Laser diode bar

A laser diode bar is a single edge-emitting semiconductor chip, typically about 10 mm wide and a few millimeters long in the cavity direction, that carries a row of broad-area emitters driven in parallel. The fraction of the width that emits, the fill factor, is the stripe width over the pitch and typically lies between 20% and 50% for continuous bars: 100 µm stripes on a 500 µm pitch give 20%, on a 200 µm pitch 50%. Bars at 808 nm and in the 9xx nm band emit tens to hundreds of watts continuous, and more in quasi-continuous pulsed operation, at currents of order 100 A and voltages below 2 V. Wall-plug efficiency is typically 50–60%; above 70% has been demonstrated near 970 nm.

Structure and drive

All emitters share one epitaxial structure and one pair of facets, and they are electrically in parallel, so the bar behaves like one laser diode with a very low series resistance. The driver must supply high current at low voltage, and the contacts and solder joints must carry it with little resistive loss. A bar with 55% efficiency emitting 100 W draws 182 W, about 107 A at 1.7 V, and dissipates 82 W as heat. Several bars mounted one above the other form a stack, which multiplies the power while keeping the footprint small.

The fill factor is a compromise. Gaps between emitters spread the heat and reduce thermal crosstalk, and a low fill factor eases fiber coupling; a high fill factor gives more power per bar at a higher heat load per unit area and is common for bars run in quasi-continuous pulsed operation.

Beam geometry

Each emitter is about 1 µm high and near diffraction-limited in the fast axis, and highly multimode in the slow axis, where the bar as a whole is 10 mm wide. With a slow-axis half-angle of 5° (87 mrad) and a half-width of 5 mm, the slow-axis beam parameter product is of order

$$5\ \text{mm} \times 87\ \text{mrad} \approx 436\ \text{mm·mrad},$$

a rough upper estimate, against a fast-axis diffraction limit of $\lambda/\pi$ = 0.31 mm·mrad at 976 nm: a ratio of about 1,400.

A fast-axis collimator, a cylindrical or acylindrical microlens a few hundred micrometers from the facet, collimates all emitters at once, and slow-axis lenslets at the emitter pitch usually follow (see How to collimate a laser beam). For fiber coupling, beam-shaping optics cut the beam into segments along the slow axis and restack them along the fast axis, dividing the slow-axis BPP by the number of segments $N$ and multiplying the fast-axis BPP by about $N$. In the ideal case the two become equal at

$$N = \sqrt{436 / 0.31} \approx 37,$$

which gives about 11.6 mm·mrad in each axis, the acceptance of a 105 µm core, 0.22 NA fiber. Real fast-axis BPP after the collimator is several times the diffraction limit, so practical bar-to-fiber modules couple into larger cores or higher NA.

Smile and alignment

"Smile" is the bow of the row of emitters in the fast-axis direction, produced by mounting stress or a curved heatsink and typically a few micrometers or less across the bar. The short focal length of the fast-axis collimator turns it into pointing error: 1 µm of smile behind a 0.3 mm focal-length collimator tilts that emitter's beam by about 3.3 mrad, comparable to the residual fast-axis divergence of each emitter. Smile therefore broadens the collimated fast-axis beam and lowers fiber coupling efficiency; it is measured by imaging the emitters through the collimator and controlled by low-stress, expansion-matched mounting.

Cooling and wavelength

Bars are soldered p-side down on copper or expansion-matched submounts and cooled by conduction to a water-cooled plate or through microchannel coolers, thin copper structures with water channels directly under the bar. The thermal resistance from active region to coolant is a fraction of a kelvin per watt. Assuming 0.4 K/W for illustration, the 82 W of heat above raises the active region by 33 K, which with the GaAs-based shift of about 0.3 nm/K moves the emission by about 10 nm. Absorption bands such as that of Yb at 976 nm are only a few nanometers wide, so coolant temperature and drive current are set together to hold the wavelength on the band; junction temperature also governs lifetime. Microchannel coolers need deionized water, because the coolant is often electrically connected to the bar.

Applications

Most bars are pump sources for optical pumping: 808 nm bars side- or end-pump Nd:YAG and Nd:YVO₄ in DPSS lasers, and fiber-coupled 9xx nm bars pump the inner cladding of ytterbium double-clad fiber. Direct-diode systems use bars and stacks for hardening, cladding, brazing and plastic welding, where a beam parameter product of tens of mm·mrad is acceptable.

Common questions

How is a diode bar different from a broad-area laser diode?

A broad-area laser diode is a single emitter. A bar places tens of them side by side on one chip, sharing facets, heatsink and current supply, so output and heat load scale with the emitter count.

Why do bars need fast-axis collimation so close to the facet?

The fast axis diverges at tens of degrees, so the beam is already of order a millimeter wide one millimeter from the facet. A short-focal-length lens close to the facet collimates it while it is still small.

References: R. Diehl (ed.), High-Power Diode Lasers: Fundamentals, Technology, Applications (Springer, 2000); L. A. Coldren, S. W. Corzine and M. L. Mašanović, Diode Lasers and Photonic Integrated Circuits, 2nd ed. (Wiley, 2012); M. Kanskar et al., "73% CW power conversion efficiency at 50 W from 970 nm diode laser bars," Electronics Letters 41, 245 (2005).