<title>Ultra high efficiency 1550nm multi-junction pulsed laser diodes</title>

Ultra high efficiency 1550nm multi-junction pulsed laser diodes Jean-François Boucher*a, Ville Vilokkinenb, Paul Rainbowa, Petteri Uusimaab, Jari Lyytikäinenc, Sanna Rantac a Laser Components Canada, 195 Joseph Carrier, Vaudreuil-Dorion, QC, Canada, J7V 5V5; b Modulight Inc, P.O. Box 770, Tampere, Finland, FIN-33101; c Optoelectronics Research Center, Tampere University of technology, P.O. Box 692, Tampere, Finland, FIN-33101 ABSTRACT The 1550nm wavelength region is critical to the development of next generation eye safe military applications such as range finding and friend or foe identification (FOE). So far the relatively low laser external efficiency was a strong limiting factor favoring shorter wavelength diode lasers. We report on the development of a new monolithic multiple junction pulsed laser diode offering an external efficiency of more than one Watt per Amp with high brightness. Peak optical output power of more than 37 Watts has been achieved from a single multi-junction diode laser. Divergence is narrow with less than 35 degrees (FWHM) in the fast axis direction. Starting from an AlGaInAs quantum well laser structure, we show the criticality of the design of InP based tunnel junctions to the growth of the three layer epitaxial monolithic laser. We then report on trenches employed to confine carriers under the contacting stripe and on growth strategies used to decouple the multiple light sources resulting from the multi-junction design. A full set of characterization data is presented concluding with a discussion on performance limitations and their potential causes. Keywords: Laser diodes, pulsed, 1550nm, multi-junction, high efficiency, epitaxial growth, divergence

1. INTRODUCTION Interest and demand for 1550nm pulsed lasers have increased significantly over the last years for range finding application and friend or foe identification (FOE). Various reasons could explain this. More powerful 1550nm lasers open the range of applications to those where 905nm lasers were traditionally used. Eye safety is certainly one of the biggest advantages of 1550nm lasers over 905nm as direct beam emission is much less harmful for retina. Also, for military applications the difficulty and the higher cost to detect 1550nm wavelength constitutes an advantage over lower wavelength lasers on the battlefield as 905nm emission can easily be detected with a device as simple as a silicon digital camera (like the ones present in most cellular phones after removing the filters). The benefit of the usage of 1550nm wavelength comes with a series of technical challenges that are more difficult to overcome than those with shorter wavelength pulsed laser diodes emitting at 905nm. The semiconductors used for 1550nm laser diodes are based on InP ternaries, quaternaries and other lattice-matched closely related semiconductor materials. This class of material is more difficult to grow and to process since there are less high volume applications developed outside the optoelectronics field (unlike GaAs based material with FET and amplifiers for instance). Therefore, any discrepancy from standard laser diode growth and micro-fabrication methods brings additional difficulties. Also the intrinsic external efficiency of InP based laser material is usually a little bit more than one third of the typical 1W/A for GaAs based laser diodes. We needed to take a different approach to obtain efficiency similar to the one of GaAs based lasers. One way to achieve this was known and consists in stacking physically discrete laser chips one on top of the other. This has been used for decades but the light source, which consists of several emission lines separated by the chip thicknesses, does not allow high brightness applications. Another solution consists of improving the intrinsic laser external efficiency. The best 1550nm semiconductor laser we tested gave close to 0.5W/A. While this is remarkable it is still only half of efficiency of a typical 905nm pulsed laser diode. It was also not practical to use it because of intellectual properties reasons. The third approach, which we took for this work consists of superimposing during the epitaxial growth two or three laser active areas within a few microns. This approach had already been followed successfully with 905nm laser structures and takes its origin more than 20 years ago [1], [2]. However it has not been reported with a 1550nm laser structure until this year (2009) as a commercial product. The targeted performance is

Unmanned/Unattended Sensors and Sensor Networks VI, edited by Edward M. Carapezza, Proc. of SPIE Vol. 7480, 74800K · © 2009 SPIE · CCC code: 0277-786X/09/$18 · doi: 10.1117/12.829925

Proc. of SPIE Vol. 7480 74800K-1

to obtain 3 times more output power than with the single laser while keeping the vertical divergence the same as the single laser. All the other parameters including the wavelength spectra and the reliability should remain the same as the single laser. The only exception would be forward voltage which is increased by the presence of two or three junctions instead of one. In section 2 we present the building blocs used to make the multi-junction 1550nm pulsed lasers. We describe basic constituents of the reference laser structure active area which will be superimposed. We also describe the work done in order to develop a tunnel junction making the link between the laser active areas. Section 3 reports on the first double and triple laser structures grown, processed and tested. It shows the degree to which our approach was successful. It also describes the arising problems we were faced with; carrier confinement, reliability, source coupling and competing / circulating modes. It finally describes how we could deal with the first two of these issues. Section 4 reports on how we attempted to fix the source coupling issue in taking a step back and in working with double laser structures. It was mostly successful and allowed us to offer one final product configuration to the market. It also paves the way to the next steps that should resolve the remaining problems and provide us with a device meeting the targeted performances. This is discussed in the conclusive section 5

2. MULTI-JUNCTIONS BUILDING BLOCKS 2.1 Basic 1550nm multi quantum well structure and tunnel junctions A 1550 nm laser structure normally consists of InP and compounds, which can be lattice-matched to InP. In general, achieving high doping concentrations required for tunnel junction operation is not straightforward and for InP materials this tends to be even more challenging than corresponding GaAs materials, which are commonly used for 630-1100 nm lasers. Nevertheless, tunnel junctions based on the InP materials have been demonstrated as reported in [3], [4],[5]. As a multi-junction laser structure includes several laser junctions and at least one tunnel junction, the overall thickness of the structure and therefore also the growth time increase significantly. The growth time and growth temperature must be carefully selected and well-controlled to be able to keep the process conditions stable and in order to avoid the diffusion of the doping atoms, which could lead to the reduction of the active doping within the tunnel junction layers and therefore deteriorate the tunnel junction performance. The laser junction used in this work is based on a multi-quantum well active region made of AlGaInAs compounds. This material combination is used, as it is known that AlGaInAs lasers have better temperature characteristics than conventional InGaAsP lasers [6]. Therefore, AlGaInAs/InP materials were also a starting point for the tunnel junction optimization done in this study. Calculations based on models described in [7], [8], [9] assisted the experimental work. Different variants of the InGaAs/InP tunnel junction test structures were grown with varying tunnel junction layer thickness in the range of 25-75 nm and varying the doping atom concentration from 1×1019 to 8×1019 cm-3. A process including a mesa-etching step through the tunnel junction layers for the current confinement and contact metal depositions were done using normal lithography and device processing steps. In addition to the tunnel junction development work, the performance of the laser junction was optimized while keeping in mind as easy as possible integration to a complete multi-junction laser structure. From the performance point of view the most important parameters are the efficiency of the lasers and the divergence of the beam. From a growth perspective the laser junction needs to be simple enough to minimize the growth time and growth temperature and thereby avoid the possible instability and diffusion problems. The structure of a tunnel junction is very close to that of a normal pn-junction or diode. The main difference is that ptype and n-type layers forming the pn-junction are highly doped, typically >1019 cm-3, which leads to the tunneling effects which is utilized in the device to link the various laser actives. In addition to the good electrical conductivity in reverse bias mode the tunnel junction diodes exhibit a region of negative resistance in the forward bias region [10]. It was found that the tunnel junction layer thicknesses did not have a major influence on the performance of the tunnel junction, whereas the active doping concentration is very critical. Even though theoretically speaking a higher doping concentration results in better tunnel junction characteristics, in practice the compensation effects limit the active doping at very high concentration levels and therefore can have a negative effect on the performance of the tunnel junction. An example of a measured IV curve for a tunnel junction is shown in Figure 1 for tunnel junction with a nominal doping concentration of ~4×1019 cm-3. The characteristics demonstrate that the pn-junction conducts current well in reverse biased mode. Furthermore, a drop in the current as the voltage increases is seen in forward bias mode indicating a

Proc. of SPIE Vol. 7480 74800K-2

negative region of resistance is also seen in the IV curve. These two features are clear evidence of a working tunnel junction.

Figure 1. Measured IV properties of a typical tunnel junction grown.

2.2 Electro-optical results of the reference laser We proceeded to the electro-optical characterization of the reference laser structure as a standalone element. This is the laser structure building block planned to be used as the repeating lasing active area in the multiple laser designs. We found it worth measuring in depth the fundamental electro-optical characteristics of this single laser diode here since they would constitute a comparative basis for assessing the double and triple laser performances. Processing was done using a standard 2 photolithography step technique where metal semiconductor contact is applied only to the emitting stripe. That type of laser operates in gain guided mode. Processed quarters were thinned and N metal contact was in turn deposited. Quarters were then cleaved in 2 different cavity lengths of laser bars i.e. 1000um and 750um which is the reference cavity length used. Bars were then separated into laser chips. 5

Output power (W)

4

3

2

1

0 0

2.5

5

7.5

10

12.5

15

17.5

20

Drive current (A)

Figure 2. PI curves of the reference lasers processed with 100um large stripe and with a cavity length of 1000um.

The lasers were driven with high peak current (10s of Amps) and short pulses (100ns typical). An InGaAs detector attached to the port of an integrating sphere was used on the detection side. Figure 2 shows the PI curve of 100um large stripe lasers having 1000um cavity length. The slope efficiency measured in the linear part of the curve at relatively low peak current was 0.43W/A in average and the threshold current was approximately 900mA. Practically speaking what most users are interested in is the peak power achievable before the curve rolls over and approaches a point of diminishing returns. Since the lasers are usually operated at high peak current a good reference point for measuring PI curves, given the laser dimensions used, is the optical output power of 3.5W at 10A. Lasing wavelength was on target at

Proc. of SPIE Vol. 7480 74800K-3

1551nm. Cone angle measurements told us that we can still collect 20% of the total peak power into a F/5 aperture. When using an imaging card to look at the far field we could see a classical ellipsoidal shape typical of Gaussian divergence curves in both fast and slow axis. The 1.15V forward voltage, at 500mA is similar to the one of regular 1550nm pulsed lasers from Laser Components of similar geometry. Finally driving the laser for 100 hours continuously with pulses of 15A and 150ns width at 3 kHz left the optical output power unchanged.

3. INITIAL DOUBLE AND TRIPLE LASER 3.1 Laser structure and processing The basic concept of a laser structure with multiple junctions is presented in Figure 3. Multiple optically active junctions generating laser light are placed in sequence and separated by a tunnel junction. The tunnel junction is used in between the lasers junctions, since it allows a good electrical conductivity when reverse biased and therefore is essential for good electrical properties of a multi-junction laser to overcome the effect of reversed biased junction that would occur between the laser structures.

Figure 3. A conceptual structure of a laser containing multiple (2) laser junctions in sequence.

Laser structures with 2 and 3 laser junctions were grown and studied as reported in this section. A double laser structure includes one tunnel junction and two laser junctions and a triple laser structure has two tunnel junctions and 3 laser junctions. All the growths were carried out throughout this work were done on InP substrates using an MBE reactor equipped with material sources for AlGaInAs-InP compounds. All the multi-junction structures consist of an active region with multiple compressively (~1 %) strained AlGaInAs quantum wells tuned for emission to about 1550 nm wavelength. The quantum wells are surrounded by AlGaInAs (λ ~ 1165nm) barrier layers and AlGaInAs (λ ~ 950 nm) wave guiding layers. Next to the waveguide layers there are AlInAs layers and 1.57 µm thick InP cladding layers with opposite doping. No intentional doping is used for the layers in the active region, whereas the doping concentration in the cladding layers is varied in the range of 2.5×1017 cm-3 - 1.5×1018 cm-3. In between two consecutive laser junctions, an InGaAs/InP tunnel junction is used. The whole laser structure is completed with a highly doped p-InGaAs contact layer. To study the device performance the laser structures were processed using both a shallow etching process and a deep etching process. The processing was started with a mesa formation using dielectric etch mask and standard lithography steps to define a 100 µm wide active stripe. In the shallow etching process approach only the topmost contact layer was etched next to the active stripe to avoid current spreading in that layer. For the deep etching process, the etching was continued through the tunnel junction layer(s) in order to prevent current spreading in those highly conductive layers. In the case of the triple junction laser the etching needs to progress through both tunnel junctions resulting in an over 8 µm etch depth. A combination of wet etching and dry etching was used to reach the required etch depth. This approach was

Proc. of SPIE Vol. 7480 74800K-4

selected in order to control the etching without excessive undercutting effects. The mesa etching steps were followed by the deposition of dielectric material on the top of the wafer and on the etching sidewalls to provide passivation and isolation. Again standard lithography methods were used to remove the dielectric material from the top of the active stripe and a lift-off technique for forming a metal pattern protecting the etched mesa sidewalls. A p-metal including TiPt-Au multiple layers was evaporated on the topside of the wafer to provide an electrical contact to the device. Next, the wafer was attached to a carrier substrate and the backside of the wafer was thinned down to 100 µm thickness approximately. A 2-step n-metal including Ni-Ge-Au and Ti-Pt-Au layers was evaporated on the thinned backside of the wafer. In between the n-metals, contact annealing for the n-contact was performed. N-metal evaporation concluded the wafer processing. The next steps were bar cleaving and coating. Different cavity lengths (750 µm, 1000 µm and 1200 µm) were cleaved and coated with AR and HR coating. A single layer AR coating with < 5 % reflectivity was used and the HR coating was a multi-pair Bragg reflector providing > 90 % reflectivity around 1550 nm wavelength. Preliminary testing was performed at bar level to confirm the functionality of the devices. 3.2 Electro-optical results From the first iteration of double and triple laser full cycle processing we initially tested four distinct lots of devices from each double and triple laser design: two lots with shallow etching process and two lots with deep etching process. The first lot of each etching method had lasers with 750um cavity length and the second lot had lasers with 1000um cavity length. Each lot was made of 20 lasers in average. Most of the devices showed good lasing behavior. Figure 4 shows typical PI plots of double and triple lasers measured from that first iteration. On the same graphs are also reproduced PI curves of single lasers. The data presented here comes from lasers all having the same geometry with 100um wide contacting stripes and 1000um cavity length. Note also that measurements are from laser diodes processed with the deep etching process which is the method we selected for the subsequent work as explained below. 15

12.5

Output Power (W)

10

Double laser

Triple laser 7.5

5

2.5

Reference laser

0 0

2.5

5

7.5

10

12.5

15

17.5

20

Drive current (A)

Figure 4. Comparison between PI curves of single, double and triple laser from the first iteration processed with 100um large stripes and with a cavity length of 1000um.

From the analysis of the PI curves we could extract the threshold current and the slope efficiency of the lasers. A summary of the data is reproduced in the table 1. In general we can state that slope efficiency of the double laser is approximately 1.5 times that of the single. The slope efficiency of the triples is in turn 2.5 times that of the single. This is less than what we were originally targeting with 2 times and 3 times the slope of the single laser but this is still an acceptable result, especially in the case of the triple. When looking into more details and when comparing shallow etched devices with deep etched devices we can see that the slope efficiency results of the deep etched ones are systematically better than the shallow etched ones. In looking at the threshold current of the shallow etched devices we can see that in going from single active to triple active it is increased by 44%. On the other hand it remains strictly at the same level for the deep etched devices. A close investigation of the threshold of the shallow etched devices showed 2 to 3 distinct sub-thresholds which suggests that the three individual junctions are not turning on at the same current level. This could happen for two reasons: first the three laser actives would be dissimilar. This is not the case since the deep etched devices have a sharp threshold. Second, the three laser active areas are not seeing the same current density. This

Proc. of SPIE Vol. 7480 74800K-5

could happen if the current spread underneath the first laser junction. This second hypothesis was confirmed in looking at the near field of some of those lasers. While the top junction showed a bright emitting line relatively well confined under the contacting stripe, the emitting line of the middle and of the bottom one were extremely pale and extending up to the edges of the chips. We formulated the hypothesis that while allowing current to flow adequately the tunnel junctions were showing a resistance high enough to make the carriers spread all across the laser chip. From that point we decided to focus only on the triple laser processed with the deep etching method. Figure 5 shows one side of the near field of a deep etched triple laser device. In this case trenches are etched down below the bottom tunnel junction. The charge carriers are then physically confined by the trench for the two upper laser actives. The bottom laser active therefore works as the only single laser where current is restricted by the contacting stripe dimension. We proceeded to another 2 iterations of growth, process and characterization with structure and process improvements intended to correct or improve performances.

Reference laser

Double Laser

Triple Laser

Ith (mA) SE (W/A) Ith (mA) SE (W/A) Ith (mA) SE (W/A)

Shallow etched 750um 1000um 700 900

Deep etched 750um 1000um NOT APPLICABLE

0.454

0.427

983

1100

700

933

0.681

0.61

0.703

0.629

1010

1200

700

811

1.007

0.919

1.157

1.134

Table 1. Comparative data of lasers processed with shallow etching and deep etching methods

Figure 5. Partial near field image of a triple laser processed with the deep etching method.

Far field plots in both fast axis and slow axis revealed multiple lobes and peaks. Figure 6 shows examples of that behavior on the second iteration of triple laser wafers. Note that all the triple laser material studied showed this type of profile. When observed with an imaging card it is possible to see a network of dots with diminishing intensity from the center region to the edges. Upon observation and analysis it was found that the pattern of the bright spots in the vertical

Proc. of SPIE Vol. 7480 74800K-6

direction looked like the interference of three coupled sources. Therefore it was deduced that the modes of light emitted by one given source were interacting with the modes of the light emitted by the other two sources. Interpretation was not so obvious for the emission pattern in the horizontal direction. Also as discussed in the next paragraph, the irregular shape of the trenched walls would potentially also induce some effects on the beam. 250

Vertical divergence 200

Intensity (a.u.)

150

Horizontal divergence 100

50

0 -80

-60

-40

-20

0

20

40

60

80

Angle (deg)

Figure 6. Far field curves of a triple laser.

Reliability of the triple lasers was assessed. It was found that the optical output power was decreasing strongly on many devices after only 100 hours approximately. Also the initial testing yield was low with 50% of the device giving lower power. Observation of the side walls with scanning optical microscope revealed very irregular shapes on the sidewalls trenched. Figure 7 shows a typical view of such sidewalls. We can see undercuts resulting from the chemical etch part. They were caused by the selectivity of the etchant used. From observation it is clear that oxide is not conformal all along the wall. Also metal is likely to have been in contact with semiconductor at several places of some chips. This would explain the reduced yield we obtained in various lots i.e. metal semiconductor contact would short the device so that not all three active would be pumped. This would also explain the lower reliability. When a short path is created with time under high peak current pulse conditions the resistance decreases. Therefore for the same voltage applied from the driver more current will flow through the remaining laser actives leading to premature degradation. We have also grown, processed and characterized devices with larger contacting stripes in other full cycle iterations. The largest stripe was 350um wide. As the stripes were larger we started to see competing / circulating modes issues. This is a well known problem with broad area lasers and is often alleviated by techniques that disperse photons and prevent them from being reflected back into the gain region by the side walls of the laser. In the case of the triple lasers the bottom junction acts normally since the arising competing modes are seeing rough sawn edges. However the top and middle junctions are seeing etched walls covered by a thin oxide layer and metal layers which act as a mirror. Therefore it reflects some laser cavity modes and enhances internal circulation. This is obvious on figure 8 where we can see three lasers with competing modes arising at current levels between 20A and 30A, diverting carriers from the longitudinal lasing mode. This was also confirmed by observation of the output power pulse shape in the time domain where the pulses show a pronounced shoulder. Despite this effect a 350um wide stripe triple laser that was less affected gave up to 37W at 70A.

Proc. of SPIE Vol. 7480 74800K-7

Figure 7. Sidewall of a deep etched triple laser from the first iteration. 30

25

Output power (W)

20

15

10

Competing / Circulating modes

5

0 0

10

20

30

40

50

60

Drive current (A)

Figure 8. PI curves of laser with 350um large contacting stripe showing competing / circulating modes.

The wavelength spectrum of triple lasers from the first iteration showed presence of more than one peak. This is highly likely to be due to small differences in quantum well composition and/or thickness from one active area to the other. Calibration works made prior to the subsequent iterations helped us to fix this problem. Figure 9 shows typical wavelength spectra obtained with lasers from the last iteration.

Proc. of SPIE Vol. 7480 74800K-8

-60

Intensity (dB)

-65

-70

-75

-80 1450

1475

1500

1525

1550

1575

1600

Wavelength (nm)

Figure 9. Wavelength spectra of triple lasers from third iteration.

4. BACK TO BASICS As indicated by the results shown in the previous section the beam profile of the multi-junction lasers differed from normal laser beams. The optical modes coming from the different laser junctions are interacting and causing the vertical beam to split into two main intensity peaks plus satellites. There are several alternatives to improve the beam shape. The first one consists in moving the laser junctions further away from each adding a spacer layer. This approach is relatively straightforward, but obviously adds to the overall thickness of the structure and hence complicates the manufacturing process. The second alternative is to change the wave guiding properties of the laser structure and thereby to minimize the interaction. However, this approach would most likely result in an undesired effect in the width of the vertical beam and is therefore not preferred. Finally the third option is to make the coupling between the modes stronger and take advantage of it to produce a normal vertical beam profile. Due to the obvious advantages from the manufacturing point of view this last option was studied to some extent; however, this approach also results in some disadvantages related to the overall device performance and it is not considered a viable solution for the problem. Therefore, the first approach was found to be the best approach for improving the beam shape. The concept was studied in more details and double laser structures, where the distance between the laser active regions was doubled, were grown with only minor design variants between each, processed and characterized. 250

200

Intensity (a.u.)

150

Vertical divergence 100 Horizontal divergence

50

0 -80

-60

-40

-20

0

20

40

60

80

Angle (deg)

Figure 10. Far field curves of double laser with increased spacing between actives.

Proc. of SPIE Vol. 7480 74800K-9

From the far field measurement we found that increasing the spacing between the actives on the double immediately fixed the multi lobe problem of the emission beam in the vertical direction in the far field regime. The spacing selected was sufficient to have the two light sources acting independently. The absence of interaction between them implies that the far field curves are now the sum of the two individual light sources. This is depicted in figure 10. From that figure we can also see that the emission beam in the horizontal direction in the far field regime is also single lobe and typical of what we would observe with a single laser. This can be partly due to the decoupling of the light source. This could also be due to the processing work on the deep etching process as it was believed to cause issues to the emission beam also. The work consisted of improving the etch control in eliminating the severe and unpredictable undercut effects. We were able to have straighter walls with no more of those enclosures that were difficult to coat in a conformal way. Figure 11 shows the sidewalls shape we were able to obtain. We also took the opportunity of that process work to introduce lift off metal process so that no more metal would be present into or around the trenches. Both the new trenching method and the metal lift off process allowed us also to fix the reliability problem. Several lots of double laser have been life tested for hundreds of hours with no degradation. This formed the basis for the first commercial device; a double junction with a 50um aperture. Competing modes remains an issue for lasers with larger aperture.

Figure 11. Sidewalls of a deep etched double laser with improved method.

5. CONCLUSION Overall this development program has been a success. We have demonstrated the concept of growing three laser active regions separated by tunnel junctions. Key results here are the slope efficiency which has three times the value of the reference laser and the current threshold that remains essentially constant irrespective of the number of active areas. We have been able to find a way to confine physically the charge carriers through a deep etching process so that current spreading is limited to an acceptable level on the bottom junction. Refining the etching process has rendered highly reliable the final devices when life tested. We have also solved the multi lobe problem encountered with the divergence in the horizontal direction or slow axis. In order to address the multi lobe issue of the vertical divergence we reverted to double laser structures with a bigger separation between the two active areas to make sure that the sources were decoupled. Figure 12 shows PI curves of this device. On the same plot are also shown PI curves of reference single laser and triple laser. The 1550nm double active laser diode with a 50um wide aperture is now available as a viable commercial product.

Proc. of SPIE Vol. 7480 74800K-10

10

Triple laser Double laser

Optical output power (W)

7.5

5

Single laser

2.5

0 0

5

10

15

20

25

Drive current (A)

Figure 12. PI curves of single, double and triple lasers with 50um aperture.

6. FUTURE WORK Current lasers are limited to double junctions with a 50um aperture and there are several challenges remaining to extend the technology to larger source sizes with higher power while maintaining Gaussian beams. Research is concentrating on frustrating the competing / circulating modes inside the cavity by modifying the etched interfaces adjacent to the laser cavity. The technique employed to prevent the coupling of the lasers in the double laser will be optimized for the triple laser devices thereby significantly extending growth times to concerns over diffusion of dopants in the layers. Work will continue on the basic structure to improve efficiency and further reduce divergence in the fast axis

REFERENCES [1] [2] [3]

[4]

[5]

[6]

[7] [8] [9] [10]

Van der Ziel J.P., Tsang W.T., “Integrated multilayers GaAs lasers separated by tunnel junctions’’, Appl. Phys. Lett., 41(6), p.499-501 (1982) Jain F.C, Laser diode assembly with tunnel junctions and providing multiple beam, United States Patent 5212706, (1993) Vilela M.F., Medelci N., Bensaoula A., Freundlich A. and Renaud P., “First time demonstration of InP p++/n++ tunnel junction,” IEEE Photovoltaic Specialists Conference - 1994, 1994 IEEE First World Conference on Publication Date: 5-9 Dec 1994, Vol. 2, 1946-1948 (1994) Freundlich A., Vilela F., Bensaoula A. and Medelci N., “Very High Peak Current CBE Grown InGaAs Tunnel Junction for InP/InGaAs Tandem Cells Fabricated on InP, GaAs and Si Substrates,” Photovoltaic Specialists Conference, 1993, Conference Record of the Twenty Third IEEE Volume, Issue , 10-14, 644 – 649 (1993) Mehta M., Feezell D., Buell D.A., Jackson A.W., Coldren L.A. and Bowers J.E., “Electrical Design Optimization of Single-Mode Tunnel-Junction-Based Long-Wavelength VCSELs,” IEEE Journal of Quantum Electronics, Vol. 42, No. 7 (2006). Zah, C.-E., Bhat, R., Pathak, B.N., Favire, F., Lin, W., Wang, M.C., Andreadakis, N.C., Hwang, D.M., Koza, M.A., Lee, T.-P., Wang, Z., Darby, D., Flanders, D., and Hsieh, J.J.: ‘High-performance uncooled 1.3-mm AlxGayIn1-xyAs-InP strained-layer quantum-well lasers for subscriber loop applications’, IEEE J. Quantum Electron, Vol. 30, (2), 511–523 (1995). Beji et. al., phys. stat. sol. (a) 183, No. 2, 273–279 (2001) Bedair et. al., Photovoltaic Specialists Conference, 2000. Conference Record of the Twenty-Eighth IEEE, Volume , Issue , 2000 Page(s):1154 - 1156. Li et. al., Semicond. Sci. Technol. 5 (1990) 408-413 Esaki L., “Long Journey into Tunneling,” Review of Modern Physics, Vol. 46, No. 2, 237-244 (1974).

Proc. of SPIE Vol. 7480 74800K-11