Dipolar relaxation effects in Al/SiO2/Si structures investigated by Transient Capacitance Spectroscopy

2014 Transient Capacitance Spectroscopy (DLTS) in Metal-Oxide-Semiconductor structures is usually analysed assuming that the only contribution to the variations in the semiconductor surface capacitance originates in detrapping of carriers from defects at the Si/SiO2 interface or in the Si or SiO2 bulks. Using original pulse configurations, we have detected, in all samples investigated, capacitance transients which cannot be accounted for by such classical emission mechanisms but are in accordance with a dipolar relaxation mechanism. After each pulse, the relaxation follows the electric field direction with a distributed time constant, almost independently of temperature and applied electric field (from 35 K up to 200 K). We further explain why these dipolar relaxation effects could not be observed easily via the usual capacitance or conductance measurements. The physical origin of these dipolar relaxations is briefly discussed. J. Physique Lett. 45 (1984) L-493 L-499 15 MAI 1984, Classification Physics Abstracts 73.40Q 73.60H Transient Capacitance Spectroscopy consists of detecting the capacitance transients of a semiconductor junction as a function of temperature in a selected time interval [1-2]. In a MetalOxide-Semiconductor (MOS) structure,the junction is at first biased in accumulation (excitation bias Vex) for a time interval Tex generally very short (100 ns-100 ~s) in order to fill the impurity states with majority carriers. The junction is then biased in depletion (relaxation bias Vr) for a time interval T, much longer than Tex. The charge state variation of the electronic traps is detected through the capacitance variation of the junction during the second phase by means of either a boxcar integrator [1-2] or a lock-in amplifier [3]. In a MOS device, these traps are located either Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphyslet:019840045010049300 L-494 JOURNAL DE PHYSIQUE LETTRES Fig. 1. Band diagram and transient behaviour of a MOS structure due to : a) the emission of trapped majority carriers; b) dipolar-like relaxation. in the silicon or silica bulks (Q) and (D respectively in figure la), or at the Si/Si02 interface ((2) in figure la). In all cases, this emission of majority carriers leads to a shrinkage of the depletion layer, i.e., to an increase in the junction capacitance during the transient (see Fig. I a). We have evidenced in MOS structure the opposite behaviour, i.e., a decrease in the capacitance during the transient. From a phenomenological point of view, we are in the presence of dipolar relaxation, the origin of which is still undetermined [4]. Several Al/Si02/Si structures having different technological origins were investigated and all exhibited the same behaviour. In this Letter, we report results for a MOS device described in the following. A four inch, n-type, 100 > oriented silicon wafer is used, having a resistivity of 13 Q. cm. Its backface is overdoped by ion implantation in order to yield a good backside ohmic contact. A 100 nm oxide is thermally grown in wet ambient at 950 ~C for 1 h. The structure undergoes several annealing steps in order to lower the defect concentration in the structure. An aluminum layer is then evaporated onto the top surface and annealed at 450 ~C in forming gas. Gates 1 mm in diameter are defined using standard photolithographic techniques. The device is then mounted in a circulating helium cryostat. The voltage cycles are provided by a pulse generator : the pulse durations (Tex and Tr) and biases (Vex and Vr) are controlled by a computer. The capacitance transients are detected via a home made impedance bridge : a gating circuit protects the detection set-up during high capacitance variations, e.g., while driving the MOS device in accumulation. This is particularly important in the experiments we present, in which drastic changes in capacitance are to be applied over relatively long periods of time (several milliseconds) L-495 DIPOLAR RELAXATION IN MOS STRUCTURES without saturating the detection system, which is not possible with usual commercial set-ups. The resulting time constant of the apparatus is then in the 20 ~s range. The time constant window is defined by a double gate boxcar integrator providing the capacitance transient amplitude C(~) 2013 C(tl) at each temperature. The data are collected by a computer which controls the whole experiment. In a first step, DLTS spectra of the MOS device are recorded for different filling pulse durations LeX" Curve 1 in figure 2 shows a typical DLTS spectrum of interface traps [3, 5]. No variations in the DLTS spectrum are observed when Lex is increased from 1 ~s to 100 ~s. This result is in accordance with an interface state filling time LS given by :