HC-07 1 Magnetic Field Dependence of the Noise in a Magnetoresistive Sensor having MEMS Flux Concentrators Arif Ozbay, E.R. Nowak, A. S. Edelstein, G. A. Fischer, C. A. Nordman, and Shu Fan Cheng One technique to mitigate the effects of the sensor’s 1/f Abstract— We report the DC and AC magnetic field noise is to modulate the incoming magnetic signal and thus dependence of the low frequency noise in a MEMS flux shift the operating frequency of a GMR sensor . This can concentrator device containing a giant magnetoresistance spin be accomplished by depositing flux concentrators (high valve. The noise is dominated by resistance fluctuations having a magnetic origin. Under nominally zero magnetic field biasing permeability magnetic material such as Ni-Fe) on MEMS conditions, the noise power is large and varies rapidly with small structures that oscillate at kHz frequencies. Depending upon changes in magnetic field. Metastability between distinct resistive the sensor, shifting the frequency reduces the sensor’s 1/f noise states is observed and can be suppressed with the application of a by one to three orders of magnitude at 1 Hz. However, low moderate longitudinal field. Stationary flux concentrators do not frequency magnetic noise in the flux concentrators (FCs) must contribute excess noise, rather the dominant source of noise is the be negligible else signals of interest may be obscured. In this spin valves themselves. This result indicates that the device is likely to increase the sensitivity of many magnetic sensors at low manuscript, we show that GMR elements themselves are the frequencies by orders of magnitude. dominant source of noise and that stationary FCs in our MEMS device do not introduce excess noise over a wide range Index Terms— field sensor, flux concentrator, of AC and DC field biasing conditions. magnetoresistance, magnetometer, noise II. DESIGN AND MATERIALS I. INTRODUCTION The inset of Fig. 1 illustrates the concept of our AC MEMS M AGNETORESISTIVE materials are candidates for developing low-power, miniature magnetic field sensors for the detection of sub-nanoTesla magnetic fields at flux concentrator device. A GMR spin valve (SV) sensor is sandwiched between two trapezoidal-shaped Permalloy FCs that are deposited on MEMS flaps. The FCs have, roughly, a milliHertz frequencies. Spin-electronic devices are of interest 80 µm front face, 150 µm rear face, 100 µm height, and a due, in part, to their compatibility with standard silicon 0.25 µm thickness. The MEMS flaps are driven to oscillate at microelectronics processing and their large magnetoresistance frequencies of order of 10 kHz by electrostatic comb drives. [1, 2]. For example, giant-magnetoresistance (GMR) and Silicon springs couple the two flaps and ensure the oscillations tunneling-magnetoresistance (TMR) devices offer much larger are in-phase. Further details of the device can be found in signals in low fields (< 10 G) compared to anisotropic Refs. [5, 6]. In our study, the FCs were stationary. magnetoresistive (AMR) sensors [1, 3]. However, signal Two sets of samples were investigated: one set consisted of resolution at sub-Hertz frequencies is often limited by low only SVs and the other of SVs with adjacent Permalloy FCs. frequency noise having a power spectrum that varies inversely The SVs were 2 µm x 90 µm and had the following with frequency, namely 1/f noise. Although GMR and TMR composition: Si/Si3N4//Ta/35 NiFeCo/50 Ta/ 42.5 devices can have more sensitivity, they also tend to exhibit NiFeCo/12.5 CoFe/ 27.5 Cu/43.5 CoFe/325 CrPtMn, where larger 1/f resistance noise. For this reason, AMR sensors the numbers denote layer thickness in Angstroms. The double currently provide the highest magnetic field resolution at free-layer is used to reduce the hysteresis and improve the frequencies below 1 Hz . linearity . A transverse (x-axis) field, Ht, and a longitudinal (y-axis) field, Hl, could be applied using our apparatus. For sensing, the signal is applied along the x-axis, transverse to the Manuscript received March 10, 2006. This work was supported in part by the National Science Foundation under grant #0405136, the donors of The free layer’s easy axis which lies along the length of the SV. American Chemical Society Petroleum Research Fund, and the Cottrell Scholar Program of the Research Corporation. III. RESULTS AND DISCUSSION Arif Ozbay and E. R. Nowak are with the Department of Physics and Astronomy, University of Delaware, Newark DE 19711 USA (phone: 302- The main panel of Fig. 1 shows the normalized 831-1087; e-mail: email@example.com, firstname.lastname@example.org). magnetoresistance of SVs with and without the FCs. The A. S. Edelstein and G. A. Fischer are with U.S. Army Research resistance is greatest when the free and fixed layer are Laboratory, Adelphi, MD 20783 USA. C. A. Nordman is with NVE Corp., Eden Prairie, MN 55344 USA antiparallel (AP) and lowest in the parallel (P) orientation. Shu Fan Cheng is with U.S. Naval Research Laboratory, Washington, DC FCs enhance the linear magnetoresistive response near Ht = 0 20375 USA. HC-07 2 G. When the MEMS flaps are in motion, the separation Further details on the field dependence of the noise will be between the FC changes, and the enhancement factor oscillates presented in a subsequent paper. Here, we are interested in between 1.8 to 3.0. The magnetoresistive response and the comparing noise properties of SVs with and without FCs. noise properties of the SVs were closely matched among many SVs both within and between the two sets of samples: sample to sample variations were of order ±10% for magnetoresistance and ±100% for the 1/f noise at Ht = 0. Fig. 2. Transverse magnetic field dependence of the normalized resistance noise power in a GMR spin-valve. Data is for f = 34 Hz. Inset shows an example of random telegraph noise observed at sharp spikes in the noise power. Telegraph noise is attributed to metastability in the free layer. For clarity, Fig. 3 shows noise data taken in a longitudinal Fig. 1. Normalized resistance as a function of magnetic field for GMR spin- field, Hl = 35 G. The longitudinal field highlights the valve sensors with (open symbols) and without (solid) flux concentrators. background 1/f noise by suppressing the occurrence of spikes Insets show the concept of the AC MEMS flux concentrator device: flux concentrator (a), sensor (b), comb drive (c), and silicon springs (d). In the resulting from the telegraph noise. The 1/f noise is largest in upper inset the sensor is between the electrodes labeled by (b). the AP state and decreases monotonically as the free and fixed layers become parallel at positive field values. This trend was In a GMR sensor, the dominant sources of intrinsic low- observed in all SVs, independent of FCs. Moreover, the frequency noise are Johnson (Nyquist) noise and resistance magnitude of the noise is equivalent to within statistical fluctuations . In general, the resistance noise has both variance between the two sets of samples. The curve for the electronic and magnetic contributions [9, 10]. Under constant FC data is compressed along the field axis due to flux current bias, resistance fluctuations in the SV give rise to concentration. The conclusion is that FCs contribute negligible voltage fluctuations having a power spectral density, SV ( f ) , 1/f noise under DC magnetic field conditions. that scales approximately as I 2/f, where I is the dc current bias. In practice, a sensor will be exposed to AC fields that could All reported noise data is for I L 1 mA. introduce additional low-frequency noise by affecting the Fig. 2 shows the dependence of the noise power on Ht in a magnetization dynamics in the SV and/or the FCs; for SV without FCs. Similar behavior is observed in SVs with example, if there is hysteresis on some magnetic field scale. To FCs (not shown). The field dependence of the noise is address this issue, we investigated to what extent the characterized by a smoothly varying background and a series background 1/f noise is affected by a large sinusoidal magnetic of narrow spikes superimposed on this background. The field perturbation. spectrum of the noise is 1/f except at fields corresponding to the occurrence of a spike in noise power where a Lorentzian- like spectrum is observed. Lorentzian spectra are associated with random telegraph noise in the time domain, see Fig. 2 inset. Noise spikes are prominent about Ht = 0 G and are absent for sufficiently high Ht at which the free layer is saturated parallel to Ht. The kinetics of the telegraph noise are highly sensitive to the applied field which explains the rapid onset and decrease in noise power in given octave frequency. The dependence on field indicates that the telegraph noise and Fig. 3. Dependence of the noise power on transverse field with a constant background 1/f noise are, in large part, magnetic in origin [8, longitudinal field, Hl =35 G. The noise has a 1/f-like spectrum. Behavior is 9]. The precise fields at which noise spikes occur varies among similar for SVs with (a) and without (b) flux concentrators. samples and depends somewhat on the magnetic history of the SV. However, the general location of their occurrence in relation to the magnetoresistance curve is highly reproducible. HC-07 3 Fig. 5. AC transverse field dependence of low-f under nominally zero magnetic field biasing conditions for a SV without (a) and with (b) flux concentrators. The detailed dependence varies among different samples. Fig. 4. AC transverse field dependence of low frequency noise under various Symbols represent noise power at different frequencies. dc magnetic field biasing conditions. The noise is independent of Hac for SVs without (a) and with (b) flux concentrators. contributing excess noise. These results establish benchmarks for comparing future noise measurements on oscillating FCs in Fig. 4a shows that the noise power in a bare SV is the AC MEMS device. If moving FCs do not contribute unaffected by field perturbations at 85 Hz and as large Hac L 5 excess noise then the operating frequency can be shifted into a Grms along the x-axis. Fig. 4b shows data for a SV with FCs. regime where sensor noise is dominated by Johnson rather Again, a dependence of the noise on Hac is not evident. The than 1/f noise. This crossover occurs at roughly 10 kHz when biasing field conditions for these data were Ht = 0 G and Hl = these SVs are operated at a 1 mA bias current. 35 G, and Hl = 0 G and Ht of order ±75 G. The latter case corresponds to saturation of the free layer. Varying the REFERENCES perturbation frequency between 60 Hz and 3 kHz had no  J. M. Daughton, “Spin-dependent sensors”, Proc. of the IEEE 91, 681 discernible effect under these field biasing conditions. (2003). The situation is very different if the external biasing fields,  S. A. Wolf, “Spintronics: A spin-based electronics vision for the Ht and Hl, are small (< 5 G). Fig. 5 shows that SVs with and future”, Science 294, 1488 (2001). without FCs do exhibit an enhancement in noise power at  J. M. Daughton, “GMR and SDT sensor applications”, IEEE Trans. certain values of Hac. The symbols on the curves in Fig. 5 Magn. 36, 2773 (2000). denote different octave frequencies, f < fac. All octaves show  N. A. Stutzke, S. E. Russek, and D. P. Pappas, “Low-frequency noise the same general trend indicating that the enhancement in measurements on commerical magnetoresistive magnetic field sensors”, noise is broadband. We note that the detailed dependence on J. Appl. Phys. 97, 10Q107 (2005). Hac and the size of the enhancement varies greatly among  A. S. Edelstein and G. A. Fischer, “Minimizing 1/f noise in magnetic samples. Based on a limited number of SVs, we could not sensors using a micromechanical system flux concentrator”, J. Appl. identify a general trend for the dependence on Hac nor could Phys. 91, 7795 (2002). we distinguish a difference in behavior between SVs with and  A. S. Edelstein, G. A. Fischer, M. Pedersen, E. R. Nowak, S. F. Cheng, without FCs. and C. A. Nordman, “Progress toward a thousand-fold reduction in 1/f The noise properties illustrated in Fig. 5 are likely due to the noise in magnetic sensors using an AC MEMS flux concentrator”, J. occurrence of telegraph noise under the nominally zero field Appl. Phys. (to be published). biasing conditions. Because the telegraph noise is so sensitive  Z. Qian, J. M. Daughton, D. Wang, and M. Tondra, “Magnetic design and fabrication of linear spin-valve sensors”, IEEE Trans. Magn. 39, to field, a plausible explanation is that AC field perturbations 3322 (2003). affect the stability of the magnetic configuration in the SV,  N. Smith, A. M. Zeltser, D. L. Yang, and P. V. Koeppe, “Very high driving it toward and away from the metastabilities that gives sensitivity GMR spin-valve magnetometer”, IEEE Trans. Magn. 33, rise to telegraph noise. 3385 (1997). For sensor applications it is desirable to suppress telegraph  L. Jiang, E. R. Nowak, P. E. Scott, J. Johnson, J. M. Slaughter, J. J. Sun, noise. Better field biasing schemes can stabilize the free layer et al., “Low-frequency magnetic and resistance noise in magnetic tunnel into a single domain state and promote coherent rotation, but junctions”, Phys. Rev. B 69, 054407 (2004). usually at the expense of decreased sensitivity . Even  R. J. M. van de Veerdonk, P. J. L. Belien, K. M. Schep, J. C. S. Kools, without telegraph noise, 1/f noise is substantial, suggesting that M. C. de Nooijer, M. A. M. Gijs, et al., “1/f noise in anisotropic and optimization of the SV should focus on reducing noise rather giant magnetoresistive elements”, J. Appl. Phys. 82, 6152 (1997). than increasing sensitivity. As demonstrated, stationary flux  D. I. Flynn, “A new technique of noise reduction for large aspect concentrators can be used to increase sensitivity without magnetoresistors”, IEEE Trans. Magn. 30, 1263 (1994).
Pages to are hidden for
"Magnetic Field Dependence of the Noise in a Magnetoresistive"Please download to view full document