Robustness of n-GaAs Carrier Spin Properties to 5 MeV Proton Irradiation
Brennan C. Pursley, Xinlin Song, R. O. Torres-Isea, Eiman A. Bokari, Asghar Kayani, Vanessa Sih
TThe following article has been accepted by Applied Physics Letters.After it is published, it will be found at http://apl.aip.org/ . Robustness of n-GaAs Carrier Spin Properties to 5 MeV Proton Irradiation
Brennan C. Pursley, a) X. Song, R. O. Torres-Isea, E. A. Bokari, A. Kayani, and V. Sih
1, 2 Applied Physics Program, University of Michigan, Ann Arbor, MI 48109 Department of Physics, University of Michigan, Ann Arbor, MI 48109 Department of Physics, Western Michigan University, Kalamazoo, MI 49008 (Dated: February 03, 2015)
Modern electronic devices utilize charge to transmit and store information. This leaves the informationsusceptible to external influences, such as radiation, that can introduce short timescale charge fluctuationsand, long term, degrade electronic properties. Encoding information as spin polarizations offers an attractivealternative to electronic logic that should be robust to randomly polarized transient radiation effects. As apreliminary step towards radiation-resistant spintronic devices, we measure the spin properties of n-GaAs asa function of radiation fluence using time-resolved Kerr rotation and photoluminescence spectroscopy. Ourresults show a modest to negligible change in the long-term electron spin properties up to a fluence of 1x10 (5 MeV protons)/cm , even as the luminescence decreases by two orders of magnitude.The vast majority of modern technology relies on con-trolling electronic charge within a circuit. Timing, lo-cation, and quantity of the charge are the fundamentalparameters for logic operations. Anything that can dis-rupt control of these parameters is an information pro-cessing hazard. Radiation filled environments are a dif-ficult challenge for electronic logic as particle collisionscan randomly introduce large quantities of charge in theshort term and degrade circuit properties in the longterm. Spin based logic has been proposed as an alterna-tive that would offer novel functionality and the trans-mitted information should be inherently robust to shortterm charge effects. A radiation-resistant spintronic de-vice should account for bursts of induced electrical cur-rent so that it would not be damaged, while accuratelymeasuring the quantity of spin current. An ideal devicewould use a pure spin current such that charge and spinbehavior are completely decoupled.In order to fabricate a radiation-resistant spintronicdevice, a material must be chosen that is relevant forspintronics applications and largely maintain its spin de-pendent properties after irradiation. In this paper, weexplore the effects of irradiation on the spin propertiesof bulk Si-doped n-GaAs samples cleaved from an off-the-shelf wafer. We expose several samples to protonirradiation, and then characterize them using photolu-minescence (PL) and gamma spectroscopy. We thenperform resonant spin amplification (RSA) using pump-probe Kerr rotation to extract the spin dependent param-eters. Our results show that the spin lifetime and g-factorof bulk n-GaAs, doped near the metal-to-insulator tran-sition, is largely unaffected by proton irradiation. Werecommend n-GaAs for further study as a candidate forradiation-resistant spintronic devices.All samples were cleaved into 4 mm x 4 mm x 0.5 mm a) Electronic mail: [email protected] chips from the same bulk Si-doped n-type GaAs wafer.The parent wafer was 2 in diameter x 0.5 mm thick, withthe following manufacturer specifications: carrier concen-tration of (4.3-6.2)x10 cm − , mobility of (3450-3880)cm /V · s, and resistivity of (2.8-3.9)x10 − Ohm · cm. Onesample was set aside as a reference. Six samples were irra-diated at Western Michigan University’s 6.0 MV Van deGraaf accelerator facility with a selection of 5.0 MeV H+ion fluences: 2.5x10 , 1x10 , 1x10 , 1x10 , 1x10 ,and 1x10 protons/cm . For comparison, equipmentin a satellite monitoring ocean features operating at a1334 km orbit above the earth, with a 63 deg angle ofinclination and 100 mil Aluminum shielding, will experi-ence a fluence of approximately 10 protons/cm in oneyear with energies distributed from 0.1 MeV to 1 GeV,peaked around 10 MeV. It should be noted that dif-ferent proton energies lead to different forms of damage.Low energy protons ( 1 MeV) can become lodged withina material inducing swelling near a surface, while highenergy protons can induce nuclear reactions. We performed PL measurements on all samples usingthe 1.96 eV emission of a HeNe laser at 1 W/cm intensityfor excitation. All samples were mounted in a liquid-Hecooled cryostat with data collection at 10 K. A liquidnitrogen cooled charge-coupled device array (CCD) andgrating spectrometer were used to detect the PL. Fig-ure 1 shows PL collected from samples up to the 1x10 protons/cm fluence. We observed negligible PL signalfrom samples exposed to higher fluence.In Fig. 1a, we label the observed PL peaks P1 throughP6 and plot the relative intensity of each peak and theirdependence on fluence in Fig. 1b. PL maxima P1, cen-tered at 1.512 eV, is attributed to the band-to-band tran-sition. This is supported by the spin dependent datashown in Fig. 3a where the strongest negative polariza-tion and general change in spin dependent behavior oc-curs at ∼ a r X i v : . [ c ond - m a t . m e s - h a ll ] F e b FIG. 1. a) Photoluminescence (PL) measurements of irra-diated and reference samples. b) Fluence dependent trendsof PL maxima labeled P1-P6 in Fig. 1a, plotted using thefunction ( I /I ) − I is the reference, non-irradiated PLintensity. I is the intensity at a given fluence. Black lines areguides to the eye. nor P2 shift their center position with increasing fluence.PL maxima at P3, P4, P5, and P6, centered at 1.443eV, 1.408 eV, 1.326 eV, and 1.297 eV respectively beforeirradiation, shift to lower energy with increasing fluence.P3 and P4 are potentially phonon replicas of P1 and P2while P5 and P6 are likely dominated by impurities. P4also appears to split into two separate peaks with in-creasing fluence, possibly from the introduction of newdefect states or the unequal degradation of multiple PLtransitions. For P4, we only analyzed the lower energypeak.Figure 1b follows the degradation of the PL maximawith fluence, comparing the intensity before and after ir-radiation, using the equation [( I /I ) −
1] = Kφ m , where I is the intensity prior to irradiation, I the intensity af-ter, φ the corresponding fluence, and K the degradation FIG. 2. Gamma spectra taken from the 1x10 protons/cm fluence sample. Spectral peaks are identified with their cor-responding energy in keV. coefficient. The exponent m is determined by the typeof damage introduced to the sample. The introduction ofmid-gap states leads to a linear dependence on fluence,while a high concentration of radiation-induced complexformations leads to a quadratic dependence. Figure 1bshows that the behavior for all peaks ranges from linearto modestly super-linear, as expected for 5 MeV protonirradiation in n-type GaAs. After proton irradiation, the samples exhibited in-creased radioactivity, which persisted in the sample ex-posed to the highest fluence, 1x10 p/cm , even afterseveral weeks. In order to determine the source of ra-dioactivity, we collected gamma spectra on that sample,at room temperature, at the University of Michigan Ad-vanced Physics Teaching Laboratory using a high purityGe solid state detector. The data, shown in Fig. 2, re-veals the likely culprit to be decay of Se. The onlystable isotope of Arsenic, As, has 100% natural abun-dance. There exists a possible nuclear reaction betweena proton and an As neutron, with an energy barrierof only 1.6 MeV and reaction product Se. The half-life of Se is 120 days and decays back to As. Ourgamma spectra match what was reported, for the sameenergy range, in Ref. 8 and measured spectral peak en-ergies were compared to the NuDat 2.6 database fordecay radiation. The vast majority of the signal can beaccounted for by Se decay to As.We performed Kerr rotation measurements utilizing atunable Ti:Sapphire pulsed laser. Each laser pulse was3 ps full-width at half-max (FWHM) in time and gen-erated at a 76 MHz repetition rate. A beam splitterdivided each pulse onto two paths: 1) a pump path withmechanical delay line and 50 kHz photo-elastic modula-tor (PEM); and 2) a linearly polarized probe path withan optical chopper operating at 500 Hz. Upon reflectionfrom the sample surface, the orientation of the probe’s
FIG. 3. a) Fitting parameters as a function of laser energy and fluence, extracted from Eq. 4. θ values are normalized andlines are guides to the eye. The inset is a zoom of the τ dependence on laser energy for clarification. b) Data from the 1x10 protons/cm fluence sample as a function of applied magnetic field (green circles) with fits (black lines). Each curve is labeledwith the corresponding laser energy on the left-hand side of the figure. linear polarization was detected by a Wollaston prism,photo-diode bridge, and two cascaded lock-in amplifiers.The first lock-in was synchronized to the PEM and thesecond was synchronized to the optical chopper. The me-chanical translation stage with a retroreflector delayedthe probe pulse relative to the pump pulse allowing forpicosecond time resolution over a 13 ns scannable range.For all spin-dependent measurements reported here, a 0.4mW probe was delayed by 12.96 ns relative to a 3.0 mWpump. Both beams had 30 µ m FWHM spot diametersand overlapped at the sample surface. All samples weremounted in a liquid-He cooled cryostat with data collec-tion at 10 K.Spin ensembles were generated by exploiting the op-tical selection rules of GaAs. The PEM modulated thepump between right- and left- circular optical polariza-tion to photo-generate the spin polarization of carrierensembles along, or against, the optical path. We per-formed resonant spin amplification (RSA) in the Kerrrotation geometry on samples up to the 1x10 p/cm fluence, extracting the dephasing time and g-factor. RSA is the byproduct of cumulatively measuring sev- eral non-interacting spin ensembles, each generated bytheir own pump pulse, separated in time by the laserrepetition period. Each spin ensemble undergoes Larmorprecession from an externally applied magnetic field anddephasing from various spin scattering mechanisms. Ifthe ensemble dephasing time is comparable to, or longerthan, the laser repetition period, multiple ensembles willbe detected by the probe pulse. If the precession fre-quency is an integer multiple of the laser repetition fre-quency, the collective measurement of the spin ensembleswill yield a constructive interference maxima. By fixingthe delay time and tuning the applied magnetic field, sev-eral instances of constructive maxima can be observed.The derivation of a quantitative model for RSA israther straightforward: utilize a model for single ensem-ble spin precession and dephasing, then sum over a largenumber of ensembles generated at the laser repetitionrate. The governing equation used for a single spin en-semble is ∂ S ( t ) ∂t − Ω × S ( t ) + S ( t ) τ = S δ ( t ) (1)where S ( t ) is the spin polarization per unit volume, S is the polarization at time 0, τ is the dephasing time and δ ( t ) is the Dirac-delta function. Ω is the Larmor pre-cession frequency defined as Ω = µ B g B / ¯ h where µ B isthe Bohr magneton, ¯ h is the reduced Planck’s constant, g is the Lande g-factor, and B is the applied magneticfield. For our experimental setup, we define B = B ˆ x and S = S ˆ z , where ˆ z is the optical axis. We connectthe spin polarization to the amount of Kerr rotation us-ing the relation θ ∝ S z , where θ is the angle betweenthe initial and final linear probe polarizations and S z isthe component of the spin polarization along the opti-cal axis. The resulting equation for the time dependentKerr rotation is θ ( t, Ω) = θ H ( t ) e − t/τ cos (Ω t ) (2)where H ( t ) is the Heaviside step function. We take Eq.2 and then sum over an infinite number of pump pulsesgenerated at the laser repetition rate, t R , to arrive at Eq.3. θ ( t, Ω) = θ ∞ (cid:88) n =0 H ( t n ) e − t n /τ cos [Ω t n ] (3)where t n = t + n t R . By assuming that t >
0, the sumcan be evaluated as a complex geometric series yieldingEq. 4 (an alternate form of Eq. 10 in Ref. 12) which wasused to fit all RSA data. θ ( t, Ω) = θ e − t/τ (cid:0) cos [Ω t ] − e − t R /τ cos [Ω ( t − t R )] (cid:1) sin [Ω t R ] + (cid:0) cos [Ω t R ] − e − t R /τ (cid:1) (4)If τ (cid:28) t R , then Eq. 4 reduces to Eq. 2 as expected.Figure 3 shows selected RSA data with fitting curves,along with the laser energy dependence of all fit values.Each data point was collected at least four times and thenaveraged. Our fits assume a single spin species which, forour samples, should be electrons as the holes are expectedto have sub-picosecond spin lifetimes and our pumpmodulation minimizes dynamic nuclear polarization. There is a distinct, and expected, dependence on laserenergy for the maximum rotation value θ , stemmingfrom the difference in indexes of circular refraction gen-erated by the injected spin polarization. There is alsoan increase in the magnitude of the dephasing time τ with decreasing laser energy below the bandedge. Refer-ence 14 showed that such behavior could be attributed tochanges in absorbed probe power, independent of wave-length, whereby probe-induced photo-excited holes re-duce the spin lifetime via the Bir-Aranov-Pikus (BAP)mechanism. Unexpectedly, the measured spin properties exhibitnegligible change with H+ fluence. The magnitudes of g , and τ , appear to increase for the lowest two fluencesby less than 2% and 40% respectively, relative to the ref-erence sample. | g | and τ then return to their referencevalues at the 1x10 p/cm fluence. The normalized am-plitude of θ shows no clear trend and the raw valueswere all within a factor of 2 of each other. This shows that the ability to optically pump and measure the cir-cular birefringence is robust to the effects of irradiationfor our experiment conditions. The modest behavior ofthe spin properties is in contrast to the orders of magni-tude monotonic decrease in measured PL intensity (seeFig. 1). For comparable fluences of 5 MeV protons tothose measured here, electrical resistivity is reported toincrease by several orders of magnitude. It is possiblethat the measured spin behavior stems from inconsisten-cies in the dopant level of the n-GaAs wafer, though morelikely irradiation played some role as we discuss below.Figure 3a shows that three data points for τ of thereference sample deviate from the rest of the data set.At the three lowest laser energies, the reference sample’sspin lifetime appears to be significantly longer than theirradiated sample’s with substantially higher error bars.As all the data shown was taken with identical settings,including magnetic field resolution, subsequent runs wereconducted to determine the validity of those data points.Doubling the magnetic field resolution did shrink the er-ror bars on the points in question, but it did not appre-ciably shift the central value or other fitting parameters(data not shown). A plausible irradiation mechanismfor the change of low energy lifetimes is the placementof defects within the bandgap. Such defects can formcharge traps, thereby changing the sample’s carrier den-sity and, in some cases, modestly enhancing electricalproperties as has been reported for low fluences of neu-tron irradiation. Some of these charge traps could beintroduced near the valence band, leading to an enhance-ment of the BAP spin dephasing. This would explain theabrupt reduction of spin lifetime in the irradiated sam-ples, compared to the reference sample, below the band-edge.In conclusion, we report the robustness of (4.3-6.2)x10 cm − Si-doped n-GaAs spin properties to 5MeV H+ irradiation. We confirmed sample damage fromirradiation using PL and gamma spectroscopy that areanalogous to data found in the literature. There is anabrupt, though modest, drop in the spin lifetime belowthe bandgap, relative to the reference sample data, whichmay be accounted for by radiation induced charge trapsnear the valence band enhancing the BAP dephasingmechanism. Our results show that GaAs, already a tech-nologically valuable material, has promising attributesfor the development of radiation-resistant spintronic de-vices.This work was supported by the Defense Threat Re-duction Agency Basic Research Award No. HDTRA1-13-1-0013, the National Science Foundation Materials Re-search Science and Engineering Center program DMR-1120923 and under Grant No. ECCS-0844908, the Officeof Naval Research, and the Air Force Office of ScientificResearch. B.P. was supported in part by the Gradu-ate Student Research Fellowship under Grant No. DGE1256260. I. ˇZuti´c, J. Fabian, and S. D. Sarma, Reviews of Modern Physics , 323 (2004), arXiv:0405528 [cond-mat]. B. Huang, D. J. Monsma, and I. Appelbaum, Applied PhysicsLetters , 072501 (2007). D. Awschalom and N. Samarth, Physics , 50 (2009). B. Behin-Aein, D. Datta, S. Salahuddin, and S. Datta, Naturenanotechnology , 266 (2010). H. Dery, Y. Song, P. Li, and I. Zutic, Applied Physics Letters , 082502 (2011). I. ˇZuti´c and H. Dery, Nature materials , 647 (2011). J. Lee, R. Oszwadowski, C. Gø thgen, and I. ˇZuti´c, PhysicalReview B , 045314 (2012). K. Vogt, F. Y. Fradin, J. E. Pearson, T. Sebastian, S. D. Bader,B. Hillebrands, A. Hoffmann, and H. Schultheiss, Nature com-munications , 3727 (2014). J.-Y. Chen, T.-M. Wong, C.-W. Chang, C.-Y. Dong, and Y.-F.Chen, Nature nanotechnology , 845 (2014). A. Johnston,
Reliability and radiation effects in compound semi-conductors (World Scientific Publishing Company, Singapore,2010) pp. 213–217. C. Ascheron and A. Schindler, Nuclear Instruments and Methodsin Physics Research Section B: Beam Interactions with Materialsand Atoms , 163 (1989). L. F. Zakharenkov, V. V. Kozlovskii, and B. A. Shustrov, physicastatus solidi (a) , 85 (1990). S. Khanna, A. Houdayer, A. Jorio, C. Carlone, M. Parentean,and J. Gerdes, IEEE Transactions on Nuclear Science , 2601(1996). A. Sonzogni, “National Nuclear Data Center, information ex-tracted from the NuDat 2 database (ver 2.6),” (2014). J. Kikkawa and D. Awschalom, Physical Review Letters , 4313(1998). R. Giri, S. Cronenberger, M. Vladimirova, D. Scalbert, K. V. Ka-vokin, M. M. Glazov, M. Nawrocki, A. Lemaˆıtre, and J. Bloch,Physical Review B , 195313 (2012). M. M. Glazov and E. L. Ivchenko, Semiconductors , 951(2008). C. J. Trowbridge, B. M. Norman, Y. K. Kato, D. D.Awschalom, and V. Sih, Physical Review B , 085122 (2014),arXiv:1405.0342. B. M. Norman,
Electrical Generation of Spin Polarization inStrained III-V Semiconductors , Ph.D. thesis, University of Michi-gan, Ann Arbor (2014). S. A. Crooker, L. Cheng, and D. Smith, Physical Review B ,035208 (2009). G. L. Bir, A. G. Aronov, and G. E. Pikus, Zh. Eksp. Teor. Fiz , 1382 (1975). V. Brudnyi, M. Krivov, and A. Potapov, Solid State Communi-cations , 117 (1980). V. Brudnyi and M. Krivov, Russian Physics Journal , 45(1980). A. Jorio, C. Rejeb, M. Parenteau, C. Carlone, and S. M. Khanna,Journal of Applied Physics74