Thermodynamic and experimental studies of ALD (Atomic Layer Deposition) of TaN and of its organometallic precursor PDMAT, Ta[N(CH3)2]5, used in microelectronics

Perrine Violet
Language of the thesis
Thesis name in original language
Etude thermodynamique et experimentale du depôt ALD (Atomic Layer Deposition) de tan et de son precurseur organometallique PDMAT, Ta[N(CH3)2]5, utilise en microelectronique
Abstract & Cover

The continued miniaturisation of transistor components is confronting scientists with increasingly difficult technological nodes to solve. In particular, the integration of the copper diffusion barrier in interconnects requires a change of process associated with the use of organometallic precursors, which are more reactive at low temperatures (of the order of 250 °C), to solve compliance problems. Thus, the present work aimed to advance the understanding of the growth of tantalum nitride films by ALD using the organometallic precursor PDMAT, Ta[N(CH3)2]5, and ammonia. The ambition to understand the deposition processes is not new. This understanding requires knowledge of the gaseous species that are transported through the lines and also those that reach the substrate. Many studies on halogenated precursors used in ALD and CVD exist. The particularity of the ALD process over the classical CVD process is the non-interaction of the precursors with each other, which aims at "eliminating" the gas phase reactions between precursors and facilitates part of the study. However, the use of organometallic precursors such as PDMAT makes the understanding of the growth mechanisms significantly more complex due to the complex structure of the precursor and the small number of studies and therefore data reported in the literature. We have conducted parallel modelling/development/characterisation actions combining experimental thermodynamics, modelling and experimentation on a prototype ALD reactor. Different models - ab-initio, statistical, dimensional - have been tested and allow in some cases to estimate the missing thermodynamic data for organometallics. In particular, the molecule Ta[N(CH3)2]4 could be studied more seriously in order to deduce an enthalpy of formation independently. The importance of calorimetric measurements of the enthalpy of formation of solids or liquids should also be noted, provided that their chemical "purity" and/or molecular composition can be attested at the same time. Thermodynamics is fully useful in understanding the mechanisms and allows us to know the state of equilibrium. Indeed, the vaporisation of organometallic precursors in bubblers is at thermodynamic equilibrium. However, given the injection times of the reagents used in ALD reactors, intermediate states may exist due to slower gas decomposition or solid precipitation kinetics, not predicted by thermodynamics. In order to study the behaviour of the precursor PDMAT in the gas phase as a function of temperature, we used the SIMAP localized mass spectrometer.

As PDMAT is an extremely reactive molecule in contact with the atmosphere, a new sealed effusive reactor adapted to the mass spectrometer was developed. This reactor allows the study of saturation vapour pressures when mounted with a single effusion cell. It also allows the analysis of the thermal cracking of the vapours of the organometallic precursor studied when it is set up with "tandem" cells, consisting of an evaporation cell and a cracker. The particularity of the effusion/evaporation cell is that it can be loaded with precursor under a glove box (controlled atmosphere), closed and kept tight during the assembly of the reactor in the desired configuration and the evacuation of the reactor. The cell is then opened remotely and allows the spectrometric study. This cell can also be weighed for calibration by mass loss. The operation of the reactor was validated by performing saturation vapour pressure measurements of a well known organometallic, Y(tmhd)3 and spectrometric measurements with mercury in both configurations against physical flow models. The present spectrometric study of the vaporisation and thermal cracking of the precursor showed that: - PDMAT in solid form at room temperature vaporises as 3 gaseous species Ta[N(CH3)2]5, Ta[N(CH3)2]4 and O-Ta[N(CH3)2]4 - the Ta[N(CH3)2]4 molecule remains by far the most stable gaseous species up to 400°C during vaporisation and cracking. - the decomposition products, mainly the amine compound HN(CH3)2, are derived from the cracking of PDMAT to Ta[N(CH3)2]4 or TaN3C6H18. - The deposits observed in the cracker are mixtures of Ta, O, N and C but we cannot certify that they are defined compounds or solid solution. All this allowed us to perform some thermodynamic simulations of the ALD process. It then appeared that the amine HN(CH3)2 in question observed only exists because its decomposition step has slow kinetics (>1s) and that it is therefore a transition species. These simulations are still incomplete because species such as TaN3C6H18 are not taken into account in the calculations due to a lack of thermodynamic data. Prospects for improving the spectrometric reactor would be to integrate an additional gas introduction to study specific and targeted reactions. A cracker operating at thermodynamic equilibrium, e.g. by increasing the residence time, would be valuable for studying reactions at thermodynamic equilibrium. Crackers with intermediate residence times would allow the study of different kinetic regimes and in particular the consequences of the pulse times used in ALD deposition. This thesis also describes in detail the ALD reactor designed by AST, a rather complex prototype reactor using a virtual valve. The geometry of this reactor leads to the elaboration of thin films of heterogeneous thickness due to an inhomogeneous distribution of the reagent inputs on the substrate surface. This geometry needs to be improved in order to obtain perfectly conformal thin films (homogeneous in thickness) typical of ALD deposits. Despite the inhomogeneity of the thickness of the deposits, the latter have allowed us to understand the technical and chemical issues related to the implementation of ALD processes. Leads on process optimisation were obtained thanks to spectrometric studies, deposition with PDMAT alone and PDMAT and NH3 and flow simulations. All this explains the difficulties encountered for the deposition of pure TaN, the incorporation of impurities such as O and C, which are difficult to avoid. The negative effect of this contamination on the diffusion barrier performance of these layers remains to be assessed. More surface-based approaches - study of the reactions between the precursors and the substrate - deserve to be explored in greater detail by means of deposition carried out by ALD or CVD, followed in situ by microbalance and/or by qualitative quadrupole spectrometric monitoring of the reactive species. The results show the interest of using the spectrometric tool to understand the growth of thin films by ALD from organometallic precursors. Thermodynamic simulation, spectrometric studies and the elaboration and characterisation of deposits are perfectly complementary techniques for the understanding and control of these emerging and complex processes


SIMaP - Science et Ingénierie des Matériaux et Procédés
(Grenoble, France)
External Link
Read Thesis
linkedin invite