Computational Science Research

Movies/H2Jet-n.pdfCombustion is and will remain a key component of countless engineering systems. It is central to propulsion, materials processing, fuel conversion, and power generation. Combustion physics in most present-day systems involves the mixing of fuel with oxidizer, chemical conversion often involving thousands of reactions of hundreds of intermediate species, and the subsequent interaction of the heat released with the device and flow. Engineering objectives, such as efficiency and power density, typically depend upon achieving rapid ignition, fast burn rates, and particular combustion products and temperatures, which are sought by adjusting the fuel-oxidizer mix, geometry, and inflow conditions.

Our target application is emblematic of this: a turbulent fuel jet mixes with a cross-flowing oxidizer and burns. In this canonical flow, there are limited options for altering or enhancing combustion. It is particularly noteworthy that none of these standard approaches offer direct control of key combustion mechanisms: there is no way to affect molecular diffusion, particular chemical reaction pathways, or local temperature. Plasmas, can manipulate all these mechanisms directly and thus fundamentally advance control of combustion and even enable previously infeasible applications.

Most directly, ionization in a plasma creates radicals that can short-circuit stages of the standard reaction pathway to reduce ignition times and increase burn rates.  Transport can also be manipulated. In non-premixed combustion, diffusion is a key mediating factor, leading to thin flame structures. The electric fields (E-fields) used to generate a plasma can bias diffusion of charged radicals in and near flames in ways that affect both flame structure and stability. Joule heating within a plasma can also alter the chemical reactions, all of which depend strongly upon temperature. In short, plasmas offer a little-explored set of ‘knobs’ that can be ‘turned’ to tune combustion to meet objectives. The long-term potential impact on combustion engineering is tremendous.

However, engineering design is limited by the complexity of the coupled electrode-plasma-combustion-turbulence system.  Important length scales include the device scale (say ∼ m), the Kolmogorov-scale of the turbulence (~ mm), the flame thickness (< mm), and thick- ness of plasma sheaths and filaments (∼ μm). Results also depend on the E-fields and electron fluxes of the electrodes, which themselves are well-known to degrade in time due to atomic-scale (∼ nm) mechanisms that alter quantum-mechanical work functions and thereby limit the electron flux. Important time scales range from the nanoseconds of plasma discharges, to the microseconds of reaction kinetics, to the milliseconds of turbulence mixing, to the minutes or longer of electrode aging.

Predictive simulations that faithfully integrate physics across this range of scales will fundamentally advance the science and technology of plasma-coupled combustion. Experiments have unambiguously demonstrated its potential and can provide sufficient data for validation, but are limited in terms of diagnostics. It is infeasible to measure all the mechanisms of importance in the overall process, and challenging to explore new configurations. The complexity also makes it challenging to anticipate the full range of potential phenomenology that can be exploited in applications or even what sort of approaches might lead to optimal performance.

For predictive simulation, multiple, high-fidelity physical models must be integrated: turbulent mixing, chemical kinetics and transport, plasma physics, and electrode-surface electrochemistry  Each of these might be considered a petascale computational task. Indeed, several of these single-physics applications have been among the first to utilize petascale computational resources. These will demand expert input across traditional disciplines. For predictive simulations of the proposed system, all components must work in concert and with assessed uncertainties. Reliable predictions will require special tools and techniques to efficiently harness the heterogeneous environment of anticipated trans-petascale and exascale computer platforms.