Investigating the effects of contacting functional oxide films

Advanced oxides are the current focus of intense research activity, driven by their numerous attractive properties for next generation microelectronics. However, reliable strategies must be developed for the electrical contacting of such films without compromising their functionality. We explore the effect of depositing both noble and oxidising metals onto the ferromagnetic La0.7Sr0.3MnO3 (LSMO) in terms of its structural and magnetic properties. Whilst noble metal overlayers have negligible impact, the metals typically used as adhesion layers, such as Ti, can drive a structural phase transition in the LSMO, producing a Brownmillerite phase and impairing the magnetisation.


Introduction
Functional oxides are promising candidates for next generation microelectronics owing to their wide range of ferroelectric, ferromagnetic and potential emergent interfacial properties such as interfacial superconductivity (within heterostructures) [1]. In addition to room temperature spintronic applications, functional oxides are also highly valuable in the investigation of spin dynamics [2] where the principal focus lies at the interface between ferromagnetic oxides and metals. Critically, all metaloxide systems rely on smooth interfaces, good electrical contact and a lack of interfacial chemical reactions, therefore a complete understanding of the metallisation of functional oxides is essential for metal-oxide hybrid devices to be fully realised.
Here, we present a systematic study of the magnetic and structural effects of depositing metal contacts onto La 0.7 Sr 0.3 MnO 3 (LSMO) thin films using transmission electron microscopy (TEM) and superconducting quantum interference device (SQUID) magnetometry. We find that complete quenching of the magnetisation of >10 nm oxide can occur after the deposition of just ~1 nm of a reducing metal. This degradation is important in the context of forming reliable metal contacts to functional oxides, especially where adhesion layers such as titanium are conventionally used.

Experiment Details
Pulsed laser deposition (PLD) [3] was used for epitaxial growth of LSMO, using (100)-oriented SrTiO 3 (STO) substrates and stoichiometric polycrystalline targets. The PLD system employed a KrF excimer (248 nm) laser and the growth was monitored in-situ by high pressure reflection high energy electron diffraction (RHEED) [4]. Ten nanometre and 20 nm thick LSMO films were deposited using 3 To whom any correspondence should be addressed. Scanning transmission electron microscopy ( of the metallised films and was spectroscopy (EELS) was used for chemical analysis of the interfaces and was Gatan Quantum 965 spectrometer, typically using a dispersion of 0.5 eV/channel. Figure 1 provides an overview of the effect of depositing Au, Pt, Cu, CoFe, Cr and Ta ele contacts on the crystal structure of LSMO films. (HAADF) images of an LSMO film with a gold Au/Pt layer appearing brightest. resulting from the deposition of either Au or Pt neither of these elements is particularly reactive. SQUID measurements ( substantial reduction in the magnetic contrast, the SQUID measurements indicate a and the LSMO crystal structure appears appearing at the Cu/LSMO interface (figures 1e,f) produce a more prominent magnetisation. In each case, the dark band is consistent with the formation of an interfacial oxide, with the prospect of yielding an electrically insulating barrier layer between metal and LSMO. discernible effects for Pt and Au, however a significant interface region forms CoFe, Cr and Ta, all of which are commonly used as electrical contacts or (for CoFe) magnetic contacts. SQUID measurements (f) indicate a substantial reduction in magnetisation following the deposition of Ti and Cr compared to the uncoat magnetisation slightly and Au and Pt have no detrimental effect. The magnetization upon Au deposition is ascribed to sample , a repetition rate of 5 Hz, an O 2 atmosphere of 0.2 mbar and a substrate C. Following deposition of the oxide layer(s), samples were transferred to the chamber under UHV for deposition of the metals. ransmission electron microscopy (STEM) was used for structural characterisation and was performed on a JEOL ARM 200cF instrument. used for chemical analysis of the interfaces and was Gatan Quantum 965 spectrometer, typically using a dispersion of 0.5 eV/channel. an overview of the effect of depositing Au, Pt, Cu, CoFe, Cr and Ta ele contacts on the crystal structure of LSMO films. Figures 1a and b are high angle annular dark field of an LSMO film with a gold and platinum capping layer, respectively, layer appearing brightest. There are no discernible effects on the LSMO crystal structure resulting from the deposition of either Au or Pt (figures 1a,b), which is perhaps unsurprising given that neither of these elements is particularly reactive. SQUID measurements (figure 1 magnetic moment of the LSMO with either a gold o SQUID measurements indicate a slight reduction in magnetic moment for a Cu contact appears more diffuse (figure 1c), with a dark narrow band of intensity interface in HAADF imaging. Similarly, the deposition of both Cr and Ta more prominent dark interfacial band and an almost complete loss of , the dark band is consistent with the formation of an interfacial oxide, with the prospect of yielding an electrically insulating barrier layer between metal and LSMO.

Results
High angle annular dark field (HAADF) image summary of the effect of different metal contacts on LSMO films: (a) Au, (b) Pt, (c) Cu, (d) CoFe, (e) Cr and (f) Ta. discernible effects for Pt and Au, however a significant interface region forms , all of which are commonly used as electrical contacts or (for CoFe) magnetic SQUID measurements (f) indicate a substantial reduction in magnetisation following the deposition of Ti and Cr compared to the uncoated LSMO (red trace), whilst Cu reduces the magnetisation slightly and Au and Pt have no detrimental effect. The apparent magnetization upon Au deposition is ascribed to sample-to-sample variations in LSMO quality.
atmosphere of 0.2 mbar and a substrate C. Following deposition of the oxide layer(s), samples were transferred to the TEM) was used for structural characterisation Electron energy loss used for chemical analysis of the interfaces and was conducted using a an overview of the effect of depositing Au, Pt, Cu, CoFe, Cr and Ta electrical are high angle annular dark field respectively, with the on the LSMO crystal structure , which is perhaps unsurprising given that figure 1g) indicate no gold or platinum cap. In reduction in magnetic moment for a Cu contact a dark narrow band of intensity the deposition of both Cr and Ta and an almost complete loss of , the dark band is consistent with the formation of an interfacial oxide, with the prospect of yielding an electrically insulating barrier layer between metal and LSMO. A small degree of intermixing is apparent at the interface but is ascribed to the data deriving from the projection through a rough interface, in addition to possible beam-broadening effects. Whilst the LSMO is unaffected by the deposition of a gold contact, a dramatic effect is observed following the deposition of Ti and explains the quenched magnetisation indicated in figure 2g. A HAADF image is given in figure 2b where the LSMO now appears bright with respect to the Ti. A clear cell-doubling effect is observed in the LSMO layer, which is consistent with a phase change to a brownmillerite structure [7,8]. The dark planes in figure 2b, highlighted by the red arrows, coincide with the MnO 2 planes of the LSMO, indicating a deficiency in either manganese or oxygen. EELS was subsequently carried out to determine the origin of the phase change in the LSMO and results are summarised in figure 2c. It is immediately clear that the oxygen signal extends some distance into the Ti capping layer, decaying at a slower rate than the Mn at the interface. This indicates that the Ti acts as an oxygen getter and subsequently forms a TiO x phase at the interface, consistent with an analysis of the Ti L 2,3 ELNES spectrum (not shown) [9]. These results can account for the complete loss of magnetisation in the LSMO since the sensitivity of LSMO magnetisation to structural defects and oxygen deficiency is well known [10].
In the next set of experiments, we explore the effect of Ti thickness on the magnetic properties of LSMO. Figure 3a illustrates that complete quenching of the LSMO magnetisation occurs for just 1 nm of Ti. What is interesting is that the crystal structure of the LSMO appears completely intact, as illustrated in figure 3b. Nevertheless, EELS reveals substantial oxygen deficiencies,   indicating that the LSMO is able to support a significant number of oxygen vacancies prior to reconstruction. The data presented in figure 3c show the Ti L 2,3 , O K, Mn L 2,3 and La M 4,5 intensity profiles across the LSMO layer following the deposition of ~1 nm of Ti. It may be observed that there is a significant oxygen signal (red line) in the Ti layer, whilst the LSMO layer is substantially oxygen depleted, suggesting an interfacial reaction and transfer of oxygen from LSMO to Ti, with the oxygen deficiency becoming more pronounced towards the Ti interface.

Discussion and Conclusions
Given that Ti and Cr are often used as adhesion layers for noble metal electrical contacts, these results indicate that great care must be taken to protect oxygen sensitive materials from damage. Whilst an interfacial reaction is commonly exploited to improve the adhesion of ohmic contacts in microelectronic devices, it is clear that in the case of functional oxides such as LSMO, that reaction can have a detrimental effect on functionality. A suitable candidate for use as a protective barrier sandwiched between LSMO and Ti was found to be the electrically conductive perovskite oxide SrRuO 3 . Alternatively, we found that SrTiO 3 could also be used as a suitable mask for subsequent dry etching if an ohmic contact is not required; these results will be described elsewhere. The results presented here provide valuable insight into the effects of metal contacting and ultimately enhance our understanding towards the development of multifunctional oxide devices.