Optimizing a CVD Process for a High-Performance Tungsten Material

High-performance fusion reactors call for high-performance materials. To optimize a production process for a tungsten material used in a fusion reactor divertor, researchers at Forschungszentrum Jülich GmbH (FZJ), Institute for Energy and Climate Research, and Max Planck Institute for Plasma Physics in Germany turned to multiphysics modeling.


By Brianne Christopher
November 2020

In order to make fusion power not only physically possible but economically possible, we need to develop high-performance fusion reactors. However, these reactors call for high-performance materials in their own right. Consider one of the many parts of a reactor, the divertor, as an example.

Divertors (Figure 1) divert ash and other plasma contaminants out of the fusion vessel. These components must be able to withstand the harshest environment in the entire reactor setup. What material, then, is a good option for these parts? Tungsten offers divertors a reasonable operational lifetime and can withstand huge particle and heat fluxes, being heavily bombarded by neutrons, and undergoing plasma erosion and thermal cycling. Tungsten has a high thermal conductivity, and it does not produce radioisotopes with a long half-life from transmutation or trap too much hydrogen, unlike some other material choices for the divertors.

A fusion reactor with an inset giving a detailed view of the divertor. Divertor
Figure 1. A divertor in a fusion reactor.

Tougher than Tungsten

Tungsten also has downsides. It is usually brittle, and coupled with exposure to neutron bombardment and overheating, it can experience even further embrittlement over the operational lifetime of a fusion reactor. One solution to its brittleness is to produce a material called tungsten-fiber-reinforced tungsten (Wf/W), a tougher material that, through its composite structure, offers crack-dissipating mechanisms that give it a pseudoductile composite behavior, as in a fiber-reinforced ceramic.

When producing Wf/W, one of the current methods of choice is chemical vapor deposition (CVD), also a popular production process in the semiconductor industry. In this process, gas molecules adsorb on the surface of, and then react in, a reaction chamber that contains a heated substrate (Figure 2). Their interaction causes a thin, highly pure material film (here, W) to deposit onto the substrate. To ensure that the Wf/W produced by this process can be used in a fusion reactor, the CVD process itself needs to be optimized to ensure that the material produced has the right relative density and fiber volume fraction. Researchers from Forschungszentrum Jülich GmbH (FZJ), Institute for Energy and Climate Research, and Max Planck Institute for Plasma Physics in Germany aimed to investigate this process and how it could be optimized.

Reaction chamber that is part of the CVD production. CVD production device, outer view
An inside look at the heated substrate in a CVD production device. CVD production device, inner view
Figure 2. Outer (left) and inner (right) view of the CVD production device.

Developing a Complete Model for CVD Production of Wf/W

One of the key factors of the CVD process for Wf/W production is the tungsten deposition rate, which depends on the temperature and partial pressures involved. The tungsten deposition rate is hard to predict because it involves a lot of different parameters, including the surface temperature and partial pressure at the reaction sites, which depend on the reactor geometry, heater temperature, gas flow rates, and gas composition.

One important motivation for predicting the CVD process is to avoid the formation of pores in the tungsten material (Figure 3). During the CVD process, gas flows through the fiber substrate and tungsten is deposited between fibers. The area between the fibers is supposed to be filled up with the solid W; however, some gaseous domains can become isolated from fresh reactants when the path from the bulk of the gas phase is closed, or obstructed, by the W deposits. In other words, the pores do not have access to the reactants needed to fill them with tungsten, thus they remain pores throughout the process.

In order to reduce or avoid material strength-reducing pore formation, the substrate geometry and the parameters of the CVD process need to be carefully adjusted.

A grayscale image of tungsten-fiber-reinforced tungsten material with pores. Tungsten-fiber-reinforced tungsten
Figure 3. Pore formation in the Wf/W.

The goal of the FZJ research was to reduce porosity in Wf/W. For this, Leonard Raumann, material engineer at FZJ, needed to find the W deposition rate equation as a first step. Existing literature about CVD for tungsten is controversial and incomplete, because the equations and values for tungsten deposition kinetics often contradict each other from study to study. Raumann found a new rate equation for the CVD process, putting the smaller pieces from literature together as a whole (Ref. 1). But how?

He designed an experimental single-fiber setup with very well-known boundary conditions. With the help of the COMSOL Multiphysics® software and a parameter study, he found the rate equations. He then used the equations to model the Wf/W production with multiple fibers. For this, Raumann applied COMSOL Multiphysics® again, followed by a parameter optimization. The resulting parameters were also applied in reality with success.

Developing and Validating a Multiphysics Model

The single-fiber setup to develop the new models for the chemical vapor deposition rate of tungsten is shown in Figure 4, including a preheater and a main heater. The researchers wanted to see how fast the tungsten would grow and how this rate of growth was affected by the temperature and partial pressure. They then adjusted the tungsten hexafluoride (WF6) reaction order between one and zero, depending on the temperature and the WF6 partial pressure. To do so, they used numerical modeling to study the fluid dynamics of the gas mixtures, heat transfer of the thermal losses, and chemistry and rate equations for the chemical reactions at the deposition surface.

A simplified lab setup of a preheater and main heater and the corresponding model geometry with an inset to show the details of the upper fiber holder and W fiber. Experimental setup and model geometry
Figure 4. Model geometry based on a simplified experimental setup. W fiber is shown to the right (thin gray vertical line).
2D simulation results side-by-side, showing temperature in red-to-white color gradient on the left and partial pressure in a rainbow color table on the right. CVD process simulation results
Figure 5. Temperature (left) and partial pressure (right) during the CVD process. Fiber surface at radius r = 0.075 cm and inner tube surface at r = 0.4 cm).

A macroscaled CVD reactor model returned the partial pressures as input for microscale transient simulations. For this, Raumann modeled the W coatings growing onto multiple adjacent W fibers as well as the surface-to-surface contact of the W coatings and the corresponding potential pore formation. In Raumann's dissertation (Ref. 1), he validated these models successfully by comparing experiments for the deposition rate, pore structure, and relative densities of the CVD process of Wf/W (Figure 6). In a third step, the multifiber model was used for a CVD process parameter optimization to successfully improve the simulated and later also experimental material density.

A comparison of experimental CVD process results to simulation results, where the overlaid image shows that the two agree; the pore formation creates a caterpillar-like impression. Experiments vs. simulation
Figure 6. Experimental results (top), simulation results (center), and an overlay of both results (bottom) of pore formation during the CVD process.

Scaling Up Fusion Research

The FZJ-IPP team is currently planning to apply the validated model to a 3D geometry to scale up Wf/W production even further. They aim to develop a new approach that would involve one coil delivering the W fabric (CVD substrate) to another, with one coil unbound and the other coiled and heated up. This allows the fabric layer stacking to take place with the chamber closed, so that all layers can be deposited in one CVD process (there is also a lower risk of contamination this way).

Scaling up the production process for tungsten-fiber-reinforced tungsten means new possibilities for fusion power. Before this research, producing one layer of the tungsten material took around 5 hours, but by optimizing the CVD process parameters, it can take just 30 minutes to produce one layer of Wf/W — which is 10 times faster! By optimizing production processes for high-performance materials for fusion reactors, we can ensure that fusion power is both possible and cost efficient.


Reference

  1. L. Raumann, Modeling and validation of chemical vapor deposition for tungsten fiber reinforced tungsten, dissertation, Energy & Environment, Schriften des Forschungszentrums Jülich, 2020.