Understanding and Designing Precursors for Chemical Vapor Deposition

Understanding and Designing Precursors for Chemical Vapor Deposition

Author: Catharina Goedecke, Ralf Tonner

Dr. Ralf Tonner leads a junior research group for theoretical surface chemistry at the University of Marburg, Germany. He talks to Catharina Goedecke for ChemViews Magazine about his recent research on precursors for the functionalization of semiconductors and how he works at the interface between chemistry, physics, and materials science.

The group looked at the decomposition of group 13 and group 15 compounds, such as gallanes or phosphines, during their deposition on semiconductor surfaces. The resulting insights into the reaction mechanisms allow the rational design of new precursors with desirable properties.

Would you please introduce your research?

I am a computational chemist, which means I use quantum mechanical methods to solve chemical problems. We are working on surface chemistry, and our main interests in this rather broad research area are chemical bonding on surfaces, at interfaces, and in materials, as well as interesting reactivity in these systems. This is rather unusual, because in the field of extended systems, i.e, surfaces and solids, there are mainly people with a physics background. They have a very different approach compared with chemists, who often take more of a local view on these systems.

Could you give an example of the chemistry problems you tackle using these methods?

One main part of our research concerns the functionalization of semiconductors. We look at how we can change semiconductor surfaces to give them new features for novel applications. We are doing this work at the interface between chemistry, physics, and materials science. We are concerned with the elementary steps and the chemistry in processes which deposit inorganic molecules on semiconductor surfaces. Our colleagues in physics and in materials science can then build on these results.

What kinds of applications do these functionalized semiconductor materials have?

Our materials science colleagues are, for example, working on improved solar cells with higher efficiency and on optoelectronic devices based on silicon. Usually, silicon uses electrons for transferring information, but an optoelectronic device uses light instead. This is interesting because light is much faster than electrons and also produces less heat.

How are the semiconductor surfaces modified?

They are covered with a thin film of so-called III/V materials. These are composed of group 13 and group 15 elements, an example is gallium phosphide. The materials are grown on top of a silicon wafer using metal-organic vapor phase epitaxy (MOVPE). This is a growth procedure where precursor molecules, usually in liquid form, are evaporated and brought to the heated silicon surface. The molecules then decompose and form the desired film. Common precursors are triethylgallane and tert-butylphosphine. We investigated the decomposition reactions of these molecules [1].

Until few years ago, people used the hydrides, gallane or phosphane, to do this deposition, but they are toxic and hard to handle. Additionally, the hydrides decompose only at very high temperatures, and our colleagues want to grow metastable III/V materials, which require low growth temperatures. “Traditional” MOVPE is done at roughly 1000 °C, but if you want to grow these optically active materials, the temperature should be between 400 °C and 675 °C. This is where we come into play, because we can look at the decomposition channels of the precursors in the gas phase and can get an idea of which reaction pathways are important and at which temperatures these molecules decompose.

The network of pathways for a given reaction can be very complex. How do you as computational chemist make sense of these many possibilities and find the most likely path(s)?

You really have to think about which kind of reactions are possible. You can narrow the pathways down by considering the experimental conditions. We know, for example, that the precursor is present only in a very low concentration. That means that it’s very unlikely that two of the precursor molecules meet, so we only consider unimolecular reactions. As a second experimental condition, we know that hydrogen is a carrier gas in the experiment, so we also consider reactions with H2.

As a next step, we can, of course, look at the literature and get an idea of the possible pathways. This is where the chemist’s experience really helps and where we can make valuable contributions. For chemists, it is often much easier to think in terms of reactivity and decomposition than for people with a physics background.

We know from the literature which classes of decomposition reactions can play a role, for example, homolytic dissociation of bonds to form radicals. Heterolytic dissociation does not usually happen in the gas phase, because generating ions is unfavorable. Lastly, certain elimination reactions can occur, as well. Once you picked the likely reactions, you actually have to go and calculate the thermodynamics for the whole reaction network.

So you look at reaction energies and activation energies to figure out what is feasible at the reaction temperature?

Yes. We apply a two-step procedure here, which we have used quite successfully for a number of precursors. First, we look at the full reaction network by the thermodynamic approach. That is rather straightforward, because we only look at ground state properties. We just compute the thermodynamics, and then we select the reactions which are exergonic at the experimental conditions. If we have a reaction which has an enthalpy of, say, +200 kJ/mol, it is unlikely to ever happen, so we don’t need to search for a transition state. We then search for transition states and calculate activation energies only for the exothermic reactions, which are surprisingly few. This provides us with an efficient approach to explore the likely reactions.

Which mechanisms are important for the decomposition reactions?

We found that homolytic bond cleavages, i.e., cutting a bond and forming radicals, are unfavorable. This was a bit unexpected because in the literature there were many propositions and indirect evidence that these homolytic dissociations may play a major role. Secondly, we found that especially β-hydrogen elimination reactions are very important because they proved to have the lowest barrier for different precursors.

How do β-hydrogen eliminations work at an element such as phosphorus?

That is what we wondered, as well, when we found this reaction pathway. For group 13 elements, β-hydride elimination is textbook knowledge: You have an empty orbital into which a C-H bond can donate electron density in a synchronous reaction. For gallium, for example, the gallium-hydrogen bond forms simultaneously with the breaking of the gallium-carbon bond.

For the group 15 elements, β-hydrogen elimination was the dominant reaction channel, as well, even though there is no empty acceptor orbital. We tried to understand that and performed an analysis of this reaction for the phosphorus precursor, which we later extended to other group 15 elements [2]. Since there is no acceptor orbital, the phosphorus-carbon bond has to break first and then the phosphorus-hydrogen bond can be formed in a later part of the reaction.

What is the impact of these results?

We can use the results of our chemical bonding analysis and make predictions that are useful for finding good precursors. For example, our calculations showed that the phosphorus-carbon bond breaking is the rate-determining step. Thus, the carbon atom needs to be able to stabilize the positive charge occurring in the transition state.

We looked for a way to stabilize this positive charge at this carbon atom better than with the tert-butyl substituent which is used experimentally. We used the insight from the chemical bonding analysis and the calculated reaction mechanism and tested different alkyl substituents. We increased the inductive effect to improve the ability of the carbon atom to stabilize the positive charge.

We looked for a relationship between the stability of the carbocation and the transition barrier, and found that there is a linear correlation [3]. This means the decomposition barrier can be predicted simply by the charge at this one atom. After this, we thought: Why only use inductive effects? Why not also use orbital effects to stabilize this carbocation? In organic chemistry, the β-silyl effect is well known and so we did what a computational chemist can do: We simply substituted a hydrogen atom by a silyl group. We could show that the β-silyl effect actually stabilizes the carbocation very strongly and lowers the decomposition barrier, so the result was a computational chemistry-inspired precursor design.

How long did the investigation take?

It took essentially two thirds of a Ph.D. thesis, so between two and three years of research. The work has been done by my student Andreas Stegmüller, who is first author on all three publications on the topic. This was the first time that we investigated this type of chemistry, so we had to establish the methodology and find all the important reactions. After Andreas’ initial work, we had several further publications and now even undergraduate students can do follow-up work based on the project.

How did you follow up on these results?

We continued to work on the topic of III-V materials on silicon. We were also approached by the Swedish group of Henrik Pedersen, an experimentalist, and just had a joint paper on how to deposit boron carbide films on surfaces published [4]. Additionally, we are collaborating more closely on new precursors with our colleagues in inorganic chemistry and materials science at the University of Marburg. Together, we proposed a new single-source precursor for arsenic and nitrogen, also based on this type of decomposition chemistry. We have also organized a workshop on the topic of chemistry-driven growth processes, which will bring together experts on the topic from different communities in November 2015.

Thank you very much for the interview.



Dr. Ralf TonnerRalf Tonner studied chemistry at the Universities of Marburg, Germany, and Auckland, New Zealand. He received his Ph.D. in theoretical chemistry from the University of Marburg in 2007 under the supervision of Gernot Frenking. After a postdoctoral stay with Peter Schwerdtfeger at Massey University, Auckland, New Zealand, Dr. Tonner returned to Marburg in 2010 as Group Leader in theoretical chemistry.

His research interests include the ab initio description of chemical bonding and reactivity at surfaces, using mainly density functional but also wavefunction based methods in periodical and non-periodical approaches.

Selected Publications


Also of Interest


Leave a Reply

Your email address will not be published.