How to pass Jacobian to IDA via NDSolve
- To: mathgroup at smc.vnet.net
- Subject: [mg131467] How to pass Jacobian to IDA via NDSolve
- From: Cheol <cheol_w_lee at hotmail.com>
- Date: Thu, 11 Jul 2013 02:03:08 -0400 (EDT)
- Delivered-to: l-mathgroup@mail-archive0.wolfram.com
- Delivered-to: l-mathgroup@wolfram.com
- Delivered-to: mathgroup-outx@smc.vnet.net
- Delivered-to: mathgroup-newsendx@smc.vnet.net
According to the IDA manual, when solving DAEs, there is an option to supply J=df/dx+c=E2=88=97df/dx=E2=80=B2(where d is for partial derivatives) to the IDA module. How can I do the same via NDSolve? I tried the following just to see what happens: vars = { alphan0, alphan1, alphan2, alphan3, alphan4, alphan5}; tvars = { alphan0[t], alphan1[t], alphan2[t], alphan3[t], alphan4[t], alphan5[t}; tvarsp = D[tvars, t]; Jdae1 = D[dae1[[All, 1]], {tvars}] - D[dae1[[All, 2]], {tvars}]; Jdae2 = D[dae1[[All, 1]], {tvarsp}] - D[dae1[[All, 2]], {tvarsp}]; jfun[t_, x_, xp_, c_] := Module[{j, jj}, j = (Jdae1 /. tvars -> x) + c (Jdae2 /. tvarsp -> xp); Function[# jj] /. jj -> j]; sol3 = First@ NDSolve[{dae1, ic}, vars, {t, 0, 10}, WorkingPrecision -> 25, AccuracyGoal -> 5, PrecisionGoal -> 5, Method -> { IDA, "ImplicitSolver" -> { "Newton", "LinearSolveMethod" -> {"Dense", "Jacobian" -> jfun} } } ] Mathematica did not generate any error messages, however I am certain I am not doing it correctly since no difference in output can be observed. Many thanks!