Magneoresitance (MR) measurement - 鎳鐵圖形化微結構中自旋傳輸及磁化翻轉行為之研究

Chapter 3 Experiments

3.3 Magneoresitance (MR) measurement

The MR measurements were carried out by introducing a constant current with sensing voltage at an external magnetic field. The current was usually less than 0.1 mA for local measurement, and 0.15~1 mA for nonlocal measurement. The field dependence of resistance (R-H loop) was recorded by sampling a value of resistance at a stable magnetic field, and then changed the field to the next one, sampling the next value of the resistance.

Electric measurement For DC measurement, the current source was provided by a Keithley Model 2400 sourcemeter, and the voltage meter was Keithley Model 2000 multimeter whose resolution is ±0.1 μV. As the devices whose physical signal less than 5 mΩ such as NLSV devices, it is hard to obtain clear spin signals by only using Keithley Model 2000 multimeter. Instead, the Keithley Model 2182A Nanovoltmeter whose resolution can reach ±1 nV is suitable for low signal measurement. Another signal was converted into digital by the Gauss meter and collected by computers.

The generation of magnetic field The magnetic field was generated by an Tamagawa model TM-YSV5509C-031 electromagnet with provided current -10 ~ +10 A by a Kepco power supply model BOP 72-14MG.

CHAPTER 4 4.1 Quantitative analysis of magnetic reversal in patterned strip wire by magnetic force microscopy

4.1.1 Fabrication and measurement of the Permalloy (Py) strip

We fabricated the patterned Py thin film composed of two strip wires with different widths as shown in Figure 4.1-1. The widths of the narrower wire and the wider one were 187 and 463 nm, respectively, and the lengths were 4500 and 5500 nm, respectively, while the thickness was 30 nm. Measurements were performed by AFM/MFM tapping mode to detect the topographic and magnetic images of the pattern at a simultaneously applied magnetic field of -200 Oe ~ 200 Oe aligned to the longitudinal direction of the strip wires.

Figure 4.1-1 SEM image of the pattern with different widths of strip wire.

4.1.2 AFM/MFM image at a positive field

Figure 4.1-2(a) shows AFM (upper diagram) and MFM (middle diagram) images, and the profile (lower diagram) of the phase of the line scan in the MFM image as well.

This image was detected at a 140 Oe field which was initially at 250 Oe and varied to –250 Oe. The configuration of magnetization in the pattern (as shown in 4.1-2(b)) is briefly considered as macroscopic uniform, and in the same direction of magnetizations in both parts with different widths. Since only the vertical component of the interactively magnetic force between the MFM tip and the sample contributes to the signals of the MFM image, the relatively darker or brighter signals in the MFM image can only appear at the regions in which magnetic flux strays out of the film plane. In our case, these regions are at both ends of the pattern as well as in the section in which the width changes abruptly. Although the strong anisotropy in long strip wire causes a briefly uniform configuration, this section forms a geometric discontinuity (indicated as b in 4.1-2(b)) in the pattern, thus causing a locally different magnetic configuration and resulting in a stray field out of the film plane. It is reasonable to consider that the magnitude (absolute value) of the phase at the wider section (labelled c) is stronger than that at the narrower one (labelled a), because the quantity of the magnetic moment is proportional to the volume of the magnetic material, and the more moment the material contains, the stronger magnetic flux it radiates into an equal volume of space near the material. Furthermore, in this magnetic configuration, since the section b connects the wider and the narrower parts, which are supposed to have negative and positive values of phase, respectively, the totally combined value is negative (as shown in 4.1-2(a), labeled b). Although it can be reasonably understood that peak c is stronger than peak b (in 4.1-2(a), lower diagram), both in experimental results and theoretical prediction, here we still could not estimate or predict the certain relationship between peak a and peak b without

further quantitative investigation of the relationship between the value of the phase and the magnitude of the stray field.

Figure 4.1-2. AFM/MFM image and MFM magnitude of phase. (a) AFM image (upper), MFM image (middle) and a profile (lower) of line scan in MFM image which was observed at a 140 Oe field which was initially at 250 Oe and varied to –250 Oe.

(b) A schematic diagram of the magnetic configuration in accordance with that in (a).

4.1.3 Observation of the phase magnitude for full loop

By measuring the completed magnetic reversal loop of the pattern, we analyzed the values of the phases. As shown in Figure 4.1-3(a), the magnetic reversal behaviors in the three sections can be individually separated. During the process of changing in magnetization from field 200 to 15 Oe, the phase intensity remains roughly unchanged. During this process, the value of the phase in section c is positive, that of b is negative, and that of a is also negative. They retain a relationship of peak c > peak b > peak a. This also means that a certain configuration of magnetization which is in accordance with that in 4.1-2(b) is retained. While in the process of -20 to -200 Oe, the phase values respectively reverse to opposite signs, i.e., c negative, b positive, and a positive (shown in Fig. 3(c), lower diagram), but their absolute values still retain the same relationship mentioned in the previous process. The schematic diagram and MFM image of magnetic configuration during this process are shown in 4.1-3(b) and (c), respectively. This configuration is reasonably the opposite direction of that in the previous process.

Figure 4.1-3 Phase magnitude for full loop and MFM image at a negative field. (a) Magnetic hysteresis loops (upper) for different local sections presented in values of phase; its zoom in (lower). (b) Schematic diagram of the configuration of the magnetization in the magnetic process of –20 to –200 Oe. (c) MFM image (upper) at –200 Oe and profile (lower) of the line scan in the MFM image.

4.1.4 Evaluation of the individual sections

The absolute values of the phases were presented with the longitudinal distance in the pattern, as shown in Figure 4.1-4. As we mentioned above, phase-shift can obviously present only in the regions in which magnetic flux strays out of the film plane. Since sections c, b, and a radiate relatively strong magnetic fluxes, the phase intensities (absolute values) appear more evidently at the centers of these regions, and rapidly decrease at areas away from the centers. The result of analysis of phase intensity shows that the magnitude is about 0.05 in section a, 0.09 in b, and 0.117 in c.

After normalization (choosing 0.117 as 1), we have 0.427 in a, 0.769 in b, and 1 in c.

In addition, the width in section a is about 200 nm, and that in section c about 450 nm, hence the volume ratio of about 0.44. That means the ratio of the quantity of magnetic moment contained in sections a to c might be near 0.44, slightly in accordance with the value 0.427 (the ratio of phase intensity in sections a to c).

Figure 4.1-4 Phase intensity (absolute value) varying with distance in the pattern near the magnetic saturation (20 ~ 200 Oe and -20 ~ -200 Oe).

Furthermore, the switching field (H_C, or coercive force) for each individual local section is also analyzed with the distance in the pattern, as shown in Figure 4.1-5. The diagram shows that Hc in sections a and c are 9 and 8 Oe, which is much larger than that in section b, in which Hc is almost zero. Additionally, at the areas away from the centers of these sections, there is no so-called switching field, since at these areas there is not even an identifiable hysteresis loop observed. Our results are quite consistent with previous reports [105-107]. Due to the shape anisotropy, the value of the Hc should strongly depend on the width of the wire; therefore, we have quantitatively determined the Hc in sections a, b, and c.

Figure 4.1-5 The switching field (or coercive force) Hc for each individual local section.

4.1.5 Lists of the phase magnitudes and MFM images

We have demonstrated that the magnetic behaviors in different local sections of a patterned strip wire can be individually separated and compared with each other. The intensity of the phase-shift in the wider end is stronger than that in the narrower one.

In contrast, the coercive force (which is defined by the reverse in the signs of the values of phase-shifts) in the narrower end (9 Oe) is larger than that in the wider one (8 Oe). This is due to a strong anisotropic effect, and thus the H_C in the neck section (i.e., section b) could become strongly affected by the competition of the head-to-tail magnetic configurations in the two parts of the strip wire. This results in a small Hc in the neck section. Furthermore, with a simple neck shape connection in a strip Py wire, we can easily to change a single domain configuration to a two single domain magnetic configuration. Finally, we list the phase magnitudes and the MFM images with its corresponding AFM images below.

Figure 4.1-6 List of phase magnitudes at different field (a) ~ (d) for +250, +230, +200, and +180 Oe.

Figure 4.1-6 List of phase magnitudes at different field (e) ~ (l) for +160, +140, +120, +80, 40, 0, -10, and -15 Oe.

0 2000 4000 6000 8000 10000 12000 -0.10

-0.05 0.00 0.05 0.10

(q)

-45 Oe

Figure 4.1-6 List of phase magnitudes at different field (m) ~ (q) for -20, -25, -30, -35, and -45 Oe.

Figure 4.1-7 List of AFM/MFM images at different field: +250, +230, +200, and +180 Oe.

Figure 4.1-7 List of AFM/MFM images at different field: +160, +140, +120, and +100Oe.

Figure 4.1-7 List of AFM/MFM images at different field: +80, 40, 20, 0, -10, and -15 Oe.

Figure 4.1-7 List of AFM/MFM images at different field: -20, -25, -30, -35, and -45 Oe.

4.2 Thermal effects on magnetoresistance of magnetic tunnel junction

4.2.1 Structure of MTJ and measurement

The MTJ cell with elliptic shape were studied (Figure 4.2-1). Standard commercial manufacturing processes including photolithography, oxidization, etching, and metallization were used to prepare the samples. SiO2/Si were chosen as the substrate, and multiple metal layers were deposited on substrates by sputtering. For structure 2, the first layer is Ta (20nm) used as a bottom electrode. PtMn (15nm) serves as an antiferromagnetic layer which pins the following tri-layered synthetic antiferromagnet (SAF) structure, CoFe(2nm)/Ru(0.8nm)/CoFe(3nm). During the depositing of PtMn layer, an external magnetic field was also applied to define the direction of magnetization of antiferromagnetic layer. Due to the exchange coupling effect, the two CoFe layers in tri-layered SAF structure can be aligned at anti-parallel with each other, and hence the tri-layered SAF structure can be used as a pinned layer. The insulator layer was made of Al-Ox (~1.2 nm) which was manufactured by pre-depositing an Al thin layer (~9 Å ) followed by introducing O₂ gas for 25~30 seconds to oxidize the Al layer. Then CoFe(1nm)/NiFe(3nm) bilayer is deposited to serve as a free layer. Finally, Ta (60nm) was deposited as the top electrode layer. In order to ensure the function of the PtMn layer, a process of magnetic annealing was applied to the wafer. The condition of magnetic annealing was at 275 ^oC, and 8000 Oe for 5 hours. After thin film depositing, the process of photolithogrophy and dry etching were used to pattern the MTJ structure and testing circuits. Electrical properties including TMR and current-voltage relation (I-V curve) were measured by using a DC source at various temperatures from 25^oC through 140^oC and magnetic field aligned to long axis of ellipse of the MTJ cell.

Figure 4.2-1 Layered structure and lateral size of the MTJ cell.

4.2.2 Temperature-dependent TMR measurement

Typical TMR minor loop with only free layer switching at room temperature is shown in Figure 4.2-2. The applied voltage is 50 mV and MR ratio about 33.25%.

Temperature dependence of TMR loop of MTJ is shown in Figure 4.2-3. The whole loop shifts to low resistance with temperature climbing up. MR ratio behaves in a similar tendency to the resistance (Figure 4.2-4). To more clearly inspect the MR behavior, we extract the resistance for P and AP configuration separately, and compare them with each other (Figure 4.2-5). It indicates that the resistance for AP configuration decreases faster than that for P configuration with temperature, and hence resulting in the decline tendency of MR ratio. Dependence of difference in resistance (Δ R) between the two configurations on temperature is shown in Figure 4.2-6 to be compared with MR ratio. Both roughly exhibit linear relation with temperature from 25 to 140℃.

Coercivities of free layers were also extracted out from MR loops. The coercivity

decreases by 35% from room temperature to 140℃ (Figure 4.2-7). Another important thermal effect is the annealing effect. Compared to the intentionally annealed process of MTJ devices, in which the annealing temperature is usually higher than 200℃ to about 350℃, 130℃~ 140℃ is considerably low temperature and could not crucially influence the material structure in MTJ. It is still our interest to explore the non-intentionally annealing effect on MTJ devices, i.e., the encountered temperature higher than room temperature could more or less affects the performance of the devices during operation or other post-stepped manufacturing. Figure 4.2-8 shows the resistance variation with annealing temperature for P and AP configurations. Not surprisingly, the resistance behaves no obvious changes and the MR ratio varies by

Figure 4.2-2 Typical TMR minor loops measured at 50 mV and room temperature.

Figure 4.2-3 Temperature dependence of TMR loop.

Figure 4.2-4 Temperature dependence of TMR ratio.

Figure 4.2-5 Temperature dependence of resistance for P and AP configuration.

Figure 4.2-6 Comparison of Δ R and MR%.

Figure 4.2-7 Coercivities variation with temperature. Left: original value. Right:

normalized value.

Figure 4.2-8 Annealing effect on resistance.

Figure 4.2-9 Annealing effect on MR ratio.

4.2.3 Bias-dependent TMR measurement

MR loops of the MTJ device was measured with changing bias voltage from 5 mV to 1.2 V. For detailed inspection of each loop, the data are separated into several diagrams (Figures 4.2-10 (a) ~ (j)), and the loops of full voltage range are shown in Figure 4.2-10 (k) to observe the tendency of bias-dependent MR loop. In the same trend with the temperature dependence, the whole MR loop also shifts towards low resistance with increasing bias voltage. The current-voltage measurement (I-V curve) was also carried out, and the dynamic conductance (dI/dV) was derived to further understand the relation between MR ratio and the density of states (DOS) of the two ferromagnetic layers on both sides of the tunnel barrier.

Figure 4.2-11 shows the typical TMR I-V curves for P and AP configuration at room temperature. It is easily to obtain the dynamic conductance, dI/dV, by

differentiating I with V. The dI/dV-V relation in Figure 4.2-12 shows an obvious difference between AP and P configurations. For P configuration, a conductance minima shifts (CMS) appears at around 250 mV which is significantly larger than that for AP configuration. This obvious asymmetry about zero bias ranges between -0.9 V

~ + 0.9 V. On the contrary, the AP configuration exhibits more symmetric about zero bias than P configuration. The CMS for AP is less than 5 mV. Such a spin-dependent asymmetry is reasonably considered as the contribution from the variation of DOS of ferromagnetic layers on both sides of the tunnel barrier (oxide layer).

Figure 4.2-10 Series of bias-dependent MR loops. (a)~(d) for bias voltage of 5 mV ~ 250 mV.

Figure 4.2-10 Series of bias-dependent MR loops. (e) ~ (j) for bias voltage of 300 mV

~ 1.2 V.

Figure 4.2-10 (k) full range.

Figure 4.2-11 I-V curve for the MTJ device measured at 25℃.

Figure 4.2-12 Dynamic conductance dI/dV. Derived from Figure 4.2-11 by differentiating I with V.

By extracting the resistance from Figure 4.2-11, the resistances for AP and P configurations are derived (Figure 4.2-13). Consequently, the asymmetry also exists in the bias dependence of resistance, and MR ratio, too. However, unlike the conductance or resistance dependence on bias, the highest value of MR ratio (Figure 4.2-14) is roughly at zero bias, although the MR-V curve for the full voltage range behaves slightly asymmetry. We then further measured a series of I-V curves with changing temperature (Figure 4.2-15), and the dynamic conductance, resistance, and MR ratio are also derived and shown in Figures 4.2-16, 4.2-17, and 4.2-18, respectively. The results indicate that the shapes of dI/dV-V and R-V curves remain unchanged with increasing temperature, but just shift towards higher (for conductance) and lower (for resistance). For MR%-V curve, however, the shape is sharper at low temperature than that at high temperature in the range from 25 ~ 140℃. In order to confirm the asymmetry of dynamic conductance, the polarities of applied bias were reversed, and the resulted dI/dV-V curves also reversed about zero bias (Figure 4.2-19). Finally, we depict an MR %(T,V) function by using the measured data to give an image describing the MR dependence on temperature and bias voltage (Figure 4.2-20).

Figure 4.2-13 Bias dependence of resistance.

Figure 4.2-14 Bias dependence of Δ R and MR ratio. Derived from Figure 4.2-17.

Figure 4.2-15 I-V curve dependence on temperature. (a) Full range (b) zoom in range from 0.4 V to 0.8 V.

( a)

( b)

Figure 4.2-15 (c) zoom in range from -0.4 V to -0.8 V.

Figure 4.2-16 Temperature dependence of dynamic conductance.

( c)

Figure 4.2-17 Temperature dependence of resistance.

Figure 4.2-18 Temperature dependence of MR ratio.

Figure 4.2-19 Comparison between positive and negative voltage polarities of dynamic conductance.

Figure 4.2-20 MR ratio dependence on temperature and bias voltage.

4.3 Magnetic reversal process and current-induced domain wall motion in ferromagnetic curve- and ring- shape structure

4.3.1 AMR behavior and current-induced domain wall motion of single-layered FM ring

We explored the AMR behavior of the silngle-layered Py ring (Figure 4.3-1) by introducing the current from I+ to I- and the sensing voltage between V+ and V- (lableled in the inset). Here I+ and V+ are in the same electrode. Whith this probing arrangement, the AMR signal would be more senstive than that with only two terminals since the ratio of domain wall area to the region between the two voltage probes is larger, and hence such arrangement is especially useful for domain wall detection. The whole MR loop exhibits a typical AMR behavior of an FM ring as mentioned previously in Chapter 2.

Figure 4.3-1 AMR measurement of a single-layered Py ring with thickness 40 nm and width 225 nm. The measureing current was 10μ A at room temperature. The inset is the SEM image of the sample and the direction of applied field is indicated by the double-head arrow.

The minor loop is shown in Figure 4.3-2. After switching from onion state to vortex state (around -120 Oe), the applied field was swept back to the zero field, and the resistance maintains in the same high level with the vortex state. Onion state is before -120 Oe. After magnetization transformed to vortex state, the votex state exists ranging 0 ~ -400 Oe in the present result. Thus the vortex can be examined as a stable state after its formation. Actually, it is believed that vortex sate can exist in +550 ~ -550 Oe according to the previous data (Figure 4.3-1).

Figure 4.3-2 The AMR minor loop. The arrows indicate the direction of field sweeping.

We then observed the current-induced domain wall motion, by introducing a pulse current with duration time 5 ms in advance, which is considered far from the dynamic scale (less than tens of ns). The magnitude of pulse current is from 100μ A to several mAs. Each pulse current was followed by a resistance record by applied a measuring

current of 10μ A. Figure 4.3-3 shows the results of current-induced measurement dependent on the magnitude of pulse current. The state was initially from onion at a constant applied field, and then switched to vortex by pulse curren. For each constant field, the pulse-current dependence of resistance was recorded. For the present case, the onion-to-vortex switching field is around -120 Oe. The critical current required to switch the state increase with the increasing difference between the held filed and the switching field. The results are shown in Figure 4.3-4. The critical current density J_c is of the order 10⁷ A/cm² which is in agreeemt with the typical results [80-85].

Figure 4.3-3 Field-dependent current-induced domain wall motion measurement.

Figure 4.3-4 Field dependence of critical current density Jc.

4.3.2 MR behavior in FM/N/FM tri-layered-ring spin valve

The MR measurement of a tri-layered spin valve in ring shape with structure Py(10nm)/Cu(9nm)/Py(20nm) is shown Figure 4.3-5. The width is 200 nm and diameter is 3 μm. The applied field transverses to the current contacts (I+ and I-). The full MR loop (inset in Figure 4.3-5(a) ) exhibits a roughly symmetry in macroscopic scale.

The main panel shows a zoom in part (-100 ~ +200 Oe) of half loop from negative field to the positive. The resistance gradually increase from the larger magnitude of field to the smaller one (before zero) without a drastic change. This behavior can be reasonably understood with the AMR effect, since both Py layers are from saturation

在文檔中鎳鐵圖形化微結構中自旋傳輸及磁化翻轉行為之研究 (頁 48-0)