In this work, we study the effects of mobile oxygen ions in a synthetic antiferromagnet structure gated by a sputtered SiO2 dielectric layer for memory and logic applications. Our devices utilize electrochemical reactions between dielectric reactive species and magnetic elements to create irreversible changes in magnetization. We analyzed the dependence of ion velocity on the gate dielectric properties such as the lattice parameter, oxygen migration energy barrier, and electric field (E-field). Hall bar devices were patterned and tested to determine the interlayer exchange coupling between the CoFeB and [Co/Pd]n layers. The anomalous Hall effect (AHE) of the CoFeB layer at different gate voltages (Vg) was measured to determine the Vg dependence. A sharp change in the behavior of the CoFeB layer with respect to negative Vg results in a non-reciprocal decrease in the coercivity and magnetization and an increase in exchange bias. The observed change in exchange bias field and magnetization allows us to measure the change in the effective thickness of the CoFeB layer. This led us to conclude that the source of such behavior is the negatively charged mobile oxygen ions from the SiO2 gate.

One-time-programmable (OTP) systems have become increasingly important in future computing and storage applications.1–5 OTP systems are designed to write a single bit of data permanently or execute a specific set of instructions. These systems either contain pre-encoded devices from the manufacturing stage or allow for one-time programing. In neuromorphic devices, it is used to maintain synaptic plasticity, which is central to neuromorphic computing.4 In the recent few years, we have seen a tremendous rise of spintronics for magnetic random access memory as well as other semiconductor applications (spin logic, etc.). It is only a matter of time before magnetic read-only memory, field programmable gate arrays, application specific integrated circuits, etc., start dotting the research landscape. The key to making OTP devices is to use magnetic materials along with non-reciprocal, irreversible changes.

Synthetic antiferromagnetic (SAF) structures have garnered considerable attention for spintronic applications such as magnetic tunnel junctions (MTJs), spin values, and spin–orbit torque (SOT) devices because the SAF structure can reduce the stray field for free layers, enhance the thermal stability of MTJs, and induce the field-free switching of SOT devices.6–11 Meanwhile, the interlayer exchange coupling (IEC) of the SAF structure can be tuned [ferromagnetic (FM) coupling or antiferromagnetic (AFM) coupling] by choosing the different FM and spacer layers or changing their thicknesses based on the Ruderman–Kittel–Kasuya–Yosida interaction theory.12–14 Recently, the E-field has been proposed as a promising way to mediate the properties of the spintronic structures, such as interlayer exchange coupling (IEC), exchange bias, and magnetic properties.15,16 More recently, Zhang et al. reported that an E-field can tune the sign of the SAF structure based on Bruno's model, realizing that the bidirectional switching of perpendicular MTJ devices through the E-field tuned the exchange coupling transition between FM and AFM couplings.17,18 In addition, their respective magnetic properties of the SAF structure can be detected through the anomalous Hall effect (AHE), while the E-field is applied through an independent top/bottom Vg. Under the application of appropriate Vg, the E-field can push the ions into the SAF structure and change the effective thickness of the FM layers, inducing the exchange coupling change.18–21 The modification of the magnetic moment of the FM layer is also observed by applying the E-field through the ionic gate.19–24 One key difference found in some of these experiments is the reversibility of the change in magnetization under the application of voltages of opposite polarity. The reversibility of the changes depends on the strength of the chemical bonds formed between the reactive species, entering from the gate dielectric, and the elements of the magnetic layer. If the chemical bonds are weak, this can lead to reversibility of the magnetization. However, if the chemical bonds between reactive species and magnetic elements are strong, it can give rise to irreversibility. Therefore, ionic movement when combined with electrochemical reactions can lead to non-reciprocal, irreversible changes of the SAF structure.

In this paper, we study the effects of ionic oxygen movement in a gate controlled SAF structure. We analyze the changes in ion velocity with respect to lattice constants, and the oxygen migration energy barrier of the SAF structure consisted of perpendicular magnetic anisotropy (PMA) CoFeB and [Co/Pd]n layers coupled via a Ta spacer layer. The AHE minor loops as a function of Vg show a strong non-reciprocal decrease in interfacial PMA of CoFeB, which is attributed to ionic movements and electrochemical reactions at the CoFeB/oxide interface.

All the films with a stack of Ta(5 nm)/Pd(10 nm)/[Co0.3/Pd0.7]3/Ta(1.0 nm)/Co20Fe60B20(1.3 nm)/MgO(2 nm) were deposited on thermally oxidized Silicon substrates using a magnetron sputtering system with the base pressure of 1 × 10−7 Pa at room temperature (RT). After deposition, the stack films were capped by a 3-nm thick Ta layer to prevent oxidation. The stack was annealed at 200 °C for 30 min using a rapid thermal annealer. The magnetic properties were characterized at room temperature (RT) using a Physical Property Measurement System (PPMS) with the Vibrating Sample Magnetometer module. These films were then fabricated into Hall bars by photolithography and ion milling. A small window was created to deposit 100 nm sputtered SiO2, as the gate dielectric followed by liftoff. The contact deposition was done by e-beam evaporation of Ti (10 nm) and Au (100 nm). The magneto-transport properties were measured using a PPMS with a Keithley DC supply (6221) and a Keithley nanovoltmeter (2182). The application of Vg and the measurement of leakage current were done by a Keithley voltage source (2400).

Figure 1(a) is the schematic of our SAF structure, consisting of a single SAF structure with two FM layers (CoFeB and [Co/Pd]n) with PMA, separated by a NM (Ta). [Co/Pd]n has a bulk PMA, whereas CoFeB has an interfacial PMA, which originates from a careful hybridization between the p-orbit of oxygen (from MgO) and the d-orbit of Co and Fe (from CoFeB). In order to investigate the ionic velocity of our device, we follow the physics-based model for ionic movement in the SiO2 gate dielectrics.25,26 The ionic migration through the gate dielectric can be written by using a drift-diffusion model to analyze the migration of oxygen ions and vacancies through the dielectric. The ion–vacancy generation-combination is given by the exponential dependency,

Pg=f0expEgγeE/(kBT),
(1)
Pr=Cionf0expErγeE/(kBT).
(2)

Here, Pg and Pr are the generation and recombination rate of ion–vacancy pairs, respectively; f0 is the vibration frequency of ions surrounding the vacancy; Cion is the ion concentration; Eg and Er are the formation and recombination energy, respectively; E is the applied electric field; and γ is the contribution of bond polarization to the local electric field. The rate of drift and diffusion in the oxide layer is dependent on the oxygen migration barrier (Em) and the lattice constant (a) of the SiO2 layer (Table I). The drift velocity is given by

Vion=af0expEm/(kBT)sinh(eaE/(2kBT)).
(3)
FIG. 1.

(a) Sample stack structure and typical device schematic. [(b)–(d)] Analysis of ion velocity as a function of lattice parameter and oxygen migration energy barrier for different E-fields. The color bar represents log |Vion(m/s)|.

FIG. 1.

(a) Sample stack structure and typical device schematic. [(b)–(d)] Analysis of ion velocity as a function of lattice parameter and oxygen migration energy barrier for different E-fields. The color bar represents log |Vion(m/s)|.

Close modal
TABLE I.

Values of different parameters used in the analysis.

ParameterParameter nameParameter value
f0 Vibration frequency 1013/s 
Em Oxygen migration barrier 0.1–0.5 eV 
a lattice parameter 2–8 Å 
ParameterParameter nameParameter value
f0 Vibration frequency 1013/s 
Em Oxygen migration barrier 0.1–0.5 eV 
a lattice parameter 2–8 Å 

Figures 1(b)–1(d) show the ionic velocity for different lattice constants and oxygen energy migration barrier for different E-fields. As expected, increasing the lattice constant and decreasing oxygen energy migration barrier increase the ionic velocity. As shown in Fig. 1(c), a small change in lattice constant creates a larger change in ion velocity at higher electric fields, whereas at lower electric fields, it is the oxygen energy migration barrier. Our sputtered gate oxide could be a low crystalline SiO2 layer, which leads to a range of lattice constants and the oxygen migration energy barrier compared to the results obtained with the parameters as given in Ref. 25.

A schematic of the SAF stack is shown in Fig. 2(a), where the program voltage to the device is applied via Vg. The AHE of the SAF structure is characterized through the measurement of passing the current through the main Hall bar and testing the Hall voltage. Figure 2(b) shows the major magnetic hysteresis (M-H) loop by sweeping the magnetic field between ±3000 Oe. This allows us to look at the magnetic properties of the CoFeB-[Co/Pd]n SAF structure, which shows a two-step switching process. We start by completely saturating the magnetic field at +3000 Oe. At this point, the magnetization of both the CoFeB and [Co/Pd]n layers is pointing upwards, i.e., the CoFeB and [Co/Pd]n layers are ferromagnetically coupled. The magnetic field is lowered and saturated in the opposite direction. The first step occurs just before zero field, which is where the magnetization of the CoFeB layer has switched to the downward direction, giving an AFM state. As we further lower the magnetic field, the second step occurs around −1700 Oe. At this point, both the CoFeB and [Co/Pd]n layers are saturated in the downward direction (FM state). The field is swept back from −3000 Oe to +3000 Oe, and the same behavior is repeated. Hence, at zero field, the magnetization of the CoFeB and [Co/Pd]n layers is always pointing in opposite directions, indicating AFM coupling as the most stable configuration in the absence of any magnetic field. The minor loop corresponds to the switching of the CoFeB layer, which has a lower coercivity, for a fixed orientation of [Co/Pd]n. The magnetic properties of the [Co/Pd]n layer are not affected by Vg (seen from Vg dependence). This is expected since the [Co/Pd]n layer is deep inside the metallic system, virtually making it robust to any external electric field. The CoFeB layer, on the other hand, has an interfacial PMA, making it very sensitive to interfacial oxygen population. The minor AHE loop measurement as shown in Fig. 2(c) clearly shows a hysteretic behavior that has been shifted to a positive magnetic field (the magnetization of the [Co/Pd]n layer is pointing up). The field shift of the minor loop denotes the interlayer exchange bias field (Hex). Under a negative Vg, the oxygen movement will be triggered. This oxygen movement will lead to electrochemical reactions with CoFeB and change the magnetization, tilting some of the magnetic domains in-plane. This can be measured in several ways, and the easiest among them would be to use the resistance of the stack, via current-in-plane (CIP) giant magnetoresistance (GMR) and anisotropic magnetoresistance (AMR). One key difference from traditional read-only memory is that the devices are initially in a high resistance state due to anti-parallel magnetizations (bit-0). After the application of programing voltage, the device will go to a lower resistance state (bit-1) due to the change in the magnetization of CoFeB.

FIG. 2.

(a) Device schematic for gate measurements. (b) M-H measurement of the major loop of the sample in the virgin state. (c) AHE measurement of the minor loop for the highest and lowest applied gate voltage.

FIG. 2.

(a) Device schematic for gate measurements. (b) M-H measurement of the major loop of the sample in the virgin state. (c) AHE measurement of the minor loop for the highest and lowest applied gate voltage.

Close modal

Figures 3(a)–3(c) show the coercivity, normalized saturation magnetization, and remanence of the CoFeB layer derived from the anomalous Hall signals. The normalized saturation magnetization is calculated from the resistance difference in the minor loop of the AHE, with respect to the virgin state. As is evident, under the application of negative Vg, we observe an irreversible change in magnetic properties of the CoFeB layer. All the key parameters (magnetization, coercivity, and remanence) show a strong decrease with negative Vg. The application of large positive Vg failed to retrace the magnetic property of the CoFeB layer. This is a strong indication of the electrochemical reaction happening at the CoFeB/MgO interface. Under the application of negative Vg, negatively charged oxygen ions drift toward the CoFeB/MgO interface. Some of them punch through the MgO layer to reach the CoFeB/MgO interface. This creates an overpopulation of oxygen at the interface. As these ions slowly diffuse into the CoFeB layer, they chemically react with the elements of the CoFeB layer (mostly Co and Fe) and hence degrade the interfacial PMA. The application of large positive Vg should, in principle, remove the oxygen ions and pull them back into the gate dielectric. However, under the circumstance of strong bond formations between transition metals and oxygen ions, the chemical reaction can be irreversible and the oxygen is trapped inside the CoFeB layer permanently. This leads to permanent degradation of the magnetic properties of the CoFeB layer, which explains the changes in coercivity, saturation magnetization, and remanence at negative Vg.

FIG. 3.

[(a)–(f)] Coercivity (Hc), normalized saturation magnetization (Ms), and remanence (Mr/Ms) (at exchange bias field), exchange bias field (Hex), and relative change in effective thickness of CoFeB contributing to PMA and the magnetization tilt angle as a function of gate voltage.

FIG. 3.

[(a)–(f)] Coercivity (Hc), normalized saturation magnetization (Ms), and remanence (Mr/Ms) (at exchange bias field), exchange bias field (Hex), and relative change in effective thickness of CoFeB contributing to PMA and the magnetization tilt angle as a function of gate voltage.

Close modal

Another key change that can be seen from anomalous Hall signal is the change in Hex at negative Vg. Hex is one of the key parameters that determines the strength of exchange coupling between the CoFeB and [Co/Pd]n layers.26 Hex between the CoFeB and [Co/Pd]n layers is given by

Hex=J/(μ0MCFBtCFB).
(4)

Here, μ0 is the vacuum permeability, J is the exchange coupling constant, MCFB is the magnetization of the CFB layer, and tCFB is the thickness of the CFB layer. Equation (4) is valid under the assumption that the [Co/Pd]n layer is not affected (rotated) by the magnetic field during AHE loop measurements. This makes the stability condition (energy minima) the same to exchange coupling between the FM and AFM layers, as depicted in the Meiklejohn–Bean formula.

The increase in Hex can be explained by a decrease in magnetization and effective thickness of the CoFeB layer. Previous works on similar physics have concluded that the change in Hex is mostly dominated by the change in thickness of the FM layer.27 This change in thickness corresponds to the movement of oxygen ions into the CFB layer, thereby increasing the effective magnetic dead layer thickness and hence decreasing the total effective thickness of the FM layer, which contributes to the decreased PMA. This decrease in the thickness of the FM layer would increase Hex between the CoFeB and [Co/Pd]n layers.

Using the values of Hex and magnetization as shown in Figs. 3(e) and 3(b), we can calculate the change in the effective thickness of the CoFeB layer based on Eq. (4). This change in the effective thickness of the CoFeB layer is approximately 22% for our device. Because the interfacial PMA of the CoFeB layer is very sensitive to its thickness change, as the effective thickness for interfacial PMA decreases, the magnetic anisotropy of the CoFeB layer shows the transition from PMA to in-plane, as shown in Fig. 2(c). Figure 3(d) shows the change in coercivity of the CoFeB with gate voltage. As is evident, we observe a strong decrease in coercivity of the CoFeB layer at negative Vg. Figure 3(f) shows the effective angle of the magnetization tilt at different Vg. The magnetization tilt angle is calculated by

θm=cos1(mr/max(mr)).
(5)

Here, mr is the ratio of the remanence magnetization to the saturation magnetization [as shown in Fig. 3(c)]. This shows that some of the magnetic domains in the CoFeB layer are now in-plane at negative Vg. The AHE allows us to measure magnetic properties of the structure in detail. However, the reading scheme for applications can be extended by using a simple readout dependent on CIP GMR and AMR of the stacks or even tunneling magnetoresistance (TMR) if the dielectric layer can be fashioned into an ultra-thin tunneling layer.

In summary, we investigated the effects of mobile oxygen ions in magneto-ionic SAF FET devices. The gate dependence showed a strong decrease in interfacial PMA of CoFeB under the application of negative Vg. The change in magnetization of the CoFeB layer along with Hex allowed us to calculate the change in effective thickness of the CoFeB layer. The ion dynamics were analyzed by using a drift-diffusion model, which allowed us to analyze the effects of various material parameters on the ion velocity. The change was non-reciprocal and irreversible, which strongly hints to the electrochemical reactions between CoFeB and oxygen ions. This structure holds strong promise for magnetic/spintronic applications.

P.S. and D.Z. equally contributed to this work.

This work was in part supported by ASCENT, one of the six centres in JUMP, a Semiconductor Research Corporation program, sponsored by MARCO and DARPA. Portions of this work were carried out in the Minnesota Nano Center, which was supported by the National Science Foundation (NSF) through the National Nano Coordinated Infrastructure, under Award No. ECCS-1542202.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

1.
A. W.
Marsman
,
C. M.
Hart
,
G. H.
Gelinck
,
T. C. T.
Geuns
, and
D. M. D.
Leeuw
,
J. Mater. Res.
19
,
2057
2060
(
2004
).
2.
W. R.
Brockhoff
,
14th International Reliability Physics Symposium
(
1976
).
3.
B.
Bhusan
,
Patent US5391518A
(21 February 1995).
4.
C.
Lüscher
,
R. A.
Nicoll
,
R. C.
Malenka
, and
D.
Muller
,
Nat. Neurosci.
3
,
545
550
(
2000
).
5.
W.
Cockerille
,
X.
Chen
,
N.
Quraishi
, and
G.
Schlottmann
,
Patent WO2006098932A1
(21 September
2006
).
6.
S. V.
Pietambaram
,
J.
Janesky
,
R. W.
Dave
,
J. J.
Sun
,
G.
Steiner
, and
J. M.
Slaughter
,
IEEE Trans. Magn.
40
,
2619
2621
(
2004
).
7.
T.
Jungwirth
,
X.
Marti
,
P.
Wadley
, and
J.
Wunderlich
,
Nat. Nanotechnol.
11
,
231
241
(
2016
).
8.
K.
Inomata
,
T.
Nozaki
,
N.
Tezuka
, and
S.
Sugimoto
,
Appl. Phys. Lett.
81
,
310
(
2002
).
9.
P. X.
Zhang
,
L. Y.
Liao
,
G. Y.
Shi
,
R. Q.
Zhang
,
H. Q.
Wu
,
Y. Y.
Wang
,
F.
Pan
, and
C.
Song
,
Phys. Rev. B
97
,
214403
(
2018
).
10.
W.
Legrand
,
D.
Maccariello
,
F.
Ajejas
,
S.
Collin
,
A.
Vecchiola
,
K.
Bouzehouane
,
N.
Reyren
,
V.
Cros
, and
A.
Fert
,
Nat. Mater.
19
,
34
42
(
2020
).
11.
Y.
Saito
,
H.
Sugiyama
, and
K.
Inomata
,
J. Appl. Phys.
97
,
10C914
(
2005
).
12.
W. X.
Xia
,
K.
Inoue
,
S.
Saito
, and
M.
Takahashi
,
J. Phys. Conf. Ser.
266
,
012064
(
2011
).
13.
D. L.
Zhang
,
C.
Sun
,
Y.
Lv
,
K. B.
Schliep
,
Z.
Zhao
,
J. Y.
Chen
,
P. M.
Voyles
, and
J. P.
Wang
,
Phys. Rev. Appl.
9
,
44028
(
2018
).
14.
X.
Zhao
,
W.
Liu
,
S.
Li
,
T.
Wang
,
L.
Liu
,
Y.
Song
,
S.
Ma
,
X.
Zhao
, and
Z.
Zhang
,
Nanoscale
10
,
7612
(
2018
).
15.
T.
Newhouse-Illige
,
Y.
Liu
,
M.
Xu
,
D.
Reifsnyder Hickey
,
A.
Kundu
,
H.
Almasi
,
C.
Bi
,
X.
Wang
,
J. W.
Freeland
,
D. J.
Keavney
,
C. J.
Sun
,
Y. H.
Xu
,
M.
Rosales
,
X. M.
Cheng
,
S.
Zhang
,
K. A.
Mkhoyan
, and
W. G.
Wang
,
Nat. Commun.
8
,
15232
(
2017
).
16.
V. L.
Korenev
,
I. V.
Kalitukha
,
I. A.
Akimov
,
V. F.
Sapega
,
E. A.
Zhukov
,
E.
Kirstein
,
O. S.
Ken
,
D.
Kudlacik
,
G.
Karczewski
,
M.
Wiater
,
T.
Wojtowicz
,
N. D.
Ilyinskaya
,
N. M.
Lebedeva
,
T. A.
Kamissorova
,
Y. G.
Kusyarev
,
D. R.
Yakovlev
, and
M.
Bayer
,
Nat. Commun.
10
,
2899
(
2019
).
17.
D.
Zhang
,
M.
Bapna
,
W.
Jiang
,
D. P.
Sousa
,
Y. C.
Liao
,
Z.
Zhao
,
Y.
Lv
,
A.
Naeemi
,
T.
Low
,
S.
Majetich
, and
J. P.
Wang
, arXiv:1912.10289.
18.
Q.
Yang
,
L.
Wang
,
Z.
Zhou
,
L.
Wang
,
Y.
Zhang
,
S.
Zhao
,
Z.
Dong
,
Y.
Cheng
,
T.
Min
,
Z.
Hu
,
W.
Chen
,
K.
Xia
, and
M.
Liu
,
Nat. Commun.
9
,
991
(
2018
).
19.
L.
Baldrati
,
A. J.
Tan
,
M.
Mann
,
R.
Bertacco
, and
G. S. D.
Beach
,
Appl. Phys. Lett.
110
,
012404
(
2017
).
20.
U.
Bauer
,
L.
Yao
,
A. J.
Tan
,
P.
Agrawal
,
S.
Emori
,
H. L.
Tuller
,
S.
Van Dijken
, and
G. S. D.
Beach
,
Nat. Mater.
14
,
174
181
(
2015
).
21.
C.
Bi
,
C.
Sun
,
M.
Xu
,
T.
Newhouse-Illige
,
P. M.
Voyles
, and
W.
Wang
,
Phys. Rev. Appl.
8
,
034003
(
2017
).
22.
A. J.
Tan
,
M.
Huang
,
C. O.
Avci
,
F.
Büttner
,
M.
Mann
,
W.
Hu
,
C.
Mazzoli
,
S.
Wilkins
,
H. L.
Tuller
, and
G. S. D.
Beach
,
Nat. Mater.
18
,
35
41
(
2019
).
23.
B.
Dieny
and
M.
Chshiev
,
Rev. Mod. Phys.
89
,
025008
(
2017
).
24.
X.
Ma
,
J.
Zhang
,
W.
Cai
,
H.
Wang
,
J.
Wilson
,
Q.
Wang
,
Q.
Xin
, and
A.
Song
,
Sci. Rep.
7
,
809
(
2017
).
25.
T.
Sadi
,
A.
Mehonic
,
L.
Montesi
,
M.
Buckwell
,
A.
Kenyon
, and
A.
Asenov
,
J. Phys.: Condens. Matter
30
,
084005
(
2018
).
26.
K. C.
Kao
,
Dielectric Phenomena in Solids
(
Elsevier
,
2004
), Chap. 7.
27.
J.
Zehner
,
R.
Huhnstock
,
S.
Oswald
,
U.
Wolff
,
I.
Soldatov
,
A.
Ehresmann
,
K.
Nielsch
,
D.
Holzinger
, and
K.
Leistner
,
Adv. Electron. Mater.
5
,
1900296
(
2019
).