improved by AC excitation: flipping for AMR and AC biasing for GMR. AC excitation lowers

Transcription

1 AC - driven AMR and GMR magnetoresistors P. Ripka 1, M. Tondra, J. Stokes and R. Beech. 2 1 Czech Technical University, Faculty of Electrical Engineering, Dept. of Measurement, Praha 6, Czech Republic. ripka@feld.cvut.cz 2 Nonvolatile Electronics, Valley View Road, Eden Prairie, MN , USA. markt@nve.com, Abstract. Anisotropic (AMR) and Giant (GMR) Magnetoresistive sensors are attractive for industrial applications, as they are more sensitive and stable than Hall sensors. Their performance can be improved by AC excitation: flipping for AMR and AC biasing for GMR. AC excitation lowers the hysteresis, reduces the offset and in some cases also decreases sensor noise. The sensitivity to perpendicular fields is reduced in case of AMR sensors. AC-driven magnetoresistors are competitive with miniature fluxgate sensors and they are suitable for precise applications such as compasses. Keywords: Magnetic sensors; Magnetoresistors; GMR; AMR

2 1. Introduction Ferromagnetic (AMR and GMR) magnetoresistors are superior to any semiconductor magnetic field sensors for low-field applications [1]. Recent developments in this technology allow ferromagnetic magnetoresistors to compete with miniature fluxgate sensors [2]. Magnetoresistors are used for proximity switches, speed sensors, angular and displacement measurement, compasses for automotive navigation systems, current meters, magnetic ink reading and security applications. Semiconductor magnetoresistors are less sensitive than AMR and GMR, but newly developed multiple-element InSb magnetoresistors in combination with rare-earth permanent magnets are attractive for position-sensing in automotive applications as they are also stable in required -60 to +200 C temperature range [3]. 2. Anisotropic Magnetoresistors (AMR) The Anisotropic Magnetoresistance (AMR) is observed in any ferromagnetic metal. Unlike the Hall effect, which is easy to understand from a strictly classical electromagnetic perspective, AMR is fundamentally a quantum-mechanical effect. The AMR is due to spin-orbit interactions of conduction electrons in the metal. The anisotropy in AMR refers to the fact that the resistance is a function of the angle between the current and the magnetization. The basic quadratic characteristics of a single-strip AMR sensor may be linearized by proper biasing. By AC biasing, a long-term offset stability of 20 nt/10 hours was achieved [4]. However, most of the AMR sensors are made as a bridge consisting of four meanders of permalloy. The barber

3 pole structure of aluminium strips change the current direction by 45 0 which also results in linear characteristics. A flipping field perpendicular to the sensing direction of the AMR bridge sensor was first generated by an external coil [5]. It is necessary to fully saturate the sensor with flipping pulses having opposite polarity to avoid distortion of the sensor characteristics caused by perpendicular fields. The Philips KMZ10A1 sensor bridge output was sensed by a gated integrator for intervals between flipping pulses of 10 µs duration and 100mA amplitude which caused a field of 4 ka/m. This technique reduced the hysteresis below 100 ppm FS and resulted in improved offset stability on the order of 10 nt. Significant noise reduction was not observed for single-axis flipping [6]; however, suppression of Barkhausen noise was achieved by using a rotating bias [7]. Newly developed AMR sensors such as Honeywell HMC 1002 and Philips KMZ 51 have integrated flipping and feedback coils. They are more compact, and even if they require a large amplitude flipping current such as 4 A, they still can work from a 5 V source when switched-capacitor techniques are employed. 3. Giant Magnetoresistance Like AMR, Giant Magnetoresistance (GMR) is also a quantum-mechanical effect. The simplest GMR structure is a metallic sandwich made of [ferromagnet / non-ferromagnetic conductor / ferromagnet] or [FM / C / FM]. The most common commercial class of these GMR materials is Co/Cu/Co. This basic structure may be modified by using various alloys in place of the Co and Cu, but the basic idea is the same. The physical principle of GMR is that spin up and spin down

4 electrons have different scattering probabilities in ferromagnetic layers and at interfaces between the ferromagnetic and non-ferromagnetic layers. This difference in scattering rates is observable when the relative orientation of the magnetizations in the magnetic layers changes. The resistance is lowest when the layers magnetizations are parallel and highest when their magnetizations are anti-parallel. Assuming the magnetization within each magnetic layer can be made uniform, the angular dependence of the resistance on the angle between magnetizations is 1 1 R= R + ( R R ) sinθ where R R R = ( ). The GMR is usually calculated as a percentage: R ( R R ) GMR = =. R R The GMR can routinely be made to be 10% in simple sandwich structures. More complicated structures, such as antiferromagnetically coupled multilayers, can have GMR s approaching 100%. These structures tend to require much larger fields to saturate them, though. So a useful figure of merit for GMR materials is sensitivity expressed in %/Oe which is calculated by taking the GMR and dividing by the saturation field. An important variation of the simple FM / C / FM sandwich is a pinned sandwich. This is constructed by using a high coercivity film which is exchange-coupled to one of the FM films so that it is magnetically rigid in relatively small fields. This structure, often referred to as a spin

5 valve, has resistance vs. field characteristics that looks very much like the B vs H loops of the unpinned FM layer. The new structure is PIN / FM / C / FM. Pinning one of the FM layers makes the GMR output of a device easier to interpret because only one of the FM layers is free to respond to an external magnetic field. However, adding the extra pinning layer makes the material somewhat harder to fabricate. Pioneering work in the field of Giant Magnetoresistance (GMR) was done in the middle to late 1980 s. The study of exchange coupling between ferromagnetic layers in ferromagnetic - nonmagnetic multilayer thin films led directly to the discovery of the GMR effect [8,9,10,11]. These early discoveries were typically studies done at low temperatures and were on magnetic structures which required large fields (greater than 1 Tesla) to saturate the magnetization. Subsequent work resulted in structures which have large magnetoresistive effects in low fields, with a consequently large sensitivity quotient (% change in R / saturation field) [12]. Continued development of GMR materials has resulted in magnetoresistive structures which are compatible with integrated circuits and are applicable to a wide range of commercial applications even at elevated temperatures [13,14]. 4. GMR Sensor Construction The GMR sensor measured in the present paper was developed at NVE. The magnetoresistive layer structure is: [Si / Si 3 N Å substrate] / Ta 30 Å / NiFeCo 40 Å / CoFe 15 Å / Cu 40 Å / CoFe 15 Å / NiFeCo 40 Å / Ta 200 Å. The film was deposited on a 4 Si wafer using RF diode sputtering. NVE s GMR sandwich film operates at temperatures above C, has a temperature

6 coefficient of about 1500ppm / 0 C, and has a sheet resistance above 10 Ω/square (high sheet resistance allows the final size of integrated devices to be smaller). The film is annealed at C. Fig. 1 shows the resistance vs. field response of the sandwich. A single GMR test structure on the wafer was tested with 1mA through a 2 µm wide line. The sheet resistance of the film is about 12 Ω/square, and the GMR is about 4.8%. The micromagnetic details of the operation of this type of sensor are complicated. There are several competing magnetic effects within the narrow stripe of GMR material which, when properly balanced, result in a high sensitivity film. They are: 1) interlayer exchange coupling 2) interlayer magnetostatic coupling 3) edge and demagnetization effects 4) growth-induced magnetic anisotropy 5) current-induced self field biasing. For Cu thicknesses greater than about 35 Å, both interlayer coupling effects (exchange and magnetostatic) generally decrease with increasing separation between the two ferromagnetic layers in the sandwich. For very small Cu thickness, there is a strong antiparallel coupling with an oscillatory dependence on the Cu thickness [15]. But the 40 Å Cu thickness used for this sensor is large enough that the exchange coupling is negligible.

7 The usual demagnetizing fields found in any magnetized material work to keep the magnetization of the top and bottom layers along the stripe. But same demagnetizing fields also result in a field being applied to the other layer around the side of the GMR sandwich. This moves the sensor more towards an antiparallel across-the-stripe configuration. The sensor lines are fabricated so that the induced easy axis of the ferromagnetic layers is perpendicular to the length of the stripe. And any current passed through the sensing strip results in across-the-stripe fields being applied to the top and bottom layers, though in opposite directions. The sensor cross-section is shown in Fig. 2. Using standard semiconductor processing techniques, the material is patterned into resistors that are 1.8 µm wide and have a total resistance of about 1kΩ. The resistors are connected into a bridge configuration using a patterned 0.5 µm thick Al film. Over the top of the bridge are the integrated biasing straps made of patterned 1µm thick Al film. The biasing strap has the shape of a flat coil with individual resistors positioned in the peripheral part so that they are biased in opposite directions. The field from this biasing strap in the location of magnetoresistors is approximately 0.1 mt / ma in the plane of the sandwich. Without a biasing field, the four magnetoresistors are symmetrical, so the bridge has, in the ideal case, zero output, independent of the external field. An on-chip biasing field causes a shift of resistor characteristics so that the bridge output is proportional to the measured field.

8 5. Measured values Large field characteristics of the described sensor are shown in Fig. 3 for two values of bias current: sensitivity is increasing with the current value, but at currents over 12 ma the sensitivity changes are small. However, bias current is limited to approx. 15 ma due to the heating of the chip. Fig. 4 shows that the characteristic is reversed for opposite bias polarity. Unlike the flipping in an AMR sensor, the bias field must be present during the measurement period of the GMR sensor. The DC biased GMR sensor has 5% hysteresis in the 300µT range as seen from the Fig. 5. The noise properties were measured by SR 770 spectrum analyzer. The sensor was in a six-layer cylindrical mu-metal magnetic shielding [4]. Fig. 6 shows the noise spectrum at low frequencies and an example of the sensor output time plot. The measurement was performed for 12 ma bias current. The noise has 1/f spectrum with power density of 43 nt rms/ Hz at 1 Hz. MR sensors may be also biased by an AC current. For the first experiments the sensor was supplied from built-in source of SR 830 Lock-in amplifier and the bridge output was measured using the same instrument. The resulting sensitivity was independent of the bias frequency up to 100 khz. An example of the measured characteristics is in Fig. 7 for 10kHz/5mA rms bias. Although the sinewave bias used in this first study is far from the optimum waveform, we observed reduction of the hysteresis to 1% and offset to 1µT. Fig. 8 shows the noise properties of the AC biased sensor: the noise was reduced to 16 nt rms/ Fig. 9 shows the large-field charactristics of the AC biased sensor.

9 6. Conclusion AC excitation of magnetoresistors improve their characteristics. In the case of AMR the short flipping pulses of large amplitude cause changes of the sensor remanent magnetisation and the sensor output is measured between these pulses. GMR sensors can be biased by much lower current but during the whole measurement cycle. By using sinewave bias the hysteresis was reduced to 1%, offset to 1 µt and we also observed reduction of the sensor noise by the factor of 3. Use of bias waveforms other than a sinewave, which was used in the present study, would be advantageous: squarewave bias current with short duration may allow use of higher field amplitudes and keep the chip temperature low at the same time. Acknowledgement: This work was supported by the Grant Agency of the Czech Republic under No. 102/96/1251 References [1] P. Ripka: Magnetic sensors for industrial and field applications, Sensors and Actuators A 42 (1994), [2] S. Kawahito, H. Satoh, M. Sutoh, Y. Tadokoro: High-resolution micro-fluxgate sensing elements using closely coupled coil structures, Sensors and Actuators A 54 (1996),

10 [3] J. Heremans: Magnetic field sensors for magnetic position sensing in automotive applications, Proc. 16 th Conference on Properties and Applications of Magnetic Materials, Chicago, IIT 1997, unpaged. [4] D. Flynn: Demodulation of a magnetoresistive sensor signal to achieve a low-cost, stable and high-resolution vector magnetometer, Sensors and Actuators A 50 (1995), [5] P. Ripka: AC - excited magnetoresistive sensor, J. Appl. Phys. 79 (8), 1996, pp [6] P. Ripka: Noise and stability of magnetic sensors, Journ. Magn. Magn. 157/158 (1996), [7] E. Paperno, B.Z. Kaplan: Suppression of Barkhausen Noise in Magnetoresistive Sensors Employing AC Bias, IEEE Trans. Magn. 31 (1995), [8] P. Grunberg, R. Schreiber, Y. Pang, M.B. Brodsky, and H. Sowers: Layered magnetic structures: Evidence for antiferromagnetic coupling of Fe layers across Cr interlayers, PRL 57 (1986), pp [9] M.N. Baibich, J.M. Broto, A. Fert, F. Nguyen Van Dau, F. Petroff, P. Etienne, G. Creuzet, A. Freiderich, and J. Chazelax: Giant magnetoresistance of (001)Fe/(001)Cr magnetic superlattices, PRL 61 (1988), pp [10] G. Binash, P. Grunberg, F. Saurenback, and W. Zinn: Enhanced magnetoresistance in layered magnetic structures with antiferromagnetic interlayer exchange, PRB 39 (1989), pp [11] A. Barthelemy, A. Fert, M.N. Baibich, P. Etienne, S. Lequien, and R. Cabanel: Magnetic and transport properties of Fe/Cr superlattices, JAP 67 (1990), pp

11 [12] B.Dieny,V.S.Speriosu,S.S.P.Parkin,B.A.Gurney,D.R.Wilhoit,andD.Mauri: Giant magnetoresistive in soft ferromagnetic multilayers, PRB 43 (1991), pp [13] J.M. Daughton, P.A. Bade, M.L. Jenson, and M.M.M. Rahmati: Giant Magnetoresistance in Narrow Stripes, IEEE Trans. Mag. 28 (1992), pp [14] J.M. Daughton, Weakly Coupled GMR Sandwiches, IEEE Trans. Mag., 30, #2, March 1994, pp [15] S.S.P. Parkin, N. More, and K.P. Roche: Oscillations in exchange coupling and magnetoresistance in metallic superlattice structures: Co/Ru, Co/Cr, and Fe/Cr, PRL 64 (1990), pp

12 Figure captions Fig. 1 Resistance vs. field characteristics of a single GMR test structure. The device was tested with 1mA through a 2 µm wide line. The sheet resistance of the film is about 12 Ω/square, and the GMR is about 4.8% Fig. 2 Cross-section of the GMR sensor: the biasing strips (forming the flat coil) are on the top. Fig. 3 Large field characteristics of the complete GMR sensor. The bias current was 4 ma and 8 ma. Fig. 4 Large GMR characteristics for +/- 12 ma bias Fig. 5 Low-field characteristics for +/- 8 ma bias current Fig. 6 Noise spectrum of the GMR biased by 12 ma DC current Fig. 7 Low-field characteristics of the sensor biased by sinewave current Fig. 8 Noise spectrum of the GMR biased by 10 khz/ 5 ma sinewave current Fig. 9 Large field characteristics of the sensor biased by 10 khz/ 5 ma sinewave

13 Biographies Pavel Ripka Received an Ing. degree in 1984, a CSc (equivalent to PhD) in 1989 and Doc. degree in 1996 at the Czech Technical University, Prague, Czech Republic. He works at the Department of Measurement, Faculty of Electrical Engineering, Czech Technical University as a lecturer, teaching courses in Electrical Measurements and Instrumentation, Engineering Magnetism and Sensors. His main research interests are Magnetic Measurements and Magnetic sensors, especially Fluxgate He is a member of Elektra society, Czech Metrological Society, Czech National IMEKO Committee and Eurosensors Steering Committee.. Mark Tondra received a B.S. in physics and mathematics, with honors, from the University of Wisconsin- Madison in 1989 and Ph.D. in 1996 from the University of Minnesota in solid state physics. He is with NVE as a physicists involved in the development of low field magnetic sensors. He is currently managing programs for the development of Spin Dependent Tunneling Devices and Spin Transistors. His research is focused on the magnetotransport properties of thin films, including AMR, GMR, and the extraordinary Hall effect. Russell S. Beech - received BS degrees in physics and chemistry from Bemidji State University in 1988 (summa cum laude), and a PhD degree in electrical engineering from the University of Iowa in Till 1993 he was teacher and researcher at the University of Iowa. Since that he is with NVE as Director of Engineering. John Stokes received a BS degree in Electrical Engineering from the University of Minnesota in From 1980 till 1995 he was with Honeywell: his last position was as a Senior Research Scientist at Honeywell s Research Center in Minneapolis, MN. While at Honeywell, he developed a wide range of integrated circuits in bipolar, CMOS, Bipolar Enhanced CMOS, and Gallium Arsenide technology. Since 1995 he is with NVE as senior circuit designer.

14 Voltage Field (Oe)

15 X X X X X X Silicon Nitride GMR Film Silicon Substrate

16 B(mT) -50 output (mv) 8mA 4mA

17 -12 ma output (mv) 12 ma B(mT)

18 40 30 output (mv) ,3-0,2-0,1 0 0,1 0,2 0,3 B(mT)

19

20 output (mv) 0-0,3-0,2-0, ,1 0,2 0,3-20 B(mT)

21

22 output (mv) B(mT)