Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision
Previous revision
rwp_jumps_continue_after_convergence_etc [2017/02/06 15:23]
johnsoevans
rwp_jumps_continue_after_convergence_etc [2017/02/06 15:30] (current)
johnsoevans
Line 3: Line 3:
 For a well behaved refinement you should normally find that the R-factor that you get when you restart a refinement is identical to the R-factor from the previous convergence. ​ If this is not the case it normally means there is an error in your input file.  ​ For a well behaved refinement you should normally find that the R-factor that you get when you restart a refinement is identical to the R-factor from the previous convergence. ​ If this is not the case it normally means there is an error in your input file.  ​
  
-The most common cause is using the "​@"​ symbol in a multi-data set refinement for a parameter that feeds into multiple data sets.  For example if you have a str that is present in two data sets and write "beq @ 1.37" the temperature factor for that site gets assigned a different parameter name for each data set and two separate values are refined ​but only one is stored in the .inp file.  The easiest way to spot this is to count the number of parameters you think you're refining with the number Topas reports.+The most common cause is using the "​@"​ symbol in a multi-data set refinement for a parameter that feeds into multiple data sets.  For example if you have a str that is present in two data sets and write "beq @ 1.37" the temperature factor for that site gets assigned a different parameter name for each data set and two separate values are refined.  Only one value gets stored in the .inp file so your refinement will restart from a different place each time.  The easiest way to spot this is to count the number of parameters you think you're refining with the number Topas reports.
  
-If you start a refinement with unit cells a long way from correct values (e.g. 4.65 A when it should be 4.50 A) then the Rwp you get on convergence may not be identical to the value you get when you re-run ​the file.  In some cases the first Rwp can actually be lower than the value you get subsequently. ​ This happens as there are a number of quantities ​that only get calculated at the start of a refinement For example, atomic scattering factors are 2-theta dependent and only calculated when a refinement is started.  ​if peaks move a lot then there values will be slightly incorrect at convergence. ​ In a normal Rietveld refinement these effects are minimal. ​ It is, however, normally sensible ​to run a refinement a couple of times to eliminate themSome of the peak shape parameters are also calculated only once if they are not flagged to refine and saved in the peaks buffer; their values might need recalculating if peak positions move significantly. ​ The same occurs for other aberrations.+second cause is when you have a user defined prm in the input file that refines to a sufficiently small value that it gets rounded in the input file to 0.000000.  ​One way round this is to just scale the prm by e.g*10^3 or *10^6 in the equations where it is used.
  
-If this is an issue then one work around is to change your input files to automatically do two convergences rather than one.  In topas v6 this can be done with:+If you start a refinement with unit cells a long way from correct values (e.g. 4.65 A when it should be 4.50 A) then the Rwp you get on convergence may not be identical to the value you get when you re-run the file.  In some cases the first Rwp can actually be lower than the value you get subsequently. ​ This happens as there are a number of quantities that topas only calculates at the start of a refinement. ​ For example, atomic scattering factors are 2-theta dependent and only calculated when a refinement is started. ​ if peaks move a lot then their values will be slightly incorrect at convergence. ​ In a normal Rietveld refinement these effects are minimal. ​ It is, however, best to run a refinement a couple of times to eliminate these issues. Similarly, peak shapes are also only calculated once and saved in the peaks buffer if they are not flagged to refine; the peaks might need recalculating if peak positions move significantly. ​ The same occurs for some other aberrations. 
 + 
 +If this is an issuethen one work around is to change your input files to automatically do two convergences rather than one.  In topas v6 this can be done with:
  
 <code topas> ​               num_runs 2 <code topas> ​               num_runs 2

Personal Tools