Transconductance and Mobility Behaviors in UTB SOI MOSFETs with Standard and Thin BOX T. Rudenko (1) , V. Kilchytska (2) , S. Burignat (2) , J.–P. Raskin (2) , F. Andrieu (3) , O. Faynot (3) , A. Nazarov (1) , V.S. Lysenko (1) and D. Flandre (2) (1) Institute of Semiconductor Physics, NAS of Ukraine, Prospect Nauki 45, 03028 Kyiv, Ukraine (2) Université catholique de Louvain, Place du Levant, 3, 1348 Louvain-la-Neuve, Belgium (3) CEA-LETI MINATEC, 17, Rue des Martyrs, 38054 Grenoble, France 1. Abstract In this paper, we analyze the effects of the front and back interfaces on the transport properties in undoped ultra-thin body (UTB) SOI MOSFETs with standard and ultra-thin buried oxides (BOX), using measurements of the transconductance, gate-to-channel capacitance and carrier mobility at various back gate biases. 2. Introduction Fully-depleted (FD) SOI MOSFETs with undoped ultra- thin silicon bodies and metal/high-k gate dielectric stacks are needed for future nano-scale CMOS devices to satisfy the ITRS specifications [1]. The use of ultra- thin Si bodies provides an effective suppression of the short-channel effects (SCEs) without using channel doping [2], [3]. An additional improvement of the SCEs can be achieved by thinning the buried oxide layer [1], [4]. The transport properties of FD UTB SOI, depending on the quality of both SOI film interfaces, have been the subject of extensive studies in the recent years [5], [6]. In this paper, we present further investigations of the transport properties of undoped UTB SOI with standard (145-nm-thick) and ultra-thin (12.5-nm-thick) BOX, using an analysis of the transconductance, gate-to- channel capacitance and carrier mobility in long- channel MOSFETs at different back-gate biases. 3. Experimental Details The devices were fabricated at CEA-LETI using UNIBOND (100) SOI wafers with 145-nm- and 12.5- nm-thick BOX. In the channel region, the Si body was thinned down to about 11 nm. Elevated source-drain structures were employed to reduce parasitic resistance. No channel doping was used. The gate stack consisted of ALD HfO 2 with EOT of 1.75 nm, and TiN gate electrode. The measured devices were n-channel MOSFETs with the channel length L=10 μm and the channel width W=10 μm. The carrier mobility was determined by split-CV, using measurements of the drain current (I d ) at a low drain voltage (V d ) and gate-to- channel capacitance (C gc ) versus the front gate voltage (V gf ) at various back gate voltages (V gb ) [5], [6]. 4. Results and Discussion Fig.1 shows I d (V gf )-characteristics for various V gb measured at V d =50 mV for t BOX= 145 nm. One can see that in the case of UTB SOI device, it is very difficult to decorrelate the front and back channel conductions using the I d (V gf )-characteristics. Indeed, as shown in the inset in Fig. 1, for high positive V gb , only one pronounced peak is observed in the d 2 I d /dV gf 2 -curve corresponding to the formation of the inversion channel at the back interface, while the onset of front channel inversion is not clearly visible. Shown in Fig. 2 are g m (V gf )-curves for various V gb in thick (full symbols) and thin (open symbols) BOX devices. g m (V gf )-curves for positive V gb feature two distinct humps related to the back and front channels. For V gb ≤0, only a single peak originating from the front channel is observed. The fact that the first peak in Fig. 2 related to the back channel is higher than the second peak associated with the front channel, suggests that the electron mobility at the back interface is higher than at the front interface. This is confirmed by g m (V gb )- characteristics for various V gf (Fig. 3). Fig. 4 presents C gc (V gf )-characteristics measured for the thick BOX device at various V gb . The capacitance plateau observed at high positive V gb is due to back inversion channel. This disappears at V gb ≤0, when only the front channel is activated [7]. Integrating the C gc (V gf )-curves yields the inversion carrier density N inv (V gf , V gb ) [5], which is then used in the determination of the effective mobility (μ eff ). Besides, the derivative of the gate-to-channel capacitance (dC gc /dV gf ) provides the values of V gf at a given V gb corresponding to the onset of inversion at the back and front interfaces, which is in essence identical to the second derivative of the drain current but offers much better resolution (inset in Fig. 4) because it is unaffected by the front channel/back channel mobility ratio. The arrow in Fig. 4 indicates V gf at which the maximum transconductance is observed. Fig. 5 shows μ eff obtained by split-CV and plotted as a function of V gf for various V gb . It can be noted that μ eff is nearly twice higher when controlled by the back interface (in a thick BOX device, this occurs at V gf <0.4