Microwave manipulation of electrically injected spin polarized electrons in silicon
aa r X i v : . [ c ond - m a t . m e s - h a ll ] D ec Microwave manipulation of electrically injected spin polarized electrons in silicon
C. C. Lo ∗ and J. J. L. Morton London Centre for Nanotechnology, University College London, London WC1H 0AH, U.K. andDepartment of Electronic and Electrical Engineering,University College London, London WC1E 7JE, U.K.
J. Li and I. Appelbaum
Department of Physics and the Center for Nanophysics and Advanced Materials,University of Maryland, College Park, Maryland 20742, USA (Dated: November 2, 2017)We demonstrate microwave manipulation of the spin states of electrically injected spin-polarizedelectrons in silicon. Although the silicon channel is bounded by ferromagnetic metal films, weshow that moderate microwave power can be applied to the devices without altering the deviceoperation significantly. Resonant microwave irradiation is used to induce spin rotation of spin-polarized electrons as they travel across a silicon channel, and the resultant spin polarization issubsequently detected by a ferromagnetic Schottky barrier spin detector. These results demonstratethe potential for combining advanced electron spin resonance techniques to complement the studyof semiconductor spintronic devices beyond standard magnetotransport measurements.
The spin degree-of-freedom of charge carriers in bulksemiconductors has been studied by electron spin reso-nance (ESR) for decades, yielding invaluable spin relax-ation information [1, 2]. These measurements demon-strated that the spin relaxation times in group IV semi-conductors, notably in silicon, are quite long. Silicon-based spintronic devices are especially relevant due tothe widespread use of the material in conventional micro-electronics, and hence using silicon as the basis for nextgeneration spintronic devices is even more attractive [3].However, while the spin degree of freedom in semiconduc-tor spintronic devices is easily accessible in many groupIII-V and II-VI compound materials with optical tech-niques, these approaches are difficult to implement withgroup IV devices due to the indirect nature of the bandgap [4], although optical detection of spin injection insilicon has been demonstrated [5]. It is because of thisdifficulty that most group IV-based semiconductor spin-tronic devices are studied by quasistatic electrical mag-netotransport techniques. These approaches include thenon-local 4-terminal detection of spin accumulation withopen-circuit voltage [6, 7] or spin current with hot elec-tron techniques [8–11], both of which reveal spin-valveand spin precession effects, and can be used to determinespin-related properties such as spin diffusion length andrelaxation times in prototypical spintronic devices.It is thus interesting to consider incorporating elec-tron spin resonance techniques to aid the study of theunderlying physics, as advanced pulse sequences will al-low the elucidation of the spin dynamics in these devices.We note that ferromagnetic resonance, a closely relatedtechnique, has been recently utilized to induce dynamicalspin injection (spin pumping) in degenerately doped sil-icon [12]. In the present work, we operate at conditionsresonant to the electrons in nearly intrinsic silicon, far ∗ [email protected]. away from the ferromagnetic resonances of the metal lay-ers in the device. We show that resonant microwave ra-diation can be used to induce spin rotation in the siliconchannel without severely affecting the device operation,thus paving the way toward pulsed ESR spin manipula-tion.The four-terminal devices studied here consist of twoferromagnetic (FM) thin films sandwiching a silicon (Si)layer in between, forming a spin-valve [8]. A schematicof the device and definitions of the voltages and currentsare shown in Fig. 1. The three components of the de- Channel DetectorEmitterCu/CoFe Al O Al/Cu NiFe/Cu V E V Si n -Si I D V/Asignal I BFM I C I A i -Si L = 225μm B B
Spin-valve Hanle
FIG. 1. Schematics and energy band diagram of the fourterminal silicon spintronic device consisting of the emitter,channel and detector. The ferromagnetic layers are shadedin gray. Definitions of bias voltages, voltage polarities andcurrent components (labelled in red) are also indicated. Inset:Orientation of the ferromagnetic layers relative to the appliedmagnetic field B in the spin-valve and Hanle measurementgeometries. vice include the emitter, the Si channel and the detector.The emitter consists of a CoFe FM cathode, Al O tun-nel barrier and Al/Cu anode for injecting spin-polarizedelectrons into the channel. The emitter covers an areaof approximately 0 .
02 mm , and is activated by a biasvoltage of V E . − . I C ) increasing quasi-exponentially for more negative voltages. At V E sev-eral hundred mV above this threshold, only ∼ L = 225 µ m. By applying differentbias voltages across the silicon channel ( V Si ), the aver-age electron transit time to the detector can be varied.The spin detector consists of a buried FM (BFM) layerof NiFe, and a Schottky junction with an n -doped (phos-phorus) Si substrate. The current measured from thedetector ( I D , “signal” in Fig. 1) has a magnitude whichis ∼ I D depends onthe polarization of the electrons reaching the detectordue to spin-dependent inelastic scattering in the BFM,and hence it is a measure of spin current. More detailsof the device fabrication and operation can be found inRef. 13.All measurements were carried out in an X-band ( ∼ . τ s ) and carrier transit times ( τ t ) across the silicon chan-nel. In all the measurements reported we used an emitterbias of V E = − . I C ≈ −
55 mA) as it was foundto give the best signal-to-noise ratio. Fig. 2(a) shows thenormalized spin-valve measurements of I D for various V Si at a cryostat temperature of T = 20 K (the actual de-vice temperature will be discussed shortly). The insetof Fig. 2(b) shows precession oscillations observed whenmonitoring I D in the Hanle configuration. These oscilla-tions unambiguously show that I D does indeed provide aspin-polarization signal [9], and are well described by aspin drift-diffusion model using current-sensing spin de-tectors [14, 15]:∆ I D ∝ Z √ πDt Lt e − L − t/τ t)24 Dt e − t/τ s cos( ωt ) dt, (1)where D is the electron carrier (and spin) diffusion coeffi-cient and ω = gµ B B/ ~ the Larmor precession frequency,with g being the Land´e g -factor, µ B the Bohr magne-ton and ~ the reduced Planck’s constant. The transit B (mT) I D ( B ) / I P D −10 −5 0 5 100.800.850.900.951.00 (a)(b)(c) V Si = 0V-2V-6V-15V-20V-10V PAP PAP PP τ t (ns) I D P − I D AP / I D P + I D AP E = | V Si | / L (V/cm) τ t / ( n s − )
0 500 1000 1500 20000.000.050.100.150.200.250.300.35 Δ I D ( n A ) τ s = 140 ± 37 ns B (mT)−80 −40 0 40 80−0.1 0 0.1 FIG. 2. (a) Spin-valve measurements carried out at a cryo-stat temperature of 20 K. (b, inset) Precession oscillationsmeasured in the Hanle configuration for V Si = −
40 V (red),with the fitted model shown in dashed lines (gray), and theextracted peak electron transit times shown in (b). (c) Spinrelaxation time τ s extraction based on the correlation betweenspin-valve signal amplitude and peak electron transit times.The blue horizontal lines indicate the FWHM of the transittime distribution. time distribution across the channel can be found by ex-amining the Fourier transform of the precession oscilla-tions [16]. Fig. 2(b) shows the peak transit time as theapplied electric field, E = | V Si | /L , is varied. In a per-fectly intrinsic silicon channel where no band bending isexpected, τ t should simply scale inversely with the ap-plied electric field in drift dominated carrier transport.The two distinct regimes observed indicate that resid-ual band bending and diffusive transport plays an in-creasingly important role in the low electric field regimeof E <
500 V/cm. We combine the spin-valve and τ t data to deduce τ s of electrons in the silicon channel as( I PD − I APD ) / ( I PD + I APD ) ∝ e − τ t /τ s , where the superscriptsP and AP correspond to the magnetic alignment of thetwo FM layers in the parallel and anti-parallel configu-rations, respectively. The results are shown in Fig. 2(c),and we obtain τ s = 140 ±
37 ns. The blue horizontal barsat each data point indicate the FWHM of the transit timedistribution, illustrating the significant diffusion-inducedbroadening of carrier transit times for small drift fields.The value of τ s , which is equivalent to the spin-latticerelaxation time T in ESR literature, is significantlyshorter than would be expected at the cryostat tem-perature of T = 20 K [17], and indicate heating inthe device up to ≈
120 K. Sample heating is also con-firmed from standard ESR measurements of the back-ground phosphorus signal in the n -Si detector, as wefound significant reduction in the hyperfine splitting ofthe phosphorus donor-bound electrons when the emit-ter bias was increased[18]. In fact, the hyperfine-splitdonors can no longer be resolved in ESR for V E . − . I C ≈ −
16 mA), indicating the sample temperature hasalready risen to above the carrier freeze-out temperatureof ≈
70 K. We attribute this large amount of heating tothe inefficient thermal anchoring of the sample on theprinted circuit board in the flow cryostat, which is espe-cially important when large currents are passed throughthe emitter for hot electron injection.To investigate the effects of resonant microwaves on thedevice, we turn on and fix the microwave excitation to9.73 GHz at 80 mW, and increase the in-plane magneticfield B to around 350 mT, which corresponds to the reso-nance condition for g ≈ I D provides a signal which is proportional to theprojection of the spin polarization along the magnetiza-tion of the injecting FM. Fig. 3(a) shows the variationsin I D as B is swept, with V Si = − I D is observed at approximately 348 mT.To confirm that the origin of the resonance signalcomes from microwave-induced variations in the electronspin polarization reaching the detector, and not due toother spin-dependent transport mechanisms or bolomet-ric detection in either the Si channel or n -Si detector, weexamine the orientation and source of spin resonance sig-nal more carefully. We also employ lock-in detection with B (mT) ∂ I / ∂ B ( p A / m T )
346 347 348 349 350−10 0 10 20 30 40 50 60 70 I BFM (x10 -2 ) I D S p i n - v a l v e H an l e I D ( p A ) (b)(a) g = 2.004(2) FIG. 3. Electrically detected spin resonance signal of the sili-con spintronic device. (a) Resonant signal detected in the DCcurrent change in detector current I D in the spin valve config-uration. (b) Lock-in detection of the buried ferromagnet cur-rent I BFM (red) and detector current I D (blue, middle trace)in the spin-valve configuration. When configured in the Hanlegeometry, the resonance signal in I D vanishes (blue, bottomtrace). magnetic field modulation, with 0 . − . I D . The current collected by the BFM, shown as the toptrace in Fig. 3(b), does not reveal any resonance signalat all, implying that the origin of the signal is not dueto spin-dependent transport mechanisms in the channellayer, but rather direct spin manipulation by the resonantoscillating magnetic field.To further eliminate the possibility that spin-dependent transport processes in the n -doped Si detectorare responsible for the resonance signal in I D , we rotatethe sample to the Hanle geometry. In this configuration,the signal disappears as shown in the bottom trace ofFig. 3(b), ruling out spin-dependent transport in the n -Sior bolometric detection as the source of the resonant sig-nal as no anisotropy is expected from such mechanisms.Indeed, most spin-dependent scattering processes, suchas those among donor-conduction electron or conductionelectron-conduction electron, are expected to vanish atdevice temperatures of T >
20 K [19, 20]. The onlystrong spin-dependent transport mechanisms in siliconat these temperatures would be spin-dependent hoppingor recombination [21], neither of which are applicablefor the present (undoped) devices. In fact, the van-ishing signal in the Hanle geometry is to be expected, − Δ I D P , r e s / ( I D P − I D AP ) E = | V Si | / L (V/cm) FIG. 4. Electrically detected spin resonance signal amplitudefor various applied electric fields at a cryostat temperatureof 20 K. The dashed line corresponds to fits to the data asdescribed in the main text. since in this case the injected spin polarized electrons arequickly depolarized by the strong out-of-plane magneticfield due to rapid Larmor precession, making the reso-nant microwave ineffective in inducing any signal vari-ation. These measurements confirm that the resonancesignal in I D in the spin-valve configuration is indeed dueto microwave-induced rotation of the electron spins inthe silicon channel, and its subsequent projection to theBFM magnetization, as detected by I D .We can now examine the effect of the applied electricfield on the resonant change in detector current. Theresonant microwave for a given power induces rotation ofthe electron spins at the Rabi frequency ω = gµ B B / ~ ,where B is the amplitude of the microwave magneticfield component. As the average τ t increases (by reducing E ), so should the electron rotation angle, and hence anincreased spin resonance signal. Neglecting spin diffusioneffects, we can show that:∆ I P , res D I PD − I APD ≈ sin ( ω τ t / , (2)where ∆ I P , resD = I P , on − resD − I P , off − resD is the integratedelectrically detected spin resonance signal amplitude af- ter lock-in detection. Fig. 4 shows the evolution of∆ I P , resD normalized in this way for various applied elec-tric fields across the channel. The dashed line is a fitwith Eq. 2 assuming B = 30 µ T. We note that due tothe relatively small B field in continuous wave measure-ments, ω ≈ × rad/s, which is slow compared withthe transit times achievable in the device, hence no Rabi-like oscillations are expected. In addition, the significantbroadening of the transit time distribution (Fig. 2(c)) di-minishes the achievable oscillation amplitudes. We alsofound increased significance of microwave rectification ef-fects at low bias electric fields; improved sample designsoptimized for microwave resonator-based ESR measure-ments can be used to reduce such detrimental rectifica-tion effects [22], and allow high microwave power pulsedESR measurements to be performed on the devices.In summary, we have applied standard magnetotrans-port measurements and combined it with electrical de-tection of spin resonance to study prototypical siliconspintronic devices. We have shown that the applied spinresonance microwaves can be used to manipulate the in-jected electron spin states in the silicon transport channelin the spin valve geometry. The vanishing resonance sig-nals in the Hanle geometry clearly rules out other spin-dependent transport mechanisms as the source of theresonance signal. This combination of electron spin res-onance and magnetotransport measurements can be aninvaluable tool for investigating other spintronic mate-rials and devices, as electron spin resonance techniquescan provide precise spectroscopic information and unam-biguous interpretation of the temporal spin dynamics inthese systems. ACKNOWLEDGMENTS
Work at UCL is supported by the European Re-search Council under the European Communitys SeventhFramework Programme (FP7/20072013)/ERC (grantagreement no. 279781). C.C.L. is supported by the RoyalCommission for the Exhibition of 1851. J.L. and I.A. ac-knowledge the support of the Maryland NanoCenter andits FabLab at U. of Maryland, and funding by the Of-fice of Naval Research under contract N000141110637,the National Science Foundation under contracts ECCS-0901941 and ECCS-1231855, and the Defense ThreatReduction Agency under contract HDTRA1-13-1-0013.J.J.L.M. is supported by the Royal Society. [1] J. H. Pifer. Microwave conductivity and conduction-electron spin-resonance linewidth of heavily doped si:pand si:as.
Physical Review B , 12:4391, 1975.[2] Y. Ochiai and E. Matsuura. Esr in heavily doped n-type silicon near metal-nonmetal transition.
Phys. Status Solidi A , 38:243, 1976.[3] R. Jansen. Silicon spintronics.
Nature Materials , 11:400,2012.[4] I. ˇZuti´c, J. Fabian, and S. Das Sarma. Spintronics: Fun-damentals and applications.
Reviews of Modern Physics , Appl. Phys. Lett. , 94:032502, 2009.[6] M. Johnson and R. H. Silsbee. Spin-injection experiment.
Physical Review B , 37(10):5326, 1988.[7] Xiaohua Lou, Christoph Adelmann, Scott Crooker, EricGarlid, Jianjie Zhang, K. Reddy, Soren Flexner, ChrisPalmstrøm, and Paul Crowell. Electrical detection of spintransport in lateral ferromagnet-semiconductor devices.
Nat. Phys. , 3:197–202, March 2007.[8] Ian Appelbaum, Biqin Huang, and Douwe Monsma. Elec-tronic measurement and control of spin transport in sili-con.
Nature , 447:295–298, May 2007.[9] B. Huang, D. J. Monsma, and I. Appelbaum. Coherentspin transport through a 350 micron thick silicon wafer.
Physical Review Letters , 99:177209, 2007.[10] Biqin Huang and Ian Appelbaum. Time-of-flight spec-troscopy via spin precession: The larmor clock andanomalous spin dephasing in silicon.
Phys. Rev. B ,82:241202, 2010.[11] P. Li, J. Li, L. Qing, H. Dery, and I. Appelbaum.
Phys.Rev. Lett. , 2013.[12] E. Shikoh, K. Ando, K. Kubo, E. Saitoh, T. Shinjo, andM. Shiraishi. Spin-pump-induced spin transport in p-type si at room temperature.
Physical Review Letters ,110:127201, 2013.[13] Jing Li, Lan Qing, Hanan Dery, and Ian Appelbaum.Field-induced negative differential spin lifetime in silicon.
Phys. Rev. Lett. , 108:157201, 2012. [14] I. Appelbaum and D. J. Monsma. Transit-time spinfield-effect transistor.
Applied Physics Letters , 90:262501,2007.[15] Jing Li and Ian Appelbaum. Modeling spin transportwith current-sensing spin detectors.
Appl. Phys. Lett. ,95:152501, 2009.[16] H. J. Jang and I. Appelbaum. Spin polarized electrontransport near the si/sio interface. Physical Review Let-ters , 103:117202, 2009.[17] J. L. Cheng, M. W. Wu, and J. Fabian. Theory of thespin relaxation of conduction electrons in silicon.
PhysicalReview Letters , 104(1):016601, 2010.[18] D. E. Dugdale, S. D. Lacey, and G. Lancaster. A temper-ature dependent hyperfine interaction in n-type silicon.
J. Phys. C: Solid St. Phys. , 4:654, 1971.[19] C. C. Lo, V. Lang, R. E. George, J. J. L. Morton, A. M.Tyryshkin, S. A. Lyon, J. Bokor, and T. Schenkel. Elec-trically detected magnetic resonance of neutral donorsinteracting with a two-dimensional electron gas.
Physi-cal Review Letters , 106(20):207601, 2011.[20] C. C. Lo.
Electrical detection of spin-dependent transportin silicon . PhD thesis, University of California, Berkeley,2011.[21] M. S. Brandt, S. T. B. Goennenwein, T. Graf, H. Huebl,S. Lauterbach, and M. Stutzmann. Spin-dependenttransport in elemental and compound semiconductorsand nanostructures.
Physica Status Solidi (C) , 1(8):2056,2004.[22] C. C. Lo, F. R. Bradbury, A. M. Tyryshkin, C. D. Weis,J. Bokor, T. Schenkel, and S. A. Lyon. Suppression of mi-crowave rectification effects in electrically detected mag-netic resonance measurements.