By emitter degradation analysis of high power diode laser bars Eric Larkins and Jens W. Tomm Outline Part I I. 1. Introduction I. 2. Experimental Techniques I. 3. Case Study 1: Strain Threshold for Increased Degradation I. 4. Case Study 2: Thermal Runaway Mechanism I. 5. Summary 1
By emitter degradation analysis of high power diode laser bars Eric Larkins and Jens W. Tomm Outline Part 2 II. 1. Introduction II.1.1Strain measurement in semiconductors and devices II.1.2Detection of defects in semiconductors and devices II. 2. Observation of defects caused by packaging induced strain II. 3. Observation of strain caused by defects II. 4. The interplay between strains and defects during device operation as monitored by by emitter degradation analysis II. 5. Conclusions 2
1. Introduction Compared to single emitter laser chips, it is well known that multiemitter laser bars degrade faster Use aging data from real single emitters Model a virtual bar consisting of 25 of these identical single emitters Compare to real aging data of a laser bar of the same batch Real bar Virtual bar Device Single Emitter Virtual Bar Real Bar Time To Failure ~39,000 hours ~27,000 hours ~7,900 hours Factor of 3.4 difference Virtual bar model and data courtesy of M. Oudart, Alcatel Thales III V Lab Quantum Well Laser Array Packaging eds. J.W. Tomm & J. Jimenez, pp. 235 239 (2007) 3
1. Introduction The virtual bar model neglects packaging induced strain, current competition and temperature gradients The virtual bar has a lifetime more than 3 times that of the real bar The more rapid degradation of laser bars as compared to single emitters appears to be related to a combination of: Increased and inhomogeneous packaging induced strain Current competition between emitters Larger and inhomogeneous thermal stress during operation Less effective heat spreading and thermal crosstalk Often little is known about the operating conditions and degradation behaviour of the individual emitters This can be studied using By emitter analysis 4
1. Introduction What is by emitter analysis? By emitter analysis is a methodology for studying the behaviour and degradation of individual emitters, which are operating in the context of a parallel connected array sharing the same physical substrate and heatsink 5
2. Experimental Techniques Conventional Aging Experiments (1) 1. Constant Power Mode 2. Constant Current Mode Catastrophic Operating Current Rapid Gradual Output Power Rapid Gradual Catastrophic Time Failure commonly defined as 20% rise in operating current Time Failure commonly defined as 20% drop in output power 6
2. Experimental Techniques Conventional Aging Experiments (2) 1. Constant Power Mode 2. Constant Current Mode Increases in current signify degradation Decreases in power signify degradation 7
2. Experimental Techniques Conventional Aging Experiments (3) P I Characterisation (typically performed before and after each aging test) Important figures of merit for the full bar can be determined: Power (W) 1) Threshold current 2) Slope efficiency Current (A) Varying efficiencies for different bars (0.62W/A & 0.78 W/A in examples shown) Threshold currents for different bars (11A & 16A in examples shown) 8
2. Experimental Techniques Conventional Aging Experiments (4) Spectral Characterisation (typically performed before and after each aging test) 980.24 nm 986.12 nm Important figures of merit for the full bar can be determined: 8.5 nm 9.7 nm 1) Peak wavelength 2) Spectral width Clear differences in peak wavelengths and spectral widths can be seen for different bars 9
2. Experimental Techniques Quantities Measureable at the By Emitter Level Many quantities relating to an individual emitter within a laser bar can be measured in situ Power Emission spectrum Near field pattern Bandgap Defect level However, it is NOT possible to determine the current of each individual emitter (parallel connected array) The true threshold currents and slope efficiencies of individual emitters can therefore NOT be determined 10
2. Experimental Techniques Apparent Threshold Current & Apparent Efficiency However, to compare the performance of individual emitters two apparent quantities can be defined Apparent threshold current Apparent slope efficiency OUTPUT P OW ER True Slope Efficiency EMITTER OUTPUT P OW ER Apparent Slope Efficiency ~ Bar η ext # Emitters True Threshold Current EMITTER CURRENT Apparent Threshold Current BAR CURRENT Single emitter laser Isolated emitter in laser bar 11
2. Experimental Techniques Measuring Emitter Beam Parameters (1) (Power, emission spectrum, near field pattern) Motion Control Unit 6 axis Positioner Sample Holder Current Supply Microscope Objective Beamsplitter Monochromator ND Filters ND Filters Lenses CCD Camera Lenses CCD Camera Simultaneous measurement of individual emitter near field images and EL spectra Lab PC IEEE 488 PC Serial Bus Optical Path SCSI 12
Near field pattern images of individual emitters EL Spectra of individual emitters 2. Experimental Techniques Measuring Emitter Beam Parameters (2) (Power, emission spectrum, near field pattern) EL Intensity, a.u. P I Bar curves for individual emitters 3.5x10 8 DIL408 (after 500h ageing) 3.0x10 8 2.5x10 8 2.0x10 8 1.5x10 8 1.0x10 8 5.0x10 7 #3 #23 #24 #2 #25 0.0 4 6 8 10 12 14 16 Current, A #1 Apparent threshold currents & apparent efficiencies for individual emitters defects self heating Relative emitter powers across a bar 10 9 DIL408 (after 500h ageing) λ shift as a function of bias current for individual emitters band filling (blue shift) self heating (red shift) EL Intensity, a.u. 10 8 10 7 10 6 5 Amps 7.5 Amps 10 Amps 15 Amps 0 5 10 15 20 25 Emitter # 13
2. Experimental Techniques Measuring Packaging Induced Strain (1) Micro Photoluminescence Spectroscopy (µ PL) µ PL setup at TRT n contact Laser diode array ~50µm p contact Active region Scanned laser beam Courtesy of TRT, Paris, France 14
2. Experimental Techniques Measuring Packaging Induced Strain (2) Micro Photoluminescence Spectroscopy (µ PL) PL spectrum measured at the centre of the substrate every 10µm along the bar Peak PL wavelength found by fitting each spectrum Record position of PL Peak along bar Peak PL Wavelength (nm) 853 852 851 850 849 848 847 Unmounted Bar 846 0 2000 4000 6000 8000 10000 Distance (µm) PL shift caused by packaginginduced strain Peak PL Wavelength (nm) 849 848 847 846 845 844 843 Mounted Bar By knowing the geometry of the bar, a peak PL value (a measure of packaging induced strain) can be assigned to each emitter 842 0 2000 4000 6000 8000 10000 Distance (µm) Courtesy of TRT, Paris, France 15
2. Experimental Techniques Measuring Strain and Defects (1) Photocurrent Spectroscopy (PCS) Excitation light excites the whole emitter Spatially resolved PC measurement at MBI System based upon a Fourier Transform spectrometer 16
2. Experimental Techniques Measuring Strain and Defects (2) Photocurrent Spectroscopy (PCS) PC spectrum measured for each emitter Derivative of PC spectrum gives QW and waveguide transition energies Spectral position of the QW transition plotted for each emitter across the bar Dil 5/4: spectral position of the QW 1.538 Peak magnitude of the PC defect band used as a relative measure of the defect concentration Photon energy (ev) 1.536 1.534 1.532 0 2 4 6 8 10 y (mm) PPDil504sca n.op j 17
2. Experimental Techniques Defect Imaging (1) Photo and Electroluminescence Microscopy (PLM/ELM) Motion Control Unit 6 axis Positioner Microscope Objective LP Filter ND Filters CCD Camera Sample Holder Lenses IEEE 488 Current Supply Ar + Laser PC Serial Bus Optical Path Lab PC 18
2. Experimental Techniques Defect Imaging (2) Photo and Electroluminescence Microscopy (PLM/ELM) Dark line defects (DLDs) observed in PLM images Reduced luminescence seen in ELM images where DLD intersects active region PLM can also reveal other defects, facet contamination and damage to a bar 19
2. Experimental Techniques Summary of By Emitter Techniques (1) Technique: Quantities measured: Sensitive to: Micro Photoluminescence Photocurrent Spectroscopy Laser Beam Induced Current Photoluminescence Microscopy Electroluminescence Microscopy Near field spectra E g (substrate) E g (quantum well) Sub bandgap absorption Defects Defects, Relative emitter power, I th_app, η ext_app Defects λ/ I Packaging induced strain Packaging induced strain Defects, Shifts in absorption edge Non radiative recombination centres Non radiative recombination centres Temperature, E g, Scattering loss, η int Non radiative recombination centres Temperature, Quasi Fermi level sep. 20
2. Experimental Techniques Summary of By Emitter Techniques (2) µ PL: Scan of 1cm bar with spectra every 10µm takes ~ 20 minutes PCS: Individual emitter spectrum takes ~ 10 minutes 2 3 hours required to measure full bar LBIC: Subset of PCS with 2 λ s (above & below bandgap) ~ 20 mins. Near field images & EL spectra: Typically measured at 10 bias currents For a 20 emitter bar, total measurement time ~ 1 hour However, setup time per bar is also ~ 30 minutes Note: In a detailed study, measurements may be repeated 3 4 times (e.g. before burn in, after burn in, after 1 st aging step, after 2 nd aging step) 21
3. Strain Threshold Defects and packaging induced strain affect degradation & lifetime Larger compressive stress Shorter device lifetime Martin et al., APL 75, 2521 (1999) V shaped facet defects observed in degraded laser bars Higher defect density in highlycompressively strained regions Andrianov et al., JAP 87, 3227 (2000) 22
3. Strain Threshold Objective: Study correlations between local strain & individual emitter degradation Micro Photoluminescence Photoluminescence Microscopy Electroluminescence Microscopy Photocurrent Spectroscopy References R. Xia et al., Synthetic Metals 127, 255 (2002) R. Xia et al., Photon. Technol. Lett. 14, 893 (2002) R. Xia, PhD Thesis, University of Nottingham (2002) 23
3. Strain Threshold Local strain for each emitter determined by µ PL Defects imaged by PLM and ELM = strain free condition = defect observed = region with several defects Increased number of defects observed in emitters with a higher level of packaginginduced strain 24
3. Strain Threshold ELM measurements reveal varying thresholds and efficiencies PC measurements reveal different levels of sub bandgap absorption 25
1.2 1.1 1.0 0.9 1.1 1.0 0.9 1.2 1.1 1.0 0.9 1.2 1.1 1.0 0.9 1.4600 1.4605 1.4610 1.4615 1.4620 1.4625 1.4630 (a) Normalized η ext (a.u.) Bar 1 Bar 2 1.2 (b) Normalized I th (a.u.) (c) PC intensity (λ=850nm) (a.u.) (d) Maximum di PC /de (a.u.) 1.4600 1.4605 1.4610 1.4615 1.4620 1.4625 1.4630 PL peak position (ev) 3. Strain Threshold Packaging induced strain shifts the GaAs µ PL peak by ~16 MPa/nm * Blue line represents strain free level Emitters with stress > 8.4 MPa (red line) show: a) a reduced (apparent) η ext b) a larger (apparent) I th c) a larger sub bandgap photocurrent d) a reduced absorption edge slope Strain threshold for degradation! * M. L. Biermann, et al., J. Appl. Phys. 96, 4056 65 (2004) 26
4. Thermal Runaway Thermal runaway refers to a situation where an increase in the temperature changes the operating conditions in a way that causes a further increase in the temperature leading to a destructive result Models for thermal runaway leading to COD Light absorption at the facet (electron hole pair generation) Electron hole pair recombination Heating Bandgap energy reduction The situation is more complex in laser arrays! Interaction between emitters must be considered: current competition thermal cross talk mechanical strain Catastrophic optical mirror damage Henry et al., JAP 50, 3721 (1979) 27
Aging Step 1 600 hours I = 60A Facet load = 10mW/µm 1.4% drop in output power By emitter measurements then performed 4. Thermal Runaway Aging Step 2 I = 75A Facet load = 12mW/µm Catastrophic degradation observed in <10 hours Sudden drops in the output power each represent failure or one or more emitters 28
4. Thermal Runaway As the bar degrades over time, the following are also observed: Decrease in wall plug efficiency Increase in threshold current Decrease in slope efficiency Remember: These measurements are of the bar as a single entity 29
4. Thermal Runaway Lower bandgap for edge emitters Causes small variations in emitter turnon voltages Also causes more significant variations in emitter operating currents λ shift determined below threshold Larger negative λ shift in edge emitters as current increases Suggests current is increasing faster in the edge emitters Edge emitters have less power (up to 60%) than those in centre Consistent with higher I th_app and lower η app observed in the edge emitters Again supports the idea that the edge emitters are hotter E g (ev) λ/ I (nm/a) Rel. Power @ 30A (a.u.) I th_app (A) ext_app (a.u.) η 1.548 1.546 1.544 0.00 0.05 0.10 0.15 0.20 1.00 (c) 0.75 0.50 0.25 0.00 18.0 16.0 14.0 1.0 (e) 0.9 0.8 0.7 0.6 0.5 1 2 3 4 5 6 7 8 9 1011121314 15161718192021222324 25 Emitter Number (a) (b) (d) 30
4. Thermal Runaway By emitter results suggest that the current is increasing faster in the edge emitters and these edge emitters are hotter However, can a temperature distribution with a minimum at the bar centre and hotter at the edges really be correct? Bulk & facet temperature measurements made on new & aged devices Typical bulk temperature profile of a high power laser bar Similar profiles are observed for both new and aged devices 31
4. Thermal Runaway Raman facet temperature measurements reveal an interesting trend Fresh Devices I = 30A I = 30A Aged ~2500h A temperature distribution that is hottest in the centre is only true of the bulk temperature and the facet temperature of new devices Facet temperature distributions can be inverted in aged devices 32
4. Thermal Runaway Possible causes of higher facet temperatures More defects in edge emitters Larger currents in edge emitters Higher surface currents at the bar edges More non radiative recombination Increased emitter currents & temperatures Positive feedback for defect generation/propagation Thermal runaway of the emitter current Onset of even more rapid degradation References: S. Bull et al., J. Mat. Sci: Mat. Electron. (2008), DOI:10.1007/s10854 008 9577 5 S. Bull et al., J. Appl. Phys. 98, 063101 (2005) S. Bull, PhD Thesis, University of Nottingham (2004) 33
5. Summary The by emitter method uses a wide range of complementary techniques Two successful examples presented: Observation of a strain threshold for increased degradation Observation of the thermal runaway mechanism Results demonstrated that a better understanding of bar degradation mechanisms can be gained by analysing individual emitters And, in Part 2: Strain measurement & detection of defects will be considered in more detail Examples of not only defects caused by packaging induced strain, but also of strain caused by defects 34
Suggested follow up reading: Jens W. Tomm and Juan Jiménez, eds., Quantum Well Laser Array Packaging, McGraw Hill, 2006 (ISBN 0071460322) M. Fukuda, Reliability and Degradation of Semiconductor Lasers and LEDs, Artech House, 1991 (ISBN 0890064652) Juan Jiménez, ed., Microprobe Characterizations of Optoelectronic Materials in M.O. Manesreh, ed., Optoelectronic Properties of Semiconductors and Superlattices, Taylor & Francis, 2003 (ISBN 1560329416 ) 35
Acknowledgements Case Study 1 Taken from the PhD work of R. Xia, University of Nottingham (2002) Case Study 2 Taken from the PhD work of S. Bull, University of Nottingham (2004) EC Projects WWW.BRIGHTer.EU (IST 035266) WWW.BRIGHT.EU (IST 51172) POWERPACK (IST 2000 29447) NODELASE (BE 1945/BRPR 0029) Further Thanks Prof. A.V. Andrianov P WERPACK 36