NANOSCALE EFFECTS OF ANNEALING ON THE ELECTRICAL CHARACTERISTICS OF HAFNIUM BASED DEVICES MEASURED IN A VACUUM ENVIRONMENT L. Aguilera 1 , W.Polspoel 2, A.Volodin 3 , C. Van Haesendonck 3 , M.Porti 1 , W.Vandervorst 2 , M.Nafria 1 and X.Aymerich 1 . 1 Dept.Eng.Elect. Universidad Autónoma de Barcelona 08193 Bellaterra Spain 0034 935813531; fax: 0034935812600; e-mail: Lidia.Aguilera@uab.es 2 IMEC Kapeldreef 75, B-3001 Leuven, Belgium 3 Katholieke Universiteit Leuven, Celestijnenlaan 200 D, B-3001 Leuven, Belgium. ABSTRACT In this work, a Conductive Atomic Force Microscope (C-AFM) built in a vacuum environment has been used to characterize the electrical properties of high k samples. In particular, the effect of the annealing on the electrical characteristics of ALD HfO 2 samples has been investigated by this technique. [Keywords: High k materials, Atomic Force Microscopy (AFM)] INTRODUCTION In the recent years, the Conductive Atomic Force Microscopy (C- AFM) has become a very useful technique in the study of the electrical properties of CMOS gate dielectrics at the nanoscale [1-4]. When scanning with the C-AFM the bare dielectric surface, the tip gets covered by contaminants (water and hydrocarbons) and can lose its conductivity and resolution. This effect is even worse with high-k devices [5]. To avoid problems involving anodic oxidation and tip contamination, CAFM measurements should be performed in a vacuum environment to allow accurate characterization of high k materials without loss of conductivity. CAFM measurements, that can analyze areas of the order of ~300nm 2 , provide essential local information. This technique has allowed to study the effect on the high k electrical properties and reliability of an annealing before or after the gate electrode deposition. EXPERIMENTAL AND RESULTS The vacuum system consists of a commercial C-AFM mounted on a chamber which can reach a pressure of 310 -5 mbar. The set-up can measure a maximum current of 10pA and allows us to perform high resolution AFM measurements. As an example, Figure 1 compares the IV mean characteristics measured in a SiO 2 sample in air and in vacuum. A reduction in the onset gate voltage (voltage needed to measure current just above the noise level) is observed between the two IVs (5V in air and 4V in vacuum). Differences between the two environments are more visible when scanning high k samples. Figure 2 shows the current map of a high k sample (provided to test the system) obtained during the 3 rd scan of a sequence of scans on the same area (500nm x 500nm) applying a voltage of 5V. The number of leaky spots (spots with a higher conductivity) observed in the map increases with every new scan but no loss of the tip-sample conductivity is observed. The same type of experiment can not be reproduced in air. In this work, the high k samples consist of ~2nm of ALD HfO 2 blanket films grown on a 1nm SiO 2 interface layer (Rapid Thermal Oxidation) on top of n-type Si. A stack of TiN/Poly capping layer was deposited to prevent oxidation. Some of the samples were annealed under nitrogen ambient at 1030°C after or before the deposition of the capping layer. This protecting layer was removed prior to the AFM measurements. MOS capacitors with the same high k and gate electrode characteristics were used to obtain macroscopic IV curves. Figure 3 shows CAFM IV characteristics measured during a Ramped Voltage Stress (RVS) on the samples with different annealing treatments. Samples annealed before the deposition of the capping layer (sample B) present a lower onset voltage. No differences between no-annealed (sample A) and annealed after deposition of the capping layer (sample C) samples can be observed from the IV curves, in agreement with macroscopic data. From the macroscopic IV characteristics (Fig.4), sample B shows a very large current at low voltages similar to a typical characteristic after Dielectric Breakdown. Only small differences in the IV characteristics of samples A and B can be observed. A sequence of RVS on the same stack location has been chosen to stress the structures (Fig. 5). Note that the first voltage at which the maximum current of 10pA is reached (V I ) is shifted towards lower voltages as the stress proceeds. The shifts in voltage during the sequence of RVS can be related to trapping/detrapping of elementary charges from already present or generated defects in the stack [6]. After the stress, statistics of V I on different stack locations shows a lower value of this parameter in sample A compared to sample C, pointing out an improved reliability after annealing. Sample B presents a lower value of V I before applying the stress (Fig.3). Current maps obtained during a 500nm x 500nm scan (Fig. 6a) while applying a constant voltage show the amount of leaky spots for every sample. Different voltage has been applied to each sample, 4V for samples A and C and 3V to sample B again pointing out the higher conductivity of sample B. The amount of leaky spots in sample B is noticeably higher than in the other samples, showing larger inhomogeneities of the stack electrical properties. Sample C reveals smaller quantity of leaky spots and larger homogeneity than sample A. The presence of large voids (darker areas) observed only on the topography maps of sample B (Fig. 6b) can be related to the higher currents and amount of leaky spots observed on the current maps. CONCLUSION A Vacuum CAFM has been used to characterize the electrical properties of different HfO 2 stacks with nanometer resolution. This set up has allowed for longer tip lifetime and lower onset voltage. In particular, this set up has been used to investigate at the nanoscale the presence of the gate on the dielectric structure during annealing. The samples annealed prior to the deposition of the TiN/Poly capping layers present more leaky spots and reveal void formation on the surface. Samples annealed after the deposition of the gate present less leaky spots and larger electrical homogeneity than the rest of the samples. The authors would like to thank to Spanish MEC (TEC2004- 00798/MIC and TEC2007-61294), the DURSI of the Generalitat de Catalunya (2005SGR-00061) and the European Union APROTHIN project. 657 978-1-4244-2050-6/08/$25.00 ©2008 IEEE IEEE CFP08RPS-CDR 46 th Annual International Reliability Physics Symposium, Phoenix, 2008