A magnetic-field-controlled non-volatile memory device is fabricated by coupling Hall effect and resistive switching effect. The non-volatile property of the device is due to the fact that the Hall voltage of the semiconductor changes the resistance state of the resistive switching unit. By changing the device configuration, the storage can be controlled by magnetic fields in different directions. The parameters of the semiconductors and the resistive switching units are experimentally investigated and simulated to optimize the performance of the devices. The key to increasing the ON/OFF ratio and reducing power consumption is finding a suitable resistance match between the semiconductor and the resistive switching unit. By enhancing the mobility of the semiconductor, the performance of the device can also be significantly improved. This hybrid device provides new insights into the manufacture of magnetic field controlled non-volatile memory devices with potential of integrating computing and storage functions.

Magnetoelectronics combines the characteristics of spin and charge to create novel multifunctional devices with broad application prospects.1 Magnetoelectronic devices can be divided into spin-based devices2–6 and magnetic-field-based devices.7–9 In magnetic-field based devices, moving carriers which have a large mobility in semiconductors will be affected by Lorentz force under the magnetic field. Based on this, diverse devices based on Hall effect, magnetoresistance effect, magnetic-diode effect, etc. have been fabricated.10–13 Among these effects, Hall effect is the most common way to convert magnetic signals into electric signals, and it has attracted much attention in scientific research. In recent years, Hall effect sensors have been widely used in microelectronics, automotive, aerospace and other fields due to their high reliability, high integration level, and low cost.14 However, magnetic-field-based devices are mostly volatile and are not suitable for magnetic memory devices. In order to solve this problem, one feasible way is to manufacture hybrid devices to introduce non-volatility. Some hybrid devices utilize a fringe field of a ferromagnetic film to create a Hall voltage by growing a ferromagnetic film on semiconductor devices.15,16 This paper reports a new method to introduce non-volatility into a Hall device and to fabricate a magnetic field controlled memory device by combining the Hall effect and the resistive switching effect.

The resistive switching effect has been found mostly in metal-insulator-metal (MIM) cells. By electrically stimulating of the MIM cell, its resistance can change between high resistance state (HRS) and low resistance state (LRS). Generally, the insulator materials in MIM structure include various oxides, chalcogenides and organic compounds.17–19 The mechanisms of resistive switching effect can be mainly divided into electrochemical metallization mechanism20,21 (ECM) and valence change mechanism22,23 (VCM) by the formation of conductive filaments. Based on the resistive switching effect, resistance random access memories (RRAMs) have been made as the next generation memory.

Incorporating an RRAM component into an integrated electronic device structure offers several advantages: i) Nonvolatility. The resistance state of the resistive switching unit can be kept at zero static power for many years. ii) Durability. The steady states can set and reset more than 105 times. iii) Speed. The switching speed of RRAM is on the order of a few nanoseconds, and theoretically much faster speed can be achieved. iv) Scalable. The resistance states of RRAM is controlled by the conductive filaments with a size of several tens of nm or even less. v) Compatibility. The materials of RRAM are common metal electrodes and oxides, which are generally compatible with silicon industry.24 

In this paper, a novel magnetic-field-controlled memory device is proposed, in which the Hall effect and the resistive switching effect are coupled. By introducing a non-volatile memory unit in Hall devices, the signals can be stored and controlled by magnetic field. These two effects are coupled in a single device without using any magnetic materials. Furthermore, the device has the potential to integrate computing and memory.

A hybrid magnetic memory device was fabricated as follows: Firstly, a 500 μm-thick lightly doped n-type silicon wafer was prepared and cut into square pieces (6×6 mm) as the substrates. Secondly, intrinsic oxidized layer of silicon was removed by diluted HF solution before being put in the sputtering chamber. Thirdly, four Ti electrodes (1×1 mm, ∼100 nm thickness) were deposited on the silicon to guarantee Ohmic contact. Fourthly, SiOx layer was grown on one of the Hall electrodes as the storage media. Last, a layer of Ag film (∼500×500 μm, 25 nm thickness) was deposited on SiOx as the top electrode. All of these films were deposited by magnetron sputtering system, SiOx films were sputtered by AC source while the rest films were deposited by DC source. X-ray Photospectroscopy (XPS) (Thermo ESCALAB 250Xi) was used for measuring the elements distribution and chemical bonding of the materials. The transport properties were measured at room temperature. A source meter Keithley 2400 was used as the current source, and a voltage meter Keithley 2182 was used to measure the voltage. Magnetic field was provided by an electromagnet from 0 T to 1.5 T.

A hybrid magnetic memory device consists of the following parts, (1) A Hall effect unit (HEU) made of semiconductor (2) Four electrodes in Ohmic contact with the semiconductor (3) Resistive switching unit (RSU) located on top of one of the Hall electrodes. In the beginning, HEU and RSU are considered separately before being combined into one overall device. On the one hand, for an ordinary HEU made of semiconductor (i.e. silicon), the trajectory of electrons is deflected due to the Lorentz force, so that the Hall voltage is proportional to the magnetic field and applied current within a certain range (Fig. 1a). On the other hand, for the RSU with an Ag/SiOx(60 nm)/Ti structure, the basic current-voltage curve is illustrated in Fig. 1b. Here, we define that the bias voltage is positive when the electric potential of Ag electrode is higher. When the positive bias voltage is larger than a specific value Vs, the device can be set from HRS to LRS. Reversely, while the negative bias voltage is larger than another specific value Vr, LRS can be reset to HRS. In RSU, Vs and Vr are unlikely to be identical in all loops, but a sufficiently large voltage can accomplish the resistance states transition process. XPS results of the SiOx layer is shown in Fig. 1c. The chemical valence of the Si is composed of three parts, Si0, Si+ and Si4+, accounting for 24.4%, 12.6% and 63.0%, respectively. XPS results indicate that the insulating layer SiOx contains a large amount of oxygen vacancies. In this structure, SiOx can be regarded as the solid electrolyte through which Ag+ ions can migrate. That is, an electrochemical reaction occurs at the Ag electrode by applying a positive bias voltage to form Ag+ ions. These ions pass through the SiOx layer under electrostatic force until they are reduced to Ag atoms. This leads to the migration of Ag nanoclusters. When the nanoclusters are joined together, conductive filaments are formed and the device resistance is reduced by several orders of magnitude. It is called the SET process in RSU. When a negative bias voltage is applied, the conductive filaments are destructed due to the Joule heating effect, and the Ag nanoclusters migrate in the opposite direction. This process is called RESET in RSU. Since the conductive filaments are made of silver, RSU can be used for non-volatile storage of electric signals.

FIG. 1.

(a) Voltage-current curves of separate n-Si Hall effect unit under different magnetic fields. (b) Current-Voltage loops of Ag/SiOx/Si resistive switching unit. (c) Chemical valence analysis of SiOx layer by XPS. (d) Element distribution of Ag/SiOx/Ti/Si structure measured by XPS with Ar ion-exfoliation.

FIG. 1.

(a) Voltage-current curves of separate n-Si Hall effect unit under different magnetic fields. (b) Current-Voltage loops of Ag/SiOx/Si resistive switching unit. (c) Chemical valence analysis of SiOx layer by XPS. (d) Element distribution of Ag/SiOx/Ti/Si structure measured by XPS with Ar ion-exfoliation.

Close modal

When HEU and RSU are combined, which means that the RSU is grown on one Hall electrode and the top electrode Ag is connected to the other Hall electrode. The element distribution measured by Ar ion peeling in an Ag/SiOx/Ti/Si structure by XPS is shown in Fig. 1d, which indicates a multilayer film structure. Since the HEU is a four-terminal device, the electrodes are labeled as 1, 2, 3 and 4 in a counterclockwise order, respectively. Due to the symmetry of the device and the magnetic field, there are two measurement configurations A and B, as shown in Fig. 2a. In configuration A(B), the RSU is grown on electrode 4(2). During the measurement, a current is applied between electrodes 1 and 3 using a source meter (Keithley 2400) and a voltmeter (Keithley 2182) is measuring the output voltage between electrodes 2 and 4. It is defined that the current applied from the electrode 1 to 3 is positive, and the voltage measured from electrode 2 to 4 is also positive. Simultaneously, a magnetic field is added perpendicular to the surface of the sample and its positive direction is defined as downward. For configuration A and B, the results of current-voltage measurements are shown in Fig. 2b and Fig. 2c, respectively. Initially, the resistance state of the RSU is HRS. When the applied current I is small, the voltage drop across the RSU is less than Vs, so the resistance state of the RSU will remain at HRS. As the applied current increases, the voltage on the RSU becomes larger until HRS turns to LRS. In this case, as the resistance of the RSU drops, the output voltage will decrease, which causes a portion of the charge to flow through the RSU instead of accumulating on the electrode 2 or 4. Since the HEU is quadruple symmetrical, the input and output resistance of the HEU can take the same value Rs. In this device, the resistance of the semiconductor is much larger than the resistance of LRS and much smaller than the resistance of HRS (RHRS>> Rs >>RLRS). Therefore, a sudden transition occurs when the Hall voltage reaches the set voltage Vs. Thereafter, even if the applied current continues to increase, the measured voltage will remain at a low value. The degree of voltage drop caused by the resistance transition is related to the resistance of the semiconductor and the RSU, which will be explained in later sections. Since the RSU is bipolar, applying a bias from the top electrode to the bottom electrode results in a transition from LRS to HRS. Thus, as shown in Fig. 2d, different configurations bring different magnetic field polarities, and make the V-I loops different in direction. The magnetic-field-controlled output voltages at a fixed current of 0.1 mA in both configurations are shown in Fig. 2e. In configuration A (B), only a positive (negative) magnetic field causes a resistance state transition. These results indicate that the resistance state in RSU is controlled by the magnetic field and the device configuration.

FIG. 2.

(a) Schematic of measurement configuration A and B. (b) Voltage-current curve of configuration A under different magnetic fields. (c) Voltage-current curve of configuration B under different magnetic fields. (d) Measured voltage varied with magnetic fields under both configurations. (e) Measured voltage at 0.1mA under different configurations and magnetic fields.

FIG. 2.

(a) Schematic of measurement configuration A and B. (b) Voltage-current curve of configuration A under different magnetic fields. (c) Voltage-current curve of configuration B under different magnetic fields. (d) Measured voltage varied with magnetic fields under both configurations. (e) Measured voltage at 0.1mA under different configurations and magnetic fields.

Close modal

To understand the physics process of a magnetic-field-controlled memory device, a Finite Element Analysis (FEA) method was used to simulate the voltage distribution in our samples. First, consider the two-dimensional conductivity tensor under a magnetic field in semiconductor,

σB=σ01+μB21μBμB1
(1)

Second, the same geometry as the experiment was constructed, and boundary conditions were set as the current source from the electrode 1 to 3. Third, an RSU was added between the electrodes 2 and 4.

After building the 2-D transport model, by solving the Laplace equation

σBU=0,
(2)

where U, σ, μ and B are the voltage potential, conductivity, carrier mobility and magnetic field, respectively, the distribution of voltage can be found. The results of voltage distribution (color maps) and current flow (red arrows) are listed in Fig. 3 for different magnetic fields and the resistance states of the RSU when the resistance relationship RHRS>> Rs >>RLRS is satisfied.

FIG. 3.

FEA simulation results of voltage distribution in HEU and RSU under different magnetic fields and resistance states.

FIG. 3.

FEA simulation results of voltage distribution in HEU and RSU under different magnetic fields and resistance states.

Close modal

Based on the above results, it can be concluded that the output voltage is dominated by the original Hall voltage and the resistance states of the RSU. Actually, the relationship of RSU resistance and the HEU resistance is critical to the output voltage. Even with the same RSU properties, semiconductors with different resistivity and mobility can have a large impact on the performance of the device. The V-I curves of the devices using HEU made of semiconductors with different resistances are shown in Fig. 4a–4d. In experiments, silicon with different doping concentrations were used as the HEU material. In some cases, the RHRS>> Rs >>RLRS relationship in FEA simulation is no longer satisfied, so non-linearity appears in the V-I curves. Here, Vinf represents the original Hall voltage because the resistance of RSU can be regarded as infinity. The difference between the Vinf (red curve) and the measured voltage (black curve) is caused by the resistance relationship of RHRS, Rs and RLRS.

FIG. 4.

V-I curves of memory devices using HEU made of semiconductors with different resistance (a) Rs=155.93 kΩ (b) Rs =864.54 kΩ (c) Rs =3.16 MΩ (d) Rs =7.81 MΩ.

FIG. 4.

V-I curves of memory devices using HEU made of semiconductors with different resistance (a) Rs=155.93 kΩ (b) Rs =864.54 kΩ (c) Rs =3.16 MΩ (d) Rs =7.81 MΩ.

Close modal

In order to further describe the characteristics of magnetic-field-controlled memory device, we introduce two important indicators. One is the ratio of the measured voltages at HRS and LRS of the RSU, calculated as r=VH/VL. The other is the ratio of the output at HRS to the original Hall voltage, calculated as k=VH/VINF. Better device performance requires both higher values of r and k. A larger r means the ON/OFF ratio is larger, and a larger k provides an easier process of transitioning the resistance state and makes the energy consumption of the device lower. For a certain RSU (RHRS=5.32 MΩ and RLRS=1.72 kΩ, Vs=0.48 V), the r and k values are a function of the output resistance of the HEU. As shown in Fig. 5a, a resistance that is too large or too small will make one of the r and k values too low to be used in application. To analyze the power consumption of the device, the Joule heating power of the resistance transition point in the magnetic field of 1T is calculated. In fact, the two parameters of the semiconductors determine the power consumption of the device, one is the resistance and the other is the mobility. The results are shown in Fig. 5b and 5c, respectively. For semiconductors with the same mobility, there is a minimum power consumption value at different resistances. However, the relationship between the power consumption and the mobility is monotonically decreasing. For the same semiconductor resistance, the power consumption is inversely proportional to the square of the semiconductor mobility (Fig. 5d).

FIG. 5.

Simulation results (a) r and k values varied with semiconductor resistance. (b) Power consumption at 1T at different semiconductor resistance. (c) Power consumption at 1T at different semiconductor mobility. (d) Inversely proportional relationship of power consumption and mobility.

FIG. 5.

Simulation results (a) r and k values varied with semiconductor resistance. (b) Power consumption at 1T at different semiconductor resistance. (c) Power consumption at 1T at different semiconductor mobility. (d) Inversely proportional relationship of power consumption and mobility.

Close modal

Furthermore, according to the analysis above, for a certain semiconductor, the r and k values can also be adjusted by the properties of RSU. Here, we investigated the influence of the thickness of the SiOx layers in RSU. The typical I-V curves of RSU of thickness of about 20, 60 and 100 nm are shown in Fig. 6a and the V-I relationship of the entire device are shown in Fig. 6b. Since the HEU shared the same parameters, the same Vinf is taken in these devices. Generally, as the SiOx layer becomes thinner, the average set voltage becomes smaller. The statistical results of these three RSU are investigated and the distribution of the set and reset voltage, the resistances of HRS and LRS are shown in Fig. 6c and 6d, respectively. According to these results, devices with thinner SiOx layer could achieve greater magnetic field sensitivity, which suggests that a smaller magnetic field can be used to transform the resistance states of RSU. However, the stability and persistence do not conform to this rule. In contrast, thicker samples have better performance here, such as the endurance and retention, which are shown in Fig. 6e and 6f. Therefore, we should consider more about choosing the proper RSU thickness.

FIG. 6.

(a) I-V loops of resistance switching unit with different thickness (b) V-I curves of hybrid device by using the same HEU. (c) Accumulation probability distribution of Vset and Vreset with different thickness. (d) Accumulation probability distribution of RHRS and RLRS with different thickness. (e) Endurance test in samples with different thicknesses. (f) Retention test in samples with different thicknesses.

FIG. 6.

(a) I-V loops of resistance switching unit with different thickness (b) V-I curves of hybrid device by using the same HEU. (c) Accumulation probability distribution of Vset and Vreset with different thickness. (d) Accumulation probability distribution of RHRS and RLRS with different thickness. (e) Endurance test in samples with different thicknesses. (f) Retention test in samples with different thicknesses.

Close modal

So far we have studied the performance of individual devices in detail. To further discuss the issue of integration, a key issue is scaling down. First, since the RSU is controlled by conductive filaments with a planar dimension of a few nm, the miniaturization of the RSU is theoretically easy to solve. Second, the RSU transition voltage can be as low as 0.3 V, which needs a longitudinal voltage Vxx=5.4 V in HEU according to the ratio of Vxx and Vxy in experiments. As the breakdown electric field Eb of silicon is ∼3×105 V/cm,25 the size of the HEU can be scaled down to Vxx/Eb = 180 nm. Therefore, the feature size of device can be as small as 30 nm. By using high mobility semiconductors, like GaAs, lower longitudinal voltage could be applied, which can lead to a smaller scalable size since they have similar breakdown field. It is worth mentioning that as resistive switching devices become more common, additional techniques can be used to improve the properties such as speed, power consumption, stability and persistence. There are many reasons to believe that by combining the resistive switching techniques with the maturity of the semiconductor industry, more attractive results will appear in the research laboratory and in industry.

In summary, magnetic-field-controlled memory devices by coupling the resistive switching effect with the Hall effect have been fabricated. Multi-functional memory of magnetic field has been achieved under different device configurations. By proper matching Rs, RHRS and RLRS, we can enhance the ON/OFF ratio and reduce the power consumption of the device. Hybrid devices that inherit the advantages of resistive random access memory devices and semiconductor technology have great potential to integrate computing and memory. The devices without any magnetic material have unique magnetoelectric properties and can make silicon-based device further advanced.

The authors thank Dr. Zhaochu Luo for profound discussions and comments. This work was sponsored by National Key R&D Program of China (Grant No: 2017YFA0206202) and National Science Foundation of China (Grant Nos.: 51471093 and 116741901).

1.
2.
A.
Ney
,
C.
Pampuch
,
R.
Koch
, and
K.
Ploog
,
Nature
425
,
485
(
2003
).
3.
P.
Xu
,
K.
Xia
,
C.
Gu
,
L.
Tang
,
H.
Yang
, and
J.
Li
,
Nat. Nanotechnol.
3
,
97
(
2008
).
4.
H.
Dery
,
P.
Dalal
, and
L.
Sham
,
Nature
447
,
573
(
2007
).
5.
Z.
Luo
,
Z.
Lu
,
C.
Xiong
,
T.
Zhu
,
W.
Wu
,
Q.
Zhang
,
H.
Wu
,
X.
Zhang
, and
X.
Zhang
,
Adv. Mater.
29
,
1605027
(
2017
).
6.
K. Y.
Guslienko
,
X.
Han
,
D.
Keavney
,
R.
Divan
, and
S.
Bader
,
Phys. Rev. Lett.
96
,
067205
(
2006
).
7.
S.
Joo
,
T.
Kim
,
S. H.
Shin
,
J. Y.
Lim
,
J.
Hong
,
J. D.
Song
,
J.
Chang
,
H.-W.
Lee
,
K.
Rhie
, and
S. H.
Han
,
Nature
494
,
72
(
2013
).
8.
Z.
Luo
,
X.
Zhang
,
C.
Xiong
, and
J.
Chen
,
Adv. Funct. Mater.
25
,
158
(
2015
).
9.
Z.
Luo
,
C.
Xiong
,
X.
Zhang
,
Z. G.
Guo
,
J.
Cai
, and
X.
Zhang
,
Adv. Mater.
28
,
2760
(
2016
).
10.
M. P.
Delmo
,
E.
Shikoh
,
T.
Shinjo
, and
M.
Shiraishi
,
Phys. Rev. B
87
,
245301
(
2013
).
11.
M. P.
Delmo
,
S.
Yamamoto
,
S.
Kasai
,
T.
Ono
, and
K.
Kobayashi
,
Nature
457
,
1112
(
2009
).
12.
C.
Wan
,
X.
Zhang
,
X.
Gao
,
J.
Wang
, and
X.
Tan
,
Nature
477
,
304
(
2011
).
13.
Z.
Luo
,
H. G.
Piao
,
A. V.
Brooks
,
X.
Wang
,
J.
Chen
,
C.
Xiong
,
F.
Yang
,
X.
Wang
,
X. G.
Zhang
, and
X.
Zhang
,
Adv. Electron. Mater.
3
,
1700186
(
2017
).
14.
R. S.
Popovic
,
Z.
Randjelovic
, and
D.
Manic
,
Sensors and Actuators A: Physical
91
,
46
(
2001
).
15.
M.
Johnson
,
B. R.
Bennett
,
M.
Yang
,
M.
Miller
, and
B.
Shanabrook
,
Appl. Phys. Lett.
71
,
974
(
1997
).
16.
M.
Johnson
,
IEEE Trans. Magnetics
36
,
2758
(
2000
).
17.
R.
Waser
and
M.
Aono
,
Nat. Mater.
6
,
833
(
2007
).
18.
D.
Kim
,
S.
Seo
,
S.
Ahn
,
D.-S.
Suh
,
M.
Lee
,
B.-H.
Park
,
I.
Yoo
,
I.
Baek
,
H.-J.
Kim
, and
E.
Yim
,
Appl. Phys. Lett.
88
,
202102
(
2006
).
19.
L. D.
Bozano
,
B. W.
Kean
,
M.
Beinhoff
,
K. R.
Carter
,
P. M.
Rice
, and
J. C.
Scott
,
Adv. Funct. Mater.
15
,
1933
(
2005
).
20.
R.
Waser
,
R.
Dittmann
,
G.
Staikov
, and
K.
Szot
,
Adv. Mater.
21
,
2632
(
2009
).
21.
Y.
Wang
,
Q.
Liu
,
S.
Long
,
W.
Wang
,
Q.
Wang
,
M.
Zhang
,
S.
Zhang
,
Y.
Li
,
Q.
Zuo
, and
J.
Yang
,
Nanotechnology
21
,
045202
(
2009
).
22.
M.-J.
Lee
,
C. B.
Lee
,
D.
Lee
,
S. R.
Lee
,
M.
Chang
,
J. H.
Hur
,
Y.-B.
Kim
,
C.-J.
Kim
,
D. H.
Seo
, and
S.
Seo
,
Nat. Mater.
10
,
625
(
2011
).
23.
Q.
Liu
,
W.
Guan
,
S.
Long
,
R.
Jia
,
M.
Liu
, and
J.
Chen
,
Appl. Phys. Lett.
92
,
012117
(
2008
).
24.
U.
Russo
,
D.
Ielmini
,
C.
Cagli
, and
A. L.
Lacaita
,
IEEE Trans. Electron Devices
56
,
193
(
2009
).
25.
S. M.
Sze
and
K. K.
Ng
,
Physics of Semiconductor Devices
(
John Wiley and Sons
,
Hoboken, New Jersey
,
2006
).