The stepwise SHELXL / XP refinement process

The stepwise structure determination procedure explained in this chapter combines two major tasks: The building of a plausible chemical model by interpretation of electron desity and, on the other hand, the mathematical refinement of the structural parameters (given by the model) to optimize the fit of the model to the experimental data. The single steps are normally carried out in the following order:

1. Assignment of all atoms except for hydrogen atoms (see previous chapter)
2. Correction of wrongly assigned atoms (if necessary)
3. Refining the anisotropic displacement of the non-H atoms
4. Identification, placement and refinement of the hydrogen atoms
5. Looking at special structural features like crystal packing and hydrogen bonds, if present
6. Documentation of the structure determination with structural pictures

After each single step of model editing in XP you will use the refinement program SHELXL. Therefore at certain stages of the refinement you will have to

1. edit the instructions file (INS) for SHELXL
2. analyze the SHELXL results

In our tutorial, the mathematical refinement of the model (just saved at the end of the previous chapter) will be the next task to do, so first the special features of a refinement instructions file will be explained.

The INS file
Your most recent file after the structural interpretation done with XP (up to a very advanced model stage in our case) is momo-new-1.ins. It includes: 
 1. all general crystal information, e.g. the cell parameters and symmetry operators (see explanations in the XPREP chapter) 
 2. the current structural information, i.e. all assigned atoms so far, found in the lines after FVAR 
 3. some special refinement commands, written automatically by XP - replacing the now obsolete phasing instruction TREF (see file momo-new.ins

Open the file with a text editor: nedit momo-new-1.ins &

The refinement instructions
The new instructions found after the UNIT line are L.S., BOND, FMAP, PLAN and FVAR. The purpose of the first four is written in the same line, as you can see in the picture. For this, we have used comments starting with the exclamation mark (!). 
Note: Comments written as a complete line start with the REM keyword, whereas comments appended to a normal intruction line are defined by the !. Generally, comments are recommended, since they can make the use and the purpose of instructions or parameters more understandable.
FVAR is an instruction managing all free variables of the refinement. More about the concept of free variables will be explained later - at the moment there is only one value of 1.00000 anyway. This first FVAR value has always the function of the overall scale factor (OSF), indicating how calculated structure factors are scaled on the experiental ones. During the first refinement calculation, the OSF will change a little.

The atom lines
The lines after FVAR start with atom names being simple labels by which the user identifies the type of element. However, whether an atom labeled e.g. 'O1' will actually be refined as an oxygen atom, is not defined by the name, but by its SFAC number given in the second column. This number refers to the SFAC line, where the element names define the scattering factors used by SHELXL. Oxygen (O) is listed in third place of that line, so number three is the SFAC identifyer for oxygen, no matter how the individual atoms are called in the respective lines. 
The fractional coordinates (x, y, z) of the atoms are found in the next three columns (3-5).
Column six contains the SHELX code for the occupancy of the atoms. What does this mean? Well, normally, you will find each unique atom exactly once in every asymmetric unit of the crystal unit cell. This is the case for our structure. However, if an atom lies exactly on a rotation axis, for example in space group P2, thus 'with one half' in one asymmetric unit and with one half in another one, the occupancy has to be 1/2 = 0.5. Such a case is called a special position of an atom. Another crystallographic phenomenon responsible for reduced occupancies is structural disorder. This means that statistically, atoms may not be found in every unit cell, but only in, say, 57% of them. For the structure determination, which leads to an 'averaged model' of the structure, this means that 57% of the atomic electron density would be present in the mean of all unit cells. The disordered atom would have the occupancy 0.57. But how are the occupancies listed in the INS file? You read 11.00000 as 1 * 1.00000 where the second part is the occupancy as explained before, and the first number is a factor that refers the occupancy to the FVAR value one, which is the overall scale factor. As long as other, user-defined free variables are not used in this context, the numbers in column six always start with one, so that an atom on a two-fold special position would get the number 10.50000 and a 65% disordered atom 10.65000.
Finally the displacement parameter U is listed in column seven. It indicates how well the electron density is defined around the point positions of the atom centers. Thermal motion of atoms, unrecognized reduced occupancies or wronly assigned atom types (heavier than they really are) can be responsible for less-defined electron density represented by larger U values than normal. Here, a global starting value of 0.05 has been given to all Qs, because they have not yet been refined as real elements.

Starting the refinement job

According to the upper picture, you should set the number of refinement cycles (L.S.) to 8. This is the only change to be made, except for the comments that you may wish to add to your instructions. Do not forget to save the INS file with the text editor option file->save, keeping its name. Then, leave the editor with file->exit.
Just like SHELXS before, also SHELXL needs two files of the same name, the INS- and the HKL file. The experimental data are of course not affected by any model changes, hence you always have to use the same HKL file. But once you update the INS file name for a new step, you also have to change the HKL name accordingly. 
There are several ways to update a HKL file name without changing the original content. The probably easiest way (also saving most disc space) is to simply rename the file, in our case type the Linux command mv momo-new.hkl momo-new-1.hkl. Now start the program SHELXL as usual with the common file name, typing shelxl momo-new-1 to the terminal. 

This picture displays a part of the terminal output messages of SHELXL. The program additionally creates a RES file (momo-new-1.res), containing the model with updated parameters (x ,y, z, U) and new difference electron density peaks as result of the refinement calculation. The second output file ist of LST type (momo-new-1.lst), in which detailed diagnostic and other information is found for every cycle of the refinement.
You can regard the terminal output as a very short summary of the LST file, listing only the most important quality criteria for each of the 8 L.S. cycles.

Parameter refinement and R-values
To rate the listed quality indicators, it is essential to understand the basic principles of least squares structure refinement The whole concept is based on the comparison of experimentally measured structure factors F(obs) with those calculated from the current model, F(calc). The latter can be calculated together with phase angles Phi(calc) by Fourier transformation from the model coordinates and atomic scattering factors. The idea is, that an optimal model should generate structure factors as similar to the measured ones as possible. The experimental data stored in the HKL file is in principle regarded as absolutely 'true' (ignoring experimental errors) and taken as reference. Hence during refinement, the model has to be fit in such a way that the difference F(o)-F(c) is minimized. This is done by the method of least squares applied to the structural parameters x, y, z and U (plus some more later).
All refinement quality indicators refer to this structure factor comparison, the most important of them, R1, is for example defined by the quite simple formula R1 = S | |Fo|-|Fc| | / S |Fo|, and would ideally become zero, once no more differences exist. R1 is the value cited after completing the structure determination to describe its final and overall quality. SHELXL lists R1 at the end of the output (5th line from the bottom - here: R1 = 0.0957 for the 4720 strong reflections with Fo being at least four times as big as its error). The indicator wR2 is defined in a similar way, but refers to squared F values (which correspond to measured intensities, as explained before). Because of the squares, wR2 is always greater than R1, normally at least twice as high, 0.2823 at the end of the current SHELXL job. A third important criterium is the GooF value. It does not only take the F difference into account, but also the number of reflections and parameters used. The GooF should have a value close to one at the very end of all refinement steps. Besides those indicators, SHELXL also lists the greatest parameter shifts of each cycle, e.g. a positional shift of 0.031 A for Atom C6' and a U-value shift of -0.016 for C6 in cycle one. This information is especially interesting in case of a problematic refinement, because it indicates atoms that have been wrongly assigned or placed. Normally, like in our case, the calculation by the L.S. method reaches a (global) minimum, and the parameter shifts should converge to zero, if enough cycles are given.

Difference electron density
The refinement of existing atom parameters does not only optimise the present model, but does also enable the calculation of an updated difference electron density map, based on the difference F(o)-F(c) at the end of all cycles. This map reveals if (and where) the model is still incomplete. At positions, where atoms are missing, the electron density calculated from the model (F(calc), phi(calc)) is evidently smaller than the density based on the experimental data (F(obs), phi(calc)) and a positive difference density results. As you know, the difference peaks are displayed as new possible atom locations Q in XP. Is, on the other hand, an atom placed at a position, where actually none should be present, a negative difference density will be calculated (which however is not displayable in XP). There is a general problem, not to be discussed in more detail within the scope of this tutorial: The model phases phi(calc) have to be also used for the calculation of the experimental density. This is of course due to the fact, that experimental phases do not exist (at least in case of a small molecule data collection) ... one could say, that the phase problem 'strikes back' at this point. Because the theoretical phases are used for both densities, the resulting difference density map is biased (over-influenced) by the model. Even a wrongly determined model therefore tends to be artificially confirmed by the difference map to some extend. The model bias phenomenon becomes especially relevant in case of experimental data lacking resolution and quality (luckily, small molecule cases are rarely affected). 

Analyzing and editing the refined model
The new RES file can be displayed and modified in XP the same way as the first one after the SHELXS job. 

Type xp momo-new-1 to the terminal. As usual, the first XP command is fmol in order to create the model connectivities from the coordinates. Next, have a first look at the model by typing proj

As you see, the connection of the new Qs to the old atoms is somewhat confusing. The main reason is that many internal bonds are assumed by XP due to very close positions. Since the current model lacks only hydrogen atoms making single bonds, it is justified to assume that the Qs should only bond to one model atom each. 
The command prun reduces the number of remaining bonds to the n shortest for all atoms named. Type prun 1 $q in order to reduce the number of bonds to one for all Qs - only the bonds to the closest non-H atom should remain. A new proj display reveals the possible hydrogen atom positions in a much clearer way: 

At many of the expected hydrogen positions you can observe Qs. Only because of the PLAN instruction (INS file) being limited to 20 peaks, you do not see them all.
Still, something else is striking: The presence of four peaks, Q1 to Q4 close to the atoms C6 and C6'. These peaks do not seem to be hydrogen atoms, because they are too close to the carbon atoms and the geometry appears to be strange. Remember also that C6 and C6' have been the two atoms with the largest parameter shifts. Both the peaks and the shifts are good arguments for a closer look at the apparent problem and its possible solution. 
Type info to list all atoms and Q's. 
The peak values for Q1 to Q4 are much higher than for the rest, so these peaks can obviously be distinguished from normal 'pseudo' hydrogen atoms. 

Looking at the atoms and their refined U-values, C6 with U = 0.007 and C6' with U = 0.008 also seem to be 'irregular', because the normal range of U-values is 0.021-0.028. 

U-values are quantities to describe the electron (density) distribution around atom centres. Thus, a better way than just giving the number is to calculate and display graphical objects from it. Using the model orientation from the last proj operation, you can show the atomic displacements graphically by creating an ORTEP plot
Type telp 0 -50 0.01 less $q to generate the ORTEP plot for all assigned atoms. A name has to be given for the picture to create, type e.g. test

As just said, ORTEP plots visualize the electronen density distribution around atoms according to the refined U values. The second parameter of the TELP instruction is -50, the minus activates the ORTEP display mode, the value 50 specifies an electron probability of 50%, which is the standard for thermal ellipsoids. The remaining two parameters are rather unimportant: 0 is for a non-stereo picture, 0.01 gives the relative bondwidth in the model display. 'Less $q' excludes the Qs from display as usual.
As you see in the picture, the electron distributions have a spherical shape. This results from the fact, that only one displacement parameter U(eq) has been refined so far. You call this a isotropic U-value refinement. U(eq) is just the radius of the sphere describing the isotropic distribution. At this stage of refinement, directional atomic displacements cannot be seen yet, but the ORTEP plot helps checking the correctness of atom assignments. Note: If a model atom is lighter, i.e. has less electrons than the real atom, then the U value attempts to compensate this by refining to an unreasonably small value, which concentrates the electron density. Similarly, too heavy model atoms lead to unreasonably big ellipsoids.
In our case C6 and C6' are responsible for unreasonably small ellipsoids, because they really are oxygen atoms (which you also would expect chemically).
Finish the TELP/ORTEP display by pressing the escape key. 
If you want to change just a few atom types by renaming the labels, it is faster to do this by single text commands than by starting a graphical pick session (see above). The XP command name [old atom name] [new atom name] is the appropriate instruction for this, it works just like the Linux mv command. Note than renamed atoms (both with pick and name) automatically get the correct SFAC number when saving the model into a new INS file.
Use the two commands name C6 O6 and name C6' O6' for a fast renaming. Then, erase all Qs ( kill $q) - this time you don't need to place new atoms from Qs. Save the new model with file momo-new-2, as RES source momo-new-1.res is given.

Anisotropic U-value refinement
The two new O-atoms are the only model changes in the new INS file. Although visible, hydrogen atoms still are not placed. Now, that all non-H atoms have been correctly assigned, it is first necessary to refine their anisotropic displacement parameters (ADPs).
Edit the new INS file: nedit momo-new-2.ins &. Insert the SHELXL instruction ANIS into the text, as shown in the following picture. In priciple, the order of instructions within the INS lines is flexible; to make sure the program runs properly, new lines should however be placed in between the instructions UNIT and WGHT.

ANIS tells SHELXL to refine six displacement parameters instead of one. These 6 parameters describe both the shape of the electron density distribution (which now can become really ellipsoidal) and its orientation relative to the 3 axial directions. The anisotropic distribution is often based on atomic vibration alone. Hence, by looking at the ORTEP plot you can tell whether an atom vibrates in a particular direction (the thermal ellipsoid has a cigar shape) and - by the orientation of the ellipsoid - also in which direction it vibrates. 
Additionally, set PLAN 20 to PLAN 40, thus enabling the display of more hydrogen peaks. Don't forget to save the INS file and to update the HKL file name to momo-new-2.hkl. Finally, start the next SHELXL job: shelxl momo-new-2

Looking at the output, you see that all quality indicators significantly improve (R1 = 0.675). This due to the much more realistic description of the model with ADPs, but also due to the corrected atom assignment. 
Now type the usual commands for XP:

xp momo-new-2
prun 1 $q


Now almost all hydrogen positions can be identified by the peaks. A look at the ORTEP display using 
telp 0 -50 0.01 less $q
confirms the oxygen atoms at positions 6/6' - the ellipsoids have a reasonable size compared to the rest. All ORTEPs still have quite a spherical shape, so there are apparently no exceptional atom vibrations. 

Bond lengths and angles
From the crystallographic point of view, bond lengths and angles are only derived quantities, resulting from the atomic positions (x,y,z). For the structural chemist however, they are the most important molecular characteristics aimed at. In our case, the bond lengths have already been helpful for the question whether carbon or oxygen is more likely at position 6/6'. Now, they can indirectly confirm hydrogen positions, because even if no Qs were at certain positions, you could prove then with help of the carbon atom geometries.
The XP command bang [atom name] lists the lengths and angles of all bonds to neighbouring atoms for any specified atom - or for all atoms if no name is given. For example it would be a good idea to apply this command on the atoms C3 and C8 to deduce the hybridization of these atoms as well as their bond partners. This will also enable us to determine all hydrogen atoms connected to the respective non-H atoms.
Type bang c3 c8 in order to list all bond lengths and angles belonging to these two atoms. 

You read the 'bang' list like a crossword table: The bond lengths are always found in the first column after the two atom names specifying the bond. In the next columns the angles between the respective atom pair and a third atom, written at the bottom of the angle column, are listed. An example:
1.523 A for the bond between C8 and C5 (1st line, 3rd column of the table for C8) and 
109.3°; for the angle between the bonds C8-C10 and C8-C5 (3rd line, 4th column).
In order to interpret the given values, you need to compare them to theoretically expected ones, as found for example in the International Tables of Crystallography. According to that source, C8-C5 and C8-C10 are single bonds between two sp3-C atoms, C8-O9 is a C-O single bond, C3-C2 is a single bond between two sp2-C atoms, C3-C4 is a C-C double bond and finally, C3-O7 is a single bond with a C-sp2 atom. All these results are also consistent with the experimentally found hydrogen peak positions. This is the chemical structure formula of our compound: 
Note: We didn't look at the angles explicitly, which are, by the way, agreeing quite well to the proposed geometry. Angle O7-C3-C4 is widened and C2-C3-C4 is narrowed compared to the expected 120 degree of a sp2 carbon atom. However, this is not unusual for a 5-membered ring.

Placement and refinement of hydrogen atoms
You would probably expect hydrogen atoms to be placed and refined the same way like the other atoms before. At least for C-H, this is not the case. In general, the 'free' refinement of H-atoms is problematic, as it is only a single x-ray active electron you are looking at. Its position cannot be determined exactly in many cases. On the other hand, the geometries of C-H-bonds are often regular (and not very interesting), so that the assumption of ideal geometry is sufficient. In other words: The Qs at this stage are only taken to indicate the presence of hydrogen atoms, but are not further used to place them. Instead, you use the SHELXL instruction HFIX in the next INS file. 

The model has not been modified at all this time, so that you can leave XP (quit) without saving. Open the latest RES file: nedit momo-new-2.res & and insert the instruction TEMP -140, the HFIX lines, and the LIST 3 instruction, as shown in the following image: 


HFIX tells SHELXL to calculate geometrically ideal positions for the hydrogen atoms (depending on the carbon atom type they are bound to). The experimental temperature given with TEMP -140 is needed for the correct theoretical C-H bond lengths, taking librational motion into account. The refinement will then be applied isotropically, using the so-called riding model. This means, the H-atoms are positionally shifted, parallel to the C-atoms they are bound to. Also the isotropic displacement parameter is coupled to the one of the C-atoms, having the 1.2- or 1.5-fold value.
The syntax of the HFIX instruction is HFIX code name, where the 'code' consists of two parts: First a number (m) for the geometrical group type, e.g. 1 for tertiary CH, 2 for methylen CH2, 3 for methyl CH3, 4 for aromatic CH. The first three types are present in our molecule. The second part (n) of the HFIX code determines the way the refinement is done, type 3 for a normal riding model is applied in most cases. So in principle, you could use the HFIX codes (mn) 13, 23 and 33. If the data are very good, it is worth using a different model for the methyl groups, additionally refining the rotation position (conformation). Due to the rotable single C-CH3 bond, the exact position of the hydrogens is not fixed and should not be pre-determined by the geometric model. Instead, a free angle refinement is combined with the riding model. If there are localizable peaks (Qs), the refinement will fit this optimal position. The HFIX for such a methyl group refinement is 137 (m=13, n=7). 
The oxygen-bound hydrogen atoms are completely ignored for now. Their distance to the oxygen atoms is more interesting, especially with regard to possible hydrogen bridges. We will take care of them later. To examine the electron density distribution around the oxygen atoms more closely in the next XP session, it is necessary to create a special file at the end of this SHELXL refinement. The instruction for the INS file is LIST 3 and the output file will be of FCF type. It closely resembles a HKL file, but also containing the reflection phases (given as complex vectors instead of angles) needed for a contoured electron density map display.
A remark about the ANIS and HFIX instructions: You maybe have wondered why the ANIS line has disappeared after the last refinement (the same will happen to the HFIX lines). This is because both are provisory instructions, changing the information content of the atom lines, starting with the first RES file after their application. ANIS extends the atom lines by 5 additional displacement parameters (they actually end with an equal sign (=) which means that they are continued in the next line). HFIX creates new atom lines for H-atoms, included in a AFIX group directly following the non-H atom they are coupled to. After the changes are made, all following refinements are automatically done anisotropically (because SHELXL finds 6 ADPs) and with the model hydrogens (because SHELXL finds the AFIX groups). The ANIS and HFIX instructions are not needed anymore. The benefit of them is that they save the user a lot of work in typing and calculating. 
Check whether the atom names given in the proposed HFIX lines are correct (I.e. compare with atom name and CH-group type in the XP model). If you agree, save the RES file as new INS file using the nedit menu: File -> Save As -> momo-new-3.ins. After updating the HKL file name as usual, start the SHELXL job: shelxl momo-new-3. After that, don't start XP imediately, but have a look at the LST file first: nedit momo-new-3.lst &
The very first part of every LST file is a complete echo of the INS lines, then the connectivity table and some data statistics are given. After that, you find the relevant part about the hydrogen atom refinement (only in LST files where HFIX has been introduced in the current refinement step):

First, the temperature-dependent libration effect is mentioned, then a difference electron density profile for the '360° cone' around the methyl carbon atoms is listed. Within the 15° steps, always 3 maxima for the density are found, e.g. 87, 59/63 and 109 for the methyl group at C19'. Indeed, these peaks are about 120° apart in projection (8 x 15°), as expected for three hydrogen atoms. Should such a profile not be observed, the simpler instruction HFIX 33 would be preferable to 137.
Close nedit, then start and setup XP as usual:
xp momo-new-3
info $q
kill q7 to q40

Creating electron density maps
You hopefully recognize that the 6 highest Qs correspond to the remaining hydrogen atoms bound to oxygen, which have not yet been assigned. To further illustrate the difference density distribution, the density map will be created from the FCF file - giving a realistic contour line profile (whereas Qs only indicate the point positions of the peaks). Unfortunately, a 3-dimensional view on the map is not possible in XP and only a 2-dimensional planar section can be displayed.
Our example takes peak Q1, which is close to O9. Also C8 shall be displayed in the resulting image. We prepare the picture in a way that you will look perpendicularly on the bond O9-Q1.
Therefore, remove all atoms except for the ones just mentioned and make the plane built by these 3 atoms the view plane (this is done with the mpln command):
kill $h $q less q1
kill $o less o9
kill $c less c8

Now, create the electron density diagram by typing the command eden and giving some necessary parameters you are asked for by the program:


Type the values shown in the picture. Most of them are defaults that can be confirmed by pressing the return key. When asked for, give the name of the FCF file (type: momo-new-3). Reading it, XP calculates the density map on which the diagram will be based. F(000) is the total number of electrons per unit cell or in case of a difference density map the sum of all electrons that have not been assigned. We assume 6 missing electrons (= 6 hydrogen atoms) per asymmetric unit, so with four of them (space group P212121) F(000) becomes 24. This number is needed to scale the following contour lines.
Now, the contour line generation starts. All the user needs to set is the height of the contours. The values should be distributed equally between 0.00 and the maximal value of a section (here: 0.54), the number of lines is free. After having finished the contours, the atoms may be labeled the same way like in a telp plot (as usual, break with 'b' or with '/'). 

Now rename the 6 Qs to H-atoms (you should assign numbers corresponding to the O-atoms). Save the same way as usual, insert the instruction HTAB into the new INS file (nedit), then carry out the SHELXL refinement: 

name Q1 H9
file momo-new-4
nedit momo-new-4.ins &
mv momo-new-3.hkl momo-new-4.hkl
shelxl momo-new-4

Hydrogen bonds and crystal packing
HTAB causes SHELXL to look for possible hydrogen bonds. Such a bond type is assumed, if O-H -- O distances are shorter than a certain value and the angle at the H is close to linear. Before looking at the hydrogen bonds, we will first check the present O-H bond lengths in XP. 
xp momo-new-4
kill $q
bang o7 o9 o11

The values are within a reasonable range from 0.85 to 0.94 Angstrom. In x-ray structures, bonds involving H-atoms are shorter than the actual atom core distances, because the single electron is usually shifted towards the bond partner (polar bonds). For O-H, values below 0.8 A are not unusual, but in the case of hydrogen bonds, the bonds are weaker and therefore longer. This is also the case here: the values found are a first hint for hydrogen bonds. If you had observed bond lengths shorter than 0.7 A or longer than 1 A, or extremely differing lengths for equivalent bonds like O9-H9 and O9'-H9', it would have been justified to apply restraints in order to force the values towards a chemically reasonable value.
After HTAB has been used, the program analysis of hydrogen bonds is found in the LST file, more precisely almost at the end, just before the list of peaks: 

The table shows that all hydrogen bonds are intermolecular, and that they do not connect two molecules of the same asymmetric unit, but always symmetry equivalent molecules. For example, the expression O9 [x+1, y, z] tells you, that the acceptor atom O9 belongs to a molecule shifted by one unit cell along the x axis direction.
There is a second way to use the HTAB instruction in the INS file. Its purpose is to include the hydrogen bonds previously found into the connective table as normal bonds. This enables the listing of those bonds in the output files and their display in pictures. Besides typing the new HTAB lines, you also have to define the symmetry equivalents involved. This is also done in the new INS file, the instruction is called EQIV
Close XP (if still open) without saving and edit the latest RES file (to a new INS file), inserting the instructions EQIV, HTAB, SIZE and ACTA, as shown in the picture below. Then, carry out the refinement. 

nedit momo-new-4.res &
SIZE 0.5 0.5 0.4
file -> save as -> momo-new-5.ins
mv momo-new-4.hkl momo-new-5.hkl
shelxl momo-new-5


Now some explanations about the new instructions: EQIV defines a symmetry operation using a variable, e.g. $1. The expression of the coordinate transformation applied is in principle analogous to the SYMM instructions of the space group or to the HTAB expressions in the LST file (as just seen). The HTAB instruction is given for each H-bond separately with the syntax HTAB donor acceptor, in case with the extension _$n to specify a symmetry equivalent atom related via operator $n.
SIZE is used to specify the crystal size in mm (long edge x medium edge x short edge). It is only needed in connection to instruction ACTA, which creates a text file of type CIF, here: momo-new-5.cif. CIF files are only generated at the end of a structure determination (which you have almost reached). They contain all structural information worth retaining for documentation and publication. The lists and tables in the CIF file are however written in a special keyworded syntax and look somehow 'encrypted'. This is because a CIF file is never printed directly, but either printable tables are generated from it with certain SHELXTL utilities or the CIF file is submitted to data bases and journals for publishing.
Call XP a last time - now it is the crystal packing we are interested in. The necessary XP command is called pack. To avoid a confusing display, remove all carbon-bound hydrogen atoms. Note also, that it may not be possible to recover the orginal molecules after having created symmetry equivalents with pack. If you want to save a certain state of work or display in XP (to return to it later), use the command save name (here: 'work'). Always use 'save' before 'pack'! 
xp momo-new-5
kill $q $h less h7 h7' h9 h9' h11 h11'
save work
The graphical window of 'pack' is similar to the one of 'proj' - again you find menu buttons on the right. The symmetry equivalent molecules within a given packing radius are shown in red-blue stereo display (used with simple 3D glasses to improve the packing depth visibility). The crystal unit cell is shown as well, and hydrogen bonds are drawn automatically. This display is not rotable, but you can include a selection from the pack mode into the connectivity list, thus diplaying it with proj later. You can use the SCAN buttons to select molecules, atoms and bonds sequentially - like in the 'pick' mode, they are blinking if selected. With [space] you proceed to the next object, with [backspace] you go back and with [return] you delete a selected object - it is often a good option to delete some molecules hiding others, so that the display gets clearer.
Click the SGEN/FMOL button (do not 'SCAN'). The current molecules are added to the atom and connectivity list. Leave the packing mode and type the following two instructions:

matr 1
proj cell


The proj window should look more or less like in the picture. Matr is a command that sets the viewpoint relative to the cartesian axes. In a rectangular spacegroup like yours, a single parameter after matr corresponds to the respective cell edge, so 'matr 1' will let you look along cell edge a. The well-known command proj does also allow the parameter cell, so that the unit cell is drawn together with the model. (The same is possible for 'telp', too).
After having displayed the packing of several molecules, the exemplary hydrogen bond H7' O9_$1 shall be displayed alone, so only two molecules have to be shown, of which one is shifted by cell edge length a (1 0 0). This task will involve some special commands and more steps than simple operations.
There are several ways to prepare the display, but the use of the SGEN command is probably the best way in a crystallographic sense. SGEN applies every allowed space group symmetry operation, including cell shifts. XP knows the allowed operations from the SYMM lines of the RES/INS file. You can list them within XP using the same command name: SYMM. For each coordinate transformation, the SGEN code is given, e.g. SYMM -> 1555 = +X, +Y, +Z. The first number (1) is the operator number, 1 is always refering to the original coordinates without transformation. The remaining number 555 is specifying the current cell, 655 would be a neighbouring cell (plus one edge length a away), 444 would be a cell shifted by minus one cell edge in all directions, and so on. For your task you need SGEN 1655, or even better SGEN -1655. The minus causes the old atom positions to be deleted, so that the number of molecules will not be doubled upon transformation. Still, a problem remains: The complete content of the asymmetric unit will be transformed, and not just one of the two molecules. Therefore you have to bother specifying all atoms that shall actually be shifted. Their labels are given as second sgen argument after the code. Luckily for the second molecule, all atom labels end with an apostroph. XP knows the ? placeholder (see linux tutorial), so you can adress all atoms of molecule two using the expressions ??' and ???'
Type in the following commands:

next work
sgen -1455 ??'
sgen -1455 ???'
join 7 h7' o9a
join 7 o9' h7a

Again, this is quite a lot of commands. Here is some explanation:
After the pack procedure, the whole work has not to be done again, because we have saved before. With next, a saved state can be recovered, 'work' in our case. If you wonder about the sgen codes, note that for your special case it is smarter to shift molecule two by -a (455) than to shift molecule one by +a (655). This is because the selection argument is much shorter for the second molecule, where you can use the placeholders.
Finally, the command join binds any pair of atoms given - no matter how far they are apart. Parameter 7 causes proj to draw the bonds in dotted lines. The result should look more or less like this: 


Creation and printing of pictures
Depending on the number of pictures you have finished with 'telp' and 'eden' (i.e. not aborted with ESC), some PLT files should already exist in the user directory, e.g. q1-fo-fc.plt. These files were created by XP automatically and can only be displayed in XP. To look at existing PLT files use the command view name (e.g. view q1-fo-fc). However, these files are not printable and have to be transformed to PostScript files (or similar formats) for program-independent use. Applying the draw command, you have to give some options (A for Adobe PostScript, next the name of the target PS file - usually identical to the one of the PLT file - and finally option C for a color picture). 
Type the following commands:

view q1-fo-fc
draw q1-fo-fc


Now leave XP, look at the PS file with program GhostView (if installed) and print it on the color laser printer:

ghostview &
lpr -P

(the last step is for internal tutorial users only!) 

Adjustment of the weighting scheme
Congratulations, you have finished (the manual part of) the structure determination! But to complete the refinement properly, one more important feature has to be taken care of: The weighting scheme. In SHELXL, the internal weighting parameters g1 and g2 affect mainly the wR2 and GooF value. These two parametes are controlled by the WGHT instruction in the INS file. Until now they have been constantly set to g1 = 0.10000 and g2 = 0.0000, in order to enable a better convergence of the parameter refinement. The recommended values for WGHT are always found in the latest RES file, directly after the END instruction: 

Note: Lines following the END instruction are ignored during the refinement.
The update of the WGHT parameters is done quite easily: Open the RES file with nedit, copy the recommended line to the place of the regular WGHT instruction and delete the old one (or comment it out writing a REM in from of it). Do not forget to save the RES file as INS file, then refine gain. Note that new file names with higher numbers for INS/HKL are not necessary for this last step. 

This procedure is repeated until the recommended value is (almost) identical to the refined one. This may take some steps (about five). You can avoid the trouble of doing all the operations over and over by applying the small utility script WFIT - (c) Thomas Pape - which automatizes the repeated update and refinement: 
wfit momo-new-5

And this is the end of the structure determination and the tutorial. Goodbye! 
back to start