4-1 Sample Preparation
In this study, we prepared ZnO films using radio-frequency magnetron sputtering by NUK Professor Hu Lab. We hoped to have good epitaxy and conductivity as bottom electrode so we chose n-type Si(100) to prepare films.
The prepared process was that we pickle substrates first. The pickling solution was HF with diluting. We took out substrates after standing in solution two minutes.
This step was in order to remove SiO2 on silicon substrates. Next, substrates were
individually and
ultrasonically cleaned in acetone,isopropyl alcohol and deionized water for 10 min. We put substrates into chamber and controlled to be equilibrium in
10-8～10-9 Torr. Then, a quartz tube is used to heat substrates to 600℃ for 50 min.
The purpose is to remove moisture and let the surface atoms of substrates rearrange neatly. Then, we poured into two working gases in 2×10-2Torr with argon 30sccm and oxygen 10sccm. Finally, we turned on RF power to 100W. After plasma was stable and pre-sputter for 10 min, we opened the shutter and coated for 2 hours. As shown in Fig. 32.
39
Fig. 32 Simple flowchart of preparing ZnO thin film
Fig. 33 Appearance of ZnO film
Because substrates was two inches in diameter, we cut samples to be 1cm×0.5cm in order to measure conveniently. The sample number and composition as shown in Table 1. Then, we analyzed sample with atomic force microscope, piezoresponse force microscopy and X-ray diffraction system, as shown in Fig. 34.
sample number sample composition
20141227-c ZnO/n-type Si(100)
Table 1 ZnO sample number and composition
40
Fig. 34 Simple flowchart of analyzing ZnO thin film
And part of cut sample, we prepared prmalloy films on ZnO films using radio-frequency magnetron sputtering in our lab. First, we put ZnO film into chamber.
We let vacuum level reach about 2×10-5 Torr by using mechanical pump and turbo pump. Next, we poured into working gas argon and let vacuum level reach about 6×
10-3 Torr. Finally, we turned on RF power to 40W. After plasma was stable and pre-sputter for 10 min, we opened the shutter and coated .
41
Fig. 35 Simple flowchart of preparing Ni80Fe20 film
In this study, we prepared three kinds of thickness with Ni80Fe20 film, 50nm, 10nm and 5nm respectively. Because no thickness detection meter was on sputtering position, we dropped correction fluid on glass, as shown in Fig. 36 . We sputtered for 30 minute in the same process parameters. After putting sample into acetone for one day, we shocked the sample about 10 second with ultrasound oscillator. The correction fluid of glass surface would peel off and surface of glass would expose out.
We could measure a height difference using AFM, as shown in Fig. 37.
42
Fig. 36 correction fluid on glass
Fig. 37 measurement of Ni80Fe20 film thickness
We obtained that the height was about 200 nm between film and substrates.
Because we coated film for 30 minute, we obtained that the ratio of coating was about 6.67nm/min. Then, we controlled the film thickness of permalloy by this ratio.
43
Finally, we analyzed these three samples by using atomic force microscope, X-ray diffraction system and Mangeto Optical Kerr Effect, as shown in Table 2.
sample number sample composition
20141227-Py50 Ne80Fe20(50nm)/ZnO/n-type Si(100) 20141227-Py10 Ne80Fe20(10nm)/ZnO/n-type Si(100)
20141227-Py5 Ne80Fe20(5nm)/ZnO/n-type Si(100)
Table 2 Ni80Fe20 sample number and composition
Fig. 38 Simple flowchart of analyzing ZnO thin film
44
4-2 Piezoresponse force microscopy analysis of ZnO
First, we could observe preferred crystal orientation ZnO(002) in 34.42∘, as shown in Fig. 39 XRD spectra. And Si(400) in 69.16∘ was primary diffraction peak of silicon substrate.
We could observe that film surface was composed of intensive and small grains in the topography (4μm×4μm) and we obtained root mean square roughness was 9.156nm, as shown in Fig. 40. It meant that the surface was not a too rough film.
Fig. 39 No. 20141227 XRD spectra
Fig. 40 No. 20141227 AFM topography (4μm × 4μm)
45
4-2-1 ZnO PFM analysis
Our conductive probe we chose was NSC36C, Our measurement parameters set point was 15nN, amplitude was 3V and frequency was 10 kHz voltage, then SR830 time const was 3ms and sensitivity was 20mV. We scanned in 1μm×1μm. Finally, we could obtain image of topography, amplitude and phase.
Fig. 41 Diagram of PFM
46
Fig. 42 ZnO PFM images of (a) topography (b) amplitude (c)phase.
From Fig. 42(b) and (c), we could observe amplitude image that it had a shadoweave in middle of two bright areas. The shadoweave showed the piezoresponse was almost zero. Then, it was a domain wall in this position. At the same position, it had two area of bright and dark in phase images. It showed that there were two different ferroelectric domain structures. It also proved the shadoweave in amplitude image.
Therefore, we could obtain ferroelectric domain of sample with PFM. It also proved this sample had ferroelectric property, and we would measure hysteresis loop and piezoelectric coefficient.
47
Fig. 43 Measured hysteresis loop in red cross
Fig. 44 (a) phase image (b) amplitude image
Next, we chose one point on ZnO film to measure hysteresis loop, position as shown in Fig. 43.
We obtained successfully hysteresis loop, and its coercivity was about ±2V, as shown in Fig. 44.
48
4-2-2 ZnO piezoelectric coefficient analysis
We used standard sample x-cut quartz (d11=2.3pm/V) to measure piezoelectric coefficient d33. We chose frequency of 10kHz and applied AC voltage Vac = 0V～3V to measure piezoelectric coefficient of ZnO and quartz respectively. We calculated the slope respectively with fitting. We obtained slope of ZnO was a’=2.27946×10-4and slope of quartz was a=1.04536×10-4. Finally, we used equation (3-7) to calculate and we could obtain d33 of ZnO was 5pm/V
d33 = d11 × (a’/ a) ．．．．．．．．．．．．．．．．．．．．．．．．．(3-7)
Fig. 45 Piezoresponse of ZnO and quartz
We used d33=5pm/V, and assumed c-axis lattice length c1=520pm and a-axis lattice length a1=325pm, as shown in Fig. 46. Therefore, hexagonal column volumes were calculated as follows.
49
This volume is 0.143nm3 we obtained. If we applied voltage 10V on this crystal in vertical direction, this crystal would deform about 50pm (ZnO d33=5pm/V). We
50
Finally, lattice deformed ratio was calculated as follows：
ratioc =|c2−c1|
c1 = 9.6%
ratioa= |a2−a1|
a1 = 5.2%
We could obtain deformed rate whether c-axis or the a-axis direction were less than 11%. In accordance with the thin film growth theory, When the crystal deformed more than 11%, it must have destroyed structure behavior.
We took 11% into equation and calculated, we would obtain maximum voltage was 11.44V. So it was within the safe range we applied voltage of 10V. And it had a reversible effect.
51
4-3 MOKE analysis
This Part measurement of mainly three samples (Table 2). The Ni80Fe20 and silicon substrate as electrode we applied voltage. First, we dropped silver plastic on Ni80Fe20 and silicon substrate and each one connected to the copper out (Fig. 49), as shown in Fig. 47. We used agilent 33220A function generator (Fig. 48) to applied DC voltage and generated a electric field to make ZnO lattice deform. In order to cause magnetic properties of Ni80Fe20 be changed. We applied voltage from 0V to 10V and measured mangeto-optical Kerr effect to observe change of coercivity and Kerr intensity.
Fig. 47 Diagram of MOKE
52
Fig. 48 agilent 33220A function generator
Fig. 49 Diagram of silver plastic and copper wires
In order to avoid film generate RRAM behavior by applying voltage too high. It made electric field weaken and the piezoelectric layer cannot cause magnetic property change of Ni80Fe20 layer. So we do a simple measurement of RRAM, as shown in Fig.
50.
53
Fig. 50 Forming Set of ZnO
We could observe the set voltage Vset=14.5V. Next, we do three times switching, as shown in Fig. 51. The set voltage were larger than 10V. So our applied voltage were in save range when measuring MOKE.
Fig. 51 three times switching
54
In order to determine hysteresis curve we measured were not contributed by ZnO films. We measured MOKE with ZnO, as shown in Fig. 52 and Fig. 53. The phenomenon could be seen no hysteresis curve so the effect were contributed mostly by Ni80Fe20 layer.
Fig. 52 Diagram of ZnO P-MOKE
Fig. 53 Diagram of ZnO L-MOKE
55
Fig. 54 ZnO cross-section SEM
In order to obtain thickness of ZnO film and calculate electric field by applying voltage on ZnO film. We measured cross-section SEM of ZnO by Student Zheng in NCKU.
From Fig. 54 we could see, the thickness of ZnO film was about 319nm. In our study, we applied voltage of 1V to 10V. We calculated by formula, E = V / d, then we obtained electric field as follows.
Voltage(V) 1 2 3 4 5 6 7 8 9 10
Field (kV/cm)
31.3 62.7 94.0 125.4 156.7 188.1 219.4 250.8 282.1 313.5
56
4-3-1 5nm thick Ni
80Fe
20MOKE analysis
We could observe that film surface was composed of intensive and small grains in the topography (3μm×3μm) and we obtained root mean square roughness was 8.872nm, as shown in
Fig. 55. Because the roughness was higher than film thickness, the film might have nanoparticles and be poor continuity.
In XRD spectra (Fig. 56), the peak of Ni80Fe20 fcc(111) was not obvious. It was possible that the thickness was too thin.
Fig. 55 Topography of 5nm Ni80Fe20
57
Fig. 56 XRD spectra of 5nm Ni80Fe20
Fig. 57 5nm Ni80Fe20 L-MOKE
58
We made y-axis Kerr intensity be normalized. It was convenience to analyze. We could observe that L-MOKE hysteresis loop didn’t have obvious magnetic coercivity.
It was hard axis. While the applied voltage was gradually increased from 0 to 10V, the hysteresis loop tended to paramagnetism, as shown in Fig. 57.
Fig. 58 5nm Ni80Fe20 Kerr intensity of various voltage
Next, we compared Kerr intensity where the maximum minus the minimum, as shown in Fig. 58. We could observe that the intensity would reduce while the applied voltage increased gradually. There was a turning point in 2V. The proportion of variation is 48.6%.
59
4-3-2 10nm thick Ni
80Fe
20MOKE analysis
We could observe that film surface was composed of intensive and small grains in the topography (3μm×3μm) and we obtained root mean square roughness was 8.605nm, as shown in Fig. 59. It was similar to 5nm Ni80Fe20. In XRD spectra (Fig.
60), the peak of Ni80Fe20 fcc(111) was not obvious. It was also possible that the thickness was too thin and we would obtain conspicuous peak in 50nm Ni80Fe20
Fig. 59 Topography of 10nm Ni80Fe20
Fig. 60 XRD spectra of 10nm Ni80Fe20
60
Fig. 61 10nm Ni80Fe20 L-MOKE
It had obvious ferromagnetism in L-MOKE measurement, as shown in Fig. 61 The magnetic coercivity Hc decreased from 3.5 to 1.3 Oe as the applied voltage increased from 0 to 10 V.
This obvious variation which we thought ZnO generated inverse piezoelectric effect by electric field to cause strain of Ni80Fe20 and change magnetic properties.
61
Fig. 62 10nm Ni80Fe20 Hc in various applied voltage
As shown in Fig. 65, it was a diagram of the magnetic coercivity Hc in different applied voltage. After fitting, we could obtain the Hc decreased gradually in 2.2V and reduced drastically in 5.7V. We could obtain a correspondence between this result and hysteresis loop (Fig. 44). The proportion of variation is 65.6%.
Fig. 63 10nm Ni80Fe20 Kerr intensity of various voltage
Next, we compared Kerr intensity where the maximum minus the minimum, as shown in Fig. 63. We could observe that the intensity would reduce while the applied voltage increased gradually. There was a turning point in 2 to 3V. The proportion of variation is 33.5%.
62
Fig. 64 10nm Ni80Fe20 L-MOKE in the negative voltage
As shown in Fig. 64, we could obtain that the magnetic coercivity Hc decreased as the applied negative voltage increased from 0 to -10V.
This result was same with previous measurement (Fig. 61). There were a symmetry with each other.
63
Fig. 65 10nm Ni80Fe20 Hc in various applied voltage
As shown in Fig. 65, there was a slowdown trend in -2V until it started to decrease drastically in -6V. It decreased from 3.4Oe to 0.7Oe and the proportion of variation is 77.6%.
Fig. 66 10nm Ni80Fe20 Kerr intensity of various voltage
Next, we compared Kerr intensity where the maximum minus the minimum, as shown in Fig. 66. We could observe that the intensity would reduce while the applied voltage increased gradually. The fitting curve was a linear distribution and proportion of variation is 25.4%.
64
Fig. 67 10nm Ni80Fe20 P-MOKE
In P-MOKE measurement, it also had ferromagnetism. But its magnetic coercivity Hc was greater than result of L-MOKE, about 59Oe.
Fig. 68 10nm Ni80Fe20 Hc in various applied voltage
Next, we observed each of magnetic coercivity Hc in various applied voltage, we obtained the magnetic coercivity Hc reduced as applied voltage increase. The magnetic coercivity Hc decrease from 59 to 27 Oe and the proportion of variation is 60.5%.
65
Fig. 69 Diagram of square wave signal
Fig. 70 10nm Ni80Fe20 relative Kerr intensity record
Finally, we measured relative Kerr intensity, we used function generator to apply two different voltage, 0V and 10V respectively. It was a square wave signal and 30 seconds for a cycle. We could obtain that the relative Kerr intensity would be modulated by voltage. It would prove previous result.
66
4-3-3 50nm thick Ni
80Fe
20MOKE analysis
We could observe that film surface was composed of intensive and small grains in the topography (3μm×3μm) and we obtained root mean square roughness was 8.386nm, as shown in Fig. 71. In XRD spectra (Fig. 72), the peak of Ni80Fe20 fcc(111) was obvious than 5nm and 10nm.
Fig. 71 Topography of 50nm Ni80Fe20
Fig. 72 XRD spectra of 10nm Ni80Fe20
67
Fig. 73 50nm Ni80Fe20 L-MOKE
Fig. 74 50nm Ni80Fe20 Hc in various applied voltage
We obtained the magnetic coercivity Hc reduced as applied voltage increase from 37Oe to 23Oe, as shown in Fig. 74. After fitting, we could obtain the Hc decreased gradually in 5.1V and reduced drastically in 8V. It was greater than previous result.
We thought that the Ni80Fe20 thickness was thicker than previous one. If we would drives Hc to change, we had to apply greater voltage. Ｔhe proportion of variation is 45%.
68
Fig. 75 50nm Ni80Fe20 Hc in various applied voltage
Next, we compared Kerr intensity where the maximum minus the minimum, as shown in Fig. 75. We could also observe that the intensity would reduce while the applied voltage increased gradually. The proportion of variation is 17.4%.
Finally, we compare the proportion of variation. It is variation of L-MOKE Kerr intensity in different Ni80Fe20 thickness. It was the maximum proportion minus the minimum proportion. We obtain that proportion decreased as thickness thickening, as the follow table.
5mm-think NiFe 10mm-think NiFe 50mm-think NiFe
ΔKerr intensity(%) 48.6% 33.5% 17.4%
69
Fig. 76 50nm Ni80Fe20 P-MOKE
The P-MOKE measurement result was in hard axis hysteresis loop. We thought the magnetic moment like to stay in horizontal. So it generated a platform in zero applied voltage and it appeared two loops in Fig. 76.
70
Fig. 77 Diagram of square wave signal
Fig. 78 10nm Ni80Fe20 relative Kerr intensity record
Finally, we measured relative Kerr intensity, we used function generator to apply two different voltage, 0V and 10V respectively. It was a square wave signal and 30 seconds for a cycle. We could obtain that the relative Kerr intensity would be modulated by voltage. We compared previous diagram (Fig. 70). It was more like square, because different thickness made it have crystal relaxation behavior in 10 nm thickness Ni80Fe20.
71
Fig. 79 List of magnetostriction constant
Element Weight%-1 Weight%-2 Weight%-3 Average ratio
Ni 2.63 3.03 2.12 50.8
Fe 2.28 2.89 2.34 49.2
Table 3 Wt% of 50nm NiFe EDS
Element Weight%-1 Weight%-2 Weight%-3 Average ratio
Ni 2.21 15.53 18.25 49.2
Fe 2.26 15.67 19.48 50.8
Table 4 Wt% of 10nm NiFe EDS
From Fig. 10[8] and 錯誤! 找不到參照來源。, if Weight% of NiFe had slight change, the magnetostriction constant (λ) would have great differences.
We executed EDS measurement with NiFe films in NPTU Professor Lai Lab, as shown in Table 3 and Table 4. We obtained the Weight% of 50nm thick NiFe was 50.8:49.2 and the Weight% of 10nm thick NiFe was 49.2:50.8. So we could conjecture the NiFe had magnetostriction effect. In previously measurement, the inverse piezoelectric effect of piezoelectric layer ZnO cause lattice strain to produce inverse magnetostriction effect of NiFe. The magnetic properties measurement results would change with the applied voltage.
72