It is a property of the compound, so you cannot do anything (OK you can try helium cryostat ;) ).

Your group would be ordered if there would be some interactions (-C-F...H-O- for example) strong enough to disallow the rotation. However if you have too much possible binding sites, you easily end up with disorder with several orientations of the group towards different sets of donors or acceptors, especially if the decrease of energy connected with such bonding is similar for several possible orientations.

Somewhat different approach is possible in those cases where you have determine molecular structure of some compound but changing solvent molecules is not an issue (relatively often when dealing with organic structures when the solvent and its interactions with yet unknown molecule is not a matter of interest). You can avoid such solvents like chloroform, hexane, petroleum ether etc. which are famous disorder creators.

Erik is right. Actually, trifluoromethyl is often disordered due to the great hardness and small dimensions of F atoms, that prevent them from being available for strong non-covalent interactions. At the same time, this may allow (almost) free rotation along the C-CF3 bond. If this is the case, you should observe dynamic disorder - that is, a further reduction of temperature should make your structure less disordered or even imply its complete ordering. On the other hand, if your disorder is static, the lowering of temperature will not have any effect: in that case, you could act on the solvent as well as on the crystallization conditions. Granted that a 'magic recipe' does not exist, if your compound is more soluble at higher temperatures you could try for example a crystallization by cooling very slowly the solution: if you are lucky (that is, your structure has a favourable thermodynamic energy minimum that corresponds to an ordered lattice) you may be able to obtain a (more) ordered structure.

There is often the misconcept that disorder is somewhat what is "bad", i.e. is the signal of "bad crystal" and possibly could be avoided. Even some coryphei of Slovak crystallography (still alive & working) have solved only a few disordered structures (I am not talking about OD structures or whole molecule disorders and similar cases, I am talking about [almost] routine static/dynamic partial disorders). The same situation is with these coryphei and twinned structures. I think it is a heritage of their long living (~ 2000 decomissioned) Syntex P21 serial diffractometer and co-temporary computers - in that time, even with with Pentium 166 15 cycles of least squares of common polyoxometalate was nearly 15 min, not 10 seconds like today. In 1978 one run of refinement of vivianite was "some hours" on mainframe computer. So they tried avoiding any "problems" (or better - time consuming jobs?) including those problems which in fact aren't problems - the researchers concerned for measurement were forced to weissenberging of candidate crystals to avoid "lower quality" and in fact non-merohedrally twinned (in opposite, I am pretty sure that nowadays "bad" mosaicities above 1° weren't discernible from the film) crystals, and to obtain cell parameters and, if possible, complete diffraction symbol, i.e. constrict possible range of space groups. At the end they were mutilating the crystal with several weeks of roasting on the Weissenberg camera (at 600 W + water consumption) before attempting the diffractometer measurements, where the same measurement with exception of non-merohedral twinning (or multi-crystal) detection, which is problematic with serial detector (they hadn't polaroid camera) can be done within a few hours (ironically, "last man standing" of these coryphei now has Siemens serial diffractometer, but indeed Weissenberg cameras he kicked out together with Syntex). At the end, almost all "bad" crystals (I am using word "challenging" instead ;) ) were rejected prior the measurement.

Funny part is, that they had something like "unsolved structures index", i.e. their own personal rate of failure to solve the structure. I haven't access to the data (because I am not from the not-named institution), but I am pretty sure that these cases are predominantly unresolved twins (merohedral, pseudomerohedral, reticular merohedral - not detected from Weissenberg photos), disorders, weak datasets (too low I/sigma) and possibly commensurately modulated structures (incommensurates were, I think, mostly considered "bad crystals"). I think the computing time was their main restraint from solving these things, and when, approximately from Pentium era, these calculations started to be quick enough, they was too old to start with?

After introduction of area detectors, younger crystallographers started to solve these types of "problems", and for example one of our older professors was curious why there are increasing numbers of disorders and twins because of the vision of "evilness" of such phenomena ("because it was always rejected before measurement, it must be something evil what we must avoid"). I think he had conceptual problem to start thinking about disorder (and often twinning) like the property of the compound, that twins (non-mero) are (mostly) easily detectable with area detectors and it is (mostly) no problem with computations (he was surprised I solved absolute structure from multiple twin) etc.

@ Rahul

Basically, the answer to your question is "learn to refine your data". Disorder of the CF3 group is one of the most common and the simplest to handle. The refinement guidelines on this particular case are easily available.

Today, when OLEX2 and SHELX2013 are available freely, one can teach an undergrad in a week to handle the simple disorder cases. Fear not, the disordered structures are accepted by all journals. And Platon/Checkcif service will notify you if you do something wrong, you do not need a human to check the refinement of a disordered CF3.

I can only recommend this book:

http://ukcatalogue.oup.com/product/9780198570769.do

with many examples on CD from routine structures to basic types of disorders (also complicated ones), twinning, protein refinement etc.

@Ivan

Thank you sir for your suggestion. I regularly use SHELXL for solving my crystal structures. I also agree with you that with correct model disorder can be take care off and disordered structures are acceptable for publication. My question was why some CF3 group doesnot show disorder while other does.

@ Eric

Thankyou sir for providing the link. I found it really very useful.

@ Rahul Shukla

1)Reason

is that you probably have molecular crystals. Free rotation along the C-CF3 bond - it is half of explanation, next half is in existing of "pocket" in crystal structure. Van der Waals forces in area of CF3-group are weak, and more polar groups dominates in building of crystal structure. In "pocket" CF3-group from two molecules often tend bring together - check distance -CF3...F3C-

2) How to avoid

Replace weak molecular crystal on more strong ionic or complex. If you have some heteroatoms, it is not a problem.

I am not an crystallographer so accept "as is" this synthetic chemist opinion

@ Rahul Shukla

It may be useful to solve your disorder problem:

1) http://edoc.unibas.ch/1459/1/MarkusNeuburger-PhD-thesis-edoc.pdf

2) http://www.canadiancrystallography.ca/cccw/files/What%20to%20do%20with%20disorder.pdf

@Andrei Rogoza

Thank you for the link. It was really useful

I have found that the best method to avoid disorder is to form the crystal at the lowest possible temperature. I had a molecule one time that had a molecular structure that was almost spherical and the crystals that were formed near room temperature were terribly disordered. I dissolved the material in diethyl ether and placed the solution in a test tube with a rubber septum top, then placed it in the freezer and forgot about it for several weeks. As the ether diffused out of the septum top, beautiful clear hard crystals formed that had no disorder in the structure. The Ortep ellipsoids for the atoms were one tenth the size as for crystals formed at room temperatures. I did not actually collect the data at low temperature however, the data collection took place at room T.