Electron Beam Evaporation of Nickel Manganite Thin-film Negative Temperature Coefficient Thermistors

Abstract

Thermistors (i.e., temperature-sensitive resistors) were introduced in the late 1940s [1] as sensors relying on the large temperature dependence of the resistivity of transition-metal oxides. These metal oxides, typically with spinel-type crystal structure, may be classi®ed in terms of chemical composition, temperature range, etc. [2]. In particular, negative temperature coef®cient (NTC) thermistors are devices whose resistance decreases with increasing temperature. They are usually ceramic materials and often possess a very high temperature dependence of electrical resistance. Generally, these materials consist of transition metal (for example, Cu, Co, Ni) spinel manganites and form an important part of the general technology of electroceramics (infrared bolometers, temperature captors, etc.). Although NTC thermistors have been produced commercially for some time, problems with stability, reproducibility and the ease of production still persist. These problems are inherent in powder processing, where control of porosity and stoichiometry are dif®cult and lead to poor run-to-run reproducibility and irregularities in the resulting devices. Such dif®culties should be less severe in dense thin ®lms, where there are no pores and where control of deposition parameters is better, at least in principle. In addition, the uniform incorporation of dopants is generally more straightforward, offering greater exibility. Nickel manganite is a signi®cant compound within the series of solid solutions formed from the solidstate reaction in air between nickel oxide and manganese oxide. Several investigations have reported on the measurements of electrical and magnetic properties [3±6], X-ray photoelectron spectroscopy [7] and microstructural studies [8]. In the literature, however, most NTC studies are for thick-®lm or ceramic pill structures [9], and there has been little reported on thin-®lm thermistors. Fau et al. reported the fabrication of the thin ®lms of nickel manganese oxide using RF reactive sputtering [10], and more recently Kao et al. reported on the preparation of (Mn0:5Co0:6Ni0:1)3O4 NTC thin ®lms by RF reactive sputtering [11]. In this study, the source powders were made by calcining a 1:1 molar-weighted mixture of NiO and Mn2O3 powders at 900 8C for 2 h. After calcination, the source material was placed into specially designed graphite crucibles, 13 mm in diameter, and mounted on a water-cooled movable pedestal in the e-beam evaporation system, comprising a variety of substrates including chemically cleaned glass, alumina with a pre-printed platinum comb electrode structure for electrical measurements and Si wafers for X-ray diffraction (XRD) studies. They were placed on a substrate holder 15 cm above the source, and during the deposition, substrate temperature was monitored using a chromel±alumel thermocouple and a Eurotherm Cal 9900 temperature controller. Deposition was carried out in a conventional oilpumped vacuum system capable of achieving pressures of 10y5 mbar. The source was heated by a beam of energetic electrons emitted from a tungsten ®lament and focused onto the surface of the source by applying a high negative voltage ( 3 kV) between the ®lament housing and the source, which was grounded. In this study, there was no intentional heating of substrates, although substrate temperatures were observed to rise to 80 8C during growth. Structural analysis of the thin ®lms was carried out using a Philips PW 2273 powder X-ray diffractometer, while a Cambridge Instruments S180 scanning electron microscope (SEM) equipped with a Link System energy dispersive X-ray (EDX) facility was used to determine the composition of the ®lms. Finally, the temperature-dependent resistance was measured between room temperature and 600 K in a purpose-built controlled oven using a Keithley 195A digital multimeter, a Eurotherm Cal 9900 temperature controller and a chromel±alumel thermocouple. The thickness of the thin ®lms, measured using a Tencor 200 alpha step pro®leometer, were found to be 1 im. The EDX analysis of the thin ®lms showed that all the deposited thin ®lms contained Mn and Ni (Fig. 1), and no peaks due to other elements were observed in any of the spectra. The ratio of the Ni to the Mn peaks, was found to be different, however, when the source material and thin ®lms were compared: the Ni=Mn EDX peak ratio for the source powder was around 0.15, while in the thin ®lms it was 0.45. Post-growth annealing did not appear to change this ratio in any of the thin ®lms. The XRD results are presented in Fig. 2, which shows XRD traces from an as-deposited ®lm, a layer annealed at 500 8C for 30 min and from a layer