G.
Varvaro
a,
A.
Di Trolio
*a,
A.
Polimeni
b,
A.
Gabbani
c,
F.
Pineider
cd,
C.
de Julián Fernández
e,
G.
Barucca
f,
P.
Mengucci
f,
A.
Amore Bonapasta
a and
A. M.
Testa
a
aCNR-Istituto di Struttura della Materia, Via Salaria Km, 29,300 00015 Monterotondo St., Roma, Italy. E-mail: antonio.ditrolio@cnr.it
bDipartimento di Fisica, Universita’ di Roma “La Sapienza”, P.le A. Moro 2, 00185 Roma, Italy
cDipartimento di Chimica e Chimica Industriale, Università di Pisa, Via Moruzzi 13, 56124 Pisa, Italy
dDipartimento di Chimica “Ugo Schiff”, Università degli Studi di Firenze, Via della Lastruccia 3, 50019 Sesto Fiorentino, FI, Italy
eCNR, IMEM, Parma, Italy
fDipartimento di Scienze e Ingegneria della Materia, dell'Ambiente ed Urbanistica – SIMAU, Università Politecnica delle Marche, Via Brecce Bianche 12, 60131 Ancona, Italy
First published on 21st November 2018
Remarkable improvements in the magneto-optical properties of Zn1−xCoxO thin films were achieved by post-growth hydrogen irradiation at temperatures of 400 °C. Hydrogen incorporation increases the magnetic circular dichroism signal resulting in large values of ellipticity and Faraday rotation at room temperature. The hysteretic behavior of the magneto-optical field dependent loops at different sub-bandgap wavelengths indicates an intrinsic ferromagnetic regime. A giant Faraday rotation of 3000 deg cm−1 was observed at 400 nm. Optical, structural and microstructural characterizations pointed out the Zn substitution by Co, ruling out the presence of noticeable Co-related secondary phases down to the nanoscale. The increased values of saturation magnetization, ellipticity and Faraday rotation have been explained based on an impurity band close to the conduction band minimum, which is induced by Co–VO (O vacancies) complexes. A phenomenological model founded on such an impurity band and electronic states induced by other Co-defect complexes in the ZnO energy gap allows a thoroughly novel, fine interpretation of the ellipticity spectra.
From the materials point of view, exploring improved MO thin film materials with a high figure of merit (i.e. high ratio of Faraday rotation (FR) per unit length to the optical loss per unit length) can pave the way for development of new integrated on-chip nonreciprocal photonic devices, such as optical circulators, isolators and directional amplifiers which are based on the Faraday effect.1–3 Garnets are the current material solution for such MO devices due to their low optical absorption and extremely high FR ranging from 3–5 × 103 deg cm−1 (yttrium iron garnet, YIG) to 8 × 104 deg cm−1 (bismuth iron garnet), at 633 nm.3–5 However, the integration of garnets in optical semiconductor circuits is a challenge due to the incompatibility between the film deposition and device fabrication processes. First of all, garnets have a complex unit cell structure and a much larger lattice parameter compared with semiconductors (YIG: 12.376 Å, Si: 5.43 Å, GaAs: 5.65 Å), which excludes the possibility of hetero-epitaxial growth of garnet films on a conventional semiconductor substrate. Secondly, the thermal expansion coefficient of YIG (10.4 × 10−6 °C−1) is much larger than that of Si (2.3 × 10−6 °C−1), GaAs (5.7 × 10−6 °C−1), InP (4.7 × 10−6 °C−1), and ZnO (5.0 × 10−6 °C−1): cracks can easily generate in the crystallized YIG films on silicon when the thickness exceeds the critical value of about 100 nm.6 Thirdly, crystallization of the garnet phase usually requires an annealing treatment at high temperature (700–800 °C): impurity phases such as YFeO3, Fe2O3 and Bi2O3 can precipitate out during the crystallization process, which may cause optical scattering loss in integrated photonic devices. Hence, such factors make the fabrication of high optical quality magnetic garnet films and also optical isolators on semiconductor substrates very difficult.7
A promising alternative to garnet films is represented by diluted magnetic semiconductors (DMS) which are a platform for device integration. In such context, wide gap ZnO-based DMS obtained by incorporating transition metal elements (e.g. Fe, Co) are suitable for the purpose. In particular, for Co-doped ZnO (Zn1−xCoxO) thin films it has been shown that a large enhancement in the ferromagnetic behavior can be induced by suitable atomic hydrogen treatment, without the formation of metal Co nanophases.8–11
In the context of DMS, the MO effect is useful to investigate the strength and the nature of ferromagnetism as arising from spin polarized carriers magnetically coupled to magnetic ions. However, due to the presence of a finite optical absorption, the polarization of the transmitted light rotates and turns from linear (i.e. superposition of left (LCP) and right circularly polarized components (RCP)) to elliptical. Hence, in general terms the MO effect can be described by a complex rotation whose real and imaginary parts are represented by FR (rotation) and ellipticity (absorption), respectively.
Such quantities provide information about the allowed optical transitions among the electronic states at energy E split by the magnetic field, according to the Zeeman effect. The MO effect was previously observed in Zn1−xCoxO films mainly at low temperatures.11–13 There are also reports of MCD at room temperature with hysteretic behavior at the band edge14 and measurements of MCD after the exposure to Zn metal vapor in a wide sub-bandgap range.15 To our knowledge the Faraday effect was not observed in Zn1−xCoxO yet, although literature results exist in the case of Ti-doped ZnO for photon energies larger than 3 eV.14
In this work, the role of hydrogen in the optical and magneto-optical properties of ferromagnetic Zn1−xCoxO films has been addressed by measuring the MCD spectra of the films irradiated by hydrogen ions. The ferromagnetic nature of the magnetic interaction has been investigated by obtaining room temperature hysteresis loops exploiting the field dependence of the MO effect at different wavelengths in the sub-bandgap region. For the samples irradiated with a high H-dose we also measured the FR, which is fundamental for MO device applications, at room temperature and visible-NIR range. Increased values of saturation magnetization, ellipticity and Faraday rotation are observed, which can be explained by the existence of an impurity band close to the conduction band minimum, induced by Co–VO (O vacancies) complexes, combined with an increase of negative carriers induced by shallow donors,8 that are, in the present case, H-related defects. A phenomenological model founded on such an impurity band and electronic states induced by other Co-defect complexes in the ZnO energy gap allows a thoroughly novel, fine interpretation of the ellipticity spectra.
D0 | LD1 | LD2 | HD1 | HD2 | |
---|---|---|---|---|---|
deposition T (°C) | 500 | 500 | 500 | 500 | 500 |
O2 pressure (mbar) | 2 × 10−5 | 2 × 10−5 | 2 × 10−5 | 2 × 10−5 | 2 × 10−5 |
Thickness (nm) | 420 | 330 | 330 | 400 | 430 |
H-dose (ions per cm2) | 0 | 1.5 × 1018 | 1.5 × 1018 | 4.5 × 1018 | 4.5 × 1018 |
XRD measurements were performed on a Bruker D8 Advance diffractometer in Bragg–Brentano geometry using the Cu-Kα radiation (λ = 1.5406 Å), with diffracted intensities collected in a θ–2θ scan. Scanning electron microscopy (SEM) analysis measurements were performed using a Zeiss SUPRA40 scanning electron microscope equipped with a field emission gun. Samples were observed without any preparation. The thickness of the Zn1−xCoxO films was measured by vertically mounting the samples on a home-made SEM holder. Transmission electron microscopy (TEM) observations were carried out using a Philips CM200 electron microscope equipped with a LaB6 filament and operating at 200 kV. For TEM cross-sectional observations, samples were prepared by the conventional thinning procedure consisting of mechanical polishing using grinding paper, diamond paste, and a dimple grinder. Final thinning was carried out using an ion beam system (Gatan PIPS) using Ar ions at 5 kV. Field-dependent magnetization measurements were performed at 300 K using a Micro Sense Model-10 VSM magnetometer with a max field of 20 kOe. Room temperature optical transmittance spectra were acquired using a Jasco V670 spectrophotometer. The MO properties were investigated by using a home-built apparatus capable of performing measurements in the wavelength range of 350–850 nm in magnetic circular dichroism configuration and 350–850 nm in the Faraday rotation configuration, under magnetic fields up to 1.3 T.16
Fig. 1 XRD patterns of D0, HD1 and LD1 films, treated with zero, high and low H-dose, respectively. Pattern of the blank substrate is also shown. Patterns are vertically shifted for clarity. |
Films thickness, morphology, and microstructure were investigated by SEM and TEM analyses. Fig. 2a and c show typical TEM bright field images of the HD2 and LD2 samples respectively. Both films consist of columnar grains having lateral size ranging from 30 to 80 nm. Selected area electron diffraction (SAED) measurements revealed an oriented growth of the Zn1−xCoxO films on the alumina substrate with the following orientation relationships: [010]ZnO//[−110]Al2O3 and (001)ZnO//(001)Al2O3, for the HD2 sample, Fig. 2b, and [−110]ZnO//[−110]Al2O3 and (001)ZnO//(001)Al2O3 for LD2, Fig. 2d. The two samples have the c axis perpendicular to the substrate in perfect agreement with the diffraction peaks in XRD spectra.
Fig. 2 Bright field TEM images and corresponding SAED patterns taken in the [−110] Al2O3 zone axis of: HD2 (a and b) and LD2 (c and d) samples. |
High resolution TEM observations, performed in different areas of the samples, did not reveal the presence of second phase precipitates confirming the XRD analysis, albeit crystal defects present. In particular, Fig. 3 shows the typical atomic structure of the [010] ZnO film in the HD2 sample. The atomic order is retained moving from one grain to another and the Fourier analysis of the images (FFT) has never revealed the presence of other phases. The inset of Fig. 3 was obtained by removing the noise from the image by using the Gatan Microscopy Suite GMS3 software [http://www.gatan.com/products/tem-analysis/gatan-microscopy-suite-software]. The typical approach consists of the FFT of the image, the application of a mask to remove the noise and an inverse FFT to reobtain the image. The atomic periodicity in the alumina substrate and the film is clearly evidenced in this way, and it is possible to observe the nature of the interface. In detail, the right side of the inset shows that there are zones in which a good match of the lattice cells is present at the interface, but due to the lattice parameters differences, the ZnO cell tends to expand in other zones during the film growth, with the formation of a disordered interface, left side of the inset. The large variation in the image contrast can be attributed to the presence of point defects connected to the Co introduction in the lattice.17,18
Fig. 3 High resolution TEM image of the HD2 sample taken in the [010]ZnO zone axis. The inset was obtained removing the image noise to evidence the sample interface. |
Fig. 4 shows the field-dependent magnetization loops at T = 300 K of the samples irradiated with the low and high H-doses. According to our previous results a high H dose can induce large values of the saturation magnetization.8 Indeed, starting from saturation magnetization (Ms) values lower than 1 emu cm−3 for the non-irradiated sample D0, Ms reaches values of 2–4 emu cm−3 for LD1 and LD2, and 10–12 emu cm−3 for HD1 and HD2. This large enhancement of Ms supports the results of a previous study8 where the ferromagnetic behavior of Zn1−xCoxO was attributed to the formation of Co–VO (oxygen vacancy) complexes and to an increase of the negative carriers induced by shallow donors represented here by H-related defects, like H at the VO site (HO) and H interstitials (HI).
Fig. 5 Transmittance ((a), upper panel) and ellipticity spectra ((b), lower panel) of the Zn1−xCoxO samples (see text). |
It is worth noting that, unlike the transmittance curves which are quite independent of the H doping, ellipticity sharply increases on increasing the hydrogen content, coherently with the magnetization measurements. In fact, starting from values of about 100 deg cm−1 for LD1 and LD2 samples, the ellipticity exceeds 1000 deg cm−1 for the HD1 and HD2 samples in the range of 400–600 nm. This large positive value directly indicates that the electronic states involved in the optical transitions are strongly influenced by the magnetic interaction leading to a large level splitting. Hence a large difference in the absorption of LCP and RCP light is expected.
In order to rule out any paramagnetic character of the ion–carrier interaction we measured the ellipticity as a function of magnetic field for the HD2 sample at different wavelengths, starting from the band edge up to 850 nm, in the near-IR region. As shown in Fig. 6 all the cycles show a clear ferromagnetic behavior. The curves are hysteretic with ellipticity values at saturation ranging from 200 to 1200 deg cm−1 and coercive fields of about 200 Oe, that well agree with the values of the magnetization measurements. In the sub-bandgap region (λ > 375 nm), the saturated ellipticity is observed to increase with λ up to 600 nm, where a value of 1200 deg cm−1 is found, keeping high values except for λ values corresponding to the optical absorption of Co2+ ions.
Fig. 6 Magneto-optic hysteresis cycles of the HD2 films at room temperature, taken for different sub-bandgap wavelengths, starting from λ = 400 nm. |
The hysteretic character of the ellipticity curves confirms the intrinsic nature of the ferromagnetic coupling arising from a spin dependent band structure with the interaction between the carriers and localized spin of the magnetic ions. In the Zn1−xCoxO case the Co ions cannot sustain magnetic interactions beyond their first neighbours,21 whereas Co–VO complexes can promote long-range ferromagnetic order, if additional n-doping is present.21–23 Accordingly, the strength of the magnetic interaction increases by filling the half-empty donor-like band due to Co–VO complexes with carriers arising from donor defects.22,23 Hence the higher MCD values for the films treated with high H-doses well agrees with the proposed model as H-irradiation induces additional carriers and favors the formation of VO defects.8
By considering that the ellipticity rapidly increases near the band edge and keeps high values in the sub-bandgap region, we associate the ellipticity values to photoionization transitions involving electronic states induced by Co-defect complexes in the ZnO energy gap.15,24
The HD1 and HD2 ellipticity spectra of Fig. 5 (bottom panel) show different shapes characterized by different relative maxima located approximately at 430 nm (2.88 eV), 600 nm (2.10 eV), and in the range of 700–800 nm (1.77–1.55 eV). The fine structures of such spectra can be explained by a phenomenological model founded on previous theoretical studies25 and on a recent Kelvin Probe Force Microscopy (KPFM) investigation of Zn1−xCoxO samples26 giving indications about the levels induced by complexes formed by Co atoms and defects like VO's in the ZnO electronic structure. Such levels are schematically reported in Fig. 7. The d levels of a Co atom substituting the Zn atom (3d7) are split by the crystal fields induced by its four O neighbors in the filled e2g(↑)2 and t2g(↑)3 majority spin states located in the valence band, a filled e2g(↓)2 minority spin about 0.6 eV above the top of the valence band (TVB), and a t2g(↓) minority spin empty state at about 3 eV above the conduction band minimum (CBM), t2g(↓)-Co, not reported in Fig. 7. When a VO defect and a Co atom form a Co–VO complex, the t2g(↓)-Co state significantly lowers in energy, about 1.5 eV, since the crystal field is weakened by the removal of an O neighbor. Moreover, when populated by electrons provided by unintentional shallow donors (responsible of the n-type character of as grown ZnO), this t2g(↓) state further descents down to the CBM and partially mix with the CBM itself, t2g(↓)-VO in Fig. 7.23 Finally, a recent KPFM study showed a significant Fermi level lowering of about 0.4 eV in Zn1−xCoxO, indicating the presence of Co-induced deep levels in the ZnO energy gap. Such levels have been related to the formation of Co-defects (Co–D) complexes enhancing the Co under-coordination already realized in the Co–VO complex, thus inducing a further lowering of the t2g(↓) state, now indicated as the t2g(↓)-D state in Fig. 7.26 Then, by disregarding the t2g(↓)-Co state too high in energy, the Co–VO and Co–D complexes may induce two electronic states close to the CBM and in the energy gap, respectively, that, under illumination with light of suitable wavelength (approximately in the range of 3.0–1.5 eV), become empty or partially occupied and therefore available for internal d–d optical transitions starting from the e2g(↓) filled states common to Co and its complexes.
Let us note that the experimental and theoretical results23 locate the t2g(↓)-VO state very close to the CBM, whereas the mentioned KPFM results permit to locate the t2g(↓)-D states only quite below the Fermi energy. They are roughly located at 0.7 eV below the CBM in Fig. 7. Moreover, the Co–D complex represents a family of complexes where the Co atom undergoes different degrees of under-coordination. This implies a dispersion of these states, as well as of the corresponding e2g(↓) filled states, in a certain range of energy as schematized by the dashed lines in the same figure. Now, a phenomenological interpretation of the HD1 and HD2 ellipticity spectra can be attempted based on the electronic states sketched in Fig. 7 and assuming different relative concentrations of the Co–VO and Co-D complexes in the HD1 and HD2 samples. In detail, in the high energy side, the ellipticity signals of the HD1 sample reach a first maximum value M1 at about 430 nm, when the transmission of light becomes high (E < 3 eV) (the MCD measurement is performed in transmission) and the photon energy corresponds to the energy required to bring e2g(↓) electrons to the corresponding t2g(↓)-VO level, about 2.8 eV higher in energy (430 nm). As soon as the energy becomes small for such transitions the ellipticity signals of HD1 decrease and then reach a second, relative maximum (M2) at about 580 nm, 2.1 eV, that may correspond to the transitions e2g(↓) → t2g(↓)-D. Then, a further signal decrease is followed by reaching the third relative maximum (M3) corresponding to a sort of plateau in the range of 700–800 nm, that is, 0.4–0.6 eV below the M2 peak. This range may still correspond to e2g(↓) → t2g(↓)-D once the small dispersion in energy of the e2g(↓) and t2g(↓)-D states, shown in Fig. 7, is taken into account. The different amplitudes of the M1, M2, and M3 peaks can be accounted for by a predominant presence of the Co–VO complexes with respect to the Co-D ones in the HD1 sample. In the case of the HD2, the ellipticity signals show, basically at the same energies of the M1, M2, and M3 maxima, three corresponding, relative maxima M1*, M2*, and M3*. These peaks correspond therefore to the same transitions explaining the HD1 spectrum, while the different MX vs. MX* amplitudes may be accounted for by different relative concentrations of the Co–VO and Co-D complexes in the two samples, the latter complex being the predominant one in the HD2 sample.
In addition, the ellipticity also receives contributions from the differential absorption of RCP and LCP light at 566, 613, and 660 nm, related to the optical Zeeman-split d*–d transitions. We observe a broad positive signal ensuing ferromagnetic nature (as it can be observed from the hysteretic behaviour of the curves of Fig. 6) although the signal decrease around 600 nm indicates competing contributions to the ellipticity from d*–d. From the inspection of Fig. 5b it can be seen that the ellipticity of the LD1 and LD2 samples at around 600 nm takes negative values. This indicates that the main contribution to the ellipticity arises from the optical d*–d transitions as H-dose and hence the carrier density in LD1 and LD2 are consistently lowered with respect to HD1 and HD2, thus reducing the photoionization.
To fully investigate the MO response of the H-irradiated Zn1−xCoxO films we also measured the complex Faraday rotation spectra at saturation (1.3 T) for unirradiated D0, and low (high) H-dose treated sample LD1 (HD2). It turns out that the MO response of D0 is quite negligible, whereas LD1 and HD2 exhibit a sizeable effect, reaching a giant value for HD2. This is a further proof of the effectiveness of the H irradiation in modifying the electronic structure of the Zn1−xCoxO and hence its MO response. Focusing on the HD2 sample both rotation and ellipticity are reported as a function of λ in the inset of Fig. 8. The MO response consists of two main regions with different characteristics: a pure rotatory regime from the band edge up to 550 nm, where a rotation of the polarization plane up to 4000 deg cm−1 is observed at around 368 nm which corresponds to the exciton wavelength of ZnO. For λ > 550 nm the ellipticity reaches its maximum overlapping the rotational value of about 1000 deg cm−1. In the last regime the polarization of the transmitted light is rotated and turned from linear to elliptical.
Finally, the efficiency of H-irradiation in producing large enhancements of ellipticity and FR in Zn1−xCoxO films may be exploited in many technological applications. In particular, it is possible to fabricate MO memories for information storage by combining H-irradiation with electron beam lithography deposition of H-opaque masks, with a fully controlled spatial modulation.27,28 The giant values of the FR may be indeed exploited for the fabrication of Faraday rotators, optical isolators1–3 and circulators taking advantage of the semiconducting nature of the Zn1−xCoxO which favours device integration. For this purpose, efforts are needed to improve the film microstructure which is generally responsible for high optical absorption loss leading to low values of the Faraday rotation over optical absorption coefficient ratio which undermines device performance.
Based on recent theoretical and experimental results giving indications on the electronic states induced by Co-defect complexes in the ZnO energy gap, a phenomenological model is proposed which permits a thoroughly new, fine interpretation of the electronic transitions at the base of the ellipticity spectra.
Finally, the values of the Faraday rotation/ellipticity of the hydrogenated films in the visible range can be tuned by a suitable choice of the H dose and are promising in the perspective of non-reciprocal magneto-optical devices integrated on semiconductors and, generally speaking, for opto-spintronic devices.29
This journal is © The Royal Society of Chemistry 2019 |