Presentation on theme: "Structure Solution and Basic Refinement. Recommended software Shelx –http://shelx.uni-ac.gwdg.de/SHELX/ WinGX –http://www.chem.gla.ac.uk/~louis/software/wingx/"— Presentation transcript:
Structure Solution and Basic Refinement
Recommended software Shelx –http://shelx.uni-ac.gwdg.de/SHELX/ WinGX –http://www.chem.gla.ac.uk/~louis/software/wingx/ Platon –http://www.chem.gla.ac.uk/~louis/software/platon/ Solution programs (see later) A good text editor –e.g. Notepad++
Chemistry Facilities Bruker Smart Apex CCD Diffractometer
From crystal to cif Get crystal Short experiment to index Long experiment to collect complete data Integration Absorption correction Identify space group Solve You start here Refine Cif
What is crystallography all about A crystal is a 3D periodic array of molecules X-rays interact with (diffract from) electrons Diffraction results in a regular pattern of spots (due to constructive and destructive interference) Intensities observed are related to atom types and positions We build a model and compare the calculated diffraction pattern with the observed
What you will get from us 2 or 3 files –ins file The shelx instruction file contains unit cell, radiation wavelength, temperature, crystal system and space group information, unit cell contents from user input –hkl file The reflection data, contains indexed intensities with associated estimated standard deviations –cif file Contains some experimental details (not essential)
Files – ins file TITL pg24 in P-1 CELL ZERR LATT 1 SFAC C H N S HG UNIT PATT HKLF 4 END
Files – hkl file
Structure Solution We need to get an initial model from which to work –This is called structure solution –We can only measure intensity but we really want to know the phase of each reflection. –Several methods of extracting initial estimated phases from our data are available.
Structure solution Direct methods –Requirements It is desirable but not required to have a centrosymmetric structure. –How it works Uses statistical relationships between the intensity of different reflections to establish phases. –Look out for… This is a very powerful method which often works very well but as it is based on statistical analysis, it will sometimes fail. Computationally demanding. Scales inversely with symmetry, i.e. low symmetry large unit cells take more time.
Structure solution Direct Methods Programs –xs from the shelx suite (works for me 99% of the time) Instruction is TREF –You may specify a number after TREF to increase the number of trials for difficult structures. –Sir software Several versions –Sir92, Sir97, Sir2002, Sir2004 –May get different results with different versions »Variety of options for structures of different size/difficulty OR »Can be set up through, e.g. WinGX for simplicity but less control.
Structure solution Patterson Methods –Requirements A heavy element, e.g. Fe, Cl, S, etc –How it works Generates a map of ‘peaks’ representing difference vectors, i.e. interatomic vectors Peak intensity is related to the product of the atomic numbers of the two atoms involved thus the heaviest elements are identifiable. –Look out for… Depending on the program you may only get the heavy atom positions. Completing the structure may take more time and effort than e.g. Direct methods
Structure Solution Patterson Methods Programs –xs from the shelx suite Instruction is PATT You will only get heavy atom positions back –Dirdif Will attempt to complete the model and guess atomic assignments. –Can be extremely useful for complicated metal clusters, etc.
Structure Solution Partial Structure Expansion –Requirements Knowledge of expected substructures and their geometries, e.g. a benzene ring –How it works Uses a Patterson map and rotates the substructure around three axes until a best fit is found to the data. There will then be attempts to complete the model using the phase information from your partial structure as a basis for phase refinement. –Look out for Not many pitfalls assuming you are confident of the unit cell contents. More time consuming to set up than other methods and unsuitable for unknown samples.
Structure Solution Partial structure expansion programs –PATSEE from the wingx suite You must provide a list of coordinates for your substructure. –Dirdif Provides a small database of common geometries e.g. benzene rings, indoles or can use your own. –Not necessarily user friendly in my opinion
Structure solution Charge flipping –Requirements Complete data (at the moment) –How it works Use random phases, ‘flip’ the sign of electron density charge below a threshold and get new phases. Use new phases with original magnitudes. Repeat. No symmetry is used for solution. It is determined afterwards. –Look out for… Unreliable results with incomplete data
Structure solution Charge-flipping software –Superflip From originators of the method Available as standalone (not recommended), via WinGX, via crystals –Flipper Available in Platon
Structure solution Has it solved? –Use your chemical knowledge… –How atoms interact E.g. expected bond distances and angles for particular arrangements Reactions or decompositions which may occur Non-bonded interaction lengths and types –E.g. +ve to +ve is not going to be common –Charge Should always be neutral for a unit cell Assignment of charge on metals, ligands –Likely patterns of motion E.g. neighbouring atoms are likely to have similar thermal motion Spinning or wagging of certain groups may be expected Groups which are likely to be rigid, move as one
Structure solution Watch out for… –If provided, don’t assume atomic assignments are correct. –Look for expected geometry, e.g. rings, octahedra –Might get messy q-peaks and need to trim back to your structure. Don’t assume that if you can’t see your compound that is isn’t there and just obscured by noisy peaks –Incorrect atomic assignment and missing atoms will affect the calculated phases and may mean some atoms don’t appear at first. Be patient –Try any and all structure solution programs you can find Different programs produce different results even using the same method. Those mentioned are not the only ones but should be sufficient.
Structure Solution example Data taken from Oxford primer: –Crystal structure determination William Clegg
Refinement Iterative process –Be patient Sometimes it takes a long time and is difficult Sometimes it is easy and quick –You must model all electron density (q-peaks) or be able to explain why modelling some peaks is not appropriate. –Structure must be chemically reasonable Pay attention to –Geometry –R-factors –Q-peaks –ADP’s –Contacts
Typical Refinement Initial solution Check atomic assignments Refine… Problems? Fourier difference map All atoms correctly identified? No Yes Model complete? No Yes Odd sized or Shaped ADP’s Large Q-Peaks Weighting scheme None Wrong atom type(s)? Disorder? Missed Atoms? All atoms anisotropic? Switch current atoms to anisotropic No Switch unusual atom(s) to isotropic refinement FINISH! START converged not converged yes Yes
Refinement Common problems –Wrongly assigned atoms –Disorder, particularly solvent* –Twinning* –Incorrect space group –Q-Peaks due to strong absorption (heavy metals present) Not all refinements will end happily –You may have to leave some atoms isotropic –You may be unable to find or place hydrogen atoms –You may have a high R-factor –You may simply not be able to finish due to the above issues or poor quality data * Workshops on these will be given later
Shelx Anatomy of a shelx file TITL SAMPLE in Pbca CELL ZERR LATT 1 SYMM 0.5-X, -Y, 0.5+Z SYMM -X, 0.5+Y, 0.5-Z SYMM 0.5+X, 0.5-Y, -Z SFAC C H N O UNIT TEMP -123 L.S. 4 BOND $H FMAP 2 ACTA CONF PLAN 20 WGHT FVAR O = C = O C HKLF 4 REM HL9005 in Pbca REM R1 = for 3267 Fo > 4sig(Fo) and for all 3421 data REM 226 parameters refined using 0 restraints END WGHT REM Highest difference peak 0.267, deepest hole , 1-sigma level Q Q Q Q Q Q
Common Shelx Commands L.S. Full least squares refinement. Number of cycles given after a space, e.g. L.S. 4 will give 4 refinement cycles CGLS Conjugate gradient least squares. Use for faster refinement with very large structures but only during initial refinement. You must switch to L.S. before generating a cif. FMAP Followed by a number requests a Fourier map. Normally you will use FMAP 2, for a difference map PLAN The number of peaks to be returned from the difference map, e.g. PLAN 20 gives 20 peaks. BOND Put Bonds into the cif file, always use BOND $H ACTA A cif will be generated WGHT This is the weighting scheme which will be used FVAR Free variables, the first is the overall scale factor. Others may be used for various purposes SADI Same distance restraint, e.g. SADI C1 C2 C3 C4 instructs the program to restrain the C1…C2 distance to be similar to the C3…C4 distance SAME Generates similarity restraints for extended geometries DFIX Distance restraint e.g. DFIX 1.54 C1 C2 puts a restraint on the C1…C2 distance to be 1.54 angstroms SIMU Similar thermal parameters will be applied to all atoms in the list following DELU Vibration restraint. The two atoms will be restrained to have similar motion along the direction of the bond AFIX Constraints. Many types available HFIX Add hydrogen atoms, many options available for different hydrogen environments
Back to our example…
Restraints and constraints A restraint allows a parameter to refine within limits –E.g. like applying a spring A constraint fixes a parameter. It is not allowed to refine. –E.g. like applying rope Restraints are treated as additional observations Restraints have an associated e.s.d. i.e. a measure of how strict the restraint should be –Most commands have reasonable defaults Use to correct poor geometry with chemical knowledge Can use temporarily to maintain reasonable geometry when identifying a problem, e.g. disorder Only use when necessary ensure e.g. distance restraints use an appropriate value, e.g. taken from CSD data Make sure there isn’t an underlying problem before resorting to R&C More on these in disorder workshop.
Atomic assignment problems Models are a picture of electron density –Different elements have different numbers of electrons –Incorrect assignments should show up in thermal parameters. –Too small an element will cause the ADP to shrink –Too large an element will cause it to grow –Why?
Atomic assignment problems Atom is actually a nitrogen which has higher z and therefore e - density than our modelled carbon. Therefore the carbon ‘shrinks’ to increase its e - density Atom is actually a carbon which has lower e - density than our model. Therefore the nitrogen ‘grows’ to spread out its e - density
Hydrogen atoms Difficult to find –Only one electron to diffract from –Heavier elements further obscure hydrogen positions Incorrectly located –Diffract from electrons not nuclei!!! –Valence electron only, located ‘in bond’
Hydrogen placement Find them if you can –They may often be visible in a difference map –You may then be able to refine, partially refine them or constrain them. Geometric placement –Uses expected (e.g. tetrahedral) angles and distances appropriate for x-ray diffraction –HFIX mn m specifies how to place e.g. tetrahedral angles n specifies how to refine, e.g. full coords or riding
Hydrogen Placement Common HFIX types
Some Quality indicators R 1 –The conventional R-factor –<10% is publishable –<5% is good Goof –Goodness of fit –Should be as close to 1 as possible More than around 0.4 away is cause for concern R int –An R-factor for data merging of equivalent reflections –If perfect would be 0 –A rough guide is to expect R 1 to be close to this value Max Shift –The max shift of all parameters from non-linear least squares refinement –Should be zero for a properly converged, finished model.
Finishing off Confident the structure is ok? Platon checkcif –IUCr web tool –Platon Does additional checks if FCF file is present Make sure it is up to date!