cm². Similar experiments with LDS 798 dye resulted in pulse energies up to 11.6 mJ, and the optical-to-optical and slope efficiencies were both determined to be ∼10%.

Owing to the bandwidth of the Ti:Al₂O₃ oscillator of Fig. 3(b), for example, resolving the contributions of spatial and temporal coherence suppression to speckle reduction for the free-running laser is problematic. Consequently, in an effort to minimize the influence of temporal coherence on testing of the multibeam laser and examine its ability to suppress speckle through an increase in angular diversity (with a corresponding decline in spatial coherence), a series of tests was conducted for which the Ti:Al₂O₃ laser resonator was modified by substituting a VBG for the output coupling mirror. As shown in Fig. 4(a), doing so results in the spectral width of the laser falling by >2 orders of magnitude [relative to that for Fig. 3(b), for example]. The black trace is the spectrum observed experimentally when the VBG is installed, and the red curve in Fig. 4(a) is a Gaussian profile having a bandwidth of 0.12 nm (∼60 GHz) FWHM, which corresponds to a calculated coherence length and time of ∼3.3 mm and ∼11 ps, respectively. These values are sufficiently large that the impact of angular diversity, and the associated suppression of spatial coherence, on speckle reduction can now be observed because the influence of temporal coherence has been minimized.

With the Ti:Al₂O₃/VBG oscillator, a set of images of a USAF binary transmission mask was acquired by placing the mask downstream of the multibeam laser. A ground glass diffuser located ∼7 cm from the target was also placed in the beam path. Images b, c, and d of Fig. 4 are representative, false-color micrographs of the speckle patterns produced by irradiating the mask with a Ti:Al₂O₃ laser array having 2, ∼10, or ∼1000 microlasers from which C is calculated to be 0.35, 0.086, and 0.022, respectively. These tests were conducted with an imaging magnification of unity, and the measured values of C are consistent with the expression

where $In̄$ is the mean intensity of the nth microlaser in the far-field and N is the total number of microlasers in the array.²¹ If each microlaser source contributes equally to the speckle pattern, then Eq. (1) reduces to C = 1/$N$⁠. Consequently, it was found that the measured variation of the speckle contrast ratio with N is consistent with C ∼ N^−1/2 scaling, and we conclude that the multi-microbeam laser described here is capable of reducing speckle to negligible levels because of suppression of the laser’s spatial coherence brought about by increased angular diversity. Since the narrow bandwidth version of the multibeam laser removes temporal decoherence as a factor in speckle suppression, the imaging experiments demonstrate clearly that reducing spatial coherence with N independent microlaser beams is responsible for the virtual elimination of speckle. It should also be mentioned that for a single, fully coherent laser beam, the theoretical value of C is unity. In practice, however, depolarization by the diffuser, camera noise, and residual amplified spontaneous emission may reduce this value to as low as ∼0.5.

When N is increased beyond ∼1000, C has been observed throughout these experiments to fall to immeasurable levels (<0.03). Consequently, for all of the experiments to follow, N was set at >1000 for the sake of maximum spatial resolution, and several images were recorded with N ∼ 4000. As a first example, Figs. 5(a) and 5(b) compare images acquired of the same sample of immobilized Spirogyra algae, recorded at 10× magnification and illuminated with either a conventional green laser or the microlaser array (and a diffuser in the beam path). With sample irradiation at 543 nm from a commercially available, He–Ne laser (fundamental transverse mode, ∼50 ms exposure), no image is discernible because the speckle pattern overwhelms the sample image. In contrast, exposure of the algae sample with a single pulse from a Ti:sapphire laser array yields a false-color image [Fig. 5(b)] having a spatial resolution of ∼3 μm. The resolution was determined experimentally from magnified images such as that of Fig. 5(b) and is in agreement with the theoretical resolution of the microscope. In particular, the double-helix structure of the Spirogyra filaments is clearly visible as are the walls separating adjoining segments of the filaments. Also visible in the foreground filament is the offset of 35 μm (along the axial coordinate) between the two, ∼85 μm diameter helices. A second Spirogyra sample was viewed in a backlit geometry with either a red LED or a red quantum dot (CdSe/ZnS) microlaser array serving as the illumination source [Figs. 5(c) and 5(d), respectively]. In the latter case, the sample was again exposed to only a single (∼8 ns) pulse from the laser, whereas the LED exposure time was 50 ms. Comparable spatial resolution was observed with both optical sources, but the exposure time again differed by almost seven orders of magnitude. In combination with Fig. 5(b), these images demonstrate that observing an object or transient phenomena at multiple wavelengths simultaneously is feasible with separate resonators having different gain media.

An extensive number of images of the motile, single-cell algae Chlamydomonas reinhardtii were recorded in vivo, and Fig. 5(e) (multimedia view) is one frame from a false-color video comprising a sequence of single-shot images obtained with the Ti:Al₂O₃ microarray laser and at 40× magnification. Obtained by combining individual images recorded every 100 ms (limited by the pump laser pulse repetition frequency of 10 Hz), this video illustrates the “run-and-tumble” motion of C. reinhardtii and shows that individual cells, depending on size, are capable of speeds up to 70 μm/s.²³ Although the contrast of this brightfield optical system is not sufficient to fully resolve the flagella by which the algae swim, probing these organisms with incoherent, near-infrared radiation allows for several prominent internal structures to be observed because the cell wall [the blue ellipse in the inset to Fig. 5(e)] and membrane are transparent in this spectral region. Specifically, the nucleus and chloroplast are identifiable, and depending upon the orientation of individual cells at any given time, the eye spot is also visible. A magnified, false-color image of a specific cell, again recorded with a single pulse of the microlaser array, is shown in the inset of Fig. 5(e). By altering the false-color scale, portions of the flagella and the eye spot (note the lower portion of the cell) are now visible, and several internal structures are evident. The examination of scores of expanded images consistently showed that the smallest identifiable features are ∼1.5 μm in at least one dimension. We wish to point out that for this optical configuration, the depth-of-field of the microscope is ∼1 μm, whereas the thickness of the aqueous region in which the cells live and move is >50 μm. Consequently, only a few cells are in focus for any given laser pulse. In addition, a number of the cells in the video of Fig. 5(e) are immobile because they are attached to an inner face of an optical window.

A more demanding test of the temporal resolution of the multibeam laser architecture is provided by Drosophila melanogaster, the common fruit fly. Figures 6(a) and 6(b) are silhouettes of Drosophila at rest and in flight, respectively. Both images were recorded by single laser pulses in a backlight geometry, and structural features in the insect’s wings and legs as small as 10 μm are discernible. Note, in addition, the segmentation of the forward-directed appendage when the fruit fly is in flight [Fig. 6(b)]. Drosophila beats its wings at frequencies up to 200 Hz, which correspond to peak linear velocities at the wingtips of several m/s. Consequently, capturing the flight of Drosophila while eliminating motion blur of the smallest features in the wings, for example, requires optical pulses ∼1 µs in duration (or less). As was the case with Edgerton’s stroboscope, the temporal resolution of the image is defined by the width of the low-coherence laser pulse. The implications of the representative images given in Figs. 5(e) and 6 are significant because they demonstrate that dynamic information can be retrieved from biological organisms and microscopic phenomena without sacrificing spatial resolution.

Transillumination imaging is generally suitable for imaging objects in close proximity to the illumination source, which places minimal requirements on the laser pulse energy. Indeed, all of the images of Figs. 5 and 6 required single pulse energies of only 1–10 μJ. Reflection imaging of objects at larger distances, however, demands far greater fluence from the optical source and particularly if a scattering medium intervenes between the object of interest and the optical source. Specifically, the inverse-square law scaling of the optical intensity backscattered from a target requires the laser intensity to increase rapidly with the laser-target distance. In this regard, the ability of the lasers of Fig. 1(a) to generate mJ pulse energies is vital. To this end, the efficacy of the multibeam laser was tested at distances of 3–5 m, and Figs. 7(a)–7(c) present images of an anodized aluminum target illuminated by three separate light sources (semiconductor laser, LED, or the narrow bandwidth laser of Fig. 4). The distance from each illumination source (and the receiver camera) to the target was set at 3 m. Figure 7(a) is the image captured when the target is illuminated by the beam from a near-infrared laser diode that is transmitted by a diffuser and collected by a 20 cm diameter condenser lens. The degree of speckle observed [Fig. 7(a)] is severe to the point that negligible information can be retrieved from the target. If the illumination is provided by a near-infrared LED with no downstream optics [Fig. 7(b)], the relatively high degree of spatial coherence (coherence area of ∼20 mm²) but low temporal coherence (coherence length ∼15 μm) of the incident light yields an image with notable speckle (C = 0.12), despite the relatively low degree of temporal coherence. Finally, employing the narrowband microlaser array as the illuminator in conjunction with a diffuser and the condenser lens produces a similar level of speckle [C = 0.13, Fig. 7(c)]. In this case, however, speckle suppression is achieved by increasing the angular diversity and decreasing the spatial coherence of the illumination. Both effects are enhanced by the large diameter of the lens but the solid angle subtended by the illumination at the target is not sufficiently large to fully suppress speckle with the present resolution.

Exemplary images of a U.S. Air Force resolution chart printed on white paper, recorded at a distance of 5 m from the laser aperture, are presented in Figs. 7(d) and 7(e). For these tests, the Ti:Al₂O₃ laser resonator represented by the spectrum of Fig. 3(b) illuminated the chart for the purpose of suppressing speckle by reducing both spatial and temporal coherence. We also note that no diffuser was present in the illumination path when acquiring these results. The image of Fig. 7(d) is representative of those obtained with a single, 1.5 mJ laser pulse. Despite the laser-to-target separation, it is evident that speckle is virtually absent (C ∼ 0.06) and the spatial resolution, limited at present by the sensor pixel pitch, is ∼250 μm. It must be emphasized that signal-to-noise ratios (SNRs) > 6.5 at such distances are realized with a standard, industrial CMOS camera and no electronic gain or software processing of the receiver signal. That is, the multibeam laser of Fig. 2(b) produces a single beam of sufficient pulse energy that neither sophisticated receiver optics nor intensified detectors are necessary for acquiring images of superior resolution. Panel (e) of Fig. 7 was recorded with the same experimental arrangement of Fig. 7(d) except for the insertion of a fog generator at the midpoint between the laser and resolution target, a configuration known to intensify speckle when a conventional laser serves as the illumination source.^8,24 The mean thickness of the fog stream was 15–20 cm, and a double-pass of the laser pulse through the fog was required. Despite the presence of dense fog in the optical path and the necessity of increasing the laser pulse energy by a factor of two (to 3 mJ/pulse) to overcome scattering losses and obtain an acceptable SNR, the image resolution is unchanged and speckle is again not evident.

Another vivid illustration of both the temporal and spatial resolution afforded by the multibeam laser is the imaging of an operating turbomolecular pump at a distance of 2 m. An optical window was installed on the 10 cm (4″) diameter pump, and the first (outermost) stage of rotor blades was imaged with 2 mJ pulses from the narrowband Ti:sapphire laser array (Fig. 4) and a diffuser and condenser lens inserted into the beam path [as illustrated in Fig. 1(b)]. Figure 8 (multimedia view) is a video for which each frame was acquired with a single, 8 ns laser pulse. Although the pump is operating at 56 000 rpm and, therefore, the rotor blades reach speeds of 300 m/s at the perimeter of the rotor stage, no motion blur is detectable. A slight degree of speckle is noticeable on the central ring of the first rotor stage in images such as those of Fig. 8 (for which C is found to be 0.11), but it does not significantly impair image quality. Most importantly, images similar to those of Fig. 8 that were acquired throughout these tests demonstrate sensor-limited spatial resolution (∼100 μm), despite the high speed of the pump rotor stages.

By modifying the classic Fabry–Pérot optical resonator with an internal array of microrefractive components such as microlenses, the operating point on the stability diagram is moved away from critical stability, and laser microbeams are produced only at the transverse position of a microlens. Generating more than 4000 mutually incoherent beams from several gain media has been achieved, suggesting that display projection, microscopy, or other imaging systems employing two or more oscillators emitting in different wavelength ranges are feasible. Because the speckle contrast ratio is observed to scale as N^−1/2, C can be tuned at will by varying the number of microresonators that are pumped. This capability is the realization of a goal suggested in 2013 by Nixon et al.,¹⁸ who, in the context of full-field imaging, suggested the tunability of C to be a desirable property for low-coherence lasers. Although the lasers reported here were designed and operated on the ∼10 ns time scale, the concept of producing a beam of low spatial coherence by way of an extended array of microbeams is applicable to both shorter and longer timescales.

In order to reduce the impact of temporal coherence on testing of the multibeam laser, the linewidth of a Ti:Al₂O₃ oscillator was reduced by >2 orders of magnitude (to ∼60 GHz) with a VBG. Imaging tests of this oscillator confirm that the speckle contrast scales as C ∼ N^−1/2, leading one to conclude that the impact of the multi-microbeam output is to virtually eliminate speckle through suppression of the laser’s spatial coherence. The peak power of the multibeam laser reported here is at least two orders of magnitude higher than that characteristic of previous laser resonators capable of suppressing speckle by generating N optical cavity modes. Furthermore, the Ti:Al₂O₃ laser bandwidth resulting from the insertion of a VBG into the resonator is among the lowest in the literature.¹⁸ Lasers designed to generate output beams of low spatial coherence generally have linewidths of hundreds of GHz to several THz,^{10,12–17,19} and, of course, those operating on the principle of reducing temporal coherence to suppress speckle typically exhibit bandwidths beyond 10 THz. In contrast, the ∼60 GHz (∼1.8 cm⁻¹) linewidth of Fig. 4(a) renders the tunable, multibeam oscillator of value for applications at the intersection of imaging and spectroscopy such as molecular LIDAR. Another recently reported application is that of reducing coherent artifacts in the recording of metasurface holograms.²⁵ 

Perhaps the most significant aspect of the results reported here, however, is the demonstration of a class of light sources adaptable for operation over an extensive parameter space, affording considerable flexibility with regard to speckle contrast, laser pulse energy and temporal profile, bandwidth, and gain medium. From a pragmatic standpoint, the oscillators described earlier are robust and compact (typically <2 cm in length and 5–8 mm in diameter), do not require precise orientation of the resonator with respect to the pump beam, and have generated single pulse energies up to 11 mJ without the benefit of an amplifier. Because speckle has been the bane of laser imaging applications in microscopy, LIDAR, and displays (for example) for decades, the scope of the illumination applications accessible with this approach to generating independent arrays of microlasers appears to be vast.

For most of these experiments, the primary microlens array (Thorlabs MLA150-7AR-M) was a monolithic, square-grid array of circular lenses fabricated from fused silica, having a nominal focal length of 5.2 mm (6.7 mm as determined by ray tracing) and seated in a plastic mount with a 25 mm diameter. The 300 μm lens array of Fig. 2(d) (Thorlabs MLA300-14AR-M) was also monolithic fused silica and had a nominal focal length of 14.6 mm (18.6 mm as determined by ray tracing). The 100 μm lens array (RPC Photonics, MLA-S100-f28) adopted for the CdSe/ZnS quantum dot experiments consisted of a polymer film of square lenses atop a glass substrate. Both Thorlabs lens arrays had broadband anti-reflective coatings (<1% reflection for both the pump and lasing wavelengths), but the RPC Photonics lens array had no such coating. The 25 mm diameter cavity mirrors were chosen to have at least 95% reflectivity at the lasing wavelength while transmitting the 532 nm pump light. With Ti:sapphire as the gain medium [producing the results of Fig. 3(d)], the optical cavity consisted of a 99% high reflector (CVI Laser Optics, PR1-800-99-IF-1025-UV) and a 5% output coupler (CVI Laser Optics, PR1-800-95-IF-1025-UV). For narrowband operation of the oscillator, a custom volume Bragg grating (Ondax Powerlocker PLR-788-95) served as the output coupler (5% output coupling) and was positioned with respect to the resonator such that the grating angle was aligned with the lasing axis. Q-switched, frequency-doubled Nd:YAG lasers (Continuum NY81C-10 and Spectra-Physics PRO-350-10H) provided the optical pump pulses, having a duration of ∼8 ns FWHM and approximately “top-hat” transverse intensity distributions (owing to unstable cavities in the oscillators). The pump light was propagated through the gain medium before reaching the microlens array so as to ensure relatively uniform pumping. Despite these precautions, non-uniformities in the pump intensity distribution in a plane transverse to the optical axis resulted in variations in the output of individual microlasers, which are particularly notable in Figs. 2(b)–2(e). The Ti:Al₂O₃ crystals were disks with diameters of 1 cm and thicknesses ranging from 500 μm to 3 mm, and the latter produced the most efficient laser. The Ti:Al₂O₃ crystalline wafers were supplied by GT Advanced Technologies with broadband anti-reflective coatings deposited on both facets and ∼0.25% Ti³⁺ doping. CdSe/ZnS carboxyl quantum dots (Qdot 655 ITK) were manufactured by Invitrogen and supplied in the form of an 8 μM solution that was deposited directly on the cavity mirror closest to the pump beam. The optical path length of the solution was 0.4–1 mm. The laser dye LDS 798 (Styryl 11) was manufactured by Exciton and dissolved in glycerol to a concentration of ∼1 mM. The dye was enclosed in a cuvette with an optical path length of 1 mm. All resonator components (except the VBG, which was mounted separately) were situated inside a Thorlabs 25 mm threaded lens tube, separated by 375 μm thick spacers (Thorlabs, SM1S01) and 2 mm thick retaining rings (Thorlabs, SM1RR) to achieve appropriate spacing, and were held firmly in place with light inward pressure exerted on both end mirrors by additional retaining rings. All of the near-field images presented here were recorded with laser pulse energies in the range of 1–10 μJ. Those recorded for objects located at ranges of 2–5 m required mJ pulses; the specific values are given in the text. The diffuser used here (Thorlabs DGUV10-220) was fabricated from a plate of fused silica and ground on one side with a grit of 220. For imaging targets at a distance of 2–3 m, a condenser lens (Edmund Optics #27-515) having a diameter and focal length of 200 and 800 mm, respectively, was employed.

The interior of the laser cavity was examined by imaging onto a camera (EO-1312C) the light transmitted through one of the mirrors with a custom-built microscope, generating the near-field images presented earlier. Far-field measurements and target illumination experiments were conducted with the laser emission exiting the output coupling mirror. For all backlight imaging results reported here, a separate camera (Sentech, STC-MC152 USB) was situated on the opposing side of the target. Both of these cameras were triggered by a TTL signal from the pump and have exposure times of ∼50 ms when imaging with the laser array. Olympus plan-achromat objectives (10× and 40×) served to collect light when imaging microscopic targets. Backscattered light was imaged onto a Lumenera Lt225 near-infrared enhanced industrial camera with a commercial telephoto lens (Canon 70–200 mm f/2.8L IS EF USM) for all tests except those conducted with a range of 5 m. For the imaging experiments at 5 m, a Sigma 150–600 mm f/5–6.3 DG OS HSM Contemporary telephoto lens (plus focal length extender: Canon, Extender EF 2× III) was adopted. The Lumenera camera was not triggered but rather was set for an exposure time of 100 ms, which corresponds with the 10 Hz pulse repetition frequency of the laser array, thus ensuring that no frames were dropped. Any native infrared-blocking filters were removed from the cameras, and the imaging systems were instead equipped with red or near-infrared longpass color filters for which “cut-on” occurred at ∼610 nm (Thorlabs, FGL610) and 721 nm (Thorlabs, FGL9), respectively. These wavelengths were chosen to match the emission wavelengths of a specific laser so as to minimize the amount of background light reaching the detector.

Measurements of the spectral and temporal characteristics of the laser arrays were designed so as to yield spatially averaged results (i.e., by viewing the array in the aggregate). This was accomplished by requiring the far-field beam to traverse an optical diffuser prior to reaching the appropriate detector. The spectra of Fig. 3(b) were recorded with an Ocean Insight FLAME-T USB spectrometer having a 1200 lines/mm grating and a 25 μm slit, resulting in a resolution of <1 nm. Also, the spectrum of Fig. 4(a) was acquired with an Acton SP2750 series spectrometer paired with a Princeton Instruments PI-MAX4 1024i intensified camera having a resolution of ∼0.02 nm. Laser waveforms and pulse energies were obtained with a Thorlabs silicon photodiode (DET10A) and calibrated Coherent EnergyMax (J-25MB-LE) pyroelectric detectors, respectively.

The support of the U.S. Air Force Office of Scientific Research (G. Pomrenke, H. Schlossberg, and J. Luginsland) under Grant Nos. FA9550-19-1-0218, FA9550-18-1-0380, and FA9550-14-1-0002 is gratefully acknowledged.

The authors are also co-inventors of U.S. Patent no. 10,620,449.

A.W.S and J.A.R. conducted the experiments; A.W.S., J.A.R., and J.G.E. analyzed the data; and A.W.S. and J.G.E. wrote the manuscript.

The data that support the findings of this study are available from the corresponding author upon reasonable request.