Re: [Libmesh-users] neighbors

 Re: [Libmesh-users] neighbors From: Roy Stogner - 2007-06-26 21:20:04 ```On Tue, 26 Jun 2007, Lorenzo Botti wrote: > It seems to me that active elements can have non active neighbors... Is it > true? Yes. > If yes, is it correct? Also yes. If you have a mesh like this: *-*-*---* | | | | *-*-* | | | | | *-*-*---* Then it has one inactive element and five active elements; the coarse active element's neighbor is the inactive element. > I know than the number of neighbors must be equal to the number of > sides but this could give some problems with discontinuos finite > elements. It gives the same problem when calculating any sort of integral over internal boundaries. Basically the solution is, on hanging nodes, to always start from the fine neighbor when doing integration. See jump_error_estimator.C for an example. --- Roy ```

 Re: [Libmesh-users] neighbors From: Roy Stogner - 2007-06-26 21:20:04 ```On Tue, 26 Jun 2007, Lorenzo Botti wrote: > It seems to me that active elements can have non active neighbors... Is it > true? Yes. > If yes, is it correct? Also yes. If you have a mesh like this: *-*-*---* | | | | *-*-* | | | | | *-*-*---* Then it has one inactive element and five active elements; the coarse active element's neighbor is the inactive element. > I know than the number of neighbors must be equal to the number of > sides but this could give some problems with discontinuos finite > elements. It gives the same problem when calculating any sort of integral over internal boundaries. Basically the solution is, on hanging nodes, to always start from the fine neighbor when doing integration. See jump_error_estimator.C for an example. --- Roy ```