Monolithic Multi-Wavelength Laser Diode Array

The present invention relates to monolithically integrated semiconductor laser diode arrays of the type producing plural output beams of different wavelengths.

Addressable monolithic multi-wavelength light sources, especially laser arrays that can simultaneously emit different wavelength light from different elements in the array are useful in a variety of applications, such as color printing, full color digital film recording, color displays, and other optical recording system applications.

In U.S. Pat. Nos. 4,925,811, 4,955,030 and 5,039,627, Menigaux et al. describe various terraced semiconductor layer structures having stacks of alternate confinement layers and active layers. Each active layer in the stack has a different composition and is capable of emitting light of a different wavelength. A desired number of laterally spaced PN junctions are formed by localized introduction and diffusion of a p-type impurity, the different PN junctions being made in the vicinity of different active layers in the stack. The resulting structures can emit, either simultaneously or independently, as desired, different wavelength light beams from the array.

In U.S. Pat. No. 4,831,629, Paoli et al. describe laser arrays that are characterized by nonuniform current pumping across the arrays. The nonuniform current distribution among the several laser emitters of the array is brought about, for example, by different current confinement or stripe widths or by some other nonuniform contact configuration. The different pump current and carrier densities across the array result in different increments of band filling in the active region's quantum well structure, thereby causing individual laser emitters to operate at different emission wavelengths.

In U.S. Pat. No. 4,993,036, Ikeda et al. describe a semiconductor laser array having a double quantum well active region in which the two quantum well layers have different energy gaps, and, consequently, different oscillation wavelengths. There are also provided two adjacent pairs of grating reflectors, each pair having a grating pitch that is tuned to reflect a different one of the two oscillation wavelengths of light. Arrays having more than two quantum well layers, each with a different energy gap, and having more than two sets of wavelength selective reflectors, are also suggested.

In U.S. Pat. No. 4,951,293, Yamamoto et al. describe a semiconductor laser and a frequency doubling optical nonlinear device, both mounted on a submount to form a visible laser source. The nonlinear device can have a plurality of parallel waveguides for receiving light beams from a semiconductor laser array.

An object of the invention is to provide monolithic multi-wavelength semiconductor laser arrays.

The above object has been met with a laser diode array having a semiconductor material body with a plurality of contiguous semiconductor layers forming a heterostructure. The heterostructure includes a composite active region made up of two or more adjacent quantum well layers of dissimilar material. The composite active region is partially disordered by an impurity free interdiffusion of atomic constituents of the quantum well layers so as to shift the emission wavelength of the active region. Different areas of the active region in the laser diode array are disordered to different degrees so that these different areas have different emission wavelengths. Each area may have one or more waveguides and distributed Bragg reflector (DBR) gratings tuned to the emission wavelength in that area for providing feedback in a resonant cavity. Further, each different area of the active region or each waveguide may be electrically pumped by a separate, individually addressable, current injection means.

The impurity free interdiffusion may be produced by a vacancy-enhanced interdiffusion technique in which a capping structure is formed on the wafer surface with native vacancies therein. In group III-V compound semiconductors, preferential formation of either group III vacancies (or group V vacancies) in the cap will cause preferential interdiffusion of group III atoms (or group V atoms) in the active region when the vacancies are caused to migrate, such as by thermal diffusion. The interdiffusion may result in gain-enhancing strain in a formerly lattice matched active region.

The object is further met in a laser diode array that includes a ferroelectric layer structure epitaxially grown on the same substrate as the monolithic array of laser oscillators to double the frequency of light emitted from at least one of the laser oscillators. A preferred embodiment comprises a Ca.sub.x Sr.sub.1-x F.sub.2 /LiTaO.sub.3 /Ca.sub.x Sr.sub.1-x F.sub.2 layer structure with an array of LiTaO.sub.3 waveguides. Those waveguides intended for frequency doubling are periodically poled, preferably using first order quasi-phase matching. A power amplifier array between the laser oscillator array and ferroelectric waveguide array amplifies the optical power from the laser for efficient harmonic conversion in the periodically poled waveguides.

FIG. 1 is a perspective view of a preferred laser diode array embodiment of the present invention.

FIG. 2 is a sectional view taken along the lines 2--2 in FIG. 1.

FIG. 3 is a side elevational view of the laser diode array embodiment in FIG. 1.

FIG. 4a is an enlarged side view of a heterostructure for the laser diode array of FIG. 1 prior to disordering in accord with the present invention, and a corresponding graph of the heterostructure's energy band structure in terms of energy gap E versus depth Z within the heterostructure.

FIG. 4b is an enlarged side view and corresponding graph, analogous to FIG. 4a, for a heterostructure that has been disordered in accord with the present invention.

With reference to FIG. 1, a monolithic multi-wavelength laser array of the present invention includes a laser active region 19 in a semiconductor heterostructure in which the emission energy has been varied laterally across the array by partially disordering the composite quantum well in selected areas, resulting in emission wavelengths that can range from 630 nm to 1.1 .mu.m. The laser array typically also includes arrays of front and rear distributed Bragg reflector (DBR) gratings 17 that have been fabricated with grating periods selected to match the various different gain peaks of the individual emission bands of the active region in the array, resulting in an array of laser oscillators 13a characterized by single longitudinal mode emission from each of the gain regions in the array. An amplifier array 13b may be used to amplify the output of the DBR master oscillators 13a to high power outputs. A patterned p-metallization 21.sub.1 -21.sub.4 individually addresses the DBR master oscillators 13a and amplifier array 13b with selected injected currents I.sub.1 -I.sub.4. The array can have a center-to-center separation as close as 10 .mu.m.

In addition to having a multi-wavelength master oscillator and power amplifier array grown monolithically on the same substrate 11, the embodiments of the invention can also include epitaxially grown ferroelectric material 15 for doubling the frequency of at least one of the outputs from the power amplifier array 13b. The frequency doubling material 15 includes an array of self-aligned waveguides 23.sub.1 -23.sub.4 therein to minimize coupling losses between the power amplifier 13a and frequency doubling material 15. Waveguides 23.sub.3 and 23.sub.4 in which frequency doubling is intended to take place are preferably periodically poled to efficiently convert the fundamental radiation received from the amplifier array 13b into the second harmonic frequency.

With reference to FIG. 4, the addressable monolithic multi-wavelength laser diode array of the present invention is based on a composite quantum well structure for the active region in which the emission band is adjustable during manufacture of the laser array by partially disordering the quantum well via the mechanism of vacancy-enhanced interdiffusion. This mechanism differs from that of impurity induced disordering in that vacancy-enhanced interdiffusion is an impurity-free technique, so the band structure of the disordered quantum well is not damaged and high quality optical emission properties are retained. In the vacancy-enhanced interdiffusion technique a capping structure is formed on the semiconductor wafer surface. This cap, depending on how it is made, results in the preferential formation of either group III or group V vacancies at the wafer surface in group III-V compound semiconductor materials. For example, if the cap is epitaxially grown with a slight excess of group V atomic species, then vacancies will be created in the group III sublattice of the compound semiconductor crystal making up the cap. Vacancies and other native point defects, such as interstitials, are important to the mechanism for atomic interdiffusion. Migration of native vacancies through one of the sublattices results in the interdiffusion of the elements in that sublattice. Heating the semiconductor wafer will cause diffusion of vacancies formed at the wafer surface into the heterostructure and will cause interdiffusion of elements across layer boundaries. Capping conditions that result in preferential formation of group III vacancies at the surface cause a much greater interdiffusion of the group III elements at the quantum well relative to the group V elements. The converse is true when the cap results in preferential formation of group V vacancies at the surface. The diffusion rates in the group III and group V sublattices have been found to differ by two orders of magnitude when the material is capped to enhance migration of one group of elements over the other.

In FIG. 4a, one possible formulation of the semiconductor heterostructure prior to disordering consists of a composite quantum well active region with both a Ga.sub.0.7 Al.sub.0.3 As layer growth and a In.sub.0.5 Ga.sub.0.5 P layer growth. The composition has been selected so that both layers are lattice matched to each other and to the GaAs substrate, and further so that both layers have an identical bandgap Eg (=1.97 eV) and an identical emission wavelength of 630 nm. Other compositions could also be selected. The composite quantum well is embedded in a separate confinement heterostructure consisting of In.sub.0.5 Ga.sub.0.35 Al.sub.0.15 P confining layers and In.sub.0.5 Al.sub.0.5 P cladding layers. The cladding layers have a higher bandgap and lower refractive index than the confining layers, which in turn have a higher bandgap Ec than the composite quantum well.

In FIG. 4b, the quantum well has been disordered via preferential interdiffusion of the group III elements. At the point where the atomic gradient for each individual specie vanishes, the Ga.sub.0.7 Al.sub.0.3 As layer has become In.sub.0.25 Ga.sub.0.6 Al.sub.0.15 As and the In.sub.0.5 Ga.sub.0.5 P layer has become In.sub.0.25 Ga.sub.0.6 Al.sub.0.15 P. The confinement and cladding layer compositions are substantially unchanged, although some intermixing may occur locally at the layer boundaries. The electron energy band structure has been modified by the interdiffusion, such that the band gap for the InGaP layer has been increased in the InGaAlP layer to near the band gap Ec for the confinement layers, and such that the band gap for the GaAlAs layer has been decreased in the InGaAlAs layer now present to a new band gap Eg'. For In.sub.0.25 Ga.sub.0.6 Al.sub.0.15 As, this band gap Eg' is about 1.32 eV, corresponding to an emission wavelength of 940 nm.

If, instead of a group III elemental interdiffusion, a capping condition is created to enhance interdiffusion of the group V elements, then at the point where the atomic gradients in the quantum well layers disappear, the layer compositions are In.sub.0.5 Ga.sub.0.5 As.sub.0.5 P.sub.0.5 and Ga.sub.0.7 Al.sub.0.3 As.sub.0.5 P.sub.0.5. The first of these two intermixed quantum well layers has a band gap of 1.38 eV corresponding to an emission wavelength of 900 .mu.m.

The relative diffusion rates of the group III (or group V) elements are determined not only by the atomic gradients for each individual specie, but also by the heat of formation of the various bonds. This latter factor has not previously been an issue for the interdiffusion of group III atoms in disordering conventional GaAlAs heterostructures, because in such cases the type of bonding is invariant with respect to lattice location. That is, in GaAlAs heterostructures, all group III atomic bonds are to arsenic atoms, the only specie present in the group V sublattice. However, in an InGaP/GaAlAs material system, such as the epitaxial layer structure in FIG. 4a, atoms migrating within the group III sublattice can exchange arsenic and phosphorus bonds, producing a change in total bond energy. If an atom gains a sufficient amount of energy by exchanging a phosphorus bond for an arsenic bond, it becomes possible to observe atomic migration or diffusion against the atomic concentration gradient. Via this mechanism, the GaAlAs layer in the composite quantum well active region can be driven to a composition of InGaAlAs in which the concentration of indium is greater than 0.25 and the concentration of gallium is less than 0.6. A band gap less than 1.1 eV and an emission wavelength longer than 1100 nm can be achieved.

Wavelengths between 630 nm and 1100 nm can be produced in areas of the active region which are only partially disordered, instead of being interdiffused to completion. Thus, by controlling the degree of disordering, intermediate emission wavelengths can be obtained. Selective disordering can produce different wavelength emission bands in adjacent areas of the array so that a monolithic multi-wavelength laser array can be constructed. Heterostructures could also be designed to promote intermixing between the quantum well layers and the adjacent confinement layers, so that partial disordering causes a shift toward shorter emission wavelengths, instead of the shift from 630 nm to longer wavelengths up to 1100 nm.

Vacancy-enhanced disordering, by interdiffusing the group III (or group V) elements in the quantum well layers, produces a lattice mismatch that induces a compressive strain in the quantum well. In the case of referential group III interdiffusion in which the constituents have been interdiffused to the point where the atomic gradients vanish, as in the example given above in FIG. 4b, the associated strain is .DELTA.a/a=1.8%. This compressive strain of the active region causes a reduction in the valence band effective mass, resulting in closer symmetry of the conduction and valence bonds, and thereby enhancing the transition probability and increasing the quantum conversion efficiency. The strain has been shown to reduce the threshold current and increase the gain of the quantum well by as much as a factor of two. The critical thickness corresponding to a 1.8% layer strain due to a lattice mismatch is approximately 12 nm. For quantum well thicknesses less than this critical thickness, the material does not develop any associated dislocations. In fact, the reliability of lasers with a strained active region less than the critical thickness has been found to be greater than those with lattice matched active regions. Consequently, the disordered and partially disordered active regions in the heterostructures of the present invention have high gain and high reliability.

Referring to FIG. 2, the heterostructure forming the multi-wavelength master oscillator power amplifier array of the present invention includes a p-type InGaAlP confinement and first upper cladding layer 31 epitaxially grown on top of the composite quantum well 19. Layer 31 may be selectively etched to form a mesa structure made up of multiple laterally spaced apart unetched mesa regions 49 separated by etched regions. A 10 .mu.m center-to-center spacing between mesas 49 is typical. An InAlP cladding layer 33 is then grown over the InGaAlP confinement layer 31, filling the regions between the mesas 49. The presence of the InAlP cladding regions 33 results in a real refractive index waveguide array that substantially confines light propagation in the quantum well active region 19 and confinement layers 29 and 31 to those areas coinciding with the mesas 49. The InAlP cladding layer 33 may be n-type material, confining current solely to the mesa regions 49. A p-type GaAs layer 35 may be deposited on top of each mesa 49. Additionally, a final InGaP cladding layer 37 and a GaAs cap layer 39 may be deposited over the InAlP cladding layer 33 and GaAs layer 35. The index guides in the resulting heterostructure stabilize each laser emitter in the array in a single spatial mode and minimize threshold current for lasing operation. Other layer structures could be used in combination with the partially disordered composite quantum well active region 19 of the present invention to produce alternative multi-wavelength laser array embodiments.

As seen in FIGS. 1 and 3, the structure also includes arrays of front and rear distributed Bragg reflector (DBR) gratings 17 for feedback of light in the associated gain region 19 of the laser oscillator array 13a. Typical reflectivities for the front and rear gratings 17 are 20% and 50% respectively. Because it is difficult to directly expose the short grating pitches needed to make first order gratings in short wavelength materials, second order gratings are preferred. The pitch of each grating 17 in the array is tuned to the emission wavelength of the associated gain region, as determined by the extent of partial disordering of the composite quantum well 19. The gratings may be fabricated between the InGaAlP confinement layer 31 and the InAlP cladding layer 33 onto portions of the confinement layer 31 at the front and rear of the mesa waveguides 49 for the laser oscillator array 13a. Preferably, these grating areas coincide with thinner etched regions of the confinement layer 31, like the filled regions on the sides of mesa structures 49 in FIG. 2, for maximum grating coupling efficiency, and are likewise filled with InAlP cladding layer material 33. A superlattice reflector 27 composed of many thin semiconductor material layers of alternating refractive indices may be formed between the substrate 11 and the n-type cladding and confining layer or layers 29. The superlattice reflector 27 reflects any light that might be diffracted by the second order gratings 17 into the substrate 11 back toward the gratings 17, where the light can be rediffracted back into the waveguides, thereby minimizing diffractive losses of the second order gratings 17. Preferably, the position of the superlattice reflector 27 relative to the gratings 17 is chosen so that the phase of the reflected light is such that it destructively interferes with the p-side radiation, further reducing the diffractive losses. Because of the multi-wavelength nature of this invention, embodiments that employ a superlattice reflector 27 require different layer thicknesses for each of the laser oscillators 13a in the array. Use of the superlattice reflector 27 allows the efficiency of the second order gratings 17 to approach that of first order gratings.

The output of the DBR laser oscillator is injected into a power amplifier array 13b. The optical power received by the power amplifier array 13b from the multi-wavelength laser oscillator array 13a is amplified high power, typically at least 300 mw, without changing the emission wavelengths. Each element of the amplifier array 13b employs a quantum well active region 19 that is disordered to the same degree as the corresponding element in the laser oscillator array 13a. Mesa waveguides 49 used by the amplifier array 13b are collinear with corresponding mesa waveguides 49 in the laser oscillator array 13a. A length of 500 .mu.m for each of the laser oscillator and power amplifier portions is typical. The front end of the amplifier array 13b is preferably antireflection coated with a silicon oxide coating layer 25 to reduce reflectivity to about 0.1% or less. The coating thickness may vary for each element of the array to maintain a 1/4 wavelength thickness for each wavelength of light emitted from the laser-amplifier array 13.

In FIGS. 1 and 3, the device includes ferroelectric frequency doubling material 15 epitaxially grown on the substrate 11 for doubling the frequency (halving the wavelength) of light emitted from at least one of the amplifier array elements. The frequency doubling material 15 is a three layer epitaxial growth consisting of a 0.15 .mu.m thick Ca.sub.x Sr.sub.1-x F.sub.2 buffer layer 41 deposited on the (111) surface of the GaAs substrate, a 0.30 .mu.m thick LiTaO.sub.3 waveguide layer 43 deposited on top of layer 41 and a 0.15 .mu.m thick Ca.sub.x Sr.sub.1-x F.sub.2 cladding layer 45 deposited on top of layer 43. The composition of the Ca.sub.x Sr.sub.1-x F.sub.2 buffer layer 41 is preferably graded to match the lattice constants of the GaAs substrate surface and the LiTaO.sub.3 waveguide layer 43. x=0.43 at the substrate 11. The upper cladding 45 need not be epitaxial, but merely be of lower refractive index than the waveguide core 43 and smooth, such as a glass film material. Other frequency doubling layer structures, such as MgO/BaTiO.sub.3 /MgO deposited on a (100) GaAs substrate surface, could also be used.

The Ca.sub.x Sr.sub.1-x F.sub.2 layers 41 and 45 have a lower refractive index than the LiTaO.sub.3 layer 43, so transverse waveguiding of the light received from the amplifier array 13b takes place, with peak optical intensity in the waveguide layer 43. The mode size for the fundamental frequency is about 0.6 .mu.m (measured at the 1/e.sup.2 points) for the layer thicknesses given above, matching the mode size defined by the layers 29, 19, 31 and 33 in the amplifier array 13b. The lateral mode of the LiTa0.sub.3 waveguides are defined by growing the layers 41, 43 and 45 over a ridge structure 47 formed in the substrate 11. The ridges 47 are formed by selective etching of the substrate 11 at the same time that the p-type InGaAlP confinement layer 31 is etched to form the mesas 49 for the waveguides of the laser-amplifier array 13. As a result, the LiTaO.sub.3 waveguides 23.sub.1 -23.sub.4 are self-aligned with the several light outputs from the amplifier array 13b. The waveguide layer 43 in the regions above the ridges 47 is further aligned with the plane of the quantum well active region 19. Together with the presence of an antireflection coating 25 between the amplifier array 13b and frequency doubling material 15, the self-alignment of the LiTaO.sub.3 waveguides 23.sub.1 -23.sub.4 maximizes coupling (about 90% or higher) into the waveguides 23.sub.1 -23.sub.4, resulting in at least 275 mW of optical power being launched into the frequency doubling material 15.

At least one of the LiTaO.sub.3 waveguides 23.sub.1 -23.sub.4 is periodically poled to cause efficient frequency conversion of the received lightwaves. The conversion is generally most efficient when TM polarized light (with the electric field oriented in the plane of the frequency doubling layer) is received from the laser-amplifier array. For LiTaO.sub.3 material, the appropriate nonlinear coefficient d.sub.33 is -17 pm/V. However, TE polarized light (with the electric field oriented normal to the plane of the frequency doubling material) can also be doubled in frequency, albeit with lower conversion efficiency. For LiTaO.sub.3 material, the appropriate nonlinear coefficient in this case, d.sub.31 is -1.47 pm/V. In FIG. 1, waveguides 23.sub.3 and 23.sub.4 are periodically poled, while waveguides 23.sub.1 and 23.sub.2 are left unpoled and act as low loss waveguides transmitting the fundamental radiation. For example, waveguides 23.sub.1 and 23.sub.2 may emit output beams with two different wavelengths .lambda..sub.1 and .lambda..sub.2 in the range from 630 nm to 1100 nm, depending on the extent of disordering in the quantum well active region 19 in the corresponding portions of the laser-amplifier array 13. Likewise, waveguides 23.sub.3 and 23.sub.4 may emit output beams with two different wavelengths .lambda..sub.3 and .lambda..sub.4 in the range from 430 nm to 550 nm. These latter wavelengths .lambda..sub.3 and .lambda..sub.4 are produced by doubling the frequency of the light emitted from the corresponding portions of the laser-amplifier array 13, such light prior to frequency doubling having fundamental wavelengths in a range from 860 nm to 1100 nm, again depending on the extent of disordering of the quantum well active region 19. While frequency doubling is generally preferred, because second harmonic generation by the ferroelectric material is relatively efficient, other frequency harmonics of the light, such as third or higher harmonics and even half-order harmonics, could also be generated to produce frequency tripling and the like, albeit with lower conversion efficiency.

The LiTaO.sub.3 layer 43 is grown on its crystal orientation aligned so that the c-axis is normal to the growth plane. Domain reversals in the LiTaO.sub.3 material are accomplished by using an electrode pattern to apply an electric field perpendicular to the c-axis at temperatures elevated to near the Curie temperature T.sub.c of the material. For LiTaO.sub.3, T.sub.c =604.degree. C. This temperature is compatible with GaAs processing. By applying the electric field, the domains become aligned and the sign of the nonlinear coefficient for light interacting with ferroelectric material is periodically reversed, enabling quasi-phase matching to be achieved.

For quasi-phase matching the periodicity of the domain reversals is ##EQU1## where p is an odd integer representing the matching order, .lambda. is the wavelength of the fundamental lightwave, and n.sub.w and n.sub.2w are refractive indices of the ferroelectric material at the fundamental and second harmonic (frequency doubled) wavelengths. For third order quasi-phase matching (p=3), L.sub.c is approximately 11 .mu.m for LiTaO.sub.3. For first order quasi-phase matching (p=1), L.sub.c is approximately 4 .mu.m for LiTaO.sub.3. First order quasi-phase matching is preferred whenever fabrication tolerances allow it to be used, because the nonlinear coefficient is three times greater in the first order as compared to the third order. Thus, the doubling efficiency is nine times greater in the first order. The principal factor limiting use of first order quasi-phase matching in frequently doubling materials is the certainty or uncertainty in the domain boundaries. In order to avoid washout of the relatively short poled domains needed for first order quasi-phase matching, the domain boundaries need to be sharp, i.e., the domain reversals need to be abrupt. The abruptness that can be achieved is a consequence of the abruptness of the opposing electric field lines applied to adjacent domains of the ferroelectric material. In order to accommodate abrupt domain reversals, the electrode patterns are deposited to form opposing field lines in adjacent domains, minimizing the electric field gradient. Further, in the three layer Ca.sub.x Sr.sub.1-x F.sub.2 /LiTaO.sub.3 /Ca.sub.x Sr.sub.1-x F.sub.2 structure monolithically integrated on the GaAs substrate, as described above, the waveguide dimension is only 0.6 .mu.m, which is small compared to the coherence length of the laser light. Thus, the abruptness of the domain boundaries can be accurately controlled throughout the entire depth of the waveguide. Together with the improvements in injected power, and the use of DBR laser oscillators 13a in the array to avoid longitudinal mode hops and maintain frequency matching to the resonance of the periodically poled waveguides, the conversion efficiency can be as much as 30%, thereby providing up to 80 mW of output power in the blue/green wavelengths (430 nm to 550 nm).