Myriam Pannetier - Claude Fermon - Gerald Legoff
The measurement of magnetic fields in the femtotesla (10-15 Tesla) range is important for applications like magnetometry, quantum computing, solid state Nuclear Magnetic Resonance, or magneto-encephalography. The only sensors capable of detecting these very small fields have been based on low-temperature Superconducting Quantum Interference Devices (SQUIDs) operating at 4.2 K. We have fabricated a magnetic field sensor based on the combination of a superconducting flux-to-field transformer with a very low noise giant magnetoresistive sensor. Niobium-based and a YBCO-basedYBCO devices have been investigated. Small size high-Tc prototypes provide the capability of measuring 30fT and the potential sensitivity of such sensor is 10-2 fT.
A GMR (Giant MagnetoResistance) spin valve sensor is made of a hard magnetic layer separated from a soft magnetic film by a thin metallic layer. The magnetization of the soft magnetic layer, usually a NiFe film coupled to a CoFe film can rotate easily in an in-plane applied field. The hard layer is composed of an antiferromagnetic layer (IrMn, MnPt…) coupled to a thin ferromagnetic layer (CoFe). The resistance of the whole stack varies with the angle between the magnetization axis of the two layers. Variations of resistance of 6-8% can be obtained in industrial production conditions on 6 inch wafers. With micron size sensors, MR variation of about 5 %/mT can be achieved. In our prototype, we get a MR variation of 2.13%/mT. The resistance R of the GMR stack is 270 W for our 5µm-wide prototype. Thermal noise at room temperature is given by
which corresponds to a sensitivity of 350pT/Hz1/2 with a sensing current I of 1mA for our sensor. At 4.2 K, the sensitivity is 40pT/Hz1/2. As we measure a resistance, the signal is proportional to the sensing current. The sensitivity can thus be increased with the use of a higher current with the limit of heating effects.
At low frequency, GMR sensors also exhibit 1/f noise which is of both structural and magnetic origin. It is globally given by
where g is the Hooge constant and NC is the number of charge carriers in the device. The magnetic noise can be completely suppressed by proper design of the GMR sensor in a yoke-shape with a good length/width ratio (typically 10), (Fig. 1, right panel). With this specific shape of the sensor, the right thickness of the soft layer and a very low roughness of each layer, we obtain in four point measurements in our samples, Hooge constants of about 5x10-3 independent of the applied field. Our samples are good metals, so the number of carriers is comparable to the number of atoms.
Flux-to-field transformation is often done with soft magnetic materials designed as flux concentrators but this solution brings an extra intrinsic noise. Therefore, we have designed a flux-to-field transformer formed of a large superconducting loop closed by a micron-sized constriction (Fig. 1). When an external low frequency field Ha is applied perpendicular to the superconducting loop, a supercurrent is established to maintain the flux through the loop. If this current flows through a constricted area, its high surface density will lead to a very high co-planar magnetic field above and below the narrow part of the loop. This field can be detected by the GMR sensor which is sensitive to the in-plane field. The supercurrent can be calculated from the inductance of the ring and the flux screened, or by effective area which screens the field. To maximize the local amplification effect, one needs to use a ring with a large outer diameter and a width of 0.7 times the radius, as well as a constriction that is as small as possible. A rough estimation of the gain is given by the ratio of the loop radius to the constriction width.
Figure 1: (A) Schematic view of the device, comprising a GMR sensor, measured with four-probe contacts, with a superconducting loop of Niobium, electrically isolated from the GMR by 400 nm of Si3N4, and being closed above the GMR by a constricted area. (B) Close-up view of the constriction area. The black line is a 20µm scale. The GMR is patterned in a yoke-type shape of 70 µm length, and Titanium contacts allow measurement of the central, domain-free, part. The superconducting constriction is located above the GMR, without covering the entire measured part. (C) The Niobium loop acts as a flux-to-field transformer. The Ha field, applied perpendicular to the plane of the sample, generates in the superconducting loop a supercurrent Is tending to expel the flux from entering the superconductor, due to the Meissner effect. The field lines (dashed lines) due to the supercurrent are locally enhanced in the constriction, and their horizontal component can be detected by the GMR sensor positioned under the constriction, according to the direction they give to the free layer magnetization, relative to the hard layer direction (bottom black arrow).
Measurements on a niobium-based device:
The first experiment is performed on a GMR/Si3N4/Nb sample. A commercial GMR stack (fabricated by IPHT-Jena, Germany) deposited on a silicon substrate has been used as a base of the device.
The GMR sensor has been measured first in a parallel field as function of temperature, for calibration (inset Fig. 2. Fig. 2 gives the resistance from -28 to 28 mT at 5 K. The MR ratio is 9 % and a maximum sensitivity of 2.13 %/mT is obtained in the middle of the linear part of the curve. The sample was cooled down below Tc in the corresponding in-plane bias field of 2 mT to insure that measurements will be done in the most sensitive region. At 4.2 K, the response of the system to a small field Ha is given in Figure 3. The effective gain of the transformer is 108, given by the ratio between the slope of the GMR in parallel field (2.13 %/mT) and the slope for Ha (231 %/mT). This ratio does not depend on temperature in the superconducting state. The supercurrent (measured through the field created) reaches a critical value for an applied field of 5.85.10-6 T limiting the parallel field and thus leading to a plateau in the GMR resistance. When one sweeps the field back, the supercurrent decreases to zero then to negative values up to the critical point in counter-rotating direction. The GMR resistance decreases with the same slope and reaches a low saturation value. The same process is reproduced as soon as the field is swept back to the high saturation value. The inset of Figure 2 gives the total MR variation as a function of temperature. It decreases with temperature, reaching zero at the critical temperature of the constriction, 6 K, which is lower than the critical temperature of the main loop (8.5 K). The effective gain leads to a sensitivity of 540 fT/ Hz1/2 with a 1mA sensing current (Fig. 4).
Figure 2: Magnetoresistance as a function of in-plane applied field at 4.2K. The superconducting loop has been cut open to avoid persistent supercurrent effect. The slope gives the reference curve to calculate the gain obtained when a supercurrent flows. (Inset) Magnetoresistance ratio as a function of temperature. At 4.2K, the ratio is 9 % to be compared with the room temperature ratio of 4.5 %.
Figure 3: Magnetoresistance as function of a perpendicular applied field at 4.9K. The upper plateau (DR/R)max (respectively lower plateau) corresponds to the point where the supercurrent reaches the critical value (respectively minus the critical value). The slope corresponds to a resistance variation of 213%/mT to be compared to 2%/mT with an in-plane applied field, i.e. an enhancement factor of 108 for the flux-to-field transformer (Inset) Variation of (DR/R)max as function of the temperature. This gives a direct measurement of the critical temperature of the constriction.
Figure 4: For the Niobium-based device, the noise spectra of the voltage output of the GMR is given with a zero current (lower curve), with 1mA at 4.2K (middle curve) and 1mA at 77K (upper curve). The field sensitivity of the device with 1mA at 4.2K is of 540fT/sqrtHz at 100Hz. At 4.2K, the 1/f noise dominates the thermal noise below 150Hz.
Measurements on a YBCO-based device:
Mixed sensors with high-Tc superconductors present several advantages: they work at 77 K with a rather good sensitivity; at 4.2 K, they offer a large critical current and very high sensing currents can be applied, up to 15 mA. In a second experiment, we measured a sample with YBa2Cu3O7-d loop (Tc = 85 K) of comparable size to the Niobium one. Preliminary results have shown that the behavior is identical to the Nb-loop sensor and the measured gain of 100, constant up to 80K (Fig. 5), is in good agreement with the theoretical predictions. Fig. 6 gives the sensitivity at 4.2K of our high Tc prototype with a 15mA, that is 35 fT/Hz1/2 (thermal noise), which is comparable to noise level of high-temperature SQUIDs.
Figure 5: Magnetoresistance as function of a perpendicular applied field at 4.2K on the YBCO device, from the field cooled state. At this temperature, the entire GMR variation is explored when applying a sufficient field. The upper plateau (DR/R)max respectively lower plateau) corresponds to the saturation of the GMR in the anti-parallel (respectively parallel) magnetization orientation of the soft and hard layer. The critical current at this temperature is reached just after the saturation and is responsible of the apparent aperture of the cycle. The maximum slope corresponds to a resistance variation of 310 %/mT to be compared to 3.11%/mT with an in-plane applied field, i.e. an enhancement factor close to 100 for the flux-to-field transformer. (Inset) Amplification gain obtained on the YBCO-based sensor as a function of temperature, calculated from the ratio between the slope measured on the GMR sensor when applying a perpendicular field and the slope of the GMR in an in-plane field at the same temperature.
Figure 6: The noise spectra of the YBaCuO mixed sensor at 4K and 77K, with respectively 15mA and 5mA of sensing current. At 4.2K and with a 15mA sensing current, the sensitivity of the YBCO-device reaches 35fT/Hz1/2, comparably to noise level of high-temperature SQUIDS.
From theser results, we can now predict theoretically the potential sensitivity of this new type of sensor. With a 1 micron constriction and 3 cm diameter loop as used in MEG the calculated gain is 7500, a factor 75 higher than our prototypes. If we use high Tc superconductors, we can use a sensing current of 5 mA at 77 K. The resistivity of the GMR system with a 1 micron constriction is about 1500 Ohms. These values lead to a sensitivity of 1 fT/ Hz1/2 at 77K (thermal noise) that should be obtained with a standard GMR sensor. The 1/f noise should dominate the thermal noise below 200 Hz. TMR (Tunnel Magneto Resistance) sensors (19), where the thin metallic interlayer is replaced by a thin insulator layer acting as a tunnel barrier can also be used. MR of 50% is currently obtained leading to a MR variation of 25%/mT. The resistance of the TMR can be also higher on a small surface. If we replace the GMR sensor by a TMR sensor we gain a factor about 5 achieving a sensitivity of 0.2fT/ Hz1/2 at 77K.
Ultimately, one could use TMR sensors with half-metallic electrodes. This kind of sensor, based on manganite compounds, exhibits low field magnetoresistance of 1800% at 4.2K. They can be epitaxially grown on the high Tc films producing a very high quality film. With the proper substrate, manganite films with extremely low 1/f noise have been produced. Such systems should be able theoretically to reach sensitivity around to 0.01fT/Hz1/2.
Femtoteslta magnetic field measurement with magnetoresistive sensors
M. Pannetier, C. Fermon, G. Le Goff, J. Simola, E. Kerr, Science, 304, 1648-1650 (2004).
Ultra-sensitive field sensors – An alternative to SQUIDs
M. Pannetier, C. Fermon, G. Le Goff, J. Simola, E. Kerr, M.S. Welling, R.J. Wiijngaarden,
Submitted to IEEEE Trans. On Appl. Supercond. ASC’04 (Octobre 2004, Jacksonville – USA).
Noise in small magnetic systems – Applications to very sensitive magnetoresistive sensors
M. Pannetier, C. Fermon, G. Le Goff, J. Simola, E. Kerr and J.M.D. Coey.
To be published in JMMM. JEMS’04 (Septembre 2004, Dresden – Allemagne).
Field line distribution in a mixed sensor
M. Pannetier, C. Fermon, P. Vedrine, M.S. Welling, R.J. Wiijngaarden.
Submitted to Sensors and Actuators. EMSA 2004 (Juillet 2004, Cardiff – UK).
Copyright CEA 25/02/06