SHELXT - small molecule structure solution

The new small molecule structure solution program SHELXT extracts the Laue group, cell dimensions and types of element present from a SHELX format .ins file, solves the structure using data expanded to space group P1, and then uses the P1 phases to find the space group. In general it produces a more complete solution than SHELXS or SHELXD and places the correct absolute structure optimally in the unit cell. It ignores the formula specified on the UNIT instruction and may even add elements if that appears to be necessary. If the data are not as complete as required for a Fourier based method like SHELXT (e.g. because a high pressure cell was used) SHELXT simply invents the missing data (the free lunch approach). SHELXT is highly parallel but the number of CPUs used may be set using the -t command line option. SHELXT requires two input files: an instruction file name.ins and a reflection file name.hkl, and outputs possible solutions to name_x.res and name_x.hkl (x=a,b,c...). A summary is output to the console and a full listing to name.lxt. SHELXT is started with the command line:

shelxt name

Exactly the same input files may be used as for SHELXS or SHELXD, but all instructions between UNIT and HKLF are ignored. Many graphical users interfaces (GUIs) and also the Bruker XPREP program may be used to create these files. shelXle and other GUIs can also call SHELXT directly. The Bruker version is called XT (or intrinsic phasing) but is otherwise identical to SHELXT.

The Laue group deduced from SYMM can be overridden with -L followed by the SADABS number for the Laue group (no intervening space); note the special options -L15, -L16 and -L17. SHELXT reads F-values rather than intensities if HKLF 3 is given in the name.ins file. Only HKLF 3 and HKLF 4 are allowed but the HKLF card may also include a reorientation matrix as for SHELXL. It is assumed that the SFAC card correctly specifies the elements present, but the numbers on the UNIT card are ignored. If the program thinks that a heavy atom has been accidentally left out it may add bromine or iodine (orignally this was an inert gas, but that would have led to a substantial increase in the number of krypton structures in the CSD and COD). The program uses the integrated electron density around an atom to assign the element type. This is not very reliable, especially distinguishing between C and N, so the assignments should always be checked carefully; the element assignment works best with high quality complete data. SHELXT is no substitute for elemental analysis!

SHELXT may stop when it thinks that it has a reasonable solution unless -a is set to force it to try all space groups in the chosen Laue group. It is advisable to input unmerged data (e.g. straight from SADABS), otherwise there may be problems with the Flack parameter. The program will automatically 'invert' a solution if the Flack parameter is greater than 0.5. Although the Flack parameter is not very reliable at this early stage, there has not yet been a report of a case where it was necessary to invert the structure later.

The basic assumption made by SHELXT is that the structure consists of resolved atoms. This is a very powerful assumption and it enables the program to work with incomplete data from a high-pressure cell or where the crystal did not diffract to high resolution. However if the assumption of resolved atoms is not appropriate - e.g. in the presence of twinning, severe or whole molecule disorder or modulation - other methods such as charge flipping may be a better choice.

Command line switches for SHELXT

The following list of command line switches is output when SHELXT is started without a data file. Default settings are given in square brackets.

-L Laue group N (SADABS code). N=15 all hexagonal and trigonal, N=16 monoclinic with a unique, N=17 monoclinic with c unique. If -L is not set, the SYMM instructions set the Laue group.
-tN use N threads, otherwise use 4 or max available, if less. This applies only to the P1 solution stage, for the other calculations all available threads are used.
-d highest resolution to be employed [-d0.8].
-e fill out missing data to specified resolution [-eX where X is max(0.9,d-0.1) and d is the resolution of the input data after possible truncation with -d].

-q structure factors Go=EqFo(1-q) [-q0.5].
-iN NGo-(N-1)Gc map in dual space recycling [-i3]
-o switch OFF Patterson superpostion (not recommended).
-kN apply random omit every kth cycle [-k3].
-fX randomly omit fraction X of atoms [-f0.3].
-z sigma threshold for P1 peak-search [-z2.5].
-uX tangent expansion for E>X after random omit [off].
-v atomic volume threshold in Ångstroms for P1 peak-search [-v13].
-m initial number of P1 dual space iterations [-m100].
-b spread factor for atom masks [-b3].
-jX CFOM = 0.01*CC - X*R(weak) [-j1].
-y CFOM = Chem*CC (alternative to default -j1) [off].
-xX accept if CFOM > X+0.01*max(20-m,0) where m is the try number [-x0.65].

Chem is a 'chemical' figure of merit that should be between 1.0 (most reasonable) and 0.0 (awful). Currently the only option (-y or -y1) is the fraction of bond angles between 95 and 135 degrees ignoring the 20% highest and 10% lowest peaks. This is only useful for organic compounds, organometallics and ligands, not for inorganics, but can be invaluable when CC and R(weak) fail to distinguish between correct and incorrect P1 solutions.

Space group determination:
-s"Name" space group (replace "/" by "_" e.g. -s"P2(1)_c") [off].
-c space group restricted to the Sohncke space groups [off].
-n space group restricted to non-centrosymmetric [off].
-w worst alpha gap for a possible solution [-w0.15].
-p maximum number of atoms in full matrix, rest are blocked [-p20].
-g smallest gap in R1 to best cent. for non-cent. SG [-g0.02].
-h halt if R1 is less that this [-h0.08].
-r radius around peak for density integration [-r0.7].
-aX search ALL space groups in given Laue group with alpha < X [off].

-a overrides -g, -h and -w, but not -c or -n; -a without a number is equivalent to -a0.3.

Problem structures:
If the default settings fail, try -y and -a. If the CC is good but the solution is a mess or if all the CC values are less than 0.87 try -m1000. Also well worth trying is truncating noisy outer data stepwise with -d. When the program produces more than one possible solution, the following criteria should be checked:

1, Is R1 low enough, taking into account that a slightly lower value may result if the space group is a lower symmetry subgroup of the true space group?

2. Has the program tried to throw in unexpected iodine or bromine atoms? This is usually a bad sign, unless they are really there of course!

3. For a non-centrosymmetric solution, a Flack parameter close to 0.5 may indicate that it is really centrosymmetric. A negative Flack parameter may be caused by a wrong element or wavelength (common for synchrotron data). However at this early stage, the Flack parameter is not always reliable.