Tunneling magnetoresistance sensors with different coupled free layers

Large differences of magnetic coercivity (HC), exchange coupling field (HE), and tunneling magnetoresistance ratio (TMR) in magnetic tunnel junctions with different coupled free layers are discussed. We demonstrate that the magnetization behavior of the free layer is not only dominated by the interfacial barrier layer but also affected largely by the magnetic or non-magnetic coupled free layers. All these parameters are sensitively controlled by the magnetic nanostructure, which can be tuned also by the magnetic annealing process. The optimized sensors exhibit a large field sensitivity of up to 261%/mT in the region of the reversal synthetic ferrimagnet at the pinned layers.


I. INTRODUCTION
The development of high-sensitivity magnetoresistive sensors involves many practical parameters, such as tunneling magnetoresistance ratio (TMR), coercivity (H C ), exchange-coupling field (H E ), that need to be optimized for the overall performance of the sensors. There are several methods to modify the magnetic properties of the free layer in magnetic tunnel junctions. These methods include using a hard magnet or magnetic field bias, using an orthogonal syntheticferrimagnet coupled layer, and using two-step or three-step orthogonal annealing processes.  The designed shape and geometry of external magnetic concentrators also affect the sensor performance. 30,31 In this study, we discuss large differences in magnetic coercivity (H C ), exchange coupling field (H E ), and tunneling magnetoresistance ratio (TMR) in magnetic tunnel junctions containing magnetic or non-magnetic coupled free layers. By optimizing the TMR ratio, H C , H E , and magnetic flux concentrators, we fabricate an ultra-high-sensitivity magnetic sensor with low dissipative power.

II. EXPERIMENT
The samples are deposited on three-inch thermally oxidized silicon wafers in a magnetron sputtering system with a base pressure of 8 x 10 9 Torr and typical deposition pressures in the range of 2-10 mTorr. A MgO tunneling barrier is obtained by rf sputtering of a MgO target at a power of 100 W. The thickness of the MgO barrier is estimated to be about 2 nm based on the resistance-area (RA) product of about 50 kΩ µm 2 . 32 The investigated MTJ layer structures (with thicknesses in nanometers) are as follows (see Fig. 1 The MTJ junctions were patterned into circles of 30 µm in diameter by photolithography and argon milling. They were then annealed at 350 • C in a magnetic field of 1 T for 1 hour in a vacuum of 2 x 10 8 Torr. The annealing magnetic field sets the pinning direction of the pinned layer, along one axis of the circle. The second annealing process was done at 320 • C in 0.1 T for 10 hours. The direction of the 0.1 T magnetic field is perpendicular (90 o ) to the pinning direction of the pinned layer (in the same plane of pinned layer) to ensure orthogonality between the free layer and the pinned layer. In order to reduce the noise of the sensors, 64 magnetic tunnel junctions were connected into a Wheatstone bridge to form a magnetic field sensor. This magnetic sensor bridge is asymmetric as described in our previous work. 3 As a result of the design, the output signal V out can be amplified by a factor of 2. During the measurement of the sensor performance, the input voltage is 1 V, supplied by batteries to minimize noise. In this study, we used Conetic alloy (Fe 16 Ni 79 Mo 5 ) for magnetic flux concentrators (MFC) that were annealed under a pressure of 10 -7 Torr at 1150 • C for 20 hours with a cooling rate of about 1 • C/min. The saturation induction (B sat ) of this annealed conetic alloy is about 0.8 T, maximum permeability (µ r ) reaches about 45,000, and the coercive field (H c ) is about 1.5 µT. The thickness of the external MFC is 1.5 mm, and the gap between two concentrators is about 0.5 mm. All measurements were taken under ambient conditions.

III. RESULT AND DISCUSSION
As shown in Fig. 2, magnetoresistance and applied magnetic field curves were measured along the direction of the easy axis of the pinned layer, which were set after the first annealing  process at 350 • C under a 1-Tesla magnetic field for 1 hour. Fig. 2(a) shows the magnetoresistance curve of MTJ-S1 with only a 1.3-nm Co 40 Fe 40 B 20 layer as the free layer. It has no hysteresis in a wide magnetic field range, and the TMR of MTJ-S1 is about 39 %. Fig. 2(b) shows the magnetoresistance curve of MTJ-S2 that has 1.3 nm Co 40 Fe 40 B 20 with additional 0.6 nm MgO added to the free layer. The TMR of MTJ-S2 is about 137% which is much higher than that of MTJ-S1. Fig. 2 20 compared with sample MTJ-S1. The TMR of MTJ-S3 is about 120%, slightly less than that of sample MTJ-S2. Both MTJ-S2 and MTJ-S3 have a much smaller saturation field than sample MTJ-S1. Fig. 3 shows the magnetoresistance curves of MTJ-S2 and MTJ-S3, which were measured in a magnetic field varying from -10 mT to +10 mT in the direction of the easy axis of the pinned layer. The coercivity (H C ) of MTJ-S2 is about 1.7 mT, and the exchange coupling field (H E ) is 5.6 mT. By comparison, MTJ-S3 has a much smaller coercivity H C (0 -1.0 mT) and a much smaller exchange coupling field (H E = 1.9 mT) than MTJ-S2. In Table I, parameters related to the overall performance of the sensors are listed, such as tunneling magnetoresistance ratio (TMR), coercivity (H C ), and exchange coupling field (H E ), where TMR is defined as 100% x (R max -R min )/R min . In general, the H C and H E are evaluated from the middle between R max and R min . For asymmetric resistance as a function of applied magnetic field (such as the MTJ-S3 curve in Fig. 3(b)), we need to evaluate H C and H E from several resistance positions (other than the middle between R max and R min ). It is clearly shown that the overall magnetic parameters of MTJ-S3 are better than those of all other MTJ samples.
It has been shown previously 30,31 that the magnetic flux concentrators (MFCs) with t-shape, tanga-shape, triangle-shape, and cone-shape are better than concentrators of bar-shape. Our concentrators were designed by combining the t-shape and the triangle-shape. From our previous study, 33 it was also found that the decrease of gap between the two concentrators results in the increase of magnetic flux density. The gap between MFCs in our design, 0.5 mm, could be achieved with a notch added at the tip of each MFC. The magnetic flux density would be amplified about 216 times. Fig. 4 shows the output voltage as a function of the applied magnetic field for the MTJ-S3 sensor device without MFC (a) and with MFC (b). The input voltage of the sensor is 1 V  supplied by battery. After the integration of MFC, there is no obvious change in the magnetic reversal behavior and the output voltage of the sensor, but the sensitivity of the magnetic sensor is increased by a factor of 216. Fig. 4(b) shows the output voltage of the sensor as a function of the applied magnetic field in the range of ±50 µT. If we use the slope in Fig. 4(b) to calculate the sensitivity, as done by many other groups, the sensitivity of the sensor is about 10,374 %/mT. 6,8,12 As shown in Fig. 5, this sensor can operate in a very low bias magnetic field (only a few tenths of mT) that corresponds to the reversal region of the pinned layer. In this field region, the magnetization of the free layer is fully saturated (the magnetic noise from the free layer will be reduced). Thus, the sensor is expected to have a smaller magnetic noise. Fig. 5(b) shows a good linear response of the sensor's output voltage as a function of the applied magnetic field in the range of 0.2 mT to 0.23 mT.

IV. CONCLUSIONS
In summary, we have investigated three designs of magnetic tunnel junctions with different magnetic free layer structures. The added layer of 0.6 nm MgO in MTJ-S2 above the 1.3 nm Co 40 Fe 40 B 20 free layer improves the crystalline quality, so that the MR loop exhibits a high TMR. The addition of 0.6 nm Ru and 1.3 nm Co 40 Fe 40 B 20 above the 1.3 nm Co 40 Fe 40 B 20 free layer in MTJ-S3 introduces a ferromagnetic-ferromagnetic exchange coupling with the free layer that improves the overall qualities of MTJ. It has small values of H C and H E with a TMR greater than 100%. We also show a new possibility of applying the magnetic sensor in the magnetic field range in which the pinned layer is reversed. The reversal behavior of the Co 40 Fe 40 B 20 in the pinned layer could be a new sensing mechanism for high sensitivity and low noise sensors.