The microscopic study of a single hydrogen-like impurity in semi-insulating GaAs
aa r X i v : . [ c ond - m a t . m t r l - s c i ] O c t The microscopic study of a single hydrogen-like impurity in semi-insulating GaAs
D.G. Eshchenko,
1, 2
V.G. Storchak, S.P. Cottrell, and E. Morenzoni Physik-Institut der Universit¨at Z¨urich, Winterthurerstrasse 190, CH-8057, Z¨urich, Switzerland Laboratory for Muon Spin Spectroscopy, Paul Scherrer Institut, CH-5232 Villigen PSI, Switzerland ∗ Russian Research Centre “Kurchatov Institute”, Kurchatov Sq. 46, Moscow 123182, Russia ISIS Facility, Rutherford Appleton Laboratory, Oxfordshire OX11 OQX, UK
The charge dynamics of hydrogen-like centers formed by the implantation of energetic (4 MeV)muons in semi-insulating GaAs have been studied by muon spin resonance in electric fields. Theresults point to the significant role of deep hole traps in the compensation mechanism of GaAs.Electric-field-enhanced neutralization of deep electron and hole traps by muon-track-induced hotcarriers results to an increase of the non-equilibrium carrier life-times. As a consequence, themuonium ( µ + + e − ) center at the tetrahedral As site can capture the track’s holes and thereforebehaves like a donor. PACS numbers: 72.20.Jv, 76.75.+i
Semi-insulating (SI) GaAs is an important materialthat forms the basis of the GaAs microwave and inte-grated circuit industries. SI substrates are used for thegrowing of GaMnAs thin films, recognised as a proto-type material for future spintronics devices [1]. The semi-insulating properties of GaAs arise from the compensa-tion of residual shallow donors, residual or intentionallyadded shallow acceptors and intrinsic or intentionally in-troduced deep centers. Deep centers may act as carriertraps, recombination centers or scattering centers, andhave a strong influence on the electronic properties of thematerial even when their concentration is much less thanthe carrier density. Hence studies of deep level defects inGaAs is of both fundamental and applied interest.Since the passivation of shallow acceptors and donorsin GaAs by hydrogen was discovered in 1986 [2], thebehaviour of hydrogen in III-V semiconductors has be-came the subject of intense research. Significant informa-tion on the microscopic properties of isolated hydrogenimpurities have been obtained from muon spin rotationspectroscopy, where the behaviour of positive muons andmuonium (Mu= µ + + e − ) are studied as an analogue ofa light hydrogen isotope (m µ ≃ / m p ). In this letterwe demonstrate that µ SR can be used to get alternativeinformation on deep centers in GaAs.Muonium in GaAs can exist in three charge statesMu + , Mu and Mu − [3, 4]. The stable position for Mu + lies on the Ga-As bond (so-called Mu +BC ), with anotherpossible location being the anti-bonding position close toAs atom (AB As ). The tetrahedral void between four Asatoms (T As ) is also considered [4] as a metastable posi-tion supporting Mu + . Negative Mu − is usially placed atthe tetrahedral position formed by four Ga atoms Mu − T Ga ,or at the anti-bonding position close to Ga atom (AB Ga ).Neutral muonium is believed to occupy the bond-centerposition (Mu ) and both T Ga and T As tetrahedral sitesand is thus denoted by Mu .In semi-insulating and slightly doped n-type GaAs, the following is the accepted picture describing the dis-tribution of muonium between the different states on ananosecond time-scale following muon implantation. Atlow temperatures (T <
100 K) the diamagnetic fractionis small, while the Mu and Mu fractions are about40(5)% and 60(5)% respectively. The Mu fraction re-mains constant up to at least room temperature, whilethe Mu fraction decreases above 100 K. This processis accompanied by a corresponding increase in the dia-magnetic fraction, and is usually described in terms ofMu to Mu +BC ionization [3] (implying the thermal emis-sion of the electron from the Mu to the conductionband). At low temperatures, electric field experiments[5, 6] have shown a gradual suppression of the Mu sig-nal by the application of an electric field of the order E char ∼ fraction isaccompanied by a corresponding increase in the diamag-netic signal; however, electric fields of up to 20 kV/cmhave no effect on the Mu signal. In this letter we showthat very small electric field of ∼ − ∼ − s) dynamics of the neutral muonium.Most of the experiments reported here were performedat the ISIS pulsed muon facility, located at the Ruther-ford Appleton Laboratory (RAL, Chilton, UK), using thenovel technique of radio-frequency (RF)- µ SR in electricfields EF-RF- µ SR[7]. RF experiments are essentially lon-gitudinal magnetic field ( LF ) measurements, where theinitial direction of the muon spin is co-linear with therelatively large (in our case about 1 . − H ∼
10 Oe) RF field applied per-pendicular to the LF at a frequency tuned to match theZeeman splitting of the diamagnetic species ( ∼
20 MHz).Diamagnetic states formed from a paramagnetic precur-sor, even on a microsecond timescale, will contribute tothe precessing RF asymmetry asymmetry, and conse-quently RF techniques have become a well establishedtool for studying the dynamics of slow muonium to dia-magnetic conversion [8]. Our previous EF- µ SR experi-ments in insulators [9] demonstrated that the muon issurrounded by its own track products, with the mainradiation damage (the majority of electron-holes pairs)being behind the stopped muon. In the flat geometry weare using, by choosing the polarity of the applied electricfield (parallel or anti-parallel to the muon track direc-tion) one can select the appropriate charge (electrons orholes) for the long range interactions with the stoppedmuon. Thus EF-RF- µ SR [7] enables the selective (elec-trons vs holes) study of the dynamics of the interaction ofmuonium with hot carriers on a microsecond timescale.Since the stopped muon is a single probe at the end ofhis own track, the dynamics of track products will be de-termined by the intrinsic properties of the sample underthe study e.g by deep centers. In this respect, µ SR ex-periments may offer a unique insight on the fundamentalnon-equilibrium properties of deep centers.For µ SR measurements, 100% spin-polarized positivemuons are implanted into the sample and the positrondecay products monitored. Due to parity violation, thepositrons are emitted preferentially along the instanta-neous direction of the muon spin. Using two sets ofpositron counters placed upstream and downstream ofthe sample (backward and forward counters) the time-differential asymmetry of the muon decay is constructedin the standard way as A ( t ) = N B ( t ) − N F ( t ) N F ( t )+ N B ( t ) , where N B ( t )and N F ( t ) are decay counts in the backward and forwarddetectors respectively. The maximum asymmetry for theapparatus depends on the positions and solid angles ofthe detectors and, in our case, is about 0.23. Each µ SRspectrum contains at least several million decays. Foursets of histograms are collected: RF field on, positiveelectric field; RF field on, negative electric field; RF off,positive electric field; RF off, negative electric field. Thepositive electric field points parallel to the initial muonmomentum. The states were changed every 1/50 s at ev-ery ISIS accelerator frame (i.e. before every muon pulse)Typical µ SR spectra measured in GaAs are presented inFig. (1). The RF off asymmetry spectrum consists ofa fast-relaxing Mu T fraction and a slow-relaxing signalwhich is the sum of the Mu and diamagnetic com-ponents. The contributions of these two components areresolved in the RF spectrum, where the diamagnetic frac-tion is seen in the precessing signal (which has a slowrelaxation due mainly to the space inhomogeneity of theRF field) while the non-relaxing Mu is observed as aconstant offset in the spectra.Studies reported here were performed using a commer-cial high resistivity GaAs substrate ( ρ ∼ (1 . − . × Ohm × cm, n-type conductivity with an electron mo-bility ∼ /V · s), 0.5 mm thick, with the < > crystallographic axis perpendicular to the surface im-pacted by the muon beam. The sample was purchasedfrom the American Xtal company, where it was grown us-ing the vertical boat (VB) technique. Silver electrodes of -0.05 0 0.05 0.1 0.15 0.2 0 2 4 6 8 10 A s y mm e t r y Time (10 -6 s) RF offRF on FIG. 1: Typical experimental spectra measured in semi-insulating GaAs after the electric field treatment measuredat T = 70 K in longitudinal magnetic field LF = 1837 Oeand zero electric field. Blue circles - RF signal is off; redtriangles - RF signal is on. ∼
80 nm thickness were deposited on both surfaces of thesample using the DC magnetron sputtering technique,thereby making two Schottky contacts. This allowed usto apply voltages of up to 1 kV in both polarities, giv-ing us a nominal electric field in the sample of up to ±
20 kV/cm.A measurement of the diamagnetic fraction in zeroelectric field is shown in Fig. (2). The triangles repre-
Temperature (K) D i a m a gn e ti c A s y mm e t r y FIG. 2: Temperature dependencies of the diamagnetic signalsmeasured at zero electric field in different states of commercialsi-GaAs. Triangles: RF diamagnetic asymmetry in the virginsample. Circles: enhanced RF signal in the treated sample.Stars: RF asymmetry measured in the process of the electricfield treatment. sent the temperature dependence of the RF diamagneticasymmetry in the virgin sample (no electric field history).This dependence is similar to that reported in the litera-ture [4, 10]. The increase in the asymmetry is describedin terms of Mu to Mu +BC conversion. Above 120 K theRF asymmetry remains constant, and is in good agree-ment with the Mu fraction measured in transversemagnetic field experiments at low temperatures [3]. Thisimplies that in this sample below 250 K there is no contri-bution to the diamagnetic fraction from conversion of theMu state, a conclusion in agreement with [4]. The pic-ture changes drastically if, at low temperatures, the sam-ple is treated by applying a large ( E >
10 kV/cm) elec-tric field simultaneously with muon implantation. Theenhanced diamagnetic RF asymmetry is shown by thecircles. Above ∼
70 K, the diamagnetic asymmetry inthe treated sample exceeds that of the high temperaturefraction measured in the virgin sample, implying an addi-tional channel for Mu muonium to a diamagnetic stateconversion is opened by the sample treatment. Measure-ments were carried out at a fixed temperature of 70 K toevaluate how the sample changes from its virgin conditionto the treated state as a function of exposure to the muonbeam. Results are shown by the stars in Fig. (1), eachcorresponding to a zero electric field measurement car-ried out after a further 30 minutes of treatment in a field E = ±
16 kV/cm. To complete the transition, about twohours of beam exposure or ∼ . − . × electron-holepairs per cubic centimeter are required. The enhancedRF asymmetry passes through a broad maximum around120–140 K (a state that persists for at least 12 hours),while at temperatures above 140 K it shows annealing be-havior and on warming to ∼
220 K the sample is returnedto the virgin state.The origin of the enhanced diamagnetic RF signal canbe understood from the electric field measurements. Typ-ical µ SR spectra measured in the treated sample in elec-tric fields ± As species. A negative electric field will pull holesaway from the stopped muon, thereby reducing the muo-nium to diamagnetic conversion, while a positive electricfield will pull out the nearest holes but push ”early” trackholes towards the stopped muon and muonium to to dia-magnetic conversion is unchanged. In the virgin sample,a small electric field does not change the RF precessingasymmetry.To identify the lattice position of the diamaneticspecies involved in the enhanced state, muon-nuclearlevel crossing measurements µ LCR experiments, wereperfomed at the Paul Scherrer Institute, Switzerland. For -0.15-0.1-0.05 0 0.05 0.1 0.15 0.2 0 2 4 6 8 10 A s y mm e t r y Time (10 -6 s) E=+1kV/cmE= - FIG. 3: Experimental spectra measured at T = 120 K inSI-GaAs after electric field treatment. Red circles: E =+1 kV/cm. Blue triangles: E = − Mu +BC the main resonance in our geometry (H parallelto < > ) is expected around H ∼ − T Ga the resonances are observed at H <
500 Oe [12].For a treated sample we failed to record any resonancein the field range of Mu − T Ga . Resonances were, however,detected at a field of ∼ T = 120 K, and results for both small positive and neg-ative electric fields are presented in Fig. (4). The reso-nance amplitude does not depend on the direction of theelectric field, suggesting that the enhanced diamagneticsignal does not belong to the Mu +BC state (at least duringthe time window of µ SR measurements ∼ − s). Wetherefore need to consider that the tetrahedral As voidmay be supporting the Mu +T As state.The semi-insulating properties of the commercial ver-tical boat (VB) GaAs substrates are achieved by thecompensation of shallow (residual or intentionally added)centers by the deep native centers. There are twomain models discribing the compensation mechanism.A widely accepted theory [13] suggests that the com-pensation can be attributed to the interaction betweenthe deep intrinsic donors EL2 (an As Ga -related defectwhich is present at the level N EL2 ∼ cm − ), resid-ual donors (usually Si) and shallow carbon acceptorswhich are intentionally introduced in GaAs at the level of n C ∼ . × cm − . However, the Fermi level (EF) is of-ten found to lie below the EL2 midgap level, implying theexistence of deep acceptor defects which can act like holetraps. Thermally stimulated (TS) current in conjunctionwith TS Hall effect measurements [14] have demonstratedthat the concentration of the deep hole traps in SI-GaAsis of the same order as the concentration of the deep elec-tron traps, and that the pinning of the Fermi level on the A s y mm e t r y Magnetic Field (Oe) A s y mm e t r y FIG. 4: Muon-nuclear level crossing integral spectra measuredat T = 120 K in SI-GaAs in the treated state. Top panel: E =+1 . E = − . f = 1 kHz. The amplitude of the resonanceat ∼ EL2 midgap may be explained as a result of the balancebetween deep electron and hole traps.The capture of electrons by the EL2 + traps occurs overa configuration barrier with energy of ∼ .
066 eV [15].The applied electric field increases the energy of the car-riers and results in enhanced phonon-assisted trappingover the configurational barrier. A similar behavior isobserved for the holes traps. Capacitance spectroscopymeasurements [16] have revealed that at low tempera-tures ( T ≃
90 K) a strong electric field ( E ≃ V/cm)causes an increase of both electron and hole capture crosssections by five orders of magnitude for the deep traps inGaAs.Based on these conclusions, the following model for theformation of the enhanced diamagnetic state measured inSI-GaAs is proposed. At room temperature, the initialstate of the sample is characterized by ionized residualshallow donors, partially ionized EL2 traps and filled byelectrons or ”ionized” deep acceptor levels (empty holestraps) and ionized shallow acceptor centers. Muon im-plantation at low temperatures results in the production of non-eqilibrium track carriers (electrons and holes) in-side the sample. If a large electric field is applied, thesecarriers are transported through the bulk of the sampleand are effectively captured by deep traps. At low tem-peratures the thermal emission of electrons/holes fromdeep traps is negligible, and consequently the sample istransfered to a metastable-like state. If the traps arefilled, the life-times of non-equilibrium carriers are in-creased and the track electrons/holes can interact withthe neutral Mu on a microsecond time scale. By con-sidering how small electric fields of opposite polarity af-fect muonium to diamagnetic state conversion one canconclude that muonium interacts with the its own trackholes. As the temperature in increased, electrons andholes are emitted from the filled deep traps, life-timesfor the excess carriers are reduced and there are no holesavailable to promote state conversion regardless of thesign of the applied electric field. The essential part ofthis model is the presence of both types of deep traps. Incontrast, if one follows the simplest model of the SI-GaAscompensation mechanism (where only one deep EL2 elec-tron trap is involved) then the electric field treatment willresult to the filling of these traps and the excess hole life-time would not increase.In conclusion, we have used the µ SR technique in com-bination with electric fields and RF measurements toprobe non-equilibrium carrier dynamics in SI-GaAs. Be-low 120 K, the simultaneous bipolar injection (from themuon beam) and treatment of the sample with high elec-tric fields results to the neutralization of both electronand hole deep traps. In the neutralized sample, excessholes from the muon track can be captured by muoniumspecied formed in the void created by the tetrahedralcage of four As atoms, where it acts as a filled donor.Our results point to the scenario where not only deepEL2 centers but also deep hole traps are involved in thecompensation mechanism in SI-GaAs.This work was partially performed at ISIS PulsedMuon Source, United Kingdom, and at the Swiss MuonSource Facility of the Paul Scherrer Institute, Villigen,Switzerland. ∗ Electronic address: [email protected][1] H. Ohno, Science , 951 (1998).[2] N.M. Johnson, R.D. Burnham, R.A. Street, andR.L. Thornton, Phys. Rev. B , 1102 (1986).[3] B.D Patterson, Rev. Mod. Phys. , 69 (1988).[4] R.L. Lichti, H.N. Bani-Salameh, B.R. Carroll,K.H. Chow, B. Hitti and S.R. Kreitzman, Phys.Rev. B , 045221 (2007).[5] D.G. Eshchenko, V.G. Storchak , and G.D. Morris, Phys.Lett. A , 226 (1999).[6] D.G. Eshchenko, V.G. Storchak, J.H. Brewer,R.L. Lichti, Phys. Rev. Lett. , 226601 (2002).[7] D.G. Eshchenko, V.G. Storchak, B. Hitti, S.R. Kreitz- man, J.H. Brewer, K.H. Chow, Physica B , 244(2003).[8] Y. Morozumi, K. Nishiyama, and K. Nagamine, Phys.Lett. A , 93 (1986).[9] D.G. Eshchenko, V.G. Storchak, J.H. Brewer, G.D. Mor-ris, S.P. Cottrell and S.F.J. Cox, Phys. Rev. B , 035105(2002).[10] B. Hitti, S.R. Kreitzman, R. Lichti, T. Head, T.L. Estle,D. Wynne, E.E. Haller, Physica B , 554 (2000).[11] B.E. Schultz, K.H. Chow, B. Hitti, R.F. Kiefl,R.L. Lichti, and S.F.J. Cox, Phys. Rev. Lett. , 086404(2005).[12] K.H. Chow, R.F. Kiefl, W.A. MacFarlane, J.W. Schnei- der, D.W. Cooke, M. Leon, M. Paciotti, T.L. Estle,B. Hitti, R.L. Lichti, S.F.J. Cox, C. Schwab, E.A. Davis,A. Morrobel-Sosa, and L. Zavieh, Phys. Rev. B , 14762(1995).[13] D.E. Holmes, R.T. Chen, K.R. Elliott, and Kirkpatrick,Appl. Phys. Lett. , 46 (1982).[14] R. Kiliulis, V. Kazukauskas, and J.C. Bourgoin, J. Appl.Phys. , 69511 (1996).[15] A. Mitonneau, A. Mircea, G.M. Martin, D. Pons, Rev.Phys. Appl. , 853 (1979).[16] V.Ya. Prinz and S.N. Rechkunov, Phys. Stat. Sol. (B)118