Application Highlights

Atmospheric chemistry. Computing sensitivities with respect to rate constants revealed a correlation between O₃-NO_x-hydrocarbon sensitivity and easily observed concentrations. 

Breast cancer modeling. High-level knowledge about program structure was used to produce derivative code as efficient as code that required two person-years to develop by hand.  Preprint MCS-P491-0195.

Computational fluid dynamics. Implicit time integration using AD-generated Jacobians reduced runtime by an order of magnitude over the original 4-stage Runge-Kutta method. Robustness was also enhanced.  Preprint ANL/MCS-P687-0997.

Mesoscale climate modeling. Sensitivity analysis revealed a numerically-induced superacoustic precursor wave that would have been undetectable using divided differences.

Network Enabled Optimization System (NEOS) server. Given a user's function definition, AD provides efficient derivatives for large problems.

Semiconductor device modeling. The use of ADIC-generated code reduced simulation time by more than 50%. Preprint ANL/MCS-P698-1097.

Water reservoir simulation. Using AD resulted in a 5-fold speedup in a yield optimization process, in large part due to the use of analytic derivatives. Preprint ANL/MCS-P585-0496.

Other applications: multidisciplinary design optimization, reactor engineering, groundwater remediation, global climate modeling, superconductor simulation, multibody simulations, molecular dynamics simulations, power system analysis, and storm modeling.