Extracting the relative dielectric constant for “high-κ layers” from CV measurements – Errors and error propagation

https://doi.org/10.1016/j.microrel.2007.01.006Get rights and content

Abstract

The paper pursues an investigation of the errors associated with the extraction of the dielectric constant (i.e., κ value) from capacitance–voltage measurements on metal oxide semiconductor capacitors. The existence of a transition layer between the high-κ dielectric and the silicon substrate is a factor that affects – in general – the assessment of the electrical data, as well as the extraction of κ. A methodology which accounts for this transition layer and the errors related to other parameters involved in the κ value extraction is presented; moreover, we apply this methodology to experimental CV results on HfO2/SiOx/Si structures produced in different conditions.

Introduction

Recent introduction of high-κ based dielectrics into production lines has demonstrated promising success [1]. Concerns about reliability issues [2], as well as the challenges put down by the ITRS in terms of screening of new dielectrics for future nodes are hot topics on the agenda of research groups worldwide [3], [4]. The process of testing new materials for replacing the SiO2 as a gate dielectric is very complex, involving not only the synthesis and characterization of thin films, but also a continuous adjustment and development of the investigation methodologies so that the results can be quantified and compared in a rigorous manner. The present paper aims to discuss how the relative dielectric constant of high-κ dielectric layers can be extracted from CV measurements, while quantifying the errors associated, for a given extraction methodology.

Section snippets

Extracting the dielectric constant from the CV measurements

One of the simplest and most appropriate way to extract the dielectric constant (κ) is from a CV plot. Before proceeding to this extraction, the raw data must be adjusted for a variety of frequency dependent parasitic phenomena such as series resistance, leakage current and lossy interfacial capacitance, and, ideally, quantum mechanical effects. Even after all these corrections are made, certain uncertainties are still present; they are related to the various extraction methods used, as well as

Experimental

We will further consider a practical example for emphasising the importance of considering the errors inherent in analysing CV data. A set of four HfO2 samples – produced by ALD, on Si substrates – were analysed by quasi-static CV; group A was deposited on Si covered with a RTO SiO2 layer. For group B, the transition layer was a SiO2 – chemical oxide. The thickness of the TL was measured by XPS; the thickness of the HfO2 layers was measured by SE [9], while the Hf content by RBS.

Electrical

Results and discussion

In a first instance, we investigated the sensitivity of εrHfO2 values to a variation in the dielectric constant of the TL; as can be seen from Fig. 1, the spread of κ-values is relatively large (10.6–19) for a variation of the εTL between 4.3 and 3.0. It can be noticed that when increasing the [Hf] concentration, the spread of results is reduced significantly from ≈7 to ≈1. Furthermore, it can be seen that for about the same thickness of TL and nearly the same concentration of Hf, the samples

Conclusions

Extraction of κ values of the dielectric layers used in MOS capacitors is a procedure that can be affected by inherent errors related to the measurement technique, as well as by errors related to other parameters (dielectric constants, thickness, area) involved in the extraction process. We demonstrate how sensitive the final results on κ are to the choice of the parameters values. A methodology for quantifying the errors related to the extraction of κ values from CV data is proposed; we also

Acknowledgements

The authors acknowledge the support of the EU FP6 network of excellence SINANO (“Silicon based Nanodevices”). OB and SH acknowledge the financial support of EPSRC (UK) and IMEC (Belgium) for the samples provided.

References (10)

  • R.J. Barlow

    Statistics – a guide to the use of statistical methods in the physical sciences

    (1999)
  • R. Chau et al.

    IEEE Electron Dev Lett

    (2004)
  • G. Ribes et al.

    IEEE Trans Dev Mater Reliab

    (2005)
  • Lee BH, Song SC, Choi R, Wen HC, Majhi P, Kirsch P, et al. In: Proceedings of the 2005 IEEE international conference on...
  • Engström O, Raeissi B, Hall S, Buiu O, Lemme MC, Gottlob HDB, et al. ULIS’06, April 20–21, 2006, MINATEC Grenoble,...
There are more references available in the full text version of this article.

Cited by (0)

View full text