Health & Medicine

High-power room temperature emission quantum cascade lasers at /spl lambda/=9 /spl mu/m

Description
High-power room temperature emission quantum cascade lasers at /spl lambda/=9 /spl mu/m
Published
of 9
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
  1430 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 41, NO. 12, DECEMBER 2005 High-Power Room Temperature Emission QuantumCascade Lasers at        m Clément Faugeras, Sébastien Forget, Elizabeth Boer-Duchemin, Hideaki Page, Jean-Yves Bengloan, Olivier Parillaud,Michel Calligaro, Carlo Sirtori  , Member, IEEE  , Marcella Giovannini, and Jérôme Faist  , Member, IEEE   Abstract— We present two different techniques for processingInP-based    m quantum cascade lasers which improve thethermal dissipation in the device. The first process is based on hy-drogen implantation creating an insulating layer to inject currentselectively in one part of the active region. The second process usesa thick electroplated gold layer on the laser ridge to efficiently re-move the heat produced in the active region. Each process is de-signed to improve heat evacuation leading to higher performancesof the lasers and will be compared to a standard ridge structurefrom the same wafer. We give evidence that the process of protonimplantation, efficient in GaAs based structures, is not directly ap-plicable to InP based devices and we present a detailed analysis of the thermal properties of devices with an electroplated gold thicklayer. With these lasers, an average power of 174 mW at a dutycycle of 40% has been measured at 10 C.  Index Terms— High power emission, InP based device, quantumcascade laser. I. I NTRODUCTION T HE performance of quantum cascade lasers (QCLs) [1]is continually improving [2], [3]. These unipolar devices emitting in the midinfrared have great potential for applicationssuch as chemical sensing and free-space communications [4],[5]. These applications will benefit greatly from room tempera-ture, high power, continuous-wave (CW) sources, and so, muchof current QCL research is working toward this goal.The main difficulty on the route to high power and CW op-eration at room temperature is the large amount of heat thatmust be dissipated in the device. Typical operating voltages forInP-based QCLs are 7–10 V and threshold currents may be of  Manuscript received August 9, 2005; revised August 18, 2005. This work supported in part by EU FP6 Grant STRP 505642 “ANSWER.”C. Faugeras and C. Sirtori are with the Pôle Matériaux et Phénomènes Quan-tiques, Université Paris VII, 75251 Paris, France (e-mail: clement.faugeras@thalesgroup.com; carlo.sirtori@thalesgroup.com).S. Forget was with the Pôle Matériaux et Phénomènes Quantiques, Uni-versité Paris VII, 75251 Paris, France. He is now with the Laboratorie dePhysique des Lasers, Université Paris 13, F9343 Villetaneuse, France (e-mail:forget@galilee.univ-paris13.fr).E. Boer-Duchemin was with Thales Research and Technology, 91404 Orsay,France. He is now with the Laboratorie Photophysique Moléculaire, Univer-sity Paris 11, 91404 Orsay, France (e-mail: elizabeth.boer-duchemin@ppm.u-psud.fr).H. Page was with Thales Research and Technology, 91404 Orsay, France.He is now with Alpes Lasers, CH2000 Neuchâtel, Switzerland (email:hideaki.page@alpeslasers.ch).J.-Y. Bengloan, O. Parillaud, and M. Calligaro are with Thales Research andTechnology, 91404 Orsay, France (e-mail: jean-yves.bengloan@thalesgroup.com; olivier.parillaud@thalesgroup.com; michel.calligaro@thalesgroup.com).M. Giovannini and J. Faist are with the Institute of Physics, Universityof Neuchâtel, 2000 Neuchatel, Switzerland (e-mail: marcella.giovannini@unine.ch; jerome.faist@unine.ch).Digital Object Identifier 10.1109/JQE.2005.858797 the order of 1 A at room temperature. If we now consider thatthe best wall-plug efficiencies reported are of the order of a fewpercent,almostalltheinjectedelectricalpowerisconvertedintoheat in the device. This means that about 10 W of power mustbe dissipated to prevent the active region from heating, whichwould lower the quantum efficiency and could block CW oper-ation. In standard processed lasers, the ridge is surrounded bya thin insulating layer (typically SiO or Si N ), and then bya thin metallic contact layer, leading to poor evacuation of theheat generated inside the active region.The most common answer to thischallenge has been the real-ization of buried heterostructures (BH) [3], [6]. In these lasers, the active region is surrounded by semiconductor material withhigh thermal conductivity (unlike the case of standard ridgewaveguide) so that heat may be easily dissipated in all direc-tions. A drawback of BH lasers, however, is the extra regrowthstep that is necessary. Recently, two alternatives have been pro-posed. They combine the simplicity of the standard ridge wave-guide with the superior heat dissipation characteristics of theburied heterostructure. These techniques are the selective cur-rent injection technique [2], [7] and the use of a thick electro- plated gold layer [8], [9]. We have applied both concepts to the same initial wafer to compare the performance obtained withthe same active region but with different processing designs.In terms of emitted power, the best result reported so far atm [10] is an emission of 150 mW measured at roomtemperature at 6% of duty cycle. In this paper, we will see howthe two techniques we have applied lead to equivalent and evenhigher performances with better thermal properties.The paper is structured as follows. Section II gives a descrip-tion of the structure of the laser. In Section III, we compare re-sults obtained on standard processed lasers to results obtainedon implanted lasers to illustrate the selective current injectiontechnique in InP-based lasers. Section IV presents the resultsobtained on a third set of lasers, which have been covered witha thick electroplated gold layer to optimize heat evacuation. Wediscuss in Section V the thermal behavior of the lasers with thethick gold layer and we describe how to determine the thermalresistance of the device. We then present our conclusions.II. L ASER  S TRUCTURE The active region of the devices is based on a four quantumwell double phonon resonance design [6], [11]. The band dia- gram of this structure is shown in Fig. 1 with an applied electricfield of 40 kV/cm. A period of the active region is composedof four quantum wells, one narrow quantum well/narrow injec-tion barrier ensuring efficient injection into the upper state level 0018-9197/$20.00 © 2005 IEEE  FAUGERAS  et al. : HIGH-POWER ROOM TEMPERATURE EMISSION QUANTUM CASCADE LASERS AT m 1431 Fig. 1. Schematic of the conduction band diagram of one period of thefour quantum-well active region together with the moduli-squared relevantwave functions. The laser transition involves levels 1 and 2 while the energyseparation between levels 2 and 3 and between levels 3 and 4 is equal toa optical phonon energy. From the injection barrier, the layer sequence innanometers is (bold layers are InAlAs and InGaAs layers are in roman)4.0/1.9/0.7/5.8/0.9/5.7/0.9/5.0/2.2/3.4/1.4/3.3/1.3/3.2/1.5/3.1/1.9/3.0/2.3/2.9/ 2.5/2.9.Fig. 2. Calculated refractive index and mode pro fi le of the laser waveguide at       m. The overlap between the electric  fi eld and the laser active region is 0     and the calculated losses are         cm . (level 1 in Fig. 1) and three larger quantum wells creating threeelectronic levels with an energy separation equal to the opticalphonon energy (36 meV). The laser transition involves levels 1and 2 in Fig. 1. Together with the bound to continuum [12] ac-tive region, this structure has proved to be one of today ’ s bestmidinfrared active region designs, allowing rapid electron es-cape from the lower level of the optical transition to achievethe population inversion necessary for laser emission. The 35period lattice matched InGaAs – InAlAs well and barrier layerswere grown on a doped InP substrate (100 m) by molecularbeam epitaxy (MBE) while metalorganic vapor-phase epitaxy(MOVPE) was used to grow the top 3.3- m InP cladding.The calculated one-dimensional refractive index and the op-tical intensity of this laser structure are presented in Fig. 2. Onboth sides of the 35 period active region, the waveguide is com-posed of 300 and 240-nm-thick GaInAs layers that enhance the Fig. 3. Schematic of the device. (a) Standard processing using a SiO layer(vertical hatched layer) for electrical isolation, the horizontal hatched regionrepresents the active region. The metallic contact layer is represented bythe dark grey layer. (b) Laser ridge of width    is de fi ned by two trenchesetched through the active layer. The injection window of width    is de fi nedby semi-insulating layers created via H implantation. The shaded regionrepresents the laser mode. average refractive index difference between the core and thecladding regions of the waveguide. This results in a calculatedoverlapfactoras highas62.1%and opticallosses of7.38 cm .Initially, QCLs were developed in the GaInAs – AlInAs mate-rialsystem,grownonanInPsubstrate[13].ThechoiceofInPas a substrate for high power emission is appealing. Its advantagesare numerous and include: 1) a low refractive index ensuringhigh optical overlap factors; 2) a high electrical conductivitymakingheavilydoped(and,thus,lossy)layersunnecessary;and3) a high thermal conductivity ensuring enhanced heat evacua-tion with respect to other material systems [1].III. S ELECTIVE  C URRENT  I NJECTION In this section, we will present and compare results obtainedon two different types of samples. For the  fi rst set of lasers,called standard processed samples, wet chemical etching wasused to de fi ne the ridge. A SiO dielectric layer was used toelectrically insulate the sides of the device from the top contact,asisshowninFig.3(a).Theridgewidthwas24 mandthelasercavity lengths were 1 or 2 mm. For the 2-mm lasers, both frontand back facets were cleaved thus resulting in a re fl ectivity of 23 [14],[15],whilethebackfacetofthe1-mmsampleswas coated with a high re fl ectivity (HR) mirror (100 nm Al O /2nm Ti/100 nm Au/100 nm Al O ) resulting in a re fl ectivity of 95 at m[14]. The samples were mounted epilayerup in order to avoid any short circuits due to over fl owing of theIn solder on the facets of the laser. Throughout this paper, thisprocessdesign will be used asa reference to evaluatethe perfor-mance of the two other process designs that will be presented.Fig. 3(b) shows the structure of the second set of lasers,named implanted lasers. Deep etched trenches de fi ne a standardridge structure, which, due to the refractive index change,con fi ne the optical mode of the laser (shown as a shadedellipse). Proton H implantation was used to produce asemi-insulating layer which de fi nes a narrow injection channel  1432 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 41, NO. 12, DECEMBER 2005 Fig. 4. Pulsed measurements of optical power and voltage as a function of current for a sample realized using standard processing. At 300 K,    is 0.68A,    is 2.8 kA/cm ,    is 655 mW/A, and maximum peak power is   550 mW. The ridge width is 24    m, with a cavity length of 1 mm, and therear facet is HR coated. in the ridge. Using this method, the width of the electricallypumped area of the device is reduced with respect to the ridgewidth . Current is then selectively injected into the center of the ridge.Typical implantation energies and doses were 600 keVand 2 10 ions/cm , respectively. With these parameters,the implanted layer is 6 m under the surface with a width of 1.14 m. The implanted ions then lie mostly in and below theactive region. After the implantation process, deep trencheswere formed by wet chemical etching, ensuring smooth wallsand, thus, avoiding optical scattering. As it is demonstrated in[2] for the GaAs – AlGaAs material system, threshold currentoptimization occurs when the ratio of the electricallypumped width to the ridge total width ranges from 0.5 to0.7. We then chose the following parameters for the device, aridge of m, an injection width of m witha laser length of 1 mm. Samples were mounted epilayer downon high quality copper bases that had been pre-evaporated with3 – 5 m of indium solder.For all experiments, the emission was collected with a ger-manium lens and focused onto the detector using aZnSe lens . The collection ef  fi ciency of this system isestimatedtobe60%.Atroomtemperatureacalibratedmercury – cadmium – tellurite (MCT) detector was used for pulse measure-ments and a thermopile for the average power measurements.All measurements are reported without any corrections due tothe collection ef  fi ciency or to the transmission of the lenses.The results for the  fi rst set of samples, processed with a SiOinsulating layer, with cavity of 1 mm and with a HR mirror, areshown in Fig. 4. These reference samples were not optimizedin terms of heat dissipation, and therefore, at high temperature,only pulsed measurements at a low duty cycle were possible.The pulsewidth used was 50 ns with a 1-kHz repetition rate,yielding a duty cycle of 0.005%. This low duty cycle value in-sures that the thermal properties (material and mounting) of thedevice do not in fl uence the measured characteristics. In partic-ular, it is possible in this regime to compare devices with dif-ferent ridge width because optical losses are not increased asthe width is bigger than the wavelength. The interaction be-tween the optical mode and the walls of the ridge, covered withan insulator ( SiO or Si N ) which is very lossy at these fre-quencies, is then minimized [16]. The active volume increases Fig. 5. Pulsed measurements of optical power and voltage as a function of current for a proton implanted sample. At 300 K,    is 0.34 A,    is 3.4kA/cm ,    is 620 mW/A, and maximum peak power is    235 mW. Theridge width is 14    m,        m with a cavity length of 1 mm, and the rearfacet is HR coated. with the width of the ridge as does the injected power. This pa-rameter will then in fl uence the characteristic of the laser whenthermal effects are considered. Fig. 4 shows peak optical powerand voltage as a function of current for  –  K. At300 K, the threshold current is 0.68 A, corresponding toa threshold current density of 2.8 kA/cm , the slope ef- fi ciency is 655 mW/A and the maximum peak poweris 550 mW. The characteristic temperature is de fi ned bythe empirical relation between the current threshold density andtemperatureas . was determined by fi ttinganexponentialtotheevolutionofthethresholdcurrentdensitywithtemperature, yielding a value of 180 K. Wall-plug ef  fi ciency,de fi ned as the ratio of the peak optical power to instantaneouselectrical power, is of 3% at 300 K.The advantages of the reduction of electrically pumped re-gion are threefold. Compared to a ridge device, thepumped region is decreased and this causes a slight decreasein the optical mode overlap with the electrically pumped re-gion and a comparable increase in the threshold cur-rent density . However, because of the area reduction, thethreshold current decreases by an amount proportional tothe ratio . The  fi rst advantage is, thus, clear: by stronglydecreasing the threshold current, the heat dissipated in the de-vices is similarly reduced. The second advantage is that thereis now semiconductor material with a high heat transfer coef  fi -cient surrounding the pumped section of the active region, thusimproving heat evacuation. The third advantage is that, by de-coupling the electrical and the optical con fi nement, the funda-mental optical mode is preferentially pumped, insuring singlemode emission [16].Pulsed measurements of optical power versus current and of voltage versus current of these devices at the same duty cycle asfor the reference lasers, are shown in Fig. 5. At 300 K,A, kA/cm , mW/A, and max-imum peak power is 235 mW.Samples processed using implantation operated CW fromliquid nitrogen to room temperature. Fig. 6 shows the opticalpower as a function of current for a typical implanted sample inCW operation on a Peltier cooler for 0, 10 and 20 C.Maximum power at 20 C was 20 mW, with values of 3.3kA/cm and 110 mW/A for and , respectively. Theinset of Fig. 6 shows the room temperature emission spectrum  FAUGERAS  et al. : HIGH-POWER ROOM TEMPERATURE EMISSION QUANTUM CASCADE LASERS AT m 1433 Fig. 6. CW optical power as a function of current for an implanted laser(      ,      , with a cavity length of 2 mm and the rear facet HRcoated) on a Peltier cooler:      , 10 and 20 C. Maximum power is20 mW, current threshold density         kA/cm and slope ef  fi ciency     mW/A at 20 C. Insert: Emission spectrum near threshold atroom temperature.Fig. 7. Threshold current in pulsed operation as a function of temperaturefor a standard processed laser of 1-mm HR coated (open squares) and aproton implanted laser of 1-mm HR coated (black dots). The characteristic   parameters are indicated. Inset: Schematic of the laser facet illustratingthe parasitic current channel (black arrows) at high temperatures owing to thebreakdown of the implanted layer. of this device measured with a high resolution (0.125 cm )Fourier transform infrared (FTIR) Nicolet spectrometer. Theemission spectrum is centered at 8.91 m. At 77 K, a similardevice but with cleaved facets in CW operation showed athreshold current density of 0.8 kA/cm , a slope ef  fi ciency of 475 mW/A with a maximum CW output power of 350 mW.The results presented in Fig. 6 for the implanted samples arecomparable to those obtained from a buried heterostructure de-vice realized in the same material [6]. This is signi fi cant asit means that the simpli fi ed process of selective current injec-tion produces devices with performances similar to BH devices.There is, however, a considerable discrepancy in the measuredvalues of characteristic temperature (80 K for the proton im-planted sample and 187 K for the standard processed sample) ascan be seen from the temperature dependance in pulsed modeof the threshold current for both processing presented in Fig. 7. Fig.8. Semilogarithmplotoftheresistanceofthestructuremeasuredbetweentwo different nonseparated devices as a function of the inverse temperature.The deduced activation energy is 75 meV. Inset: Schematic of the sample. Theimplanted layer is shown in black and the active region in gray. The surface is2 mm 2       m. As one can see, threshold current densities in the case of the im-plantedlaserandforalltemperaturesbelow240K,aresimilartothose measured on standard processed lasers. At room temper-ature, the threshold current densities suddenly increase for theimplanted sample, leading to a small value of the characteristicparameter. This is contrary to expectation, as the two devicesshould show the same temperature dependance as they are pro-cessed from the same material. This difference was not found inGaAs-based devices [2], and may be explained as follows.While the process of proton implantation is very ef  fi cient forGaAs-based lasers [2], it breaks down at high temperatures forInP-based devices. Proton implantation in GaAs creates deeplevel centers which trap carriers, rendering the layer semi-in-sulating [17]. Since the defect states are near the middle of theband gap, the charge carriers remain trapped, even at elevateddevice operating temperatures, and the layers remain noncon-ducting. On the other hand, in n-InP, the created defects leadto the pinning of the Fermi level in the upper half of the bandgap, thus allowing thermal activation of trapped carriers in theconduction band [18]. We explain the difference between thethreshold values for implanted and nonimplanted samples atroom temperature by the thermally activated breakdown of theimplantation layer. This causes a leakage current and a strongtemperature dependance.The breakdown of the insulating properties of the proton im-planted layers at high temperatures in the ridge of the laser andalso outside of the trenches (as shown in the inset in Fig. 7), canbeclearlyseenbycomparingthe  –  characteristicsofthenon-implantedandimplantedsamplesshowninFigs.4and5.Fortheimplanted lasers, three effects are apparent while the tempera-ture is varied: a  fl attening of the voltage knee, a general drop of the voltage, and a shift toward higher currents of the negativedifferential resistance point. In this speci fi c case, the electricalleaks are due to current  fl owing from the top contact to the sub-strate outside of the ridge of the laser.To clarify this breakdown of the insulating properties of theimplanted layer, we show in Fig. 8 the evolution of ln(Resis-tance) as a function of the inverse temperature measured in theregion between two different lasers that is entirely implanted  1434 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 41, NO. 12, DECEMBER 2005 Fig. 9. Microscope image of the facet of a laser with a thick electroplated goldheat spreader. Electrical isolation is provided by a SiO layer (black solid line).Note that the gold layer also  fi lls the semicircular double trenches. (seeinset inFig.8).Upto200K,theresistance oftheimplantedlayer decreases slightly while for higher temperatures, we ob-serve an important decrease of the resistance with an activationenergy of 75 meV. This means that it is not possible with protonimplantation to insulate InP based structures properly, and thus,for temperatures above K, leakage currents appear.This section proves that room temperature CW operation of m emitting QCLs is possible even without having to re-grow an InP planarization layer necessary for the realization of BH [6]. CW operation was observed up to 20 C with an outputpower of 20 mW. At even higher temperatures, the electrical in-sulation induced by the implanted ions breaks down, leading topoor electrical characteristics of the device. This is due to thetype of defects created by proton implantation in n-InP. Thus,this technique is not directly applicable to InP-based devicesand cannot be exploited, in this material system, as an alter-native to buried heterostructures. A solution to overcome thisproblem could be the use Fe-doped InP for which this isola-tion breakdown should not occur as the defect induced by Fe inthis material is known to be in the middle of the semiconductorband gap. However, as Fe ions cannot be deeply implanted inInP-based devices, it should be incorporated as a dopant duringa growth procedure.IV. E LECTROPLATED  G OLD In this section, we will present results obtained on lasersprocessed from the same initial wafer. As for the referencelasers, the laser ridge of width 24 m is de fi ned by wet chem-ical etching. Instead of proton implantation, electrical isolationis ensured by a 500-nm-thick SiO layer on the top surfacerepresented as a black solid line in Fig. 9. A window is openin this insulating layer above the ridge for electrical contact.An initial Ge – Au contact is then deposited. A 20- m layer of electroplated gold is then deposited on top of the  fi rst metalliccontact to ef  fi ciently remove the heat generated by the powerdissipation in the active layer [3], [8], [9]. As can be seen in Fig. 9, the thick electroplated gold layeralso  fi lls the trenches. Heat generated inside the active regioncan thus be evacuated in all directions and because of thehigh thermal conductivity of gold, spreads ef  fi ciently in thewhole metal layer. After growth and processing, the lasers werecleaved into 1- and 2-mm-long cavities and the rear facets of the lasers were gold coated to realize a high re fl ectivity mirror.The lasers were then soldered epilayer down or up on a goldcoated copper mount, pre-evaporated with 3 – 5 m of indiumsolder. Fig. 10. Optical power as a function of the current for temperatures between240and300 Kfor2-mm-long(solidlines) and1-mm-long(dashedlines)lasers.Voltage as a function of the current measured at 300 K for a 2-mm-long laser.Both lasers have the rear facet HR coated.Fig. 11. Threshold current density    and slope ef  fi ciency as a function of temperature for a 2-mm-long laser with rear facet HR coated (open and black dots), and for a 2-mm-long laser with cleaved facets (open and black squares).In both cases, the    parameter is 165 K.    is 247 K for the cleaved facet laserand 376 K for HR coated laser. Fig.10showstheopticalpowerasafunctionofthecurrentfordifferent temperatures and for 1 (dashed lines) and 2 -mm-long(solid lines) lasers, as well as a typical - curve measuredat 300 K for a 2-mm device. The pulsewidth used was 100 nswith 5 kHz repetition rate, yielding a duty cycle of 0.05%. Themaximum output power measured for the 2-mm (1-mm) laserswas 1.40 (0.96) W at 77 K, with still more than 900 mW (580mW) at 300 K. The wall-plug ef  fi ciency for the 2-mm deviceswas 3.7% at 200 K and 2.4% at 300 K. The threshold currentdensityin pulsedmodeat300Kwas2.5 kA/cm (3.65kA/cm )forthe2-mm-long(1-mm-long)lasers.Thecorrespondingslopeef  fi ciency is 440 mW/A (530 mW/A).Fig. 11 shows the evolution of the threshold current densityand of the slope ef  fi ciency measured for two 2-mm-long lasers,with HR coating (open and black circles) and with untreatedcleaved facets (open and black squares). Similar experiments(notshownhere)havealsobeenperformedon1-mm-lasers.Thecharacteristic temperatures are 160 and 180 K for the 2- and1-mm-long lasers. These values are in agreement with previousmeasurements on the standard processed lasers of Section III.The threshold value for the HR coated device, in this range of 
Search
Tags
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x