|
Cantera
2.5.1
|
Method to solve a pseudo steady state surface problem. More...
#include <solveSP.h>
Public Member Functions | |
| solveSP (ImplicitSurfChem *surfChemPtr, int bulkFunc=BULK_ETCH) | |
| Constructor for the object. More... | |
| ~solveSP () | |
| Destructor. Deletes the integrator. More... | |
| int | solveSurfProb (int ifunc, doublereal time_scale, doublereal TKelvin, doublereal PGas, doublereal reltol, doublereal abstol) |
| Main routine that actually calculates the pseudo steady state of the surface problem. More... | |
Public Attributes | |
| int | m_ioflag |
Private Member Functions | |
| solveSP (const solveSP &right) | |
| Unimplemented private copy constructor. More... | |
| solveSP & | operator= (const solveSP &right) |
| Unimplemented private assignment operator. More... | |
| void | print_header (int ioflag, int ifunc, doublereal time_scale, int damping, doublereal reltol, doublereal abstol) |
| Printing routine that optionally gets called at the start of every invocation. More... | |
| void | printIteration (int ioflag, doublereal damp, int label_d, int label_t, doublereal inv_t, doublereal t_real, size_t iter, doublereal update_norm, doublereal resid_norm, bool do_time, bool final=false) |
| Printing routine that gets called after every iteration. More... | |
| doublereal | calc_t (doublereal netProdRateSolnSP[], doublereal XMolSolnSP[], int *label, int *label_old, doublereal *label_factor, int ioflag) |
| Calculate a conservative delta T to use in a pseudo-steady state algorithm. More... | |
| void | calcWeights (doublereal wtSpecies[], doublereal wtResid[], const Array2D &Jac, const doublereal CSolnSP[], const doublereal abstol, const doublereal reltol) |
| Calculate the solution and residual weights. More... | |
| void | updateState (const doublereal *cSurfSpec) |
| Update the surface states of the surface phases. More... | |
| void | updateMFSolnSP (doublereal *XMolSolnSP) |
| Update mole fraction vector consisting of unknowns in surface problem. More... | |
| void | updateMFKinSpecies (doublereal *XMolKinSp, int isp) |
| Update the mole fraction vector for a specific kinetic species vector corresponding to one InterfaceKinetics object. More... | |
| void | evalSurfLarge (const doublereal *CSolnSP) |
| Update the vector that keeps track of the largest species in each surface phase. More... | |
| void | fun_eval (doublereal *resid, const doublereal *CSolnSP, const doublereal *CSolnOldSP, const bool do_time, const doublereal deltaT) |
| Main Function evaluation. More... | |
| void | resjac_eval (DenseMatrix &jac, doublereal *resid, doublereal *CSolnSP, const doublereal *CSolnSPOld, const bool do_time, const doublereal deltaT) |
| Main routine that calculates the current residual and Jacobian. More... | |
Private Attributes | |
| ImplicitSurfChem * | m_SurfChemPtr |
| Pointer to the manager of the implicit surface chemistry problem. More... | |
| std::vector< InterfaceKinetics * > & | m_objects |
| Vector of interface kinetics objects. More... | |
| size_t | m_neq |
| Total number of equations to solve in the implicit problem. More... | |
| int | m_bulkFunc |
| This variable determines how the bulk phases are to be handled. More... | |
| size_t | m_numSurfPhases |
| Number of surface phases in the surface problem. More... | |
| size_t | m_numTotSurfSpecies |
| Total number of surface species in all surface phases. More... | |
| std::vector< size_t > | m_indexKinObjSurfPhase |
| Mapping between the surface phases and the InterfaceKinetics objects. More... | |
| std::vector< size_t > | m_nSpeciesSurfPhase |
| Vector of length number of surface phases containing the number of surface species in each phase. More... | |
| std::vector< SurfPhase * > | m_ptrsSurfPhase |
| Vector of surface phase pointers. More... | |
| std::vector< size_t > | m_eqnIndexStartSolnPhase |
| Index of the start of the unknowns for each solution phase. More... | |
| std::vector< size_t > | m_kinObjPhaseIDSurfPhase |
| Phase ID in the InterfaceKinetics object of the surface phase. More... | |
| size_t | m_numBulkPhasesSS |
| Total number of volumetric condensed phases included in the steady state problem handled by this routine. More... | |
| std::vector< size_t > | m_numBulkSpecies |
| Vector of number of species in the m_numBulkPhases phases. More... | |
| size_t | m_numTotBulkSpeciesSS |
| Total number of species in all bulk phases. More... | |
| std::vector< ThermoPhase * > | m_bulkPhasePtrs |
| Vector of bulk phase pointers, length is equal to m_numBulkPhases. More... | |
| std::vector< size_t > | m_kinSpecIndex |
| Index between the equation index and the position in the kinetic species array for the appropriate kinetics operator. More... | |
| std::vector< size_t > | m_kinObjIndex |
| Index between the equation index and the index of the InterfaceKinetics object. More... | |
| std::vector< size_t > | m_spSurfLarge |
| Vector containing the indices of the largest species in each surface phase. More... | |
| doublereal | m_atol |
| The absolute tolerance in real units. units are (kmol/m2) More... | |
| doublereal | m_rtol |
| The relative error tolerance. More... | |
| doublereal | m_maxstep |
| maximum value of the time step. units = seconds More... | |
| size_t | m_maxTotSpecies |
| Maximum number of species in any single kinetics operator -> also maxed wrt the total # of solution species. More... | |
| vector_fp | m_netProductionRatesSave |
| Temporary vector with length equal to max m_maxTotSpecies. More... | |
| vector_fp | m_numEqn1 |
| Temporary vector with length equal to max m_maxTotSpecies. More... | |
| vector_fp | m_numEqn2 |
| Temporary vector with length equal to max m_maxTotSpecies. More... | |
| vector_fp | m_CSolnSave |
| Temporary vector with length equal to max m_maxTotSpecies. More... | |
| vector_fp | m_CSolnSP |
| Solution vector. length MAX(1, m_neq) More... | |
| vector_fp | m_CSolnSPInit |
| Saved initial solution vector. length MAX(1, m_neq) More... | |
| vector_fp | m_CSolnSPOld |
| Saved solution vector at the old time step. length MAX(1, m_neq) More... | |
| vector_fp | m_wtResid |
| Weights for the residual norm calculation. length MAX(1, m_neq) More... | |
| vector_fp | m_wtSpecies |
| Weights for the species concentrations norm calculation. More... | |
| vector_fp | m_resid |
| Residual for the surface problem. More... | |
| vector_fp | m_XMolKinSpecies |
| Vector of mole fractions. length m_maxTotSpecies. More... | |
| DenseMatrix | m_Jac |
| Jacobian. More... | |
Method to solve a pseudo steady state surface problem.
The following class handles solving the surface problem. The calculation uses Newton's method to obtain the surface fractions of the surface and bulk species by requiring that the surface species production rate = 0 and that the either the bulk fractions are proportional to their production rates or they are constants.
Currently, the bulk mole fractions are treated as constants. Implementation of their being added to the unknown solution vector is delayed.
Lets introduce the unknown vector for the "surface problem". The surface problem is defined as the evaluation of the surface site fractions for multiple surface phases. The unknown vector will consist of the vector of surface concentrations for each species in each surface vector. Species are grouped first by their surface phases
The unknown solution vector is defined as follows:
| C_i_j | kindexSP |
|---|---|
| C_0_0 | 0 |
| C_1_0 | 1 |
| C_2_0 | 2 |
| . . . | ... |
| C_N0-1_0 | N0-1 |
| C_0_1 | N0 |
| C_1_1 | N0+1 |
| C_2_1 | N0+2 |
| . . . | ... |
| C_N1-1_1 | NO+N1-1 |
Note there are a couple of different types of species indices floating around in the formulation of this object.
kindexSP: This is the species index in the contiguous vector of unknowns for the surface problem.
Note, in the future, BULK_DEPOSITION systems will be added, and the solveSP unknown vector will get more complicated. It will include the mole fraction and growth rates of specified bulk phases
Indices which relate to individual kinetics objects use the suffix KSI (kinetics species index).
This routine is typically used within a residual calculation in a large code. It's typically invoked millions of times for large calculations, and it must work every time. Therefore, requirements demand that it be robust but also efficient.
The solution methodology is largely determined by the ifunc parameter, that is input to the solution object. This parameter may have one of the values defined in Surface Problem Solver Methods.
The time step is determined from sdot[], so that the time step doesn't ever change the value of a variable by more than 100%.
This algorithm does use a damped Newton's method to relax the equations. Damping is based on a "delta damping" technique. The solution unknowns are not allowed to vary too much between iterations.
EXTRA_ACCURACY: A constant that is the ratio of the required update norm in this Newton iteration compared to that in the nonlinear solver. A value of 0.1 is used so surface species are safely overconverged.
Functions called:
ct_dgetrf – First half of LAPACK direct solve of a full Matrixct_dgetrs – Second half of LAPACK direct solve of a full matrix. Returns solution vector in the right-hand-side vector, resid. 1.9.1