Proton Damage in LEDs with Wavelengths Above the Silicon Wavelength Cutoff

Transcription

1 Proton Damage in EDs with Wavelengths Above the Silion Wavelength Cutoff Heidi N. Beker and Allan H. Johnston Jet Propulsion aboratory California Institute of Tehnology Pasadena, Califomia Abstrat Proton damage is investigated for EDs with wavelengths of 1050 and 1550 nm. ight output beomes nonlinear with urrent after irradiation, unlike AlGaAs EDs. Mehanisms are proposed that are related to the material properties. 1

2 Proton Damage in EDs with Wavelengths Above the Silion Wavelength Cutoff Heidi N. Beker and Allan H. Johnston Jet Propulsion aboratory Califomia Institute of Tehnology, Pasadena, California ED Manufaturer and Part Number INTRODUCTION Displaement damage effets have been studied in lightemitting diodes for many years [ 161. However, the majority of the work was done on AlGaAs or GaAs EDs emitting in the nm region. ittle work has been done on EDs that operate beyond the 1 pm utoff for silion detetors. Suh devies use different materials and are typially designed for highspeed operation in fiber opti ommuniation appliations. This paper evaluates radiation damage in two EDs in that extended wavelength range that are designed with fast response times for fiber ommuniations appliations, where they are alternatives to laser diodes. Those results are ompared with tests of an advaned 875 nm ED that uses advaned fabriation tehniques that enhane light extration. EXPERIMENTA PROCEDURE Table 1 lists the EDs in this study and some of their key properties. The Agilent HSD420 uses an advaned fabriation tehnique with a transparent substrate that eliminates absorption loss in the substrate region [7,8]. The other two devies use different material tehnologies, and are designed for fast response time. All three devies were mounted in epoxy pakages. I Agilent I 875 HSD420 Epitex Epitex 1550 I 0 Table 1. Devies Seleted for the Study Peak Wavelength (nm) Material AlGaAs TSlDouble Heterojuntion GaAs InGaAsP Optial Power 50 ma (mw) The researh in this paper was arried out at the Jet Propulsion aboratory, Califomia Institute of Tehnology, under ontrat with the National Aeronautis and Spae Administration (NASA) The devies were irradiated with 6MeV protons at Croker Nulear aboratory, UC Davis. The maximum fluene was x 10 p/m2. Five samples of eah devie were irradiated in an unbiased ondition (all leads grounded). Irradiations were performed unbiased in order to minimize the effets of reombinationenhaned annealing during irradiation. Changes in light output were measured between irradiation steps using a speial testing fixture that oupled the ED under test to a silion photodiode for the 875nm ED, or an InGaAs phototransistor (diodeonneted) for the longer wavelength EDs. The EDs were mounted on an aluminum plate that was attahed to a thermoeletri ooling (TEC) module during measurements. The TEC maintained devie temperature at 25 C f. 1 C during haraterization, reduing measurement variability beause of the sensitivity of ED light output to temperature. An Agilent 4156B parametri analyzer was used to measure hanges in optial power of the EDs at several forward urrents, up to looma, the maximum rated urrent for the EDs. The measurement program limited the amount of time and forward urrent at eah measurement step in order to minimize reombinationenhaned annealing during haraterization. Measurement repeatability was typially 1% or better. EXPERIMENTA RESUTS Figure 1 shows light output at two forward urrents, 10 and 50 ma, vs. fluene. This range of forward urrent is within the region where these EDs would be operated in a typial appliation. Data has been normalized to preirradiation values. All three types of EDs exhibit muh less degradation ompared to the highly sensitive amphoterially doped EDs that have reeived so muh attention in previous work. The 875 nm ED degraded only slightly more when the forward urrent was redued to 10 ma, whereas both of the other EDs were degraded by a muh larger amount with lower forward urrent; note in partiular the large differene for the 1550 nm ED with 10 and 50 ma forward urrent. 2

3 ! I 0.1 loma 100 't 1 in 875 nm +5Ond,n=2/ 50 m4, n=l j 1Om4, n = 1 10 *1Ond,n=2/ r < 5 =r, 1 : a' l.e+ll le12 1.Et1 6MeV Proton Fluene (p/m2) Fig. 1. Normalized degradation of ED output with forward urrent = 10 and 50 ma Although the normalized light output shown in Fig. 1 is a useful way to examine ED damage, it is possible to fit the damage to a power law that provides some insight into the degradation mehanisms as well as providing a parameter that has a well defined relationship with partile fluene. Rose and Barnes [] showed that damage in EDs with long lifetime ould be desribed by the power law relationship below: Fig. 2 Data for the Agilent 875 nm ED, using the power law with n = 2/ and n = 1. A similar analysis of the results for the 1050 nm ED is shown in Fig.. For this devie, the data fit (with a slope loser to 1) Eq. 1 far more losely with n = 2/ instead of n =l. Another important differene is that the damage is onsiderabey higher when the ED is measured with a forward urrent of 10 ma ompared to the results with IF = 50 ma. That behavior has not been observed in other studies of EDs. However, EDs operating at 1050 nm have not been investigated previously. where I, is the initial light output and I is the redued power output after irradiation, n is an exponent between 1/ and 1, z, is the initial minority arrier lifetime, K is the lifetime damage onstant, and CD is the partile fluene. For an ED that is ontrolled by lifetime damage with a uniform distribution of impurities in the bandgap, n should have the value of Amphoterially doped EDs usually fit that equation very losely, but more advaned AlGaAs EDs with narrow heterostrutures usually fit Eq. 1 far more losely with n = 1 instead of 2/ [5]. Fig. 2 ompares test results for the Agilent 875 nm ED at 10 and 50 ma. For both urrents the slope is nearly linear with n = 1 ompared to n = 2/. At 50 ma the parameter alulated from Eq. 1 at the first radiation level departs somewhat from the slope at higher fluenes. These results are similar to earlier results for other heterojuntion EDs using AlGaAs [5].. % == 1 :. 0.1 : 10 nd,n= nd, n =2/ U' 1050 nm 0.01 I 1.H11 I.E+12 1.H1 6MeV Proton Fluene (p/m2) Fig. Data for the Epitex 1050 nm ED using the power law with n = 2/ and n = 1. Analysis of the results for the 1550 nm ED are shown in Fig. 4. Just as for the previous ase, a nearly exat fit (with the slope = 1) is provided with n = 2/. The damage at lower urrents is muh

4 F higher ompared to the 875 nm ED, and the differene is greater than for the 1050 nm devie. 1 : 1 E t ReRad 1.OOEt12 1E4 t.00et12 s! J 1E5 * 0 9 0" 1E na, n=2/ + 50 na, n=l &1OnA,n==2/ I E7 &' nm.' EtI1 1.H12 I.E+1 6MeV Roton Fiuene (pld) 875 nm Fig. 5. Dependene of output power on forward urrent over an extended urrent range for a typial 875 nm ED. The 1050 nm ED behaves very differently, as shown in Fig. 6. Initially the slope is almost exatly one. After the first radiation level it inreases to 1.25, and ontinues to inrease to a value of 1.6 after the last irradiation level. Consequently the damage at high urrents is a great deal lower than the damage under lower forward urrent onditions. Aording to the manufaturer this devie is fabriated with GaAs. However, the wavelength is well above the utoff wavelength for GaAs, and it is likely that this devie uses other materials on a GaAs substrate, suh as InGaAs [lo]. 1E 1E4 i h 5. t" 1E5 i 9 0 a, 1eReRad H121 1E7 I 1, I 7 I, nm I ED forward urrent (A) Fig. 6. Dependene of output power on forward urrent over an extended urrent range for a typial 1050 nm ED. The 1050 nm ED also exhibited large differenes in damage when we ompare 10 ma and 50 ma injetion onditions, although the differene was smaller than for the 1550 nm ED. Basi material harateristis may be a fator in the hange in linearity after irradiation. The AlGaAs material system has higher heterojuntion barriers ompared to InGaAsP, and is also less sensitive to Auger reombination. Degraded linearity in InGaAsP EDs was attributed to eletron leakage through heterostrutures in one study [ 1 1, and this is one possible mehanism for the hange in linearity after irradiation in the longer wavelength EDs. Doping levels >lo'* m are required to ahieve the short risetimes of our IR ED samples. Carrier removal rates are approximately 0 m", probably too low to affet the layers in EDs of this type. However, bulk reombination enters are more important in InGaAs and InGaAsP ompared to AlGaAs beause bimoleular reombination rates are slightly lower, and devies made with the longer wavelength materials are limited to lower injetion levels beause of the importane of Auger reombination under high injetion onditions. 4

5 Annealing Annealing measurements were done on representative samples of the three types of EDs, The devies were plaed in a temperatureontrolled test fixture during the extended annealing period, with forward bias applied. Fig. 7 shows annealing for devies biased with 5 ma of forward urrent. All EDs were irradiated to a proton fluene of x 10l2 p/m*, whih dereased the power output by fators of approximately 5 to 20, depending on the ED tehnology. Some annealing ourred for all three types of EDs, but when referened to the preirradiation value it is lear that only a small fration of the damage atually reovered. In ontrast, about 40% of the damage in amphoterially doped EDs reovers after annealing under omparable onditions [ 121. Thus, all three types of EDs are relatively insensitive to annealing, just as for older results for AlGaAs EDs made with heteroj un tions. 1K n w iol irm 1 i*o1 io5 ward Seonds under blas (Q 5mA r) Fig. 7. Annealing of the three types of EDs with a forward urrent of 5 ma. The damage is referened to preirradiation optial power levels. CONCUSIONS This paper has examined the effets of proton damage on two types of EDs in the wavelength region above the silion bandgap limit. Unlike AlGaAs EDs, the optial power linearity in both types of devies hanges signifiantly after irradiation. This may be aused by the lower heterojuntion barriers assoiated with these materials. From a pratial standpoint the hange in linearity requires more extensive haraterization ompared to AlGaAs devies where linearity is only slightly affeted. However, both of the longer wavelength ED tehnologies show less degradation ompared to AlGaAs EDs. This is likely related to the short response times that require thin ative regions and high arrier densities, both of whih derease radiation sensitivity. [l] [2] [] [4] [5] [6] REFERENCES C. E. Barnes, Radiation Effets in Eletroluminesent Diodes, IEEE Trans. Nul. Si., Vol. 18, pp.221, De R. H. Hum and A.. Barry, Radiation Damage Constants of ightemitting Diodes by a owcurrent Evaluation Method, IEEE Trans. Nul. Si., Vol. 22, pp , De B. H. Rose and C. E. Barnes, Proton damage effets on lightemitting diodes, J. Appl. Phys., vol. 5, no., pp ,1982. A.. Barry, et al., The Energy Dependene of ifetime Damage Constants in GaAs EDs for 1500 MeV Protons, IEEE Trans. Nul. Si., 42(6), pp ( 1995). A. H. Johnston and T. F. Miyahira, Charaterization of Proton Damage in ight Emitting Diodes, IEEE Trans. Nul. Si., 47(6), pp (2000). R.. Reed, et al., Energy Dependene of Proton Damage in ightemitting Diodes, IEEE Trans. Nul. Si., 47(6), pp (2000). [7] D. A. Vanderwater, et al., HighBrightness AlGaInP ight Emitting Diodes, Pro. ofthe IEEE, 85(1 l), pp (1997). [8] [9] F. A, Kish, et al., Very High Effiieny Semiondutor WaferBonded Transparent Substrate ight Emitting Diodes, Appl. Phys. ett., 64, pp (1994). 0. Pursianen, et ai., Identifiation of Aging Mehanisms in the Optial and Eletrial Charateristis of ightemitting Diodes, Appl. Phys. ett., 79(18), pp (2001). [lo] S. Yamakoshi, e td, Diret Observation of Eletron eakage in InGaAsP/InP Double Heterostruture, Appl. Phys. ett., 40(2), pp (1982). [ 1 11 J. J. Coleman, Strainedayer InGaAs QuantumWell Heterostruture asers, IEEE J. on Selet. Topis in Quant. Elet., 6(6), pp ,2000. [12] A. H. Johnston, Proton Displaement Damage in ight Emitting and aser Diodes, IEEE Trans. Nul. Si., 48(5), pp (2001). 5