- Email: [email protected]
Faraday effect of polycrystalline bismuth iron garnet thin film prepared by mist chemical vapor deposition method
Journal of Magnetism and Magnetic Materials 422 (2017) 100–104
Contents lists available at ScienceDirect
Journal of Magnetism and Magnetic Materials journal homepage: www.elsevier.com/locate/jmmm
Faraday effect of polycrystalline bismuth iron garnet thin film prepared by mist chemical vapor deposition method Situ Yao, Ryosuke Kamakura, Shunsuke Murai, Koji Fujita, Katsuhisa Tanaka n Department of Material Chemistry, Graduate School of Engineering, Kyoto University, Nishikyo-ku, Kyoto, 615-8510 Japan
art ic l e i nf o
a b s t r a c t
Article history: Received 24 April 2016 Received in revised form 3 August 2016 Accepted 24 August 2016 Available online 24 August 2016
We have synthesized polycrystalline thin film composed of a single phase of metastable bismuth iron garnet, Bi3Fe5O12, on a fused silica substrate, one of the most widely utilized substrates in the solid-state electronics, by using mist chemical vapor deposition (mist CVD) method. The phase purity and stoichiometry are confirmed by X-ray diffraction and Rutherford backscattering spectrometry. The resultant thin film shows a small surface roughness of 3.251 nm. The saturation magnetization at room temperature is 1200 G, and the Faraday rotation angle at 633 nm reaches 5.2 deg/μm. Both the magnetization and the Faraday rotation angles are somewhat higher than those of polycrystalline BIG thin films prepared by other methods. & 2016 Elsevier B.V. All rights reserved.
Keywords: Bismuth iron garnet Thin film Mist chemical vapor deposition Faraday effect
1. Introduction Garnet-type ferrites such as Y3Fe5O12 (YIG) and Gd3Fe5O12 (GIG) are known to show large Faraday effect in a wide wavelength range covering the ultraviolet, visible, and infrared regions. In particular, they are highly transparent as well as possess large Faraday rotation angle in the infrared region, and consequently, those garnet-type ferrites are nowadays practically utilized as nonreciprocal photonics devices such as optical isolators and modulators in the optical telecommunications, which break the time-reversal symmetry of light propagation. It is also known that the replacement of the rare-earth ion by Bi3 þ ion at the eightcoordinated site in the garnet-type structure efficiently enhances the Faraday effect while keeping the transparency to the infrared light rather high [1]. Although the concentration of Bi3 þ which can replace the rare-earth ion in a stable phase of garnet-type ferrite is restricted, Bi3Fe5O12 (bismuth iron garnet, BIG), a garnet-type ferrite with the eight-coordinated sites occupied only by Bi3 þ ions, can be synthesized as a metastable phase, as first demonstrated by Okuda et al. [2] BIG manifests extraordinary large Faraday effect and high magneto-optical figure of merit in the visible and nearinfrared region. This metastable compound can be epitaxially grown onto only a single-crystalline garnet substrate because it forms in a non-equilibrium way. Actually, if a non-garnet material such as silica glass is used as a substrate, BiFeO3 and Bi2Fe4O9 that are thermodynamically stable phases are readily grown. Therefore, n
Corresponding author. E-mail address: [email protected] (K. Tanaka).
http://dx.doi.org/10.1016/j.jmmm.2016.08.077 0304-8853/& 2016 Elsevier B.V. All rights reserved.
it is not easy to practically apply this excellent magneto-optical material to integrated photonics system, where construction of tiny photonics devices on ubiquitous substrates such as silica glass is strongly required. To overcome this problem, buffer layers with garnet-type structure have been introduced for the deposition of BIG thin film on non-garnet-type substrates, and several attempts have been made to prepare polycrystalline BIG thin films onto silica glass substrates by using several deposition techniques including reactive radio-frequency (RRF) sputtering, reactive ion beam (RIB) sputtering, and pulsed laser deposition (PLD) [3–7]. In the previous studies, we have demonstrated that mist chemical vapor deposition (mist CVD) method, one of the emerging techniques for thin film deposition, is an efficient process to achieve high quality garnet-type ferrite thin films [8–10]. BIG thin film epitaxially grown on a single-crystallinegadolinium gallium garnet substrate by means of mist CVD method exhibits magnetooptical functionalities comparable to those of BIG thin films prepared by other deposition techniques. For instance, magneto-optical figure of merit for the mist CVD-derived epitaxial BIG thin film is 1.5 times larger than that of PLD-derived thin film, due to its excellent optical properties which stem from a relatively flat surface and better morphology [9]. In the present work, we report on the successful preparation of polycrystalline thin film composed of a single phase of BIG on a fused silica (amorphous SiO2) substrate by using the mist CVD method. Buffer layers utilized in our study are YIG and yttrium aluminum garnet (YAG, with the chemical formula of Y3Al5O12). We also discuss physical properties including optical transmittance, magnetization, and Faraday effect for the resultant thin film.
S. Yao et al. / Journal of Magnetism and Magnetic Materials 422 (2017) 100–104
2. Experimental procedure
3. Results and discussion Fig. 1(a) shows the XRD patterns of as-deposited (amorphous)
(a) Annealed
Bi3Fe5O12
Intensity (arb. unit)
Y3Fe5O12
Post-annealed
As-deposited 10
20
As-deposited
30
40
50
60
2θ (deg)
(b)
Bi3Fe5O12 Post-annealed Post-annealed
Intensity (arb. unit)
The mist CVD setup utilized in this study is a fine-channel type mist sourced film former (TOUKI Co. Ltd., Type: MSFF-FC). Fused silica substrates were ultrasonically cleaned before deposition of thin films. First, a buffer layer of polycrystalline YIG was deposited on the fused silica substrate by using the mist CVD method. The precursor solution was prepared by dissolving tris(acetylacetonato)iron(III), Fe(C5H7O2)3, and tris(acetylacetonato)yttrium(III) hydrate, Y(C5H7O2)3 nH2O, in acetone. Details of the preparation method for YIG layer were described elsewhere [8]. A buffer layer of polycrystalline YAG was also prepared by the mist CVD method to evaluate adequately magnetic and magneto-optical properties of polycrystalline BIG thin film. The precursor solution was prepared by dissolving tris(acetylacetonato)yttrium(III) hydrate, Y(C5H7O2)3 nH2O and tris(acetylacetonato)aluminum(III), Al(C5H7O2)3, in N,N-dimethylformamide (DMF). It should be noted that aluminum oxide had an extremely low deposition rate, which was about one twelfth of iron oxide and one fifth of yttrium oxide at a substrate temperature of 500 °C. Thus, when the molar ratio of Y(C5H7O2)3/Al(C5H7O2)3 in the precursor solution was equal to 0.125, a single phase of YAG without any impurity phases could be obtained. For the process, the substrate temperature and the postannealing temperature were 500 and 800 °C, respectively. For the preparation of polycrystalline BIG thin film, the precursor solution was prepared by dissolving tris(acetylacetonato) iron(III), Fe(C5H7O2)3 in DMF and mixing it with 2-ethylhexanoic acid solution containing 25 wt% of bismuth(III) 2-ethylhexanoate, Bi(C8H15O2)3. The molar ratio of Bi(C8H15O2)3 to Fe(C5H7O2)3 was kept to be 0.463 and the total concentration was 0.050 mol/L. The precursor solution was ultrasonically atomized by using a 2.4 MHz transducer, and the resultant mist droplets were transferred to a reaction vessel with nitrogen gas at a flow rate of 6 L/min. The substrate temperature was set as 300 °C and single-phase of BIG was grown at a post-annealing temperature of 530 °C for 30 min. All of these process conditions are the same as those described in Ref. [9], in which the conditions proved to be optimal. The crystalline phase was identified by using X-ray diffraction (XRD) with CuKα radiation. The composition of the thin films was determined by Rutherford backscattering spectrometry (RBS) using a 2 MeV He þ ion beam, which was produced by a Pelletrontype accelerator located at the heavy-ion accelerator facility at the Quantum Science and Engineering Center of Kyoto University. The thickness of BIG thin films and buffer layers was evaluated by a high-sensitivity surface profiler and confirmed by the analysis of Rutherford backscattering spectra. The surface morphology of the thin films was observed by atomic force microscopy (AFM). Magnetic field dependence of magnetization was measured at room temperature by using a superconducting quantum interference device magnetometer (SQUID). Optical transmittance measurements were carried out at room temperature in a wavelength range from 500 to 2000 nm. Faraday rotation angle was measured at room temperature by a polarization modulation technique with a Xe lamp as a light source. The thin film sample was placed in a static magnetic field of 15 kOe applied in a direction perpendicular to the surface of the thin film, and the rotation angle was measured in a wavelength range from 350 to 850 nm. The Faraday rotation angles of fused silica substrate with YAG buffer layer were measured before the deposition of BIG thin film and subtracted from those of the sample composed of BIG thin film, YAG buffer layer and the fused silica substrate to obtain the values for the BIG thin film.
101
10
20
Y3Al5O12
30
40
50
60
2θ (deg) Fig. 1. X-ray diffraction patterns of (a) as-deposited and post-annealed thin films grown on YIG buffer layer/fused silica substrates and (b) post-annealed thin film grown on YAG buffer layer/fused silica substrate. The closed circles, triangles, and upside-down triangles denote X-ray diffraction lines assigned to BIG, YIG, and YAG phases, respectively.
and post-annealed (crystallized) thin films. For the as-deposited thin film, all the diffraction lines are ascribable to YIG, suggesting that the YIG buffer layer is thick enough to be clearly detected by XRD; the thicknesses of the YIG layer evaluated by the high-sensitivity surface profiler is 45 nm. On the other hand, the diffraction lines observed for the post-annealed thin film are assigned to BIG in addition to YIG, indicating that the thin film is composed of a single phase of BIG without any impurity phases. The measurement by high-sensitivity surface profiler indicates that the BIG thin film is 220 nm thick. The average grain size of BIG can be evaluated by the Williamson–Hall method:
β cos θ sin θ 1 = 4ξ + , λ λ t where β, λ, ξ, and t indicate the full width at half maximum (FWHM), the wavelength of X-ray (1.54060 Å), the strain, and the grain size, respectively. By using the relation, the average grain size of the polycrystalline BIG thin film prepared in the present study is estimated to be 50.8 nm. The lattice parameter is obtainable by using the relation between the 2θ peak positions and the Nelson–Riley function: 1/2[(cos2θ/ sinθ)þ(cos2θ/θ)] [11] and extrapolating the data to θ ¼ π/2. The lattice parameter thus obtained is a¼ 12.631 Å, in excellent agreement with values reported previously [9,12–14]. XRD pattern of post-annealed thin film prepared on a YAG buffer layer/fused silica substrate is shown in Fig. 1(b). Each of the diffraction lines is ascribable to BIG or YAG, indicating that single phase of BIG is grown also on the YAG layer/fused silica substrate. The average grain size and the lattice parameter of the BIG thin film are evaluated to be 54.3 nm and 12.636 Å, respectively. These values are almost identical to those obtained for the BIG thin film
102
S. Yao et al. / Journal of Magnetism and Magnetic Materials 422 (2017) 100–104
Energy (keV) 1000
200 400 600 800 1000 1200 1400 1600
Experimental Simulated O Si Fe Y Bi
Counts
800 600 400 200 0
Si
Bi
Fe Y
O 100 200 300 400 500 600 700 800 900
Channel Fig. 2. Rutherford backscattering signals of single-phase BIG thin film deposited on YIG buffer layer/fused silica substrate. The experimental and simulated signals are denoted by dots connected by solid line and crosses, respectively. Each of the different curves represents a signal simulated for each of the constituent atoms as indicated in the figure.
on the YIG layer/fused silica substrate. Fig. 2 illustrates the result of RBS for the polycrystalline BIG thin film grown on the YIG buffer layer/fused silica substrate. From the analysis of signals ascribable to the constituent atoms, it is found that the distribution of elements is homogeneous in depth and the chemical composition of the thin film is determined as Bi: Fe ¼0.596:1, which agrees well with the stoichiometric composition of BIG. The thickness of the YIG buffer layer and the polycrystalline BIG thin film evaluated by the high-sensitivity surface profile was used in the analysis. From a point of view of practical applications, BIG thin film should be as smooth as possible. Fig. 3 depicts the AFM images of as-deposited and post-annealed thin films. The standard deviation of surface roughness evaluated from the AFM measurements is 2.723 and 3.251 nm for the as-deposited and post-annealed thin films, respectively. The present thin films have relatively flat surfaces when compared with BIG thin films prepared by other methods [15,16]. However, the value of 3.251 nm is nearly three times larger than the standard deviation of surface roughness for the epitaxial BIG thin film grown on gadolinium gallium garnet (GGG, with the chemical formula of Gd3Ga5O12) substrate under the same preparation conditions [9]. The result mainly stems from the distinct difference between the surface roughness of the single-crystalline GGG substrate and YIG buffer layer; the former has an extremely flat surface. Nonetheless, the surface roughness of the present polycrystalline BIG thin film is below 5 nm, which is small enough for the lithography processing [17]. A high transparency is an important factor to gain high device performance in the applications of magneto-optical materials. The optical transmittance spectra of the YIG buffer layer on fused silica substrate and the polycrystalline BIG thin film deposited on the substrate are given as open diamonds and triangles in Figs. 4 (a) and (b), respectively. The transmittance of the present polycrystalline BIG thin film is as high as the epitaxial BIG thin film reported in our previous study [9]. The interference fringes are clearly observed, confirming that polycrystalline BIG thin film with high transparency and flatness can be obtained by using the mist CVD method. The optical transmittance spectra calculated by the Swanepoel formula [18] are shown by the solid curve in both Figs. 4(a) and (b). In the calculations, the two-layer system composed of YIG buffer layer/fused silica substrate was first treated. Then, the calculations were performed for the polycrystalline BIG thin film deposited on the YIG buffer layer/fused silica substrate by
Fig. 3. Atomic force microscopic images of (a) as-deposited and (b) post-annealed BIG thin films grown on YIG buffer layer/fused silica substrates. The standard deviation of surface roughness is 2.723 and 3.251 nm for the as-deposited and postannealed thin films, respectively.
considering the YIG buffer layer/fused silica substrate as a single layer. The optical parameters, i.e., the refractive index, n, and the extinction coefficient, k, are assumed to be expressed in terms of the following empirical equations:
⎛ B ⎞2 n (λ ) = A + ⎜ ⎟ , ⎝ λ⎠ and
a
k( λ ) =
λ 4π
k( λ ) =
⎡ ⎛ a′ ⎞2 ⎤ ⎛ λ ⎞ ⎜ ⎟ exp ⎢ ⎜ ⎟ – b′⎥ ( for BIG), ⎝ 4π ⎠ ⎢⎣ ⎝ λ ⎠ ⎥⎦
( λ – b)1.8 +
c
( for YIG),
where λ is the wavelength, A, B, a, b, c, a′, and b′ are the empirical parameters, and t is the film thickness. The agreement between
S. Yao et al. / Journal of Magnetism and Magnetic Materials 422 (2017) 100–104
1500
1000
4πMS (G)
80 60 40
Measured Calculated
20
1000
1500
2000
Wavelength (nm)
Transmittance (%)
80 60
Measured Calculated
20 0 500
1000
1500
2000
Wavelength (nm) Fig. 4. Optical transmittance spectra of (a) YIG buffer layer (open diamonds) and (b) polycrystalline BIG thin film (open triangles). Solid curves in (a) and (b) represent spectra calculated by using the Swanepoel formula. Table 1 Parameters used in the analysis of transmittance spectrum for the YIG buffer layer and polycrystalline BIG thin film. B (nm) a (nm1.8) b (nm) c (nm1.8) a′ (nm) b′
2.3
490
1933 1550
-800 -300
-5000
0
0
5000
300
10000
Magnetic Field (Oe) Fig. 5. Room temperature magnetization as a function of magnetic field for polycrystalline BIG thin film deposited on YAG buffer layer/fused silica substrate. The inset shows an enlarged view in the low field range.
100
40
0
-500 -1000
0
YIG 2.19 244 BIG 2.42 310
800
0
-1500 -10000
500
@ 300 K
500
Thickness t (nm)
45 15.6 223
experimental and calculated transmittance spectra is rather good in a wide wavelength range. The parameters A, B, a, b, c, a′, and b′ and the film thickness t obtained from the calculations are listed in Table 1. Our results of n (λ) and k (λ) for YIG and BIG are in agreement with the reported values [7,9,19–21]. The film thicknesses of the YIG buffer layer and the polycrystalline BIG thin film are 45 and 223 nm, respectively. These values agree well with those estimated by the surface profiler. In order to eliminate the effect of magnetization and Faraday rotation angle originating from the YIG buffer layer and to deduce magnetic and magneto-optical properties inherent to the polycrystalline BIG thin film, YAG was used as a buffer layer for measurements of magnetization and Faraday effect. Room temperature magnetization as a function of external magnetic field is shown in Fig. 5. A hysteresis loop is clearly observed (as illustrated in the inset of Fig. 5), indicating that the present BIG thin film is ferrimagnetic at room temperature. The saturation magnetization 4πMS reaches 1200 G at around a magnetic field of 2000 Oe. Although the value of saturation magnetization is 25.9% smaller than that of the epitaxial BIG thin film grown by the mist CVD method
on a single-crystalline GGG substrate, [9] it is almost the same as that of polycrystalline BIG thin film grown by the RIB sputtering method [3]. The fact that the present polycrystalline BIG thin film has lower saturation magnetization than the epitaxial BIG thin film is not very surprising, because in general structural imperfections such as grain boundaries, point defects, and dislocations in a polycrystalline ferromagnetic and ferrimagnetic material hinder the motion of domain walls, and as a result, magnetic domains hardly combine with each other to form a large single domain even under high magnetic fields, and the resultant net magnetization is inevitably reduced when compared to that of a single-crystalline magnetic material. The wavelength dependence of Faraday rotation angle of polycrystalline BIG thin film deposited on YAG buffer layer/fused silica substrate is depicted as open squares in Fig. 6. For BIG, Faraday rotation angle can be calculated by considering the electronic transitions of Fe3 þ ions at octahedral and tetrahedral sites of garnet-type structure in the visible and ultraviolet region. Details of the calculations were given elsewhere [9]. The calculated spectrum is shown as a solid curve in Fig. 6. It should be noted that after multiplying by a factor of 1.4, the experimental data of Faraday rotation spectrum for the polycrystalline BIG thin film are fitted well with the spectrum derived from the calculations, which in turn agrees well with the Faraday rotation spectrum of the epitaxial BIG thin film prepared by the mist CVD method [9]. In other words, Faraday rotation angles of the polycrystalline BIG thin
60 45
θF (deg/μm)
Transmittance (%)
100
A
103
Measured 1.4 Calculated
30 15
0 -15 -30 350
450
550
650
750
850
Wavelength (nm) Fig. 6. Wavelength dependence of Faraday rotation angle for polycrystalline BIG thin film deposited on YAG buffer layer/fused silica substrate. Experimental data multiplied by a factor of 1.4 and calculated values are denoted by open squares and a solid curve, respectively. The calculated Faraday rotation angles were quantitatively obtained by considering the electronic transitions of Fe3 þ ions at both octahedral and tetrahedral sites in BIG with the garnet-type structure.
104
S. Yao et al. / Journal of Magnetism and Magnetic Materials 422 (2017) 100–104
film are 28.6% smaller than the calculated values or the experimental data of epitaxial BIG thin film. The reduction of Faraday rotation angles is roughly in accordance with the decrease in the saturation magnetizations, i.e., 25.9%, because the angle of Faraday rotation is proportional to the magnetization [22]. The Faraday rotation angle at 633 nm, i.e., the wavelength at which the values of Faraday rotation angle were most frequently reported, for the present mist CVD-derived polycrystalline BIG is 5.2 deg/μm. The rotation angle is slightly higher than those of polycrystalline BIG thin films prepared by other methods [3,7]. It should be emphasized that we can use the mist CVD method to fabricate metastable BIG thin film of polycrystalline form, which has rather high magnetization and large Faraday effect, at relatively low growth temperature such as 530 °C and for short growth duration like 30 min. This is an advantageous point when practical application of BIG thin film is considered.
4. Conclusions A polycrystalline thin film composed of a single phase of BIG was successfully prepared on a fused silica substrate by using the mist CVD method. Because metastable BIG is unable to be directly synthesized on non-garnet-type substrates, it is necessary to deposit buffer layers of YIG and YAG on the fused silica substrates before the preparation of BIG thin film. The polycrystalline BIG thin film has transparency as high as the epitaxially grown thin film, although the surface is somewhat rough as confirmed by AFM observation. The saturation magnetization at room temperature is about 1200 G, which is 25.9% lower than that of the epitaxially grown BIG thin film. Accordingly, Faraday rotation angles are smaller by about 28.6% than those of the epitaxial BIG thin film prepared by the mist CVD method. Nonetheless, the Faraday rotation angle at 633 nm for the present thin film is slightly larger than those of polycrystalline BIG thin films prepared by other deposition methods.
Acknowledgments This research was supported by a Grant-in-Aid for Scientific Research (A) (No. 25249090) of the Ministry of Education, Culture, Sports, Science and Technology (MEXT).
References [1] S. Kahl, V. Popov, A.M. Grishin, J. Appl. Phys. 94 (2003) 5688. [2] T. Okuda, N. Koshizuka, K. Hayashi, H. Taniguchi, K. Satoh, H. Yamamoto, IEEE Transl. J. Magn. Jpn. 3 (1988) 483. [3] T. Okuda, T. Katayama, K. Satoh, H. Yamamoto, J. Appl. Phys. 69 (1991) 4580. [4] H. Toraya, T. Okuda, J. Phys. Chem. Solids 56 (1995) 1317. [5] T. Okuda, A. Kudoh, S. Yoshihara, N. Adachi, H. Ohsato, J. Phys. IV Fr. 7 (1997), C1-707. [6] A. Heinrich, T. Körner, J. Simon, W. Mader, M. Knoll, A. Kalytta, S. Horn, B. Stritzker, J. Magn. Soc. Jpn. 32 (2008) 130. [7] T. Körner, A. Heinrich, M. Weckerle, P. Roocks, B. Stritzker, J. Appl. Phys. 103 (2008) 07B337. [8] S. Yao, T. Sato, K. Kaneko, S. Murai, K. Fujita, K. Tanaka, Jpn. J. Appl. Phys. 53 (2014) 05FB17. [9] S. Yao, T. Sato, K. Kaneko, S. Murai, K. Fujita, K. Tanaka, Jpn. J. Appl. Phys. 54 (2015) 063001. [10] S. Murai, T. Sato, S. Yao, R. Kamakura, K. Fujita, K. Tanaka, J. Lumin. 170 (2016) 808. [11] J.B. Nelson, D.P. Riley, Proc. Phys. Soc. Lond. 57 (1945) 160. [12] T. Okuda, T. Katayama, H. Kobayashi, N. Kobayashi, J. Appl. Phys. 67 (1990) 4944. [13] N. Adachi, V.P. Denysenkov, S.I. Khartsev, A.M. Grishin, T. Okuda, J. Appl. Phys. 88 (2000) 2734. [14] M.Y. Chern, J.S. Liaw, Jpn. J. Appl. Phys. 36 (1997) 1049. [15] S. Mino, M. Matsuoka, A. Shibukawa, K. Ono, Jpn. J. Appl. Phys. 29 (1990) L1823. [16] S. Leitenmeier, T. Körner, J. Griesbauer, M. Herbort, A. Heinrich, B. Stritzker, J. Cryst. Growth 310 (2008) 5392. [17] E. Popova, L. Magdenko, H. Niedoba, M. Deb, B. Dagens, B. Berini, M. Vanwolleghem, C. Vilar, F. Gendron, A. Fouchet, J. Scola, Y. Dumont, M. Guyot, N. Keller, J. Appl. Phys. 112 (2012) 093910. [18] R. Swanepoel, J. Phys. E 16 (1983) 1214. [19] D.O. Dzibrou, A.M. Grishin, J. Appl. Phys. 106 (2009) 043901. [20] S.I. Khartsev, A.M. Grishin, J. Appl. Phys. 101 (2007) 053906. [21] M. Deb, E. Popova, A. Fouchet, N. Keller, J. Phys. D Appl. Phys. 45 (2012) 455001. [22] A.S. Moskvin, A.V. Zenkov, E.I. Yuryeva, V.A. Gubanov, Physica B 168 (1991) 187.
Contents lists available at ScienceDirect
Journal of Magnetism and Magnetic Materials journal homepage: www.elsevier.com/locate/jmmm
Faraday effect of polycrystalline bismuth iron garnet thin film prepared by mist chemical vapor deposition method Situ Yao, Ryosuke Kamakura, Shunsuke Murai, Koji Fujita, Katsuhisa Tanaka n Department of Material Chemistry, Graduate School of Engineering, Kyoto University, Nishikyo-ku, Kyoto, 615-8510 Japan
art ic l e i nf o
a b s t r a c t
Article history: Received 24 April 2016 Received in revised form 3 August 2016 Accepted 24 August 2016 Available online 24 August 2016
We have synthesized polycrystalline thin film composed of a single phase of metastable bismuth iron garnet, Bi3Fe5O12, on a fused silica substrate, one of the most widely utilized substrates in the solid-state electronics, by using mist chemical vapor deposition (mist CVD) method. The phase purity and stoichiometry are confirmed by X-ray diffraction and Rutherford backscattering spectrometry. The resultant thin film shows a small surface roughness of 3.251 nm. The saturation magnetization at room temperature is 1200 G, and the Faraday rotation angle at 633 nm reaches 5.2 deg/μm. Both the magnetization and the Faraday rotation angles are somewhat higher than those of polycrystalline BIG thin films prepared by other methods. & 2016 Elsevier B.V. All rights reserved.
Keywords: Bismuth iron garnet Thin film Mist chemical vapor deposition Faraday effect
1. Introduction Garnet-type ferrites such as Y3Fe5O12 (YIG) and Gd3Fe5O12 (GIG) are known to show large Faraday effect in a wide wavelength range covering the ultraviolet, visible, and infrared regions. In particular, they are highly transparent as well as possess large Faraday rotation angle in the infrared region, and consequently, those garnet-type ferrites are nowadays practically utilized as nonreciprocal photonics devices such as optical isolators and modulators in the optical telecommunications, which break the time-reversal symmetry of light propagation. It is also known that the replacement of the rare-earth ion by Bi3 þ ion at the eightcoordinated site in the garnet-type structure efficiently enhances the Faraday effect while keeping the transparency to the infrared light rather high [1]. Although the concentration of Bi3 þ which can replace the rare-earth ion in a stable phase of garnet-type ferrite is restricted, Bi3Fe5O12 (bismuth iron garnet, BIG), a garnet-type ferrite with the eight-coordinated sites occupied only by Bi3 þ ions, can be synthesized as a metastable phase, as first demonstrated by Okuda et al. [2] BIG manifests extraordinary large Faraday effect and high magneto-optical figure of merit in the visible and nearinfrared region. This metastable compound can be epitaxially grown onto only a single-crystalline garnet substrate because it forms in a non-equilibrium way. Actually, if a non-garnet material such as silica glass is used as a substrate, BiFeO3 and Bi2Fe4O9 that are thermodynamically stable phases are readily grown. Therefore, n
Corresponding author. E-mail address: [email protected] (K. Tanaka).
http://dx.doi.org/10.1016/j.jmmm.2016.08.077 0304-8853/& 2016 Elsevier B.V. All rights reserved.
it is not easy to practically apply this excellent magneto-optical material to integrated photonics system, where construction of tiny photonics devices on ubiquitous substrates such as silica glass is strongly required. To overcome this problem, buffer layers with garnet-type structure have been introduced for the deposition of BIG thin film on non-garnet-type substrates, and several attempts have been made to prepare polycrystalline BIG thin films onto silica glass substrates by using several deposition techniques including reactive radio-frequency (RRF) sputtering, reactive ion beam (RIB) sputtering, and pulsed laser deposition (PLD) [3–7]. In the previous studies, we have demonstrated that mist chemical vapor deposition (mist CVD) method, one of the emerging techniques for thin film deposition, is an efficient process to achieve high quality garnet-type ferrite thin films [8–10]. BIG thin film epitaxially grown on a single-crystallinegadolinium gallium garnet substrate by means of mist CVD method exhibits magnetooptical functionalities comparable to those of BIG thin films prepared by other deposition techniques. For instance, magneto-optical figure of merit for the mist CVD-derived epitaxial BIG thin film is 1.5 times larger than that of PLD-derived thin film, due to its excellent optical properties which stem from a relatively flat surface and better morphology [9]. In the present work, we report on the successful preparation of polycrystalline thin film composed of a single phase of BIG on a fused silica (amorphous SiO2) substrate by using the mist CVD method. Buffer layers utilized in our study are YIG and yttrium aluminum garnet (YAG, with the chemical formula of Y3Al5O12). We also discuss physical properties including optical transmittance, magnetization, and Faraday effect for the resultant thin film.
S. Yao et al. / Journal of Magnetism and Magnetic Materials 422 (2017) 100–104
2. Experimental procedure
3. Results and discussion Fig. 1(a) shows the XRD patterns of as-deposited (amorphous)
(a) Annealed
Bi3Fe5O12
Intensity (arb. unit)
Y3Fe5O12
Post-annealed
As-deposited 10
20
As-deposited
30
40
50
60
2θ (deg)
(b)
Bi3Fe5O12 Post-annealed Post-annealed
Intensity (arb. unit)
The mist CVD setup utilized in this study is a fine-channel type mist sourced film former (TOUKI Co. Ltd., Type: MSFF-FC). Fused silica substrates were ultrasonically cleaned before deposition of thin films. First, a buffer layer of polycrystalline YIG was deposited on the fused silica substrate by using the mist CVD method. The precursor solution was prepared by dissolving tris(acetylacetonato)iron(III), Fe(C5H7O2)3, and tris(acetylacetonato)yttrium(III) hydrate, Y(C5H7O2)3 nH2O, in acetone. Details of the preparation method for YIG layer were described elsewhere [8]. A buffer layer of polycrystalline YAG was also prepared by the mist CVD method to evaluate adequately magnetic and magneto-optical properties of polycrystalline BIG thin film. The precursor solution was prepared by dissolving tris(acetylacetonato)yttrium(III) hydrate, Y(C5H7O2)3 nH2O and tris(acetylacetonato)aluminum(III), Al(C5H7O2)3, in N,N-dimethylformamide (DMF). It should be noted that aluminum oxide had an extremely low deposition rate, which was about one twelfth of iron oxide and one fifth of yttrium oxide at a substrate temperature of 500 °C. Thus, when the molar ratio of Y(C5H7O2)3/Al(C5H7O2)3 in the precursor solution was equal to 0.125, a single phase of YAG without any impurity phases could be obtained. For the process, the substrate temperature and the postannealing temperature were 500 and 800 °C, respectively. For the preparation of polycrystalline BIG thin film, the precursor solution was prepared by dissolving tris(acetylacetonato) iron(III), Fe(C5H7O2)3 in DMF and mixing it with 2-ethylhexanoic acid solution containing 25 wt% of bismuth(III) 2-ethylhexanoate, Bi(C8H15O2)3. The molar ratio of Bi(C8H15O2)3 to Fe(C5H7O2)3 was kept to be 0.463 and the total concentration was 0.050 mol/L. The precursor solution was ultrasonically atomized by using a 2.4 MHz transducer, and the resultant mist droplets were transferred to a reaction vessel with nitrogen gas at a flow rate of 6 L/min. The substrate temperature was set as 300 °C and single-phase of BIG was grown at a post-annealing temperature of 530 °C for 30 min. All of these process conditions are the same as those described in Ref. [9], in which the conditions proved to be optimal. The crystalline phase was identified by using X-ray diffraction (XRD) with CuKα radiation. The composition of the thin films was determined by Rutherford backscattering spectrometry (RBS) using a 2 MeV He þ ion beam, which was produced by a Pelletrontype accelerator located at the heavy-ion accelerator facility at the Quantum Science and Engineering Center of Kyoto University. The thickness of BIG thin films and buffer layers was evaluated by a high-sensitivity surface profiler and confirmed by the analysis of Rutherford backscattering spectra. The surface morphology of the thin films was observed by atomic force microscopy (AFM). Magnetic field dependence of magnetization was measured at room temperature by using a superconducting quantum interference device magnetometer (SQUID). Optical transmittance measurements were carried out at room temperature in a wavelength range from 500 to 2000 nm. Faraday rotation angle was measured at room temperature by a polarization modulation technique with a Xe lamp as a light source. The thin film sample was placed in a static magnetic field of 15 kOe applied in a direction perpendicular to the surface of the thin film, and the rotation angle was measured in a wavelength range from 350 to 850 nm. The Faraday rotation angles of fused silica substrate with YAG buffer layer were measured before the deposition of BIG thin film and subtracted from those of the sample composed of BIG thin film, YAG buffer layer and the fused silica substrate to obtain the values for the BIG thin film.
101
10
20
Y3Al5O12
30
40
50
60
2θ (deg) Fig. 1. X-ray diffraction patterns of (a) as-deposited and post-annealed thin films grown on YIG buffer layer/fused silica substrates and (b) post-annealed thin film grown on YAG buffer layer/fused silica substrate. The closed circles, triangles, and upside-down triangles denote X-ray diffraction lines assigned to BIG, YIG, and YAG phases, respectively.
and post-annealed (crystallized) thin films. For the as-deposited thin film, all the diffraction lines are ascribable to YIG, suggesting that the YIG buffer layer is thick enough to be clearly detected by XRD; the thicknesses of the YIG layer evaluated by the high-sensitivity surface profiler is 45 nm. On the other hand, the diffraction lines observed for the post-annealed thin film are assigned to BIG in addition to YIG, indicating that the thin film is composed of a single phase of BIG without any impurity phases. The measurement by high-sensitivity surface profiler indicates that the BIG thin film is 220 nm thick. The average grain size of BIG can be evaluated by the Williamson–Hall method:
β cos θ sin θ 1 = 4ξ + , λ λ t where β, λ, ξ, and t indicate the full width at half maximum (FWHM), the wavelength of X-ray (1.54060 Å), the strain, and the grain size, respectively. By using the relation, the average grain size of the polycrystalline BIG thin film prepared in the present study is estimated to be 50.8 nm. The lattice parameter is obtainable by using the relation between the 2θ peak positions and the Nelson–Riley function: 1/2[(cos2θ/ sinθ)þ(cos2θ/θ)] [11] and extrapolating the data to θ ¼ π/2. The lattice parameter thus obtained is a¼ 12.631 Å, in excellent agreement with values reported previously [9,12–14]. XRD pattern of post-annealed thin film prepared on a YAG buffer layer/fused silica substrate is shown in Fig. 1(b). Each of the diffraction lines is ascribable to BIG or YAG, indicating that single phase of BIG is grown also on the YAG layer/fused silica substrate. The average grain size and the lattice parameter of the BIG thin film are evaluated to be 54.3 nm and 12.636 Å, respectively. These values are almost identical to those obtained for the BIG thin film
102
S. Yao et al. / Journal of Magnetism and Magnetic Materials 422 (2017) 100–104
Energy (keV) 1000
200 400 600 800 1000 1200 1400 1600
Experimental Simulated O Si Fe Y Bi
Counts
800 600 400 200 0
Si
Bi
Fe Y
O 100 200 300 400 500 600 700 800 900
Channel Fig. 2. Rutherford backscattering signals of single-phase BIG thin film deposited on YIG buffer layer/fused silica substrate. The experimental and simulated signals are denoted by dots connected by solid line and crosses, respectively. Each of the different curves represents a signal simulated for each of the constituent atoms as indicated in the figure.
on the YIG layer/fused silica substrate. Fig. 2 illustrates the result of RBS for the polycrystalline BIG thin film grown on the YIG buffer layer/fused silica substrate. From the analysis of signals ascribable to the constituent atoms, it is found that the distribution of elements is homogeneous in depth and the chemical composition of the thin film is determined as Bi: Fe ¼0.596:1, which agrees well with the stoichiometric composition of BIG. The thickness of the YIG buffer layer and the polycrystalline BIG thin film evaluated by the high-sensitivity surface profile was used in the analysis. From a point of view of practical applications, BIG thin film should be as smooth as possible. Fig. 3 depicts the AFM images of as-deposited and post-annealed thin films. The standard deviation of surface roughness evaluated from the AFM measurements is 2.723 and 3.251 nm for the as-deposited and post-annealed thin films, respectively. The present thin films have relatively flat surfaces when compared with BIG thin films prepared by other methods [15,16]. However, the value of 3.251 nm is nearly three times larger than the standard deviation of surface roughness for the epitaxial BIG thin film grown on gadolinium gallium garnet (GGG, with the chemical formula of Gd3Ga5O12) substrate under the same preparation conditions [9]. The result mainly stems from the distinct difference between the surface roughness of the single-crystalline GGG substrate and YIG buffer layer; the former has an extremely flat surface. Nonetheless, the surface roughness of the present polycrystalline BIG thin film is below 5 nm, which is small enough for the lithography processing [17]. A high transparency is an important factor to gain high device performance in the applications of magneto-optical materials. The optical transmittance spectra of the YIG buffer layer on fused silica substrate and the polycrystalline BIG thin film deposited on the substrate are given as open diamonds and triangles in Figs. 4 (a) and (b), respectively. The transmittance of the present polycrystalline BIG thin film is as high as the epitaxial BIG thin film reported in our previous study [9]. The interference fringes are clearly observed, confirming that polycrystalline BIG thin film with high transparency and flatness can be obtained by using the mist CVD method. The optical transmittance spectra calculated by the Swanepoel formula [18] are shown by the solid curve in both Figs. 4(a) and (b). In the calculations, the two-layer system composed of YIG buffer layer/fused silica substrate was first treated. Then, the calculations were performed for the polycrystalline BIG thin film deposited on the YIG buffer layer/fused silica substrate by
Fig. 3. Atomic force microscopic images of (a) as-deposited and (b) post-annealed BIG thin films grown on YIG buffer layer/fused silica substrates. The standard deviation of surface roughness is 2.723 and 3.251 nm for the as-deposited and postannealed thin films, respectively.
considering the YIG buffer layer/fused silica substrate as a single layer. The optical parameters, i.e., the refractive index, n, and the extinction coefficient, k, are assumed to be expressed in terms of the following empirical equations:
⎛ B ⎞2 n (λ ) = A + ⎜ ⎟ , ⎝ λ⎠ and
a
k( λ ) =
λ 4π
k( λ ) =
⎡ ⎛ a′ ⎞2 ⎤ ⎛ λ ⎞ ⎜ ⎟ exp ⎢ ⎜ ⎟ – b′⎥ ( for BIG), ⎝ 4π ⎠ ⎢⎣ ⎝ λ ⎠ ⎥⎦
( λ – b)1.8 +
c
( for YIG),
where λ is the wavelength, A, B, a, b, c, a′, and b′ are the empirical parameters, and t is the film thickness. The agreement between
S. Yao et al. / Journal of Magnetism and Magnetic Materials 422 (2017) 100–104
1500
1000
4πMS (G)
80 60 40
Measured Calculated
20
1000
1500
2000
Wavelength (nm)
Transmittance (%)
80 60
Measured Calculated
20 0 500
1000
1500
2000
Wavelength (nm) Fig. 4. Optical transmittance spectra of (a) YIG buffer layer (open diamonds) and (b) polycrystalline BIG thin film (open triangles). Solid curves in (a) and (b) represent spectra calculated by using the Swanepoel formula. Table 1 Parameters used in the analysis of transmittance spectrum for the YIG buffer layer and polycrystalline BIG thin film. B (nm) a (nm1.8) b (nm) c (nm1.8) a′ (nm) b′
2.3
490
1933 1550
-800 -300
-5000
0
0
5000
300
10000
Magnetic Field (Oe) Fig. 5. Room temperature magnetization as a function of magnetic field for polycrystalline BIG thin film deposited on YAG buffer layer/fused silica substrate. The inset shows an enlarged view in the low field range.
100
40
0
-500 -1000
0
YIG 2.19 244 BIG 2.42 310
800
0
-1500 -10000
500
@ 300 K
500
Thickness t (nm)
45 15.6 223
experimental and calculated transmittance spectra is rather good in a wide wavelength range. The parameters A, B, a, b, c, a′, and b′ and the film thickness t obtained from the calculations are listed in Table 1. Our results of n (λ) and k (λ) for YIG and BIG are in agreement with the reported values [7,9,19–21]. The film thicknesses of the YIG buffer layer and the polycrystalline BIG thin film are 45 and 223 nm, respectively. These values agree well with those estimated by the surface profiler. In order to eliminate the effect of magnetization and Faraday rotation angle originating from the YIG buffer layer and to deduce magnetic and magneto-optical properties inherent to the polycrystalline BIG thin film, YAG was used as a buffer layer for measurements of magnetization and Faraday effect. Room temperature magnetization as a function of external magnetic field is shown in Fig. 5. A hysteresis loop is clearly observed (as illustrated in the inset of Fig. 5), indicating that the present BIG thin film is ferrimagnetic at room temperature. The saturation magnetization 4πMS reaches 1200 G at around a magnetic field of 2000 Oe. Although the value of saturation magnetization is 25.9% smaller than that of the epitaxial BIG thin film grown by the mist CVD method
on a single-crystalline GGG substrate, [9] it is almost the same as that of polycrystalline BIG thin film grown by the RIB sputtering method [3]. The fact that the present polycrystalline BIG thin film has lower saturation magnetization than the epitaxial BIG thin film is not very surprising, because in general structural imperfections such as grain boundaries, point defects, and dislocations in a polycrystalline ferromagnetic and ferrimagnetic material hinder the motion of domain walls, and as a result, magnetic domains hardly combine with each other to form a large single domain even under high magnetic fields, and the resultant net magnetization is inevitably reduced when compared to that of a single-crystalline magnetic material. The wavelength dependence of Faraday rotation angle of polycrystalline BIG thin film deposited on YAG buffer layer/fused silica substrate is depicted as open squares in Fig. 6. For BIG, Faraday rotation angle can be calculated by considering the electronic transitions of Fe3 þ ions at octahedral and tetrahedral sites of garnet-type structure in the visible and ultraviolet region. Details of the calculations were given elsewhere [9]. The calculated spectrum is shown as a solid curve in Fig. 6. It should be noted that after multiplying by a factor of 1.4, the experimental data of Faraday rotation spectrum for the polycrystalline BIG thin film are fitted well with the spectrum derived from the calculations, which in turn agrees well with the Faraday rotation spectrum of the epitaxial BIG thin film prepared by the mist CVD method [9]. In other words, Faraday rotation angles of the polycrystalline BIG thin
60 45
θF (deg/μm)
Transmittance (%)
100
A
103
Measured 1.4 Calculated
30 15
0 -15 -30 350
450
550
650
750
850
Wavelength (nm) Fig. 6. Wavelength dependence of Faraday rotation angle for polycrystalline BIG thin film deposited on YAG buffer layer/fused silica substrate. Experimental data multiplied by a factor of 1.4 and calculated values are denoted by open squares and a solid curve, respectively. The calculated Faraday rotation angles were quantitatively obtained by considering the electronic transitions of Fe3 þ ions at both octahedral and tetrahedral sites in BIG with the garnet-type structure.
104
S. Yao et al. / Journal of Magnetism and Magnetic Materials 422 (2017) 100–104
film are 28.6% smaller than the calculated values or the experimental data of epitaxial BIG thin film. The reduction of Faraday rotation angles is roughly in accordance with the decrease in the saturation magnetizations, i.e., 25.9%, because the angle of Faraday rotation is proportional to the magnetization [22]. The Faraday rotation angle at 633 nm, i.e., the wavelength at which the values of Faraday rotation angle were most frequently reported, for the present mist CVD-derived polycrystalline BIG is 5.2 deg/μm. The rotation angle is slightly higher than those of polycrystalline BIG thin films prepared by other methods [3,7]. It should be emphasized that we can use the mist CVD method to fabricate metastable BIG thin film of polycrystalline form, which has rather high magnetization and large Faraday effect, at relatively low growth temperature such as 530 °C and for short growth duration like 30 min. This is an advantageous point when practical application of BIG thin film is considered.
4. Conclusions A polycrystalline thin film composed of a single phase of BIG was successfully prepared on a fused silica substrate by using the mist CVD method. Because metastable BIG is unable to be directly synthesized on non-garnet-type substrates, it is necessary to deposit buffer layers of YIG and YAG on the fused silica substrates before the preparation of BIG thin film. The polycrystalline BIG thin film has transparency as high as the epitaxially grown thin film, although the surface is somewhat rough as confirmed by AFM observation. The saturation magnetization at room temperature is about 1200 G, which is 25.9% lower than that of the epitaxially grown BIG thin film. Accordingly, Faraday rotation angles are smaller by about 28.6% than those of the epitaxial BIG thin film prepared by the mist CVD method. Nonetheless, the Faraday rotation angle at 633 nm for the present thin film is slightly larger than those of polycrystalline BIG thin films prepared by other deposition methods.
Acknowledgments This research was supported by a Grant-in-Aid for Scientific Research (A) (No. 25249090) of the Ministry of Education, Culture, Sports, Science and Technology (MEXT).
References [1] S. Kahl, V. Popov, A.M. Grishin, J. Appl. Phys. 94 (2003) 5688. [2] T. Okuda, N. Koshizuka, K. Hayashi, H. Taniguchi, K. Satoh, H. Yamamoto, IEEE Transl. J. Magn. Jpn. 3 (1988) 483. [3] T. Okuda, T. Katayama, K. Satoh, H. Yamamoto, J. Appl. Phys. 69 (1991) 4580. [4] H. Toraya, T. Okuda, J. Phys. Chem. Solids 56 (1995) 1317. [5] T. Okuda, A. Kudoh, S. Yoshihara, N. Adachi, H. Ohsato, J. Phys. IV Fr. 7 (1997), C1-707. [6] A. Heinrich, T. Körner, J. Simon, W. Mader, M. Knoll, A. Kalytta, S. Horn, B. Stritzker, J. Magn. Soc. Jpn. 32 (2008) 130. [7] T. Körner, A. Heinrich, M. Weckerle, P. Roocks, B. Stritzker, J. Appl. Phys. 103 (2008) 07B337. [8] S. Yao, T. Sato, K. Kaneko, S. Murai, K. Fujita, K. Tanaka, Jpn. J. Appl. Phys. 53 (2014) 05FB17. [9] S. Yao, T. Sato, K. Kaneko, S. Murai, K. Fujita, K. Tanaka, Jpn. J. Appl. Phys. 54 (2015) 063001. [10] S. Murai, T. Sato, S. Yao, R. Kamakura, K. Fujita, K. Tanaka, J. Lumin. 170 (2016) 808. [11] J.B. Nelson, D.P. Riley, Proc. Phys. Soc. Lond. 57 (1945) 160. [12] T. Okuda, T. Katayama, H. Kobayashi, N. Kobayashi, J. Appl. Phys. 67 (1990) 4944. [13] N. Adachi, V.P. Denysenkov, S.I. Khartsev, A.M. Grishin, T. Okuda, J. Appl. Phys. 88 (2000) 2734. [14] M.Y. Chern, J.S. Liaw, Jpn. J. Appl. Phys. 36 (1997) 1049. [15] S. Mino, M. Matsuoka, A. Shibukawa, K. Ono, Jpn. J. Appl. Phys. 29 (1990) L1823. [16] S. Leitenmeier, T. Körner, J. Griesbauer, M. Herbort, A. Heinrich, B. Stritzker, J. Cryst. Growth 310 (2008) 5392. [17] E. Popova, L. Magdenko, H. Niedoba, M. Deb, B. Dagens, B. Berini, M. Vanwolleghem, C. Vilar, F. Gendron, A. Fouchet, J. Scola, Y. Dumont, M. Guyot, N. Keller, J. Appl. Phys. 112 (2012) 093910. [18] R. Swanepoel, J. Phys. E 16 (1983) 1214. [19] D.O. Dzibrou, A.M. Grishin, J. Appl. Phys. 106 (2009) 043901. [20] S.I. Khartsev, A.M. Grishin, J. Appl. Phys. 101 (2007) 053906. [21] M. Deb, E. Popova, A. Fouchet, N. Keller, J. Phys. D Appl. Phys. 45 (2012) 455001. [22] A.S. Moskvin, A.V. Zenkov, E.I. Yuryeva, V.A. Gubanov, Physica B 168 (1991) 187.