The growing number of laser diodes in use has focused attention on the design of the optics for this type of device. Understanding the special characteristics of diode radiation, including beam divergence and angular asymmetry, is fundamental to practical design. However, the size, weight, and cost of the optical elements must also be considered.
Special characteristics of diode radiation must be accommodated by the optics. This diagram shows the divergent laser output first collimated, then converted from elliptical to circular cross section (in the anamorphic prism pair), expanded, and focused to a small spot.
Laser diodes have many advantages: they are small and can be directly modulated, and the power requirements are modest. Their use is increasing, and they are expected to replace gas lasers in many applications. Compared with other lasers, however, the output radiation from laser diodes is highly divergent. As a result, the use of a laser diode, whether in a new application or as a replacement for another laser, usually requires that some sort of collimating or refocusing optics be specified.
This article will acquaint readers with the types of optical systems needed for various laser diode applications and show how a realistic set of specifications can be developed. We begin with a brief review of laser-diode operating principles and show how laser-diode construction gives rise to the unusual output characteristics. Next, we shall examine some specific collimating and focusing systems; special attention will be paid to the way in which each specification affects the complexity and, ultimately, the cost of the optics.
FIGURE 1. Typical double-heterostructure laser diode
Laser-diode operation
A schematic of the essential elements in an injection laser diode is shown in Figure 1. The active medium, where lasing
occurs, is sandwiched between two cladding layers of both higher bandgap and lower refractive index. This structure is on an appropriate substrate; a contact layer to facilitate electrical connection is deposited on top of the stack. In operation, the p-n junction is forward biased; this causes holes and electrons to be injected from the cladding layer into the active region, where they are confined by the energy barrier between adjacent layers. Carrier recombination in the active region produces light, which is confined to this region by the surrounding regions of lower index. A photon produced by recombination can cause other carriers to recombine and emit photons coherent with the first (stimulated emission); as current increases in the active region, the gain due to stimulated emission eventually exceeds the intrinsic losses in the cavity due to absorption, and the device lases. Mirror facets at the ends of the cavity provide optical feedback, the other critical element needed to make the device a laser.
Laser diodes can be divided into two generic types depending on the method of confinement of the lasing mode in the lateral direction. Gain-guided laser diodes work by controlling the width of the drive-current distribution; this limits the area in which lasing action can occur. Because of different confinement mechanisms in the lateral and vertical directions, the emitted wavefront from these devices has a different curvature in the two perpendicular directions. This astigmatism in the output beam is one of the unique properties of laser-diode sources. Gain-guided injection laser diodes usually emit multiple longitudinal modes and sometimes multiple transverse modes. The optical spectrum of these devices ranges up to about 2 nm in width limiting their coherence length. This feature limits the ability to collimate or refocus emitted light to the diffraction limit. Damage at the output facet is directly related to the optical power density there; gain-guided lasers typically have spots of several micrometers at their output faces and are therefore capable of producing continuous wavepower in the tens-of-milliwatts range.
Index-guided laser diodes use refractive index steps to confine the lasing mode in both the transverse and vertical directions, so that these lasers are free of astigmatism. Index guiding also generally leads to both single transverse-mode and single longitudinal-mode behavior. Typical linewidths are on the order of 0.01 nm. The spot size at the output facet is typically a couple of micrometers, limiting the amount of power available from these devices to about 15 mW or less. Recent developments, such as large-optical-cavity (LOC) and non-absorbing-mirror (NAM) devices have allowed an increase in the lasing spot size without sacrificing single-mode characteristics. These techniques have pushed output power past 20 mW continuous wave, while maintaining reliability and lifetime.
A feature common to all laser diodes is asymmetric output: the beam divergence is different in the planes parallel and perpendicular to the emitting junction. Because of diffraction, the narrower vertical dimension of the output facet causes greater beam divergence than in the transverse direction. Ratios between the angles typically range from 2:1 to 6:1.
The output from most diodes being manufactured today falls in the wavelength range between 780 and 905 nm or in the range between 1.2 and 1.7 µm. The longer-wavelength range has been developed to take advantage of the high optical-fiber transmission and low dispersion achievable at these wavelengths, so that these devices are used mostly in telecommunications.

Collimating optics
Because the highly divergent output of the injection laser diode is difficult to work with, the first task usually facing the designer of a laser-diode-based system is specification of collimating optics. Performance of the collimator determines spot size, beam divergence, resolution, etc. of the rest of the system, so it is imperative that the collimator quality be sufficient. On the other hand, to limit the cost of the optics, excessively stringent performance requirements should be avoided. Many possible applications for injection laser diodes currently use helium-neon or other gas lasers. It would be expeditious to use as much of existing optical designs as possible, so in these cases we would like our laser-diode source to mimic the characteristics of a gas laser; this means a small-diameter beam with diffraction limited collimation quality. Once achieved, this package can be inserted into a pre-existing optical system, which may have to be reoptimized for the new wavelength and beam diameter. Typical situations where this might occur are in laser printers, barcode scanners, or laser measurement and alignment systems.
In designing a high-quality collimator for laser-diode sources the first consideration is collection angle, or numerical aperture. Numerical aperture is the sine of the half angle of collection. Divergence angles for laser diodes in the direction perpendicular to the junction range as high as 60 degrees FWHM (full width at half maximum), typical values are in the 40 degree to 60 degree range. For a beam with a Gaussian intensity profile, the 1/e
One important reason for maintaining a reasonable working distance is that many of the laser diodes currently available have a coverglass over the package. Coverglasses are typically 0.17-mm to 0.30-mm thick and are usually spaced anywhere from 0.7 to 1.5 mm from the emitting surface of the laser diode. Obviously, the imposition of a coverglass limits mechanical access to the laser diode and dictates a certain minimum working distance. The other major factor introduced by a coverglass is wavefront aberration. A plane parallel plate introduces spherical aberration into a diverging beam, so if diffraction-limited collimation is desired, this effect must be compensated in the collimating optics. In Figure 2 the wavefront distortion introduced by a window is plotted as a function of window thickness, for specified values of refractive index; and numerical aperture. As can be seen, the index; of refraction (window material) is relatively unimportant; for collimators with a numerical aperture of 0.3 or less, the effect of a window, unless very thick, will introduce less than 1/10 of a wave aberration and can usually be ignored.

In general, the design of a diffraction-limited laser-diode collimating objective should correct for spherical aberration, coma, astigmatism, and spherochromatism. Chromatic aberration and field curvature can usually be ignored. Correction for spherical aberration is essential for good performance on-axis, while correction for coma and astigmatism allows for some tolerance in the positioning of the diode relative to the lens. Ignoring field curvature allows for greater correction of the other off-axis aberrations; the effect of field curvature can be compensated by adjusting the focus of the collimating system. Single-mode injection laser diodes are essentially monochromatic sources, and therefore no correction for chromatic aberration is needed. However, there may be batch-to-batch wavelength differences in laser diodes or temperature-dependent wavelength changes that necessitate correction for spherical aberration over a certain wavelength range.
Circularizing the elliptical beam


Figure 3 shows the design of a 6.5-mm focal length, 0.62-numerical aperture collimating lens; the working distance for this design is 1.21 mm. The lens consists of a cemented doublet followed by two singlets, the last of which is a thick plano-convex lens. This particular lens form allows as great as possible a working distance for such a high numerical aperture. Using a plano-convex lens as the final element provides some economy in manufacture, since flat surfaces require no special tooling, and there is no limit to the number of lenses that can be processed at once; but the main advantage of this lens form is that it makes correction for a coverglass particularly easy.
To compensate for the presence of a coverglass on the diode, we simply change the thickness of the final element; no other changes in radius or thickness are needed. To first approximation, the amount of thickness to be removed from the final element is simply the thickness of the coverglass itself. Furthermore, removing this material increases the working distance; all that is necessary is that the distance from the laser diode surface to the inside of the coverglass, plus the distance from the outside of the coverglass to the rear lens element, be the same as the original working distance (before modification). As a result of this feature, a single collimator design can be used for a variety of different diodes and configurations with only one simple modification, namely the thickness of the final element.
If system design or diode specifications change after the original design has been completed and built, retooling is not needed to take changes into account. Even existing lenses can easily be retrofitted to work under the new conditions, if necessary. From the point of view of a lens supplier, one basic design can be used to satisfy a wide range of customer requirements, finished elements can be kept on the shelf, and only polishing of the final plano side of one element is needed to complete a collimator that will be tailored for the customer's exact configuration.(5)
Objective lenses for optical recording

One application for injection laser diodes is in laser-videodisc and optical data-storage systems. The primary concerns in designing focusing objectives for these recording applications are spot size and working distance. Storage capacity is directly related to the size of the "pits" on the disc surface. Since longer wavelength means a larger focused spot, these systems typically use the shortest-wavelength laser diodes commonly available -- currently about 780 nm. With this wavelength and an objective in the 0.45 to 0.50 numerical-aperture range, submicrometer-sized pits can be formed on the disc surface. Figure 4 shows one possible design for a videodisc objective; it is an 8-mm focal length, 0.50 numerical aperture lens with a working distance of 1.56 mm. This lens embodies several essential features we would want to include in any recording objective. Notice that it has only three elements; weight is a concern since the lens assembly must be mounted on a voice coil and moved axially to correct for focus errors caused by movements of the disc surface. To correct for movement of the disc in the perpendicular direction during playback, the lens must be designed to work well over a small field of view, usually about 80 µm in diameter. There is no significant degradation in spot size for this lens over this range. As in the case of the collimator, field curvature correction is unnecessary, since the focus servo can correct for movement in the axial direction. Correction for spherochromatism is important because of possible diode-to-diode wave-length variations mentioned before.
The rear element of this lens is a plane parallel plate. Once again, the use of plane surfaces allows for economy during manufacture and simplifies correction for the presence of a disc. As before, an amount of material approximately equal to the thickness of the disc is removed from the final element to correct for the aberrative effects of the disc. Since the plate is 4-mm thick in this design we can easily accommodate the standard 1.2-mm disc, or a thicker disc if required. This one design can be readily adapted to a variety of conditions by the lens manufacturer or easily modified by a user to accommodate changes in disc specifications.
Because the linewidth single-mode lasers is so narrow, they have a long coherence length; light reflected by the disc that re-enters the laser can cause feedback problems and output power instability. One solution to this is to build an optical isolator using a polarizer and a quarter-wave plate. Another solution is to use a multi-mode gain-guided laser, which emits over a wavelength range of about 4 nm. If this type of source is used it will also be necessary to introduce some correction for chromatic aberration. A good rule to use is that the secondary spectrum of the lens should be less than the depth of field; an estimate of useful depth of field is given by

=780 nm and a numerical aperture (NA) of 0.50, we get a depth of field of 1.6 µm. The collimator we have discussed has a focal-length variation of about 2.5 mm over a 4-nm range, so it would require some additional color correction and the addition of more elements, to work optimally under these conditions.
Focusing lenses for fiber optics
A typical application in the fiberoptics/telecommunications field might be focusing already collimated light into a fiberoptic cable. This situation might arise in a system that uses wavelength multiplexing or requires sampling of the beam. Because of the working-distance restrictions on collimators already mentioned, it is desirable to collimate first, and then insert any beamsplitters or other components into the collimated beam. This light then needs to be refocused into the fiber core. Typical single-mode-fiber core diameters range from 2 to 8 µm, with numerical apertures of 0.12 or less, and multimode fibers have core dimensions of 50 µm upwards, with typical numerical apertures of about 0.25 to 0.50.(7) These combinations of spot size and numerical aperture make this application much less demanding than the videodisc focusing objective. At wavelengths above1.3 µm, the availability of higher-index materials such as silicon open up the possibility of using single-element focusing lenses, even for single-mode applications. As an example, consider a single-mode fiber for 1.55-µm wavelength with a core diameter of 6 µm and a numerical aperture of 0.12. At this wavelength and numerical aperture the diffraction-limited spot size is about 8 µm; this is absolutely the best that can be achieved. This can easily be accomplished with a single meniscus lens made of silicon. Because of its higher index of refraction, silicon can provide a higher level of correction per element than optical glass. Using glass, a two-element system would be needed to match the performance of the silicon singlet just mentioned.
The relatively large core diameters of multi-mode fibers make coupling collimated light into them fairly easy. Even though the numerical apertures of most multimode fibers are much larger than in the single-mode case, the lens systems for this application do not need to be even remotely diffraction limited. In general, two elements are the most that are needed to focus into a multimode fiber, and in many cases a singlet suffices (this assumes diffraction-limited input collimation). One exception might be the situation in which two or more wavelengths must simultaneously be coupled into the fiber.