Chapter 9 Summary and outlook

Transcription

1 Chapter 9 Summary and outlook During the course of the present study, new precursors were developed and tested for the deposition of tantalum and niobium nitride thin films via Metal Organic Chemical Vapor Deposition (MOCVD) and Atomic Layer Deposition (ALD). These films find their application in the fabrication of microelectronic devices, namely as metal gate and diffusion barrier materials. Both materials, TaN as well as NbN, feature high thermal stability, good conductivity and excellent barrier properties, when deposited in the face centered cubic phase. The challenge is to deposit carbon free and dense films that withstand post oxidation processes and show good values for the work function of the material. The experimental work and the results of this dissertation can be divided into two parts. The first part deals with the chemical realization of new ligand concepts for MOCVD and ALD precursors. These concepts comprise the increase of complexity of the metal organic complexes compared to simple compounds, (mixed amido/imido complexes) having lower coordination numbers and simple monodentate ligands. The influence on the chemical as well as thermal behavior has been investigated. The second part consists of MOCVD and ALD experiments with newly synthesized, as well as already known, but not fully tested precursors. I. Synthesis and characterization of new precursors All synthetic routes of new compounds were based on amido/imido complexes of tantalum and niobium as starting compounds (S4-S9, figure 88). These four coordinated tris(dialkylamido)-alkylimido-tantalum and niobium complexes were synthesized starting from the metal chloride via a pyridine stabilized chloro/imido intermediate (S1-S3). The synthetic routes were improved, in terms of yields and batch sizes (up to 40 g per batch). In the following, these compounds were modified by transamination reactions with hydrazine derivatives or the insertion of alkyl substituted carbodiimide derivatives. Thus, two new classes of precursors were successfully synthesized: Mixed amido/imido/hydrazido complexes of tantalum (A1-A3) as well as mixed amido/imido/guanidinato complexes of tantalum and niobium (B1-B8; C1-C4). In the context of these experiments, some side products of reactions were characterized and discussed e.g. hydrolyzed species or decomposition products. The compounds were analyzed by means of 1 H-NMR, 13 C-NMR, EI- Mass spectroscopy, CHN analysis, TG/DTA analysis and - when possible - by single crystal X-ray diffractometry. 147

2 Figure 88. Reaction scheme depicting the syntheses of the complexes A1-A3 by transamination reactions (one amido group is replaced by one hydrazido ligand) and of the guanidinato complexes B1-B8 and C1-C4 by insertion reactions (two equivalents of carbodiimide insert into the metal nitrogen bond of the amido group). The hydrazine containing compounds A1-A3 are all liquid at room temperature and by their nature sufficiently volatile for application as precursors. Repeated quantitative distillation without any sign of decomposition is possible. TG/DTA analysis showed that the decomposition temperatures of A1, A2 and A3 are above 280 C which is 90 C higher than the temperatures that lead to decomposition of the amido/imido compounds S4-S6. Along with the higher thermal stability goes a significantly lower chemical reactivity, e.g. towards hydrolysis. The crystal structure of A1 exhibits the η 2 -coordination of hydrazido ligand and consequently a coordination number of five for tantalum in the solid state. Further exchange of more than one amido group was not possible. The six coordinated guanidinato complexes (B1-B8; C1-C4, figure 89) show a large window for the variation of ligand substituents (R 1 - R 4 ). However, not all carbodiimide derivatives could be inserted into the metal amido bonds as a result of steric constraints. All guanidinato complexes are colorless solids that form needle-like, agglomerated crystals from solution at reduced temperatures. Due to the chirality of the molecules and the multiplicity of ligand types, the NMR spectra of these complexes displayed high complexity. Crystal structures for eight of the twelve compounds could be determined. The complexes are structurally very similar, with bond lengths and angles of the solid state structures showing only minor variations. TG/DTA measurements and isothermal 148

3 studies revealed the low volatility of the cy-cdi based complexes. This was rather unexpected, considering the fact that the structural and chemical behaviors of these classes of compounds are very similar. Thus, cy-cdi based complexes are in general not suitable for classical MOCVD purposes due to a supposedly weak precursor transport. Still, application of the compounds derived from cy-cdi in Liquid Injection MOCVD is feasible. The choice of substituents of the amido or imido groups does not seem to have a significant impact on the thermal behavior of the resulting complexes. Finally, thermal decomposition experiments of B1-B3 clearly showed that highly complex pyrolysis reactions take place which nonetheless lead to a clean decomposition of the selected precursors. II. MOCVD and ALD experiments MOCVD and ALD experiments were performed using different s. Preliminary, but meaningful MOCVD experiments were conducted with a selfbuilt for small substrates. Precursors, showing promising properties were sent to our cooperation partner in Erlangen (large-wafer instrument) or were tested with the industrial Aixtron 200 RF MOCVD or the Aixtron 200 FE Liquid Injection MOCVD s, situated in our research group (table 24). ALD experiments were performed in the commercially available ASM F 120. MOCVD experiments applying mixed amido/imido compounds (S5, S6, S8) as single source precursors led to the formation of conducting films that showed high contaminations of carbon. On one hand, addition of ammonia reduced the amount of carbon drastically (most probably due to transamination reactions), but on the other lowered the quality of the films by means of lower density and deteriorated conductivity. Experiments using the guanidinato complexes clearly showed that an increase of the complexity of a precursor (higher coordination number, higher atomic percentage of carbon, heterolepticety) for TaN-depositions does not necessarily conflict with the thin film quality regarding carbon incorporation and the formation of cubic, conductive and carbon free tantalum nitride. Interestingly, ammonia seems not to be a stringent requirement for the formation of conductive cubic TaN and NbN using guanidinato containing complexes (B1- B4, C1, C2) as precursors. A possible explanation is that the blocking of free coordination sites at the metal center, due to the introduction of bidentate ligands inhibits carbon incorporation caused by the formation of intermediate metal-carbon bonds during decomposition. Thus, guanidinato complexes may have quite a potential for the MOCVD of metal nitrides, although sufficient mass transport of the precursor to the still remains a challenge. LI-MOCVD experiments were performed with B2 to enhance the precursor 149

4 transport into the that indicated the formation of carbon free films. However, problems occurred in terms of oxidation and the formation of amorphous TaN x, because the was originally designed for the deposition of oxide materials. Table 24. Compounds that were used for MOCVD and ALD including a short summary of the results. Formula Reactor type Chapter Qualitative results S5 [Ta(NMeEt) 3 (N-t-Bu)] Selfbuilt MOCVD Cubic TaN was deposited having high levels of impurities (O, C). S6 [Ta(NEt 2 ) 3 (N-t-Bu)] Aixtron 200 RF Cubic, conductive TaN was deposited. Low level of impurities (with ammonia as reactive gas). S6 [Ta(NEt 2 ) 3 (N-t-Bu)] Large wafer (Erlangen) S5 [Ta(NMeEt) 3 (N-t-Bu)] Large wafer (Erlangen) A3 [Ta(NEt 2 ) 2 (tdmh)(n-t-bu)] Selfbuilt MOCVD A2 [Ta(NMeEt) 2 (tdmh)(n-t-bu)] Large wafer (Erlangen) B2 B2 [Ta(NMeEt){η 2 -(N-i- Pr) 2 C(NMeEt)}(N-t-Bu)] [Ta(NMeEt){η 2 -(N-i- Pr) 2 C(NMeEt)}(N-t-Bu)] Selfbuilt MOCVD LI-MOCVD Aixtron (200 FE) S8 [Nb(NMe 2 ) 3 (N-t-Bu)] Selfbuilt MOCVD C1 [Nb(NMe 2 ){η 2 -(N-i- Pr) 2 C(NMeEt)}(N-t-Bu)] Selfbuilt MOCVD S6 [Ta(NEt 2 ) 3 (N-t-Bu)] ASM F 120 ALD S9 [Nb(NEt 2 ) 3 (N-t-Bu)] ASM F 120 ALD Heavily oxidized TaN was deposited. Still, the films showed good electrical properties Heavily oxidized TaN was deposited Amorphous and insulating TaSi x N y was deposited Heavily oxidized TaSi x N x was deposited. The films were completely insulating Cubic TaN with almost no impurities was deposited without ammonia. (fair electronic properties) Oxidized and insulating, but carbon free TaN x was deposited. 7.1 Cubic NbN was deposited having high levels of impurities (O, C). 7.2 Cubic NbN with almost no impurities was deposited without ammonia Amorphous and insulating films with low densities were deposited Amorphous and insulating films with low densities were deposited. MOCVD-experiments in the presence of ammonia using hydrazido containing precursors in a selfbuilt (A3) gave amorphous, insulating, but carbon free tantalum nitride films with significant amounts of silicon, incorporated in the film. The presence of silicon in the films concurrent to the absence of carbon is a very interesting finding due to the very high thermal stability and excellent barrier properties of TaSi x N y films according to the literature. The intrinsic incorporation of Si into the layers is interesting as such and points to a special ligand fragmentation of the hydrazido moiety of A2, which warrants further studies. 150

5 On the other hand, this property rules out the deposition of phase-pure crystalline cubic TaN using A1-A3. Testing this class of precursors in a large scale (Erlangen, A2) at least showed that - in principle - this type of precursor is suitable for covering large wafers in terms of precursor transport and geometry. However, we failed to avoid undesired heavy oxygen contamination with the setup in Erlangen. Within this work, the ASM F120 ALD was installed in Bochum. Initial problems with uniformity of the depositions were overcome by changing the setup from a high volume to an almost zero-volume. Experiments using S5, S6 and S9 as precursors showed very moderate results. The films were less dense and completely insulating. In addition even at a substrate temperature of 200 C thermal CVD-like decomposition contributed to the film growth. Thus no ideal ALD-behavior was observed for TaN. The oxidation of all films that can occur either during the deposition (e.g. by leakages or impurities in the reactants) and/or after deposition represents a general problem when dealing with oxophilic metal nitride films. Due to the absence of possibilities to measure the composition of the films or their electrical properties contemporaneously to the fabrication of the films, the source of oxygen could not be certainly identified. Figure 89 summarizes the findings for the conductivities of all films, deposited within this work. For each experimental row, the substrate temperature of the film having the lowest resistivity is displayed. Based on these results amido/imido precursors still are the best alternative for fabricating diffusion barriers or metal gate stacks. The fact that the resistivities of the samples prepared in Bochum were not measured directly after deposition may lead to the impression that films, deposited in Erlangen have better electrical properties (direct measurement after deposition). Regardless of this issue, guanidinato complexes have a high potential due to their excellent ability to act as single source precursors. No additional ammonia is required that would support the formation of insulating N-rich Ta 3 N 5. All results have to be taken with care, as it holds for all publications regarding the MOCVD and ALD of thin nitride films. Many factors have an effect on the electrical properties of a film: grade of crystallinity, grain boundaries, metal to nitrogen ratio, and oxidation of the whole film or the surface of the film. These factors in turn depend on the type of (e.g. volume/design) and other parameters (e.g. flow rates, pressure, precursor transport rate etc.). Therefore conclusions, drawn from these results, about the suitability of a precursor have limited universal validity. 151

6 Figure 89. Diagram including the depositions experiments of this work. From each experimental row, the data set of the film with the lowest resistivity is displayed. The green box marks the expectations of the industry that have to be met. Yellow boxes show experiments using the precursors as SSP. Blue boxes indicate the use of ammonia during deposition. Grey boxes show ALD-experiments with additional ammonia. 152

7 Outlook While the part of this work, dealing with the precursor synthesis was elucidated to a very satisfying extent, further experiments are required to identify the decomposition pathways of the precursors. So far the underlying mechanisms, e.g. the incorporation of silicon by using hydrazido containing precursors have not been resolved. It is desirable to couple MOCVD s or TG/DTA instruments with other analytical methods, e.g. EI-MS spectroscopy in order to further analyze the decomposition products during the deposition of the films. Furthermore detailed CVD studies for a use of the less volatile, but soluble guanidinato complexes in liquid injection MOCVD can identify the potential of this precursor. However, the has to be designed for oxygen free depositions. Variation of the solvents (e.g. usage of aliphatic instead of aromatic hydrocarbons) probably will have an effect on the properties of the resulting films. ALD experiments, though less successful within this work, can be performed at higher substrate temperatures using the guanidinato complexes. The higher thermal stability of these complexes and the interesting decomposition behavior may lead to ALD-like growth at higher temperatures. It is questionable, whether an increase of the deposition temperature to about 250 C is sufficient for the formation of dense and stable electrically conductive metal nitride thin films. 153