Low voltage control of ferromagnetism in a semiconductor p-n junction

The concept of low-voltage depletion and accumulation of electron charge in semiconductors, utilized in field-effect transistors (FETs), is one of the cornerstones of current information processing technologies. Spintronics which is based on manipulating the collective state of electron spins in a ferromagnet provides complementary technologies for reading magnetic bits or for the solid-state memories. The integration of these two distinct areas of microelectronics in one physical element, with a potentially major impact on the power consumption and scalability of future devices, requires to find efficient means for controlling magnetization electrically. Current induced magnetization switching phenomena represent a promising step towards this goal, however, they relay on relatively large current densities. The direct approach of controlling the magnetization by low-voltage charge depletion effects is seemingly unfeasible as the two worlds of semiconductors and metal ferromagnets are separated by many orders of magnitude in their typical carrier concentrations. Here we demonstrate that this concept is viable by reporting persistent magnetization switchings induced by short electrical pulses of a few volts in an all-semiconductor, ferromagnetic p-n junction.

The concept of low-voltage depletion and accumulation of electron charge in semiconductors, utilized in field-effect transistors (FETs), is one of the cornerstones of current information processing technologies. Spintronics which is based on manipulating the collective state of electron spins in a ferromagnet provides complementary technologies for reading magnetic bits or for the solidstate memories. 1 The integration of these two distinct areas of microelectronics in one physical element, with a potentially major impact on the power consumption and scalability of future devices, requires to find efficient means for controlling magnetization electrically. Current induced magnetization switching phenomena 2 represent a promising step towards this goal, however, they relay on relatively large current densities. The direct approach of controlling the magnetization by low-voltage charge depletion effects is seemingly unfeasible as the two worlds of semiconductors and metal ferromagnets are separated by many orders of magnitude in their typical carrier concentrations 3 . Here we demonstrate that this concept is viable by reporting persistent magnetization switchings induced by short electrical pulses of a few volts in an all-semiconductor, ferromagnetic p-n junction.
To establish the physics behind this main result of our work, we organized the paper as follows: After describing the basic structure of the device we present direct electrical measurements of simultaneous charge depletion and "depletion" of the magnetization, i.e. of the shifts in the ferromagnetic Curie temperature, by applying a few volts. This provides the ultimate test of the ability to manipulate the magnetic state by low-voltages. We point out that the method we introduce for accurate electrical measurement of T c in the microdevice is essential as standard magnetometry is not feasible at these small sample dimension.
We then proceed by discussing the nature of the magnetotransport responses in our system. They are remarkable on their own for their large magnitude and voltage-dependence and provide the means for detecting the magnetic moment reorientations induced by lowvoltage pulses, again by electrical measurements. The main body of the paper is devoted to presenting and discussing the phenomenology of the electrically controlled magnetization switchings measured as a function of the magnitude and angle of the external magnetic field.
The microscopic physical interpretation of our experiments, based on semiconductor theory modelling, is discussed in the concluding paragraphs.
The schematic cross-section of the III-V heterostructure used in our study is shown in Fig.1(a). It is a semiconductor p-n junction FET specially designed to accommodate ferromagnetism in the p-type region and its efficient depletion by low voltages. From the top, the structure comprises a 5 nm thick approximately 2.5% Mn-doped GaAs capped by 2 nm of undoped GaAs to prevent oxidation of the underlying transition metal doped semiconductor film. These two top layers were grown by low-temperature molecular-beamepitaxy (MBE) to avoid Mn precipitation. The 2.5% doping was chosen to pass the insulatorto-metal transition threshold which for the moderately deep Mn Ga acceptor is between 1-2% and to achieve robust ferromagnetic state with Curie temperature T c ≈ 30 K, while still minimizing the number of unintentional interstitial-Mn impurities. 4,5,6 (The interstitial Mn is highly mobile at the growth temperature and its diffusion into the p-n junction would result in detrimental leakage currents.) The Curie temperature measured by SQUID in an unpatterned piece of the wafer is comparable to maximum T c 's achieved at the same Mn-doping in thicker films, indicating a very good quality of our ultra-thin ferromagnetic semiconductor epilayer.
The n-type gate electrode is formed by a highly Si-doped (2 × 10 19 cm −3 ) GaAs grown by high-temperature MBE. The large electron doping is required in order to achieve appreciable and voltage dependent depletion of the ferromagnetic p-region with hole doping ∼ 10 20 cm −3 .
The built-in electrostatic barrier due to the depletion effect at the p-n junction is further supported by inserting an undoped AlGaAs spacer with a large conduction band off-set to the n-type GaAs; for the same reason a large valence band offset AlAs spacer is placed next to the p-type (Ga,Mn)As.
Self-consistent numerical simulations, 7 shown in Fig. 1(b), confirm that sizable depletions are achievable by gating our heterostructure with less than 4 Volts. Measurements discussed below were done at voltages between −1 V (forward bias) to +3 V (reverse bias) for which the leakage currents between the n-GaAs gate and p-(Ga,Mn)As channel were more than two orders of magnitude smaller than the channel currents. The (Ga,Mn)As channel was lithographically patterned in a low-resistance Corbino disk geometry with the inner contact radius of 500 µm and the outer radius of 600 µm.
In Fig. 2 as predicted by the simulations in Fig. 1(b) and with the vicinity of the metal-insulator transition which causes the superlinear increase of R with V g . At 4 K, the increase of R between -1 and +3 V is by more than 100%.
In Fig. 2(b) we show the voltage-dependence of the Curie temperature in the ferromagnetic p-n junction. Our measurement technique is distinct from previous studies which relied on approximate extrapolation schemes based on Arrot plot measurements at finite magnetic fields. 8,9,10 Recent observation and interpretation by the authors 11 of the peak in the zerofield temperature derivative of the resistance at the Curie point in good quality (Ga,Mn)As materials has provided the tool for direct transport measurements of T c in microdevices without relying on any extrapolation schemes. In Fig. 2 directions. Although the specific responses to these symmetries can be very different for the AMR and for the magnetic anisotropy, the presence of the cubic and uniaxial AMR terms and their sensitivity to the gate voltage observed in Fig. 2(d) suggest that the in-plane magnetization orientation itself can be switched at weak magnetic fields by the low voltage charge accumulation or depletion.
A variable width of hysteretic magnetization loops measured at different constant gate voltages, shown in Fig.3(a), is the prerequisite for observing electrically assisted magnetization switchings. Note that electrical measurements of magnetization reorientations utilized    To discuss the detail phenomenology of these persistent low-voltage induced magnetization switchings we present in Figs. 4(a) and (b) field-sweep measurements at fixed field angles spanning the whole interval from 0 to 180 • in 5 • steps. In panels (a) and (b) we show color-maps of the resistance as a function of the field magnitude and angle for -1 V constant voltage and for the +3 V peak-voltage measurements, respectively. The main effect observed in these plots is the overall suppression of the magnitude of the switching fields by depletion. Additionally, the relative suppression is stronger at θ = 0 than at 90 • , as highlighted in Fig. 4(c). This indicates that both the magnitude and ratio between the uniaxial and cubic anisotropy fields is modified by the gate voltage. To quantify the depletion induced modification of the magnetic anisotropy we extracted the anisotropy constants from fitting the measured θ = 0 and 90 • switching fields to a single domain anisotropy energy model, We now discuss the key experimental observations by employing the k · p semiconductor theory approach combined with the mean-field kinetic-exchange model of hole mediated ferromagnetism in (Ga,Mn)As. 4,5 Calculations for 2.5% local moment doping and hole density p ∼ 1 × 10 20 cm −3 , for which the simulations in Fig.1(b) predict hole depletions consistent with the measured variations of the channel resistance at temperatures near T c , yield T c ∼ 20 K and dT c /dp ≈ 1 × 10 −19 Kcm 3 . Both the absolute value of the Curie temperature and the few Kelvin suppression of T c at a ∼ 20% hole depletion predicted by the theory are consistent with our p-n junction simulations and the measured gate-dependent T c values.
The semiconductor theory modelling which includes strong spin-orbit coupling effects in the host semiconductor valence band captures also the sensitivity of magnetocrystalline anisotropies in (Ga,Mn)As to hole density variations. The cubic anisotropy is included by accounting in the k · p model for the zincblende crystal structure of GaAs. The additional weak uniaxial anisotropy is often present in (Ga,Mn)As epilayers but its microscopic origin is not known and we will therefore focus only on the stronger cubic anisotropy term. As shown in Fig. 4(f) tion directions, consistent with the experimental data. The typical magnitudes of K c of ∼ 10 mT are also consistent with experiment and considering the large gate action seen at low temperatures we can also associate, semiquantitatively, the decreasing magnitude of the experimental K c at depletion with the behavior of the theoretical K c at low hole densities.
To conclude we have reported low-voltage control of magnetic properties of a p-n junction FET via depletion effect in the ferromagnetic semiconductor channel. We have shown