How the Chemical Vapor Infiltration process can be optimized for the production of advanced composite and porous ceramics

Chemical Vapor Infiltration (CVI) is a specific materials processing technique allowing the insertion of a refractory material inside a porous medium at temperatures well below its melting point. It is a variation of the Chemical Vapor Deposition (CVD) process involving a porous substrate. Among other applications we find (i) the ceramization of porous or architectured foams, turning them capable to withstand very high temperatures encountered in Concentrated Solar Power heat exchangers [1], (ii) the preparation of porous catalyst supports  or burners [2], (iii) the deposition of photocatalytic ceramics in a porous medium [3], (iv) the modification of porous membranes [4],  and, (v) very importantly, it is one of the important steps in the manufacturing of Ceramic-Matrix Composites (CMC) and of Carbon/Carbon Composites (C/C) parts designed to integrate aircraft or spatial propulsion systems, thermal protection systems for space objects, braking systems, etc. [5]. Currently, CVI is the industrial process by which all brake disks for aircraft with more than 100 passengers are manufactured throughout the world.

The process consists in letting a gaseous precursor (say, hydrocarbons for carbon infiltration) penetrate the porous preform by diffusion and react with the solid, yielding a deposit that progressively fills the porous space. Conditions are set for gas pyrolysis and heterogeneous reaction, especially temperature (high), pressure (low, usually) and a carefully chosen gas composition. Fig. 1 gives a schematic description of the main phenomena taking place in the reactor.

This process enables the production of the best quality of matrices in CMC, albeit at the expense of usually very long processing times: indeed, in its basic version (ie. isothermal, isobaric), the gas diffusion through the porous media has to be fast enough as compared to the heterogeneous reaction rate, which therefore has to be limited. Manufacturing a batch of parts can last several weeks. 

Hence, slashing down these processing times is an important source of cost reduction for the resulting materials. Trial-and-error optimization being extremely expensive, there is a great interest in setting up modelling approaches, either to find the best operating parameters for a setup, or to assess the pertinence of some variations, like Thermal-Gradient CVI (TG-CVI), or Pulsed CVI (P-CVI) and others.

Figure 1: Multi-scale view of CVI. From the left, at the bottom: fibre scale; at the top: Atomistic scale. On the right: Furnace + part scale. Source : CEM-WAVE H2020 project website)

Modelling this process involves several length scales : 

1. at the part scale, one has to represent simultaneous transfers of heat (possibly including electromagnetic sources like RF induction of micro-waves), of gas species transfer, of chemical reactions and of the porous medium evolution.
2. at the pore scale, heat and gas transfer have to be evaluated from the architecture of the solid phase and from its physical characteristics, as well as for the gas. It is very frequent to fall in the intermediate regime between continuum (Darcian) gas flow and the rarefied gas flow, represented by Knudsen diffusion or by the Klinkenberg correction to Darcy’s law. This is a challenge for the determination of effective transfer coefficients by change-of-scale methods.
3. in eg. woven fibrous composite preforms, the pore space may be multi-scale itself, with clearly distinct intra-bundle and inter-bundle pore sizes.
4. gas pyrolysis and deposition are the result of a large number of homogeneous and heterogeneous elementary reactions, the result of which can yield non-linear deposition kinetics that also can depend on diffusion to some extent. 

These different scales are treated with distinct modelling tools. Finite Element solvers are quite adapted to the large-scale simulations, whereas several types of image-based modelling tools are suitable for pore-scale representations of heat & mass transfer. These may start from X-ray CT scans of the porous preforms and include change-of-scale techniques like homogenization, volume averaging or more original Monte-Carlo Random Walks methods, associated to surface modification algorithms in order to produce estimates of transfer properties evolutions during infiltration.

Additionally, simpler analytical tools may provide valuable insights into the process control. For instance, a simple criterion for a correct inside-out infiltration front has been set up and tested successfully in several situations.

The CVI process, already mature in several industrial applications, has as of now a great potential for further developments, pushed by the growing interest of high-performance materials manufacturers in producing new Ceramic-Matrix Composites and tailored porous ceramics in varied applications, some of them led by the need to adapt the industry to greener sources of energy.