Blake A.J.(ed.) Crystal Structure Analysis

Подождите немного. Документ загружается.

82 Practical aspects of data collection

with such a correction, the noise created by thermal excitation imposes

a practical upper limit on the exposure time for each frame.

The steps involved in setting up and collecting data with an area

detector are broadly similar irrespective of the technology used.

6.6 Crystal screening

Although area-detector data are quite tolerant of poorly centredcrystals,

it is obviously best to centre crystals as accurately as possible. Initial

exposures can be recorded in a matter of a few seconds to give an almost

instantaneous indication of the quality and intensity of diffractionby the

crystal. It is a good idea to record frames at one or two different φ angles

as this can detect crystals that look promising in one direction but show

serious problems in another. Such exposures are taken with the crystal

in a random orientation, and either stationary or oscillated through a

small angle (Figs. 6.3 and 6.4).

At this stage, some obvious problems such as poor (or no) crystallinity,

splitting of reﬂections and overall weak diffraction can be identiﬁed (see

Figs. 6.5 to 6.9): an obviously unsuitable crystal can be quickly discarded

and a hopefully better crystal selected. Note that such exposures are

essentially two-dimensional and may not indicate all possible reﬂection

splitting or other problems: these may only be detected when a series

of frames have been collected for unit cell determination (and indexing

may have failed – see below). The rapid screening allows more crystals

to be investigated from each sample, increasing the chance of ﬁnding

a useable one, or conversely increasing your level of certainty that no

acceptable crystals exist in that sample. However, it should be noted that

because of their greater sensitivity and efﬁciency overall, CCD systems

Fig. 6.3 Single frame by rotation through a small angle (0.3

◦

6.6 Crystal screening 83

Fig. 6.4 A pseudo-oscillation photograph generated by superposition of a small number

of frames.

35000

–33–34–35

35000

30000

25000

20000

15000

10000

5000

Counts

Fig. 6.5 A broad peak (FWHM ∼1.3

◦

can often handle samples of apparently rather poor quality and still give

an acceptable structural result, so it may be worth persevering with less

promising samples: experience will tell you when to give up on your

own crystals.

A common problem occurs when you are faced with a crystal that

is of poorer quality than you would like: do you remove it from the

cold stream of the diffractometer so that you can try another crystal,

potentially losing the best available crystal, or carry on with it? This

can be important if either time or the supply or crystals is limited. One

84 Practical aspects of data collection

Fig. 6.6 Limited diffraction.

Fig. 6.7 A weak, non-single crystal.

solution is to store each crystal over liquid nitrogen after removal from

the diffractometer until it is clear which one is best. However, if you

have, even temporarily, access to twoinstrumentsyou can try aless risky

procedure whereby only the poorer of each sequential pair of crystals

is removed from the diffractometer: after a few cycles of comparison it

should be possible to identify the best crystal.

6.6.1 Unit cell and orientation matrix determination

Collecting a series of frames covering two or three small regions of

reciprocal space takes a few minutes. After obtaining the positions of

6.6 Crystal screening 85

1800018000

327

326.6 329.6

Omega (Deg)

328 329

16000

14000

12000

10000

8000

6000

4000

2000

Counts

Fig. 6.8 A split/broad/poorly shaped reﬂection.

the reﬂections, including interpolation between successive frames to

obtain precise setting angles, these are stored in a list by the control

program. This yields co-ordinates of observed reﬂections in reciprocal

space, referredto goniometer axes. With reasonably sized structuresand

moderate intensities, there may be upwards of a hundred reﬂections. It

may be necessary in some cases to adjust the criteria for the inclusion

of reﬂections in this list: this is most commonly done on the basis of I

or I/σ for more weakly diffracting crystals, where default settings may

not yield enough reﬂections. An incorrect (or uncertain) initial matrix

and cell at this stage are not a problem, as they can be revised following

the full data collection without loss of data. One advantage in obtaining

a good initial matrix and unit cell is that it is possible to check whether

the latter corresponds to a known phase: another is that it gives some

encouragement that you will ultimately be able to index and process the

full dataset.

Fig. 6.9 A powder pattern from a poly-

crystalline sample originally thought to

be a single crystal. The variation in the

intensities of the individual lines indicates

preferred orientation within the sample.

The theory behind indexing has already been covered. Indexing of the

reﬂections and determination of the crystal orientation matrix and unit

cell parameters have the advantage of usually having many reﬂections

available and hence a lower probability of obtaining an incorrect cell.

The number of reﬂections can also make it easier to obtain a result from

twins and other samples that are not single crystals, though the outcome

then needs to be examined particularly carefully to see which reﬂections

do not ﬁt the proposed cell and whether there is a clear explanation for

this. Assessment of the mosaic spread is usually also part of this step,

86 Practical aspects of data collection

and is important for decisions on how to collect the full dataset and how

to extract intensities from the raw data.

6.6.2 If indexing fails

Ifindexingfails,youshould ﬁrst examine the indexing parameters: these

place limits on a number of factors that control the indexing, including

(i) the indices that can be assigned, (ii) the lengths of the axes of the

original primitive cell, (iii) how far the indices are allowed to deviate

from integral values and (iv) the minimum fraction of data that must

be indexed by any candidate cell. Parameters (i) and (ii) can be used to

exclude ridiculously large cells, but must be reduced with care unless

you know the cell beforehand. I would suggest starting off with upper

and lower limits on cell axes of 3 Å and 60 Å, respectively, and indices

of 15 or 20. Parameter (iii) allows less well centred reﬂections to be

included: a value of 0.1 may be too low in many cases, but more than

about 0.3–0.4 is likely to generate multiple false cells. Parameter (iv)

is useful if a minor twin component is suspected and you are trying

to index the major component, but there are more sophisticated and

controllable methods for handling such cases. If indexing fails the ﬁrst

thing to try is increasing the tolerance on parameter (iii); a visual survey

of the frames may indicate whether increasing (i) or (ii) is likely to help.

6.6.3 Re-harvest the reﬂections

If manipulation of the default list of reﬂections fails to provide a plausi-

bleindexing, it may be worthwhileharvestingreﬂectionsfromthestored

frames using your own criteria: these could include modiﬁed limits on

I or I/σ , resolution or other factors. Indexing on this list may prove

more successful. If indexing still fails, have a closer look at the frames,

as examination of the rocking curves may show splitting or other effects

not obvious on individual frames, which may mean that an initially

promising crystal is in fact unsuitable. It is probably time to try another

crystal if one is available. Even if indexingdoes work,you shouldalways

examine the rocking curves: indexing routines are robust and can often

cope with split or broad reﬂection proﬁles, so a successful indexing

does not automatically qualify a crystal for data collection. At the same

time, you can check whether there is a good correspondence between

the diffraction pattern and the reﬂection positions predicted from your

orientation matrix: if the correspondence is poor the indexing must be

regarded as questionable. Too many predicted spots may indicate that

the cell is too large or that centring has been missed; too few could mean

the cell is too small. Carrying out these comparisons throughout all the

collected frames ensures that you can detect any crystal movement or

instrumental problem that might have occurred during data collection.

Such an occurrence is not usually a disaster: the data can be processed

in batches using different orientation matrices, but failure to detect the

6.6 Crystal screening 87

problem will give puzzling effects ranging from high merging residuals

to a failure of the structure to solve or reﬁne.

6.6.4 Still having problems?

If you are faced with a persistent failure to index, and the reasons for

this are not obvious, you may be able to call on a program that can

display a reciprocal lattice plot of all the reﬂections harvested from the

frames. By rotating the lattice you should be able to decide whether the

crystal is single or not. At any signiﬁcant sign of non-single character

the crystal should be discarded and another one examined; if all the

available crystals show twinning the procedures outlined later can be

adopted.

Even if indexing seems to have succeeded you should be wary of

unexpectedly large, often triclinic, cells with large uncertainties on their

cell dimensions. If a cell is allowed to be large and uncertain enough it

will be able to accommodate almost any list of reﬂections. If indexing

appears to succeed but some reﬂections have indices at or near special

values such as 0.50 or 0.33, this may indicate the cell is too small in that

direction and should be increased by a factor of two or three, respec-

tively. If such a situation is unclear, it does not need to be resolved now,

but can be investigated when the full area detector dataset is available.

If indexing has not worked, but there is no obvious reason for this, it

may help to acquire some additional frames, preferably from a region

of reciprocal space that has not yet been sampled. Because a valid ori-

entation matrix is not mandatory for acquiring area-detector data, these

additional frames can consist of a full data collection, in the hope that

indexing will be successful when a selection of all available reﬂections

can be used, but runs the risk that the indexing will still fail, giving a set

of frames that cannot be processed.

6.6.5 After indexing

Successful indexing is followed by least-squares reﬁnement of the ori-

entation matrix and a Bravais lattice determination. The reﬁnement also

acts as a check on whether the indexing is valid: a plausible indexing

that fails to give a satisfactory reﬁnement must be discarded. In order

to be accepted a reﬁnement must end with the vast majority of signif-

icant reﬂections having (near)-integral indices, reasonable s.u.s on cell

parameters and suitable values for various quality indicators. It may

not always be possible to be certain about the Bravais lattice.

6.6.6 Check for known cells

There is probably little point in continuing if the cell from the index-

ing is known, either in the literature or within your own group or

department.Anexcellent facility (CrystalWeb) forsearching the relevant

crystallographic databases is available to UK academic researchers via

88 Practical aspects of data collection

the EPSRC Chemical Database Service at STFC Daresbury Laboratory

(Fletcher et al., 1996; e-mail uig@dl.ac.uk; http://cds.dl.ac.uk/cweb/).

In October 2006, EPSRC summarily announced that parts of this Service

would close at the end of March 2007, although access to CSD and ICSD

has now been secured until at least 2011.

To avoid repeating an in-house determination you should be able to

search a database of your unpublished unit cells, for example by using

a program such as LCELLS (Dolomanov et al., 2003). Finding that your

unit cell is already known is not exactly good news, but at this stage

you have invested only a few minutes in the crystal. If it is undetected

until after data collection and structure solution such duplication could

waste several hours of valuable instrument time.

6.6.7 Unit cell volume

You can calculate whether your cell volume is compatible with the

expected molecular formula, using 18 Å

per non-hydrogen atom.

This value is valid for a wide range of organic, metal-organic and co-

ordination compounds, although some adjustment might be required

for special cases (from 14 Å

for some highly condensed aromatic com-

pounds to 23 Å

for some organosilicon compounds). It does not work

for inorganic compounds like NaCl or FeTiO

. A signiﬁcant discrep-

ancy may indicate that the unit cell is incorrect, that the compound is

not as proposed, or that solvent molecules are present. The number

of molecules present in the unit cell may warn you that, if the pro-

posed molecular formula is correct, symmetry considerations mean that

disorder must be present.

6.7 Data collection

It is necessary to set a small number of parameters to control the full

data collection: their values are obtained from the preliminary screening

measurements and unit cell determination. The most important factors

affectingdecisions concerning theseparameters arethe overall intensity

level on the frames, the mosaic spread and the crystal symmetry.

6.7.1 Intensity level

The intensity level will help determine the frame measuring time,

which should be long enough to give diffraction to adequate resolu-

tion: as a very crude rule of thumb, diffraction maxima should be clearly

detectable to at least

2 –

3 of the required 2θ

max

. For strongly diffracting

crystals, too long an exposure time may result in detector overloads: if

this happens the time should be reduced provided this does not result

in the loss of higher-angle data. For special problems it may be feasible

to collect low-angle data relatively rapidly, while using longer exposure

times for the higher-angle data and allowing signiﬁcant overlap of the

6.7 Data collection 89

two regions. However, this requires two settings of the detector and data

collection will therefore take much longer.

6.7.2 Mosaic spread

The mosaic spread is established from the widths (x, y) of the reﬂections

on individual frames, plus an estimate of z from the width of the rocking

curves. While there is little point in taking small steps through broad

reﬂections, this may be highly advantageous for a crystal giving narrow

reﬂections. Broad reﬂections are not necessarily problematic, provided

they can be separated from neighbouring maxima, and a greater scan

width giving fewer frames may be appropriate and deliver the dataset

more rapidly. However, it is important to check that the combination of

exposure time and scan width gives the required intensity. The actual

scan mode and width chosen will depend on the software (and acqui-

sition philosophy) being used, with narrow (e.g., 0.3

◦

) frames being

typical for Bruker SMART instruments, while wider 1−3

◦

frames are

used on Nonius KappaCCD diffractometers. In the former most or all

reﬂections will be partial, and integration is carried out as part of data

reduction; in the latter the detector integrates a good fraction of the

reﬂections on each frame. As with the other factors, it is important to

choose frame widths appropriate to theindividual crystal being studied.

A related issue occurs with larger unit cells (and usually Mo Kα radia-

tion), where you may have to move the detector further away from the

crystal to avoid reﬂection overlap. Unless you have quite broad reﬂec-

tions, with MoKα radiation you should not have tothink about thisuntil

you have axes of over 30 Å, possibly longer if the cell is centred. If you

have to move the detector back you should make sure that hardware

and software settings are compatible, although this is a problem only if

the detector distance cannot be changed under program control. A mis-

match of these settings can lead to inexplicable indexing failures. As the

crystal-to-detector distance is increased, the maximum value of 2θ that

can be recorded is reduced, and you need to make sure that data can

still be collected to the resolution you require: this may require two dif-

ferent 2θ settings. After you have collected your dataset, it is courteous

to reinstate the normal detector settings for the next user.

6.7.3 Crystal symmetry

It is useful, but not essential, to establish crystal symmetry at this stage.

The symmetry determines the minimum fraction of the whole-sphere

diffractionpattern to be measuredinorderto achieve a complete dataset.

If there is any doubt as to the correct diffraction symmetry, the lower

symmetryshould always beassumed.Althoughitispossible to calculate

the optimum set of frame runs to achieve this completeness as efﬁciently

as possible, for routine work it is sensible to collect the whole sphere of

data for triclinic crystals and at least a hemisphere for all other crys-

tal systems. Non-routine work might include higher-symmetry crystals

90 Practical aspects of data collection

that diffract weakly or are prone to decay in the X-ray beam: in such

cases it is important to use a data-collection strategy capable of deliv-

ering a unique set of data as quickly as possible. Another consideration

is that inefﬁciency in achieving the unique set will give a dataset with

higherredundancy(an alternative term is ‘multiplicity ofobservations’).

Despite the negative connotations of the term in other contexts, as far

as data collection and processing are concerned redundancy is a very

good thing: equivalent and duplicate reﬂections are valuable in correct-

ing data, for example for absorption, and they can be merged to provide

a unique dataset containing more precise intensity measurements.

6.7.4 Other considerations

Setting these parameters incorrectly can cause problems but area-

detector systems are generally tolerant of such mistakes. In particular,

collecting area-detector data without a valid orientation matrix is rarely

catastrophic provided valid indexing is eventually possible. Although

other data-collection parameters such as the rotation axis (e.g., ω or

φ) can also be varied, a typical data-collection setup procedure is

fairly simple and straightforward. Area-detector systems can there-

fore be operated by less experienced workers with minimal risk of

compromising data quality or completeness.

Re-recording of some of the initial frames at the end of the main data

collection and comparison of the integrated intensities allows detec-

tion of and correction for any decay, although signiﬁcant decay is rare,

particularly at low temperature.

At some point, record the crystal colour, shapeand dimensions. If a list

ofindexedcrystal faces is requiredfor a numerical absorptioncorrection,

these will have to be measured. If you expect even very minor icing of

your crystal during the datacollection, make these measurements before

you start. Finally, you could check that the X-ray generator settings are

correct, that the ﬂow of cooling water is adequate and stable and that

any low-temperature device is operating at the desired temperature and

has sufﬁcient cryogenic ﬂuid to last through the data collection.

References

Allan, D. R., Blake, A. J., Huang, D., Prior, T. J. and Schröder, M. (2006).

Chem. Commun., pp. 4081–4083.

Dawson, A., Allan, D. R., Parsons, S. and Ruf, M. (2004). J. Appl.

Crystallogr. 37, 410–416.

Dolomanov, O. V., Blake,A. J., Champness, N. R.and Schröder, M.(2003).

J. Appl. Crystallogr. 36, 955.

Fletcher, D. A., McMeeking, R. F. and Parkin, D. (1996). J. Chem. Inf.

Comput. Sci. 36, 746–749.

Görbitz, H. (1999). Acta Crystallogr.B55, 1090–1098.

McMillan, P. F. (ed.) (2006). Chem. Soc. Rev. 35, 847–854.

Exercises 91

Exercises

1. Assuming both are available, which of Cu or Mo

radiation would you use to determine the following

problems, and why?

a) C

;

b) C

;

c) C

;

d) absolute conﬁguration of C

;

e) absolute conﬁguration of C

2. A crystal indexed to give a metrically orthorhombic unit

cell.After processing the frameset, the reﬂection ﬁle was

examined in order to establish the true diffraction sym-

metry, and the measurements below are representative

of the pattern found. There were no signiﬁcant absorp-

tion effects. Is the crystal system really orthorhombic?

hklIntensity

10 2 4 258.2

−10 2 4 187.4

10 −2 4 267.4

10 2 −4 216.4

−10 −2 −4 245.2

10 −2 −4 200.9

−10 2 −4 264.6

−10 −2 4 208.3

3. A compound C

crystallized from tetrahy-

drofuran(C

O) solution gives a primitive monoclinic

unit cell of volume 1850 Å

. What are the likely unit cell

contents?

4. Estimate the range of absorption correction factors for

the following crystals with μ = 1.0 mm

−1

a) a thin plate 0.02 × 0.4 × 0.4 mm

b) a tabular crystal 0.2 × 0.4 × 0.4 mm

c) a needle 0.06 × 0.08 × 0.40 mm, mounted parallel

to the ﬁbre

d) a needle 0.06 × 0.08 × 0.40 mm, mounted across

the ﬁbre

Repeat the calculations with μ = 0.1 and 5.0 mm

−1

5. Two estimates were made of a set of unit cell parameters

a ...γ.

a) 8.364(12), 10.624(16), 16.76(5) Å, 89.61(8), 90.24(8),

90.08(6)

◦

b) 8.327(4), 10.622(6), 16.804(8) Å , 90, 90, 90

◦

The ﬁrst estimate was derived from the original orien-

tation matrix reﬁnement using 67 reﬂections, while the

second was obtained by a ﬁnal constrained reﬁnement

using 5965 reﬂections from the entire frameset. Esti-

mate the approximate contribution in each case to the

uncertainty in a C–C bond of 1.520 Å.