Design optimization of flux concentrators for magnetic tunnel junctions-based sensors

This paper presents a design optimization scheme for magnetic tunnel junction-based sensors that balances flux concentrator gain and magnetic noise through finite element simulations and analytical modeling, ultimately achieving a three-orders-of-magnitude performance improvement over single-junction designs.

The Gist

Imagine you are trying to hear a single whisper in a crowded, noisy stadium. That's what scientists face when they try to detect ultra-weak magnetic fields (like those from the human brain or distant planets). The "whisper" is the tiny magnetic signal, and the "crowd noise" is the electrical and magnetic static that naturally exists in electronic sensors.

This paper is about building the ultimate ear for these whispers. The researchers are designing a special sensor called a Magnetic Tunnel Junction (MTJ) and trying to make it so sensitive it can hear a whisper from across the stadium.

Here is how they did it, broken down into simple concepts:

1. The Problem: The "Whisper" vs. The "Noise"

The sensor works like a tiny switch that changes its electrical resistance when a magnetic field hits it.

• The Goal: Make the switch react strongly to the whisper (High Sensitivity).
• The Problem: To make the switch react stronger, you usually have to make it bigger or run more electricity through it. But doing that also makes the sensor "shout" its own internal noise louder. It's like turning up the volume on a radio to hear a faint station, but the static gets louder too.

2. The Solution: The "Magnetic Funnel" (Flux Concentrator)

To solve this, the researchers added a Flux Concentrator (FC). Think of this as a giant, magnetic funnel made of a special soft metal (Permalloy).

• How it works: Imagine the magnetic field lines are like rain falling on a roof. The funnel catches all that rain from a wide area and squeezes it into a tiny bucket (the sensor).
• The Benefit: This amplifies the "whisper" (the magnetic signal) by hundreds of times before it even hits the sensor. Crucially, because the sensor itself hasn't changed, its internal "noise" stays the same. So, you get a louder signal without getting any noisier.

3. The Dilemma: The "Crowded Room"

Here is where it gets tricky. The researchers wanted to make the sensor even better by using many tiny sensors (MTJs) packed together inside the funnel's bucket (the air-gap).

• The Logic: More sensors = more volume = less noise (averaging out the static).
• The Catch: To fit more sensors, you have to make the bucket (the air-gap) wider.
• The Trade-off: If you make the bucket too wide, the funnel stops working as well. The "rain" (magnetic field) starts leaking out the sides, and the amplification drops.

It's a balancing act: Do you make the bucket wide to fit more sensors, or keep it narrow to keep the funnel strong?

4. The Experiment: Testing the Shape

The researchers used powerful computer simulations (like a virtual wind tunnel) to test different shapes and sizes.

• They tried making the gap wider and narrower.
• They tried putting sensors side-by-side (increasing the length of the bucket) versus stacking them (increasing the width).

The Discovery:
They found that width is the enemy, but length is the friend.

• Making the gap wider (to fit bigger sensors) kills the signal amplification.
• Making the gap longer (to fit more sensors in a line) hurts the amplification very little.

So, the best strategy is to use many small sensors lined up in a long row, rather than a few big sensors.

5. The "Magic Formula"

To avoid running thousands of computer simulations, the team created a simple math formula based on "magnetic resistance" (think of it like electrical resistance, but for magnetic flow).

• They treated the magnetic funnel like an electrical circuit.
• This formula allowed them to predict exactly how much signal would be lost based on the gap size, without needing a supercomputer every time.

6. The Grand Result: The Sweet Spot

Using their formula, they calculated the perfect design:

• Shape: A rectangular funnel with no wedge (no fancy tapering).
• Sensors: About 160 tiny sensors lined up in a row.
• Size: Each sensor is very small (about 13 micrometers wide).

The Payoff:
By following this design, they improved the sensor's ability to detect weak fields by 1,000 times (3 orders of magnitude) compared to a single sensor.

• Before: Detecting a field of 55 nanotesla.
• After: Detecting a field of 55 picotesla.

The Bottom Line

The paper teaches us that to hear the faintest whispers of the universe, you don't just need a bigger microphone. You need a smart funnel that gathers the signal efficiently, and you need to pack many small microphones into a long, narrow hallway rather than a wide, shallow room. This design allows us to build sensors small enough to fit in a pocket but sensitive enough to map the magnetic fields of the human brain or explore deep space.

Technical Summary

1. Problem Statement

The paper addresses the challenge of developing miniaturized, ultra-sensitive magnetometers capable of detecting ultralow magnetic fields (picotesla range) for applications in space exploration and biomedical monitoring.

• The Trade-off: Optimizing Magnetic Tunnel Junction (MTJ) sensors involves balancing two competing factors:
  1. Sensitivity: Can be increased by integrating high-gain Permalloy flux concentrators (FC) to amplify the external magnetic field at the sensor site.
  2. Noise: The dominant noise source in MTJs at low frequencies is magnetic $1/f$ noise, caused by thermal fluctuations of magnetization. This noise is proportional to sensitivity but can be reduced by increasing the total magnetic volume (i.e., using larger or more junctions).
• The Conflict: To reduce noise, one must pack more MTJs into the sensor. However, packing more junctions requires a wider air-gap in the flux concentrator. A wider air-gap significantly reduces the magnetic flux gain provided by the concentrator, thereby negating the sensitivity improvements.
• Objective: The authors aim to find the optimal geometric trade-off between the flux concentrator's air-gap dimensions and the number/size of MTJs to maximize overall sensor detectivity (Signal-to-Noise Ratio).

2. Methodology

The authors employed a multi-step approach combining numerical simulation, analytical modeling, and system-level optimization:

• Finite Element Simulations (FEM):

  • Used COMSOL Multiphysics to simulate the magnetic behavior of NiFe flux concentrators ($\mu_r = 1500$, thickness $t=8$ $\mu$m).
  • Investigated the dependence of the concentrator gain ($G$) on air-gap width ($w$) and length ($l$).
  • Simulated external fields of 1 $\mu$T to calculate the field amplification in the air-gap.

• Analytical Modeling (Magnetic Reluctance):

  • Developed an analytical formula based on the concept of magnetic reluctance ($\mathcal{R}$) to reproduce simulation results.
  • Modeled the FC as a series of reluctances: two magnetic arms (rectangular and wedged regions) and the air-gap.
  • Derived a modified formula for gain $G(w, l)$ that accounts for the non-conservation of flux in open circuits, fitting the simulation data with high accuracy without free parameters (except a prefactor $G_0$).

• Sensor Detectivity Optimization:

  • Defined the sensor detectivity ($D$) as the ratio of noise to sensitivity.
  • Modeled the noise as dominated by magnetic $1/f$ noise (Hooge model), which scales with the total junction area ($A = N \cdot \pi d^2 / 4$).
  • Established geometric constraints:
    • Air-gap width $w = d + 2b$ (where $d$ is junction diameter, $b$ is lithography margin).
    • Air-gap length $l \approx N(d + a)$ (where $N$ is the number of junctions, $a$ is spacing).
  • Formulated an objective function $F(d, N)$ to minimize, representing the inverse of the signal-to-noise ratio, subject to the constraint that the junction row must fit within the concentrator width ($N(d+a) \leq W$).

3. Key Contributions

• Reluctance-Based Analytical Model: The authors successfully adapted the magnetic reluctance concept to open flux concentrator systems. They derived a closed-form analytical expression for FC gain that accurately predicts simulation results across varying air-gap geometries, providing a rapid design tool without needing heavy FEM simulations.
• System-Level Optimization Framework: They integrated the FC gain model with MTJ noise characteristics to create a holistic optimization model for sensor detectivity.
• Counter-Intuitive Design Insight: The study challenges the assumption that maximizing gain (via wedged shapes or small gaps) is always optimal. It demonstrates that maximizing the magnetic volume (by adding more junctions) often outweighs the loss in concentrator gain.

4. Results

• Gain vs. Geometry:
  • Width ($w$): Gain decreases drastically as air-gap width increases (following an approximate $1/w$ law). Doubling the width from 5 to 10 $\mu$m nearly halves the gain.
  • Length ($l$): Gain decreases gently as air-gap length increases.
  • Conclusion: It is significantly more efficient to add junctions along the length of the air-gap rather than increasing the width.
• Optimal Configuration:
  • The optimization function $F(d, N)$ revealed that the best detectivity is achieved with a rectangular flux concentrator (no wedge) and a large number of small junctions.
  • Optimal Parameters: Approximately 160–166 junctions with a diameter of ~13 $\mu$m.
  • Geometry: This results in an air-gap width of ~19 $\mu$m and a gain of approximately 80.
• Performance Improvement:
  • Using the optimized design (assuming a Hooge parameter of $1.6 \times 10^{-7}$ $\mu$m$^2$ and sensitivity of 20%/mT), the estimated detectivity is 55 pT/$\sqrt{\text{Hz}}$ at 10 Hz.
  • This represents an improvement of three orders of magnitude compared to a single MTJ sensor without a flux concentrator (reference detectivity: 55 nT/$\sqrt{\text{Hz}}$).

5. Significance

This work provides a critical roadmap for the design of next-generation ultra-sensitive magnetometers. By quantifying the trade-off between flux concentrator geometry and MTJ array density, the authors demonstrate that maximizing the total magnetic volume (via high-density arrays) is the most effective strategy for reducing $1/f$ noise, even if it necessitates a wider air-gap and lower concentrator gain.

The proposed analytical reluctance model offers a robust, computationally efficient tool for engineers to design flux concentrators for specific sensor requirements, potentially enabling the deployment of picotesla-level sensors in compact, low-power applications such as medical diagnostics and space instrumentation.