CRUNCH

11. What to do in the case of problems?

In the case of failure of a standard run of Crunch - the structure does not come out - try the following:

  1. Increase the number of trials.
  2. Use the option 'blessing' to calculate E-values and try to solve the structure again. This is particularly useful if the quality of your data is not completely up to scratch. In one test, collecting the results of 100 trials, using Blessing's E's six solutions were found, instead of two obtained using Normal80's E's.
  3. In the subdirectory drivers of the Crunch system a file 'deter.asc' is present. In the directory were you are trying to solve your structure create a file 'deter.use', of a similar format. You need to supply only those parameters you would like to give non-default values to. It is best not to change NNRM, INMX and NINMX in this way, as Crunch allows you to specify the number of matrices used and their orders when you start your efforts or after you've cleaned up the directory. Try different values for the following parameters:
    • IBK Default this value is calculated by Deter, based on the number of 4-byte words the program has available. This number is calculated on installation, the value you've given as the total amount of Mb present in the computer is used. If you make the value of IBK too large Crunch will run out of memory. The value you specify here must be smaller than the square root of the number of words available in Deter. Sometimes smaller values give better results. Try something like 1500. Do not use values which are larger than de default used by Deter (c.f. the file deter.cout).
    • NRL Increase or decrease the value a bit.
    • ICRA You might like to try the value 2 here.
    • NOTF This value defines the weight of forbidden reflections in the matrix contruction process. The default is one, i.e. they are treated as valuable compared to very weak reflections. Sometimes treating them as statistically very weak works better: Change the value into zero. Changing the value into two is another option, increasing their importance.
    Try all this using the option
    crunch code deter 1 1
    
    Inspect deter.cout and try to maximize the average E-value in the matrices and/or the number of strong, independent reflections.
  4. For larger structures it is sometimes useful to use matrices of smaller dimensions than the default chosen by Crunch. Type
     
    crunch code clear
    
    Next, type
     
    crunch code deter 1 1 
    
    and now, when the system asks if you would like to use default values for the matrix construction, type 'u'.

    Choose a different, preferably lower, number of matrices and/or a different order for the matrices. The minimum number allowed is 1. The number of matrices you may use is prescribed, not just any positive number will do. If you type a number here which is unacceptable Crunch will substitute the nearest acceptable one. The order of the matrices you may choose to be any ODD number.

    Try to choose your parameters to fit the requirements. E-average should be high. The number of strong, independent reflections should be at least about 2.5 times the number of independent atoms you are looking for. A significant number of symmetry equivalents - including Friedels - should be present. If you succeed in finding a set of matrices which looks promising, - high average E, enough strong independent reflections while containing a sufficient amount of redundancy - try to solve the structure again using the 'options' try or 'collect'.

  5. Generate about a hundred solutions. Look in the file 'hits' for the solution which yields the lowest value for R2. Type:
    crunch code recycle n n+10
    
    or use 20 in stead of 10. n is the number of the trial chosen. If this does not work and other trials looking better than most others exist, try them in the same way.

If all this does not work give up. Crunch will not solve this structure, at least not for you. If you like, contact the authors, giving full details of what you have done so far.


Back to the Crunch page

Previous

Next

Table of Contents