Characteristics of Dark Current and Photocurrent in Superlattice Infrared Photodetectors

Transcription

1 Characteristics of Dark Current an Photocurrent in Superlattice Infrare Photoetectors Wen-Hsing Hsieh a, Chun-Chi Chen a,jet-mingchen a, Yuen-Wuu Suen b an Chieh-Hsiung Kuan a a Department of lectrical ngineering, National Taiwan University, Taipei, Taiwan, b Department of Physics, National Chung-Hsing University, Taichung, Taiwan, 40 ASTRACT The characteristics of a superlattice infrare photoetector, which has a 0-perio GaAs/AlGaAs superlattice embee between two AlGaAs current blocking layers are investigate. We propose a moel to explain the conuction of ark current an photocurrent. In particular, the interesting feature of this etector, which can also be explaine by our moel, is bias reistribution ue to the backgroun photocurrent. With the backgroun raiation incient upon the etector, the blocking layers are tilte up for 16meV at zero bias. The electrons are all confine by the tilt-up blocking layers an cannot tunnel through them. As the external bias increases, 70% of the voltage is ae on the rear blocking layer an attempts to ecrease its barrier height. specially at -0.18V, the barrier is flat an the barrier height is 16.5meV higher than the bottom of the secon miniban. In this case, only those photoelectrons with high enough energy can pass the blocking layer. As the external bias increases up to -0.33V, the rear blocking layer is tilte own for 105meV. The blocking layer seems to be transparent to all photoelectrons to tunnel through. In aition, the photo responsivity ue to the short wavelength excitation inicates that the photoelectrons o have the inelastic relaxation of the energy along the growth irection. Keywors: Superlattice infrare photoetector, current blocking layer, ark current, an photocurrent. 1. INTRODUCTION Long wavelength infrare etection using intersubban transitions in quantum well structures has been uner intensive investigation Such multiple quantum well etectors are characterize by a thick active region that contains usually 30 to 50 quantum wells. 6. In contrast to the multiple quantum well etectors, the single quantum well etector has the avantages of saving the material an reucing the operating bias. However, the bias istribution on the barrier region is not stable because of the space charge accumulation. Therefore, we propose a novel etector which is similar to the single quantum well structure but with the superlattice rather than the quantum well to absorb infrare raiation. The evice is expecte to be stable ue to the large capture rate in the superlattice than that in the quantum well, an the operation bias is much smaller than that of the conventional multiple quantum well infrare photoetector (QWIP) ue to the narrow barrier thickness of the superlattice structure. The characteristics of the ark current an photocurrent in our etector will be shown in this paper an a simple moel is propose to escribe them. One component of the ark current is the thermionic emission (T) current an the other is the thermally assiste tunneling () current The simulation for the ark current agrees well with the experiment ata in this paper. With infrare raiation, the lie voltage is almost roppe on the rear blocking layer in orer to supply the ark current plus the backgroun photocurrent. In aition, the photo responsivity ue to the short wavelength excitation inicates that the photoelectrons o have the inelastic relaxation of the energy along the growth irection. This paper is organize as follow. Section 1 is a brief introuction an section shows the sample structure an the evice process. The characteristics of the ark current an photocurrent will be shown in section 3 an section 4. In section 3, we will present a simple moel to escribe the physical mechanisms of the ark current. The fitting result matches well with the experiment. Similarly we also erive a formula to escribe the photocurrent an emonstrate that the formula can escribe the experimental results of responsivity versus bias for a monochromatic raiation an spectral response in section 4. The final concusion is given in section 5. Corresponence: mail: kuan@cc.ee.ntu.eu.tw; Telephone: ext. 439; Fax: Physics an Simulation of Optoelectronic Devices IX, Yasuhiko Arakawa, Peter loo, Marek Osinski, itors, Proceeings of SPI Vol. 483 (001) 001 SPI X/01/$

2 . SAMPL STRUCTUR 64.3meV 07.8meV n miniban 07.1meV 8.76µm 141.6meV µ 4.3meV Top 66.meV 57.meV 50nm Al 0.3 Ga 0.68 As 1st miniban 0-perio Al 0.3 Ga 0.68 As(4nm)/GaAs(6nm) 50nm Al 0.3 Ga 0.68 As ottom Fig. 1 Schematic energy ban structure iagram without exten bias. The GaAs well 17 region of the superlattice is oping with 1 10 cm 3 of Si. The schematic ban iagram of our etector is shown in Fig. 1 where the superlattice structure is esigne to absorb infrare raiation by the miniban transitions of electrons. The miniban energies are calculate with the effective mass roximation an the Kronig-Penny moel. 10. The first an secon miniban of a 0-perio GaAs(6nm)/Al 0.3 Ga 0.68 As(4nm) superlattice ranges from 57. to 66.meV an 07.8 to 64.3meV corresponingly. 11. Therefore, the esigne peak absorption wavelengths are at 5.99µm an 8.76µm corresponing to transitions at zone center an ege respectively. In aition, the barrier height of the front an rear blocking layers is both 16.5meV above the bottom of the secon miniban. The superlattice is ope with cm -3 of Si to provie electrons for miniban transition an the ohmic layers are ope with cm -3 of Si. The current blocking layer is introuce to reuce the ark current originate from the conuction of the electrons in the first miniban of the superlattice. The energy of the blocking layer is esigne to be high enough so that the ark current is suppresse more efficiently than the photocurrent that is mainly cause by the photoelectrons of in the secon miniban tunneling through the blocking layer. The samples were processe into (µm) mesa by stanar lithography an wet chemical etching. AuGeNi/Au was evaporate on top an bottom to form the ohmic, an the etector was polishe into 45 on the substrate facet for light coupling. The I-V curves for the ark current an photocurrent in our etector are measure an analyze, an the bias epenence of the responsivity at various wavelength are also measure in orer to unerstan the transport behavior of the secon miniban. 3. MODL AND XPRIMNT-PART I 3.1. Dark Current Moel We first present a simple moel to escribe the physical mechanisms of the ark current in our etector. Figure emonstrates the electron path across the etector for the ark current. The injecte electrons from the top pass the front blocking layer an enter into the secon miniban of the superlattice. The superlattice structure is assume to be long enough so that the electrons are relaxe into the first miniban. Similar to those electrons from the top, the injecte electrons from the superlattice pass the rear blocking layer an enter into the bottom. The currents cause by the electrons through the front an rear blocking layers must be equal. Although the ensities of states in the top an the superlattice are not the same, the voltage rops on the front an rear blocking layers are assume to be the same for the ark current. Due to the oping ensity, no voltage rop in the superlattice is also expecte. 490 Proc. SPI Vol. 483

3 I Top C F 0 e - e - e - e - T ε T(), WHM Max. Front blocking layer Superlattice L ottom Rear blocking layer L P Fig. The electron path across the etector for the ark current. Our emphasis is mae on the ark current generate by those injecte electrons from the superlattice passing the rear blocking layer. In the superlattice, the electron state is quartzite in the growth irection an move freely in the plane perpenicular to the growth irection. lectrons with momentum perpenicular to the growth irection may be scattere by impurities or phonons an pass the blocking layer with the transmission Probability T ( ) where is their iniviual total energy instea of the groun-state energy. In this case, we can use a simple formula to express the ark current I, D where m is effective mass of GaAs, p I D em = πh L P f ( ) AL T( ) P 1, (1) L is the length of each perio in superlattice, f ( ) is the Fermi-Dirac istribution, is the scattering time by the impurities or phonons to transfer their energy along the superlattice plane into the growth irection, an A is the cross-section area of the etector. Theoretically, the ark current consists of the thermionic emission current I an the thermally assiste tunneling current T I The former is contribute by those electrons with >, so that the transmission probability T ( ) is equal to c one. The latter is attribute to the other electrons with < an with the assistance of the electric fiel to tunnel through the c blocking layer. Using the W.K.. metho, we may write the two current components as em A I = f T c πh ( ) an I em A = k πh c em A = f 0 πh A T exp, () K T ( ) T( ) em A = πh h ln k T m ε e A exp +, (3) k T K T k T Proc. SPI Vol

4 10-180K K K I D (A) K 80K 60K T T V (V) Fig. 4 The ark-current I-V characteristics at various temperatures. where = βv is the activation energy, ε = V L A c F is the electric fiel in the current blocking layer with V an L are the external bias an the thickness of the current blocking layer, is a main factor to escribe the process. The factor βv is to explain the rise of the Fermi level ue to the accumulation of electrons at the interface between the superlattice an the blocking layer uner the external lie bias. In the erivation, we have assume that is a constant an inepenent of for simplicity. As shown in Fig., the energy istribution of those electrons passing the blocking layer with thermally assiste tunneling can be escribe with the peak magnitue an the energy banwith of the half magnitue. The total number of the electrons is roximately equal to the prouct of the peak magnitue an the banwith. The peak magnitue is equivalent to fin the T T is given by W.K.. metho as maximum value of f ( ) ( ) in q.(3) where ( ) T ( ) = exp ( ) 3 c h 3eε m 4. (4) m is the barrier effective mass. The Fermi-Dirac istribution f ( ) is roximately ( ( ) ) In the formula, exp k T, F i.e., is assume to be much lager than T. The energy banwith of the half magnitue can be foun with F ( ) T ( ) f equal to the half of the peak magnitue. The final result gives 3.. xperiment an Analysis for Dark Current k = h e ε 4m k T as. (5) xperimentally, the current-voltage (I-V) characteristics were measure by using a Keithley 36 source measurement unit. The sample was mounte on the col hea of an APD DMX-0 close-cycle cryostat to control the sample temperature. A Si-ioe temperature sensor was also mounte near the sample on the same col hea. We have recore an controlle the sample temperature by using the Si-ioe sensor. 49 Proc. SPI Vol. 483

5 V 10-6 I ark /T (A/K) V /T (K -1 ) Fig. 5 Plot the ark current (scatter curve) as a function of inverse T k at -0.1 to -0.6V bias voltages (soli curve). The ark-current I-V characteristics at various temperatures are shown in Fig. 4. For the etector at 0k, the ark current can be obviously separate into the two regimes. In the low bias regime, the ark current seems to increase less sensitively to the lie bias. This is attribute to the thermionic emission (T) process of the electrons through the blocking layer. In the secon regime, the ark current is more sensitive to the lie bias. Actually this is the primary characteristic of the thermally assiste tunneling () mechanism. To use our moel to explain the ark current I, especially the tren of the activation energy, we plot I T D D as a function of inverse T at negative-bias voltages in a semi-log scale. The result is shown as the open circle in Fig. 5. In terms of qs. ()~(5), the function can be written as ln( I A αv em A ) = + ln[1 + exp( )] + ln 3, (6) T k T 3k T πh T where α = ln k h m e k. The soli curves in Fig. 5 are the fitting results. The main fitting parameters are α an β, which are equal to 8616K 3 V 1 an 46.67meV / V erive from experimental ata. Theoretically α is equal to 8111K 3 V an agrees with the experimental results. The activation energy is equal to 154meV at V = 0 an is consistent with the esign value 150meV. A The accumulate space charge at the interface between superlattice an the rear blocking layer may cause the Fermi level to rise 3 about 8meV at V = 0.6V. This is equivalent to extra charge ensity cm there. 1 Proc. SPI Vol

6 hv Top C F 0 e - Front blocking layer e - e - Superlattice e - I P ε 0 V 0 I D ε =0 V =0 ottom Rear blocking layer Fig. 6 The electron path an the bias istribution with ark current (ash line) an photocurrent (soli line). The sensitivity of I to the external bias is ue to the factor of βv while that of I is ue to.since T proportional to V, I is more sensitive to the external bias than I. T is 4. MODL AND XPRIMNT-PART II 4.1. Photocurrent Moel In this section, we first erive a formula to escribe the photocurrent an then emonstrate the experiment results of responsivity versus bias an spectral response. In the en, we iscuss the agreement between the experimental results an our formula. With infrare raiation incient upon the sample, the electrons are excite from the first miniban into the secon miniban. There may exist some relaxation mechanism, to cause the energy of the photoelectron along the growth irection to ecrease. For example, the impurities or phonons in the superlattice may scatter the photoelectrons an transfer their energy. Therefore, the generation rate G( ) of the photoelectrons by the infrare raiation may be a function of energy where is between the secon miniban of the superlattice. For simplicity, we assume the photoelectron with energy has a lifetime an velocity v to oscillate between the two blocking layers where the photocurrent can be written as I P where T (,ε) metho gives (,ε) I P f an v are both inepenent of the energy. Uner these circumstances, ( ) v T ( ε) = eg,, (7) f is the transmission probability of the photoelectrons through the blocking layer. Again using the W.K.. T as q. (4). With only ark current existing, 50% of the lie voltage rops on the front blocking layer an 50% oes on the rear blocking layer as shown in Fig.. With aitional backgroun photocurrent, most of the lie voltage is roppe on the front blocking layer in orer to supply the ark current plus the backgroun photocurrent as shown in Fig xperiment an Analysis for Photocurrent In orer to unerstan the potential profile of the rear blocking layer observe by the photoelectrons, the spectral response an the voltage epenence of responsivity for a monochromatic raiation are both important experiments to procee. The first one f 494 Proc. SPI Vol. 483

7 µm Responsivity(mA/W) µm -0.33V -0.3V -0.3V -0.V -0.5V V Wavelength(µm) Fig. 7. Spectral response venues several bias at 30K. is to ientify the wavelength range in which the sample has a response. The latter is to observe the increment tren of photoelectrons versus bias ue to a monochromatic raiation. With the experimental results an our formula, the potential profile of the rear blocking layer expects to be unerstoo. First, the spectral responsivity was measure at various bias an temperature by using Perkin lmer 000 Fourier transform infrare (FTIR) spectrometer. The absolute value of the responsivity was etermine by measuring the blackboy responsivity R with a blackboy source. In the experiment, the temperature of the blackboy source was set at 800K. The response R(800K) can be written as R R(800K) = P R( λ) W ( λ, 800K) 0 W ( λ, 800K) λ 0 where W ( λ, T ) is the raiance of the blackboy source at T, R( λ) responsivity an measure by using FTIR an R is the peak responsivity of R ( λ) P λ, (8) is the relative responsivity normalize by the peak, which is what we want to know. Figure 7 shows spectral responsivity venues several biases at 30K. The primary peak position varies from the short wavelength to the large one with increasing the bias. specially, the peak of the short wavelength (6.µm) responsivity is going to ear at -0.V an the peak of the long wavelength (8.8µm) responsivity is earing at -0.9V in Fig. 7. It inicates that the photoelectrons from the bottom ege of the secon miniban can just pass the blocking layer at -0.9V, i.e., the effective barrier height is about 07meV. The short wavelength peak (6.µm) an long wavelength peak (8.8µm) are close to our esigne values (6.0µman8.7µm). The voltage epenence of responsivity at various wavelengths is measure with a chopper an lock-in amplifier. The monochromatic infrare raiation is generate by passing the thermal raiation from a glow bar through a monochromator an a long-pass filter with cut-off wavelength 5µm, an the photocurrent is transforme into voltage signal by a transconuctance amplifier SR570. The measure voltage epenence of responsivity at wavelengths 6.6µm an8.8µm is shown as open circle in Fig. 8. It is observe that the long wavelength responsivity is negligible at bias smaller than 0.V in magnitue, however it increases more rapily than the short wavelength one at high bias region. Proc. SPI Vol

8 5 0 T=30K 6.6µm 10meV 50meV, x4 Responsivity(mA/W) µm 8.8µm G()(A.U.) 10meV nergy (mev) 8.8µm V (V) Fig. 8 Voltage epenence of responsivity at two peak wavelength 6.6µm an 8.8µm for experiment (scatter curve) an theory (soli curve). In orer to match the experimental ata with q. (7), we have to figure out the voltage rop on the rear blocking layer. With backgroun raiation incient upon the sample an at thermal equilibrium, no current shoul be etecte in our sample. The barriers of the two blocking layers are expecte to tilt up to block the photocurrent. Uner this situation, the voltage rop on the rear blocking layer can be consiere as negative while that on the front one as positive. The total amount of the voltage rop is zero. As the external bias increases the voltage rop on the rear blocking layer increases an is becoming positive. As a first-orer roximation, the voltage rop on the rear blocking layer can be written as V = V + b V, (9) r off where V is the negative voltage rop on the rear blocking layer at zero bias, an b represents the percentage of external off lie bias to rop on the rear blocking layer. Another important factor in q. (7) is G ( ), which involves with the energy istribution of the photoelectrons. As expecte, the photoelectrons may relax their energy an enter into the bottom state of the secon miniban. Therefore, G( ) can be roximate as a two-peak function for a monochromatic raiation except the transition between the top of the first miniban into the bottom of the secon miniban. Taken the 6.6-µm raiation into consieration. The photoelectrons transit from the 6-meV state in the first miniban to the 50-meV state in the secon miniban. In aition, relaxation may cause the photoelectrons to be accumulate in the bottom G may be a two-peak function as that shown in the insert of Fig. 8. (10-meV) state in the secon miniban. In this case, ( ) For the 8.8-µm raiation, G( ) become a one-peak function since the final state is the bottom state of the secon miniban. In the fitting, we use Gaussian function with the peak value an the associate variance as parameters to simulate a istribution in T,ε. The final results of the fitting are shown as the soli G ( ). quation (9) is use to calculate the electric fiel ε in ( ) curve in Fig. 8. The insert shows G( ) for 6.6-µm an 8.8-µm raiation. For 6.6-µm raiation, the contribution from the 50-meV state is represente as the ash line. For these photoelectrons in 50-meV state, they o not observe any barrier even at -0.15V an prouce a saturate photocurrent at all bias range shown in Fig. 8. In fact, almost 90% of the photoelectrons are relaxe into the bottom state of the secon miniban an cause the responsivity to increase with bias since they have to tunnel through the barrier just like the photoelectrons prouce by 8.8µm-raiation. 496 Proc. SPI Vol. 483

9 0. V r (ev) b =0.7 hv -0.1 hv hv (3) -0. (1) () V (V) Fig. 9 ias on the rear blocking layer with only ark current (ash line) an photocurrent (soli line) at 30K. The voltage rop on the rear blocking layer versus the external bias with backgroun raiation incient upon the sample is shown as the soli line in Fig. 9. The three inserts show the potential profile at three ifferent biases. The ash line in the same figure stans for the same voltage rop with only ark current existing. As shown in the first insert of Fig. 9, the blocking layers are tilte up for 16meV at zero bias ue to the backgroun raiation incient upon the etector. The secon miniban is confine by the tilt-up blocking layer an no electron can tunnel through it. On the contrary, there is no voltage rop on the rear blocking layer with only the ark current existing. As the external bias increases, 70% of the voltage is roppe on the rear blocking layer an attempts to ecrease its barrier height. As shown in the secon insert of Fig. 9, the rear blocking layer has a flat barrier at -0.18V. It is note that the barrier height of the blocking layer is 16.5meV higher than the bottom of the secon miniban. In this case, only those photoelectrons with high enough energy can pass the blocking layer. In particular, those photoelectrons in the bottom state of the secon miniban cannot tunnel through the blocking layer. As shown in the insert of Fig. 8, this correspons to those photoelectrons without relaxation. The photocurrent is expecte to be small since the number of the photoelectrons without relaxation is limite. As shown in Fig. 7, at -0.V, the responsivity is small an only the short wavelength raiation can cause response. This is consistent with the preiction from the potential profile of the rear blocking layer. As the external bias increases up to -0.33V, the rear blocking layer is tilte own for 105meV. The electric fiel on the barrier is V/cm. The transmission probability for the photoelectrons at the bottom state of the secon miniban through the blocking layer is about It increases ramatically for the photoelectrons with higher energy. Therefore, the blocking layer seems to be transparent to all photoelectrons, an the responsivity in Fig. 7 shows high values an wie range of respone wavelength. 5. CONCLUSION We have investigate the characteristics of the backgroun photocurrent an ark current in a superlattice infrare photoetector, which has a 0-perio GaAs(6nm)/Al 0.8 Ga 0.7 As(4nm) superlattice embee between two 50nm Al 0.8 Ga 0.7 As block layers. It is conclue from the ark-current I-V characteristics an photo responsivity versus external bias that the external bias is inee reistribute ue to the backgroun photocurrent. We successfully use a simple moel to escribe the physical mechanisms of the ark current an photo responsivity. With infrare raiation incient upon the sample, there are almost 70% of V is roppe on the rear blocking layer to ecrease the barrier height of the blocking layer. specially at zero bias, the back blocking layer is tilte up for 16meV to prevent from the backgroun photocurrent an ark current. In Proc. SPI Vol

10 aition, the photo responsivity ue to the short wavelength excitation inicates that the photoelectrons o have the inelastic relaxation of the energy along the growth irection. ACKNOWLDG This project is sponsore partially by National Science Council uner grant number NSC RFRNC 1. L. C. West an S. J. glash, "First observation of an extremely large ipole infrare transition within the conuction ban of a GaAs quantum well," Appl. Phvs. Lett., vol. 46, pp , S. R. Anrews an. A. Miller, xperimental an theoretical stuies of the performance of quantum-well infrare photoetectors, J. Appl. Phys., vol. 70, pp , F. Levine, Quantum-well infrare photoetectors, J. Appl. Phys., vol. 74, pp. R1 R81, G. Sarusi, S. D. Gunapala, J. S. Park, an. F. Levine, Design an per-formance of very long-wavelength GaAs/Al x Ga 1-x As quantum-well infrare photoetectos, J. Appl. Phys., vol. 76, pp , K.K. Choi, The physics of quantum well infrare photoetector, Worl Scientific, chap., G.Hasnain,.F.Levine, S.Gunapala, an N.Chan, Large photoconuctive gain in quantum well infrare photoetectors, Appl. Phys. Lett., vol. 57, pp , Pelve et al., Analysis of the ark current in ope-well multiple quantum well AlGaAs infrare photoetectors, J. Appl. Phys., vol. 66, pp , C. H. Kuan et al., Hot-electron istribution in multiple quantum well infrare photoetectors, Appl. Phys. Lett., vol. 63, pp , P. Man an D. S. Pan, Hot-carrier-temperature moel for the ark current of quantum-well infrare photoetectors, Appl. Phys. Lett., vol. 66, pp , D. F. Nelson, R. C. Milier, an D. A. Kleinman, an-nonparabolicity effects in semiconuctor quantum wells, Phys. Rev., vol. 35, pp , M. Helm, Infrare spectroscopy an transport of electrons in semiconuctor superlattices, Semicon. Sci. Technol., vol.10, pp , F. Levine, C. G. ethea, G. Hasnain, V. O. Shen,. Pelve, R. R. Abbott, an S. J. Hsieh, High sensitivity low ark current 10 µm GaAs quantum well infrare photoetectors, Appl. Phys. Lett., vol. 56, pp , H. C. Liu, Li Jianmeng, M. uchanan, an Wasilewski, Z.R., High-frequency quantum-well infrare photoetectors measure by microwave-rectification technique, I J. Quantum lectronics, vol. 3, pp , H.C.Liu et al., Quantum well intersubban transition physics an evices, A Harcourt Science & Technology Co., pp , Proc. SPI Vol. 483