Highly Sensitive Detection of Weak Low Frequency Magnetic Fields Using Single Nanoscale Orthogonal MgO Magnetic Tunnel Junctions under a Large Bias Field

T.N. Anh Nguyen; Q. Ngan Pham; V. Thanh Chu; K. Tung Do; T. Huong Nguyen; H. Nam Pham; Minori Goto; Miyoshi Fukumoto; Hiroyuki Tomita; Tatsuki Watanabe; Hitoshi Kubota; Akio Fukushima; Kei Yakushiji; Yoshishige Suzuki

doi:10.2320/matertrans.MT-MG2022017

Abstract

Ultrasensitive magnetic field sensors at low frequencies envisaged for applications on biosensors require the detection of superparamagnetic nanoparticles as biomarkers and weak biomagnetic fields. However, superparamagnetic nanoparticles can be magnetized and detected in the presence of a relatively large external magnetic field in which other low field sensors and flux concentrators cannot be used since those are saturated in the bias field. To overcome this problem, we made submicron-sized orthogonal magnetic tunnel junctions with an appropriate perpendicular anisotropy free layer. By applying a perpendicular external magnetic field, the magnetization in the magnetic free layer is brought into an unstable state. As a result, relatively high field sensitivity of 1.3%/Oe is achieved in the external field of about 975 Oe for the 300 nm magnetic tunnel junction. By tilting the angle of the bias field of 10 degrees from film normal, the maximum sensitivity of 4.5%/Oe was obtained at 295 Oe. We also investigated the low-frequency noise generated by the sensor for their capability to detect weak low-frequency magnetic signals. Noise equivalent signal of the device is estimated to be about 235 nT/$\sqrt{\text{Hz}} $ from 1/f noise measurements which is suitable for single superparamagnetic nanoparticles detection.

1. Introduction

Magnetic bio-sensing techniques have shown great promise for future biosensor applications. One of the techniques centers on the detection of the stray field of superparamagnetic magnetic nanoparticles (NPs) bound to target analytes (e.g. cell, proteins, DNA, etc.). Thus detection of weak low frequency magnetic fields is crucial.¹^,²⁾ There are numerous low-field magnetic sensors that have been investigated for use in biosensors including coil-based sensors,³^,⁴⁾ direct-current superconducting quantum interference devices (dc SQUIDs),²^,⁵^,⁶⁾ fluxgate magnetometers,⁷^–⁹⁾ giant magneto-impedance sensors,¹⁰^–¹²⁾ Hall effect sensors¹³^–¹⁵⁾ and recently frequency-based sensors,¹⁶^–¹⁸⁾ and magnetoresistive (MR)-based sensors.¹^,¹⁹^–²²⁾ Above all others, MR-based sensors offer significant advantages: (i) MR sensors have an inherent advantage in size and power when compared to coil-based, fluxgate, and more complicated low-field sensing techniques such as SQUID and spin resonance magnetometers;¹⁹⁾ (ii) the technique used for fabricating MR sensors is a well-established technology for information storage which is CMOS compatible, large-scale, and cost-effective to produce; (iii) MR sensors also provide high sensitivity, adjustable detected field range, and room temperature (RT) operation. To date, MR sensors have been widely employed in various applications such as magnetic recording media, read heads, magnetic random access memory (MRAM),²³^–²⁶⁾ pressure measurement,²⁷^,²⁸⁾ current sensing,²⁹^–³¹⁾ angle position sensing,³²^,³³⁾ and bio-sensing.¹⁹^–²²^,³⁴^,³⁵⁾ Among several available MR sensor technologies based on anisotropic/giant/tunneling MR effects, tunneling MR sensors with MgO barriers and MR ratios substantially higher than 100% at RT³⁶^–⁴⁰⁾ are desired for biomedical applications. Such MgO-based magnetic tunnel junctions (MTJs) are also highlighted as the most competitive sensors for ultra-low field detection due to their high sensitivity, low power consumption, and small size.⁴⁰^–⁴³⁾ For use as magnetic-based biosensors, the detection of a weak magnetic field of superparamagnetic NPs (∼nT) as biomedical magnetic tags at RT is required.¹^,²⁾ However, superparamagnetic NPs can be magnetized and detected in the presence of a relatively large external magnetic field in which other low field sensors or flux concentrators cannot be used since those are saturated in the larger working bias field. To overcome this problem, we present, in this paper, a novel submicron size MgO-based MTJ with an orthogonal magnetic geometry of a relatively weak perpendicular magnetic anisotropy free layer and an in-plane magnetic anisotropy reference layer. By applying a perpendicular magnetic field, the magnetization in the free layer is brought into an unstable state which makes it sensitive to the external field. Compared to in-plane or perpendicular MTJs, this type of orthogonal MTJ provides higher sensitivity and wider sensing range.⁴³^–⁴⁶⁾

Our purposes are to investigate and identify the possibility of the detection of superparamagnetic NPs for use as biomarkers, and to explore the technical limit of the detection of single superparamagnetic NPs using MTJ sensors. The first major issue is the size of MTJs. For a superparamagnetic NP with radius r, which is supposed to be smaller than the radius of MTJ a, scaling of the signal to noise ratio (S/N) should be,

\begin{equation} S/N = \frac{\textit{stray field}}{\textit{thermal noise field}} \propto \frac{r^{3}a^{-3}}{(a^{2}t)^{-\frac{1}{2}}} \propto a^{-2}, \end{equation}

(1)

where t is the thickness of the free layer in the MTJ. Therefore, smaller MTJs offer higher S/N which is better for detecting single superparamagnetic NPs. For this purpose, arrays of MTJs cannot be applied. Thus, submicron-sized single MTJ sensors which are close to the single magnetic domain state are suitable for single superparamagnetic NPs detection. Our previous study had reported that nano-sized MTJ-based sensors provide excellent detectivity due to the reduced magnetic 1/f noise thanks to the reduction of the inhomogeneous magnetic fluctuations.⁴³⁾

The second critical issue in detecting super-paramagnetic NPs using MTJ sensors is related to the large bias magnetic field that is necessary to magnetize the NPs. The large bias field may easily saturate the magnetization of a free layer with weak in-plane magnetic anisotropy. Therefore, a special design to facilitate a highly sensitive field sensor that can work in a large working bias field is necessary. For those purposes, submicron-sized orthogonal MTJs where the free layer has moderate perpendicular magnetic anisotropy, which is smaller than the demagnetization energy,⁴⁷⁾ and the fixed layer has in-plane magnetic anisotropy, have been designed and fabricated. By applying a perpendicular magnetic field, the free layer magnetization can be brought into an unstable state and becomes sensitive to the small transverse magnetic signals. The research suggests a guideline for MTJ designing for detecting the weak magnetic fields generated by superparamagnetic NPs.

2. Experimental Procedure

2.1 Fabrication process

MTJ films were prepared using an ultrahigh vacuum magnetron sputtering system (pressure of less than 7.5 × 10⁻¹⁰ Torr). The stack structure of MTJ Ta(3)/PtMn(15)/FeCo(2.5)/Ru(0.9)/CoFeB(3.3)/MgO(1.1)/FeB(2.1)/MgO(1.3)/TaB(2)/Ta(5)/Ru(7) was grown on a thermally oxidized Si substrate (nominal thickness in nm). FeB is employed as the material for the free layer because it is expected to exhibit large interface perpendicular magnetic anisotropy energy K_i induced by a thin MgO capping layer.⁴⁷^,⁴⁸⁾ In our experiments, however, K_i in the FeB free layer is kept moderate with the thickness of 2.1 nm and its magnetization is perpendicularly magnetized in an external perpendicular bias field. CoFeB and FeB were sputtered from Co₄₀Fe₄₀B₂₀ and Fe₈₀B₂₀ targets, respectively. The magnetization of the lower CoFe is pinned unidirectionally in the in-plane direction by an exchange-bias field from the PtMn antiferromagnetic layer. The CoFe/Ru/CoFeB is a synthetic ferrimagnet structure that acts as a reference layer in which the magnetizations of CoFe and CoFeB align in an antiparallel configuration in the in-plane direction. MTJs with a circular shape and a design diameter of 300 nm were fabricated using optical and e-beam lithography combined with an Ar ion-milling and a lift-off technique.

After microfabrication, an annealing process was carried out at 350°C for 1 h in a magnetic field of 7500 Oe to fix the magnetization direction of the reference layer and obtain a high TMR ratio due to coherent Δ₁ tunneling through the crystallized CoFeB/MgO/FeB junction.³⁹^,⁴⁷^,⁴⁸⁾

2.2 Material characterization

To determine the magnetoresistance characteristics of the MTJ sensors, the MR curves were measured in current-perpendicular-to-the-plane configuration using a commercial wafer probe station (TOEI Scientific Industrial Co., Ltd.). An external field of up to 2000 Oe and 4000 Oe was applied parallel and perpendicular to the film surface respectively.

To characterize the low-frequency noise of the MTJs, noise spectra were measured in the low-frequency range of up to 30 kHz using the home-built measurement setup, including two independent amplifier channels, which has been described in the previous work.⁴³⁾ Noise measurements at different bias voltages and hard-axis bias fields (perpendicular to the film plane) were performed. The power spectrum density (PSD) S_V of noise was calculated by a fast Fourier transformation of a cross correlation function of the two channels and averaged for 20 times. Since the amplifier noise in these two channels is uncorrelated, the cross correlation technique was used to remove the unwanted amplifier noise to extract the sensor noise.⁴³^,⁴⁹^–⁵¹⁾ To determine the detectability of the MTJ sensors, a solenoid coil was introduced close to MTJs for generating an ac magnetic signal. In order to eliminate the influence of geomagnetic background noise and urban electromagnetic noise, the solenoid coil, MTJ sample and magnet were kept in a magnetically shielded box. All measurements were done at RT.

3. Results and Discussions

3.1 Tunneling magneto resistance

Typical MR curves under in-plane and out-of-plane magnetic fields measured at RT to determine MTJ sensor performance are shown in Fig. 1(a) and (b), respectively. The saturation field in the out-of-plane field is about 2000 Oe. The minimum resistance in the parallel state was 76 Ω. The large tunneling MR of 112% and the low parallel state resistance-area product RA of 5.4 Ω µm² confirmed the high quality of the developed MTJ. In fact, the performance of the MTJ sensors is strongly dependent on the noise level and linear characteristics of their tunneling MR curve and sensitivity.

Fig. 1

TMR curve for 300 nm MTJ with composition of Ta/PtMn/FeCo/Ru/CoFeB/MgO/FeB/MgO/TaB/Ta/Ru measured under (a) in-plane and (b) out-of-plane magnetic fields.

3.2 Low-frequency noise

For sensor applications, the noise level is one of the critical parameters that needs to be determined in order to find the minimum detectable field at any frequency. The noise in MTJs usually includes amplifier noise, thermal noise, shot noise, random telegraph noise and 1/f noise. The amplifier noise is eliminated using cross-correlation technique.⁴³^,⁴⁹^–⁵¹⁾ The electric background noise (Johnson-Nyquist/shot noise) originates from the fluctuation of the charge/current in MTJ, and is independent of frequency in the low frequency range. The magnetic background noise is induced due to thermal fluctuation of the magnetization which also shows a flat spectrum in the low frequency range. The electric random telegraph noise is related to the charging and discharging process of the defect centers, and can be eliminated using a proper annealing process.⁵²^,⁵³⁾ The magnetic random telegraph noise is caused by sudden changes in magnetic domains. Both electric and magnetic random telegraphs show Lorentzian type spectrum centered at zero frequency.⁵⁴⁾ Finally, 1/f noise has both electrical and magnetic origins. However, 1/f noise in highly-sensitive MTJ sensors is dominated by magnetic noise originating from a coexistence of various magnetic fluctuations.⁵⁵^,⁵⁶⁾

In order to ascertain that the noise detected is indeed from the examined MTJ, not from the environment or peripheral equipment, we measured noise as a function of bias voltage across the 300 nm MTJ at RT in the absence of an external magnetic field. The noise spectra of the 300 nm MTJ sample for the various bias voltages are shown in Fig. 2. The data show that the zero bias noise level is around 10⁻¹⁵ V²/Hz at 10 Hz, a value 10³–10⁴ times lower than that of other voltages. Therefore, it is obvious that the noise at the finite bias voltage is dominantly induced by the MTJ. The noise spectra for all voltages (varies from 20 mV to 80 mV) in the low-frequency range clearly shows a typical 1/f noise behavior. Therefore, it is concluded that the Johnson-Nyquist noise, shot noise and random telegraph noise provide only minor contributions to the observed noise spectrum which is in good agreement with our previous report.⁴³⁾

Fig. 2

(a) Noise spectral density S_V of 300 nm MTJ sensor at different bias voltages at the zero field shows a typical 1/f noise behavior and (b) the noise amplitude at 10 Hz shows a quadratic voltage dependence, as described by Hooge’s law. The solid curve shows the quadratic fit to the data.

The voltage dependence of the noise amplitude at a frequency of 10 Hz is plotted in Fig. 2(b), the solid curve results from least squares fitting. The magnitude of 1/f noise increases as the square of bias voltage increases, which is consistent with Hooge’s empirical rule⁵⁷⁾ with S_V ∼ V², indicating that the noise is caused by resistance fluctuations in the MTJ.

In order to evaluate the noise level of the developed MTJs for use in detecting superparamagnetic NPs which can be magnetized and display high magnetization only in the presence of a large magnetic field. The noise measurements of the 300 nm MTJ in the presence of a strong external magnetic field were performed. In Fig. 3, the noise spectra of the MTJ as a function of the magnetic bias field strength (varies from 250 Oe to 4000 Oe, perpendicular to the film plane) are shown. The observed noise power follows 1/f-like behavior for the all applied fields (Fig. 3(a)) and shows a strong field dependence (Fig. 3(b)). This strong field dependence of the 1/f noise confirms the magnetic origin of the observed 1/f noise.⁵⁸^,⁵⁹⁾ The 1/f noise in the investigated MTJ should be associated with thermal fluctuation of magnetization in both the “free” and “fixed” magnetic layers.⁶⁰⁾ The noise spectral density decreases dramatically when the magnetic field increases from 250 Oe to 4000 Oe by a factor of 30. This factor is 5 times higher than that of the 400 nm MTJ reported in our previous work⁴³⁾ which suggests that the noise in such small size MTJs can be significantly reduced by applying a strong vertical field. In comparison to the 400 nm MTJ, the higher noise level at zero and low bias field in the 300 nm MTJ can be related to the contributions of domain formation and/or disordered surface spins.⁶¹^,⁶²⁾ In the common picture, magnetic nano-objects which are close to a single domain-size have a constant overall magnetic moment (homogeneous region) originating from the magnetization of the single-domain core surrounded by surface regions (faces and edges) hosting spin disorder/canting (inhomogeneous region).⁶¹^,⁶²⁾ Quantitatively, the smaller MTJ has large surface area to volume ratio, therefore, the contribution from its inhomogeneous phase at the surface is larger, leading to the higher noise at zero or low bias field. However, for higher bias fields, only the homogeneously magnetized state can exist, thus the noise level is significantly reduced.

Fig. 3

(a) Noise spectral density S_V of 300 nm MTJ sensor with 80 mV bias voltage under different perpendicular bias fields and (b) the 1/f noise amplitude at 10 Hz as a function of applied field. The solid curve shows the exponential decay fit to the data.

3.3 Sensitivity

In order to determine the sensitivity as well as detectability of the developed MTJ, a low frequency noise spectrum of the MTJ with the presence of ac signal was performed under a large perpendicular bias field. It is worth noting that the MTJ sensor sensitivity S in the linear range as high as 1.85%/Oe,⁶³⁾ 4.07%/Oe,⁶⁴⁾ 15%/Oe,⁶⁵⁾ and 18%/Oe⁶⁶⁾ under zero or very small bias field have been reported. However, the sensitivity under large bias field is common in very small value, 0.048%/Oe.⁶⁷⁾ For use in super-paramagnetic NPs detection, the sensitivity of the developed MTJ under large perpendicular bias field up to 4000 Oe was investigated. The measurement magnetic field configuration is shown in Fig. 4(a) in which a perpendicular bias field is produced by a permanent magnet and an in-plane ac signal field is generated by a solenoid coil. Figure 4(b) is a perpendicular bias field dependence of the ac output signal. The ac signal field H_ac produced at MTJ surface is proportional to the applied current I and the distance d from the AC coil center to MTJ sensor which can be calculated using Ampere’s law:

\begin{equation} H_{\textit{total_ac}} = \int_{\textit{near edge}}^{\textit{far edge}}H_{\textit{ac}} = \frac{\mu_{o}IN}{2L}\left[\frac{d + L/2}{\sqrt{(d + L/2)^{2} + R^{2}}} - \frac{d - L/2}{\sqrt{(d - L/2)^{2} + R^{2}}} \right] \end{equation}

(2)

where μ_o = 4π × 10⁻⁷ H/m is the permeability of vacuum, N is the total number of coil turns, L is the coil length and R is the radius of the coil. In this work, the ac coil with R = 2 cm, L = 1 cm, N = 100 turns and d = 3.5 cm was used. The calculated ac field of 4.3 µT is produced at the sensor surface when a 10 mA alternating current with a frequency of 100 Hz is applied to the ac coil. Dc bias voltage applied on the MTJ is fixed at 50 mV. The output signal shows several peak patterns from 550 to 1050 Oe related to complex magnetization behavior which is likely due to the coexistence of in-plane and out-of-plane anisotropies. The maximum sensitivity $S = \frac{\Delta MR}{H_{ac}} = \frac{\Delta R_{ac}}{(R_{AP} - R_{P}) \times H_{ac}} \times 100$ [%/Oe] calculated from the data is 1.3%/Oe at around 975 Oe which is much higher than the previously reported value⁶⁷⁾ in which $\Delta R_{ac} = \frac{\Delta V_{ac}}{I_{ac}}$ is determined from the ac peak detected by noise measurement, R_AP and R_AP are the high and low resistance of the MTJ in the antiparallel (AP) states and parallel (P) states, respectively. In Fig. 4(c), ac current dependence of the output signal is shown. Linear dependence on the signal field amplitude is confirmed. Since the free layer has an out-of-plane anisotropy, the angle between the direction of the bias field and the direction of the sample surface normal can also affect the sensitivity of the sensor. We investigated the sensitivity as a function of the bias field angle. The results show that by tilting the angle of the bias field of 10 degrees from film normal, maximum sensitivity of 4.5%/Oe was obtained at 295 Oe bias field (Graph is not shown).

Fig. 4

(a) The measurement configuration, (b) perpendicular bias field, H_DC, dependence of the ac signal output and (c) the detected ac field signal as a function of ac current. The solid line represents the linear fit of the experimental data.

From the calculated magnetic field sensitivity of S = 1.3%/Oe and the observed noise power density, S_v, spectrum at the same field, one may estimate the noise equivalent signal (NES) as follow;

\begin{equation} \mathit{NES} = \frac{1}{DV_{\textit{MTJ}}}\sqrt{S_{V}} \end{equation}

(3)

Obtained NES is 235 nT/$\sqrt{H_{z}} $ at 100 Hz. The result is not only much better than the usual Gauss meter but also much larger than ultra-sensitive MR sensors. However, if one is aware of the small size of the sensor and the large working bias field, one may say that the obtained NES is a good result. For example, a Fe₃O₄ super-paramagnetic NP with its diameter of 15 nm produces 1600 nT at 300 nm distant position under 500 Oe of the bias field. Therefore, the developed sensor may be able to detect 15 nm size super-paramagnetic single particles in seconds (if we vibrate the sensor at 100 Hz).

4. Conclusion

The results show that the developed nanoscale orthogonal MTJ provides excellent detectability under large bias fields owing to their high sensitivity and low noise power. Such MTJ can be considered as a promising candidate for biosensors. The capability of detecting weak magnetic field with such nano-sized MTJ sensors is hoped to enable the development of a new generation of solid-state, matrix-insensitive biosensors and flow cytometers.

Acknowledgments

This work was financially supported by the Vietnamese Ministry of Science and Technology (MOST) with grant number NĐT.88.JP/20. Part of this work was supported by JSPS Grant-in-Aid for Scientific Research (S) Grant Number JP20H05666, Japan and CREST (Non-classical Spin project, JPMJCR20C1) of the Japan Science and Technology Agency.

REFERENCES

Corresponding author

Register with J-STAGE for free!