Presentation on theme: "Sample Preparation, Data Collection and Phase-ID using Powder XRD"— Presentation transcript:
1 Sample Preparation, Data Collection and Phase-ID using Powder XRD Pamela WhitfieldNational Research Council, Ottawa9th Canadian Powder Diffraction Workshop, Saskatoon, May 2012
2 Horses for courses…Data quality required depends on what you want to do with itPhase-ID has less stringent requirements on both sample prep and data collectionQuantitative phase analysis, Rietveld analysis and structure solution require careful sample prep but can require different data collection regimesI’ll mostly cover requirements for phase-ID but will touch on considerations for other techniques.
3 Questions to ask What is in your sample? How much have you got? Organics often better collected in transmissionFluorescence can cause problems in data qualityHow much have you got?Very small quantitiescapillary or foil transmission? (not an option for many people)smear mount?We’ll assume conventional reflection geometry unless stated otherwiseWhat kind of instrument have you got access to?If you have a choice which is the best?
4 What matters for phase-ID? Peak positions most importantRelative intensities secondarybut very important for Rietveld, etc….If wanting to do search-match it is useful if the phases exist in the PDF database!
5 Where to start? What errors affects peak positions? What affects relative intensities?Preparing the samplesDifferent types of sample holders
6 Peak positions – sources of error Zero point error - is the system properly aligned?use a NIST standard periodically to check itSample displacement - sample too high/low?(0.1 mm ~ 0.045°)Note: convention is that –ve sample displacement = sample too highNot an issue for parallel beam systems
7 Peak positions – sources of error Sample transparencyif X-rays penetrate a long way into the sample can get a ‘sample displacement’ even if the height is perfectnot an issue for parallel-beam systemsif necessary use a thin sample to avoid transparency peak shiftsrelative intensities will be affectedDiffraction patterns from powdered sucrose as both deep and thin samples
8 Parallel versus para-focussing The systems don’t look that different but don’t behave the same..Parallel-beam immune to sample displacement & transparency but has worse peak resolution – twin mirror system exceptedDivergent-beam without & with secondary graphite monochromatorParallel-beam setups with long slits and secondary mirror
10 Sample-related problems Grainy samples or ‘rocks in dust’Microabsorptiona serious issue for quantitative analysis and could fill a talk by itself!Preferential orientationExtinction
11 No. of diffracting crystallites “Grainy” samplesIssue of graininess relates to particle statisticsParticle statistics is what makes a powder a true powder!600 mesh sieve = <20 mmDiameter40mm10mm1mmCrystallites / 20mm35.97 × 1053.82 × 1073.82 × 1010No. of diffracting crystallites1276038000Comparison of the particle statistics for samples with different crystallite sizesCrystallitesize range15-20mm5-50mm5-15mm<5mmIntensity reproducibility18.2%10.1%2.1%1.2%Reproducibility of the intensity of the quartz (101) reflection with different crystallite sizes
12 “Seeing” particle statistics Playing Russian roulette with a grainy sampleStacking the odds in your favour by micronizing….
13 How to improve particle statistics There are a number of potential ways to improve particle statisticsIncrease the area illuminated by X-raysDivergence angleRotate samplesUse a PSDReduce the particle size(without damaging crystallites!)McCrone mill = goodMortar and pestle = bad
14 I don’t have a 2D detector – now what? A series of phi-scans can show up problemsWith a rotation stage phi is a set angle instead of full rotationPhi-scans across 5 fingers of quartz with different samples
15 I don’t have a 2D detector – now what? Can also run repeats after reloading sample each time (get real stats as a bonus)Unmicronized : MgO only appears in 1 sample out of 3periclaseOverlay of 3 repeat patterns from un-micronized cementOverlay of 3 repeat patterns from micronized cement
16 Extreme examples… Occasionally reflections are unexpectedly split Quartz is particularly prone….Synchrotron data are not immune – in fact it can be worse due to the extremely parallel beamMain 101 reflection of ~100 micron quartz with a fuller pattern inset showing spurious intensitiesCapillary and rocked reflection data from LaB6 on a strip heater taken with the Australian synchrotron
17 MicroabsorptionMicroabsorption is the thing that causes most nightmares for analysts doing quantitative phase analysisCaused by a mixture of high and low absorbing phasesHigh absorbersbeam absorbed at surfaceonly fraction of grain diffractingrelative intensity underestimatedQPA too lowLow absorbersbeam penetrates deepermore diffracting volumerelative intensity overestimatedQPA too high
18 What can you do about it? Change radiation? Use neutrons? Absorption contrast changes with energyHigher energy X-rays often less problematicUse neutrons?Not usually practical but a ‘gold standard’Use the Brindley correction?Need to know absorption of each phaseNeed to know particle (not crystallite!) size for each phaseAssumes spherical particles with a monodisperse size distributionUsually unrealistic!
19 Effect of particle size Brindley proposed that a maximum acceptable particle size for QPA can be calculated by:m = linear absorption coefficient (LAC)corundummagnetitezirconCuKa LAC (cm-1)1251167380tmax (mm)0.80.10.3CoKa LAC (cm-1)19524057188.8.131.52
20 The scale of escalating despair! Brindley also devised a criteria for whether you should be ‘concerned’ about microabsorptionmD = linear absorption coefficient x particle diameterFine powdersmD < negligible m-absorptionMedium powders0.01 < mD < m-absorption present – Brindley model appliesCoarse powders0.1 < mD < 1 large m-absorption – Brindley model estimates the effectVery coarse powdersmD > severe m-absorption – forget it!
24 QXRD Round Robin: CPD #4 unmilled Unmilled grain sizes: Al2O3 28mm, Fe3O4 36mm, ZrSiO4 21mmCan’t fit intensitiesvery poor particle statistics? (<200 diffracting crystallites per phase)Al2O3 mD = 0.345, Fe3O4 mD = 4.15!!, ZrSiO4 mD = 0.7880.1 < mD < 1 large m-absorption – Brindley model estimates the effectmD > severe m-absorption – forget it!without Brindleycorrectionwith BrindleycorrectionCuKa with graphite monochromatorweighed amountsAl2O %Fe3O %ZrSiO %Thanks to Ian Madsen for the data
25 CPD #4 confirming what we suspected Large grains can be confirmed using 2D detector as before or using a series of scans with different phi angles (no rotation)N.B. Al2O3 may still have preferential orientationCMZAl2O3ZrSiO4Fe3O4Thanks to Arnt Kern for the data
26 Micronized CPD #4? Particle statistics no longer a problem Al2O3 and ZrSiO4 still have some orientation – correctedCoKa doesn’t help much as problem switches from Fe3O4 to ZrSiO4How about an SEM?CuKa withgraphite monochromatorCoKa withgraphite monochromatorweighed amountsAl2O %Fe3O %ZrSiO %Thanks to Ian Madsen for the data
28 Pick a number, any number… What value of particle size do we choose for the Brindley correction?Wt% Corundum Magnetite ZirconWeighedNo correction Mean BiasBrindley model, Ø = 1m Mean BiasBrindley model, Ø = 5 m Mean BiasBrindley model, Ø = 10m Mean BiasThanks to Ian Madsen for the analysis
29 What does reality matter anyway? Can fudge the particle size numbers and packing so the Brindley correction gives the right result for CoKabut mD for ZrSiO4 and Al2O3 well into coarse rangeCuKa not as good (and mDs are even worse)CoKaAl2O3 17mm 40%PDFe3O4 5mm 40%PDZrSiO4 14mm 40%PDCuKaAl2O3 17mmFe3O4 5mmZrSiO4 14mmif these numbers correct send your McCrone mill and SEM back!weighed amountsAl2O %Fe3O %ZrSiO %
30 Preferential orientation (texture…) Preferential orientation (PO) is most often seen in samples that contain crystallites with a platey or needle-like morphology.Particular culpritsPlatesmicaclayssome carbonates, hydroxides e.g. Ca(OH)2Needleswollastonitemany organicsThe extent of the orientation from a particular sample depends greatly on how it is mounted
31 Orientation of plate-like samples There’s no getting away from it – they can be a real painTop-loading is hopeless as you make it worse….Back-loading the usual approach but not always enough…Breaking up the alignment of the plates by back-loading onto a rough surface such as sandpaper can help…
32 Going the extra mile…With plate-like samples if you have a capillary stage then use it!Background-subtracted data from micronized 40S mica in a 0.5mm capillaryIf not then spray-drying the sample can be an alternative….200001Top-loaded, spray-dried 40S micaSEM of spray-dried mica
33 Just to prove the data is usable…. Micas not pleasant to deal with at the best of times and this has some messy anisotropic broadening..However, the data from the top-loaded, spray-dried sample fits a un-refined literature biotite structure very well with no orientation correctionRwp = 11.8%GOF = 1.82That’s not to say there were no corrections needed at all! Refinement of the top-loaded, spray-dried 40S mica using a literature biotite structure without orientation correction
34 Corrections for PO in Rietveld software Two different corrections exist in most software to correct orientation during Rietveld analysisMarch-Dollase (MD)Single variable but an orientation direction must be supplied by the analystSpherical Harmonics (SH)VERY powerful approach – can increase SH ‘order’ to fit increasingly complex behaviourNo orientation direction requiredNumber of variables increase with reducing cell symmetryBe very careful in quantitative analysis with severe peak overlap (e.g. cements)Negative peaks are very common and very meaningless!
35 ExtinctionReduction in the intensity of a Bragg reflection by re-diffraction by the successive planes back in the direction of the incident beamDependent on size/shape of the coherently-diffracting domainsPrimary re-diffraction within a single crystalliteEffect minimized by reducing grain size – ideally submicronNormally seen in large, ‘perfect’ crystallites such as silicon or quartzSecondary mosaic crystals, not seen in powders
36 The different preparation techniques ReflectionTop-loadingFlat plateBack-loadingSide-loadingTransmissionCapillaryFoil transmission
37 Top-loadingSimplest but most prone to inducing preferential orientationSpecial holders often in this categoryAlternative holders such as cavity zero background silicon or air-sensitive often top-loading as well
38 Flat plate aka: smear mount Used with very small samples (phase-ID , Rietveld )Sample adhered to zero background plate using some form of binder/adhesive that doesn’t have any Bragg peaksVaseline, vacuum grease, hairspray (spray ~12” from holder)Slurry with ethanol or acetone – tricky to get right consistencyN.B some quartz plates show a sharp reflection when spunSilicon zerobackground plateQuartz zerobackground plateGem Dugout a commonly used source for zero background plates (www.thegemdugout.com)
40 Side-loading I don’t have one of these! but basic principle….. plug powderglassslideholdersample
41 CapillariesProbably best way to reduce orientation in platey materialsCommercially either quartz, borosilicate or soda-glassrange in diameter from 2mm to 0.1mmOr use thin-walled polymer tubing of Kapton, PET, etcMost useful where sample absorption is low, e.g. organicsCan be extremely fiddly to fill!0.2 mm1 mm
42 Capillary instrument setup Capillary setups can be quite specializedFocussing optics specific to transmission geometryEven q-q systems better run as q-2q in capillary modeTransmission better for low anglesCapillary setup on q-q system at very high 2q angle using detector scan with focussing primary optic, PSD, radial Soller slits and primary slit setup to reduce low angle scatter reaching detector
43 Capillaries – highly absorbing samples Absorption reduces the peak intensities at low anglesCorrections exist but they have limitsSmaller capillaries and/or dilution with a ‘light’ phase will help (e.g. diamond, amorphous boron, carbon black, etc)Rietveld refinement of ~10 vol% SnO2 in diamond powderCapillary and reflection data from pure SnO2
44 Foil transmission Another approach for small samples Powders can be mounted between films of Kapton, Mylar, etcNot immune to preferential orientation – the plane is just rotated 90° so the peak intensities change accordingly!Quartz powder between KaptonTwin-mirror system set up for foil transmission
45 Foil transmissionSample can be very thin so highly absorbing samples possible without dilution1/cos(q) correction required for accurate relative intensitiesRietveld refinement of SnO2 (1400cm-1)
46 Data collection strategies Rietveld analysis guidelines published by McCusker et al in 1999Choose beam divergence so the beam doesn’t overspill the sample at low angleremember the under-scan when a PSD is used!1st datapoint may be at 10° 2q but the scan may start at 8°!(ENeqV1_0.xls very handy for working out correct divergence)(http://ig.crystallography.org.uk/spreadsh/eneqv1_0.xls)Rule of thumb - step size of ~ FWHM/5 to FWHM/8Too small = wasting time and producing noisy dataToo coarse = chopping intensity and peaks not modelled properly
47 Experiment optimization ‘Horses for courses’ – collect data fit for purposeData for phase-ID does not have to be of the same quality as for structure solution, etcMost common mistake among userstoo small step size for sample0.01º step, 1s countRwp = 15.2%0.02º step, 2s countRwp = 12.0%2 different datasets from quartz stone– both experiments took 25 secondsSmaller Rwp corresponds to a better fit.
48 Peak-to-background A number of things affect the peak-to-background air-scatter at low anglesuse air-scatter sinks if needednanoparticles have lower intrinsic peak heightsnot much you can do hereeventually Rietveld results are no longer meaningfulcapillaries always have higher backgroundsubtracting capillary blank can improve this but careful not to distort counting statisticsfluorescence is the main cause of poor peak-to-background…Rietveld refinement round robin suggested a minimum P/B value of 50 for accurate structural parameters….
49 Why does background matter? With a high background the uncertainty in the background parameters increase (often use more parameters as well)uncertainty in the extracted peak intensities increases→ greater uncertainty in structural parameters and quantitative phase analysisWhich line would you choose?
50 FluorescenceFluorescence even adversely affects phase-ID detection limitsa secondary monochromator on conventional system is an effective way to filter out fluorescenceCuKa - Li1.15Mn1.85O3.9F0.1Lin (Counts)10020030040050060070080090010001100120013002-Theta - Scale1520304050607080No monochromatorProperly alignedmonochromator/mirrorthere is a realpeak here!
51 Fluorescence – what to do about it? With a PSD a conventional monochromator not possible – data with CoKaCoKa - LiMn1.5Ni0.5O4Which dataset do you prefer?
52 Fluorescence cont.Can improve PSD data significantly by adjusting the detector’s electronic discriminator windowP/B = 13.4Rescaled to normalize backgroundP/B = 4.5Sacrifice intensity to improve P/B ratioP/B = 4.2P/B still along way off 50. Change radiation or instrument.
53 Problematic sample: quant analysis FeS + Mg(OH)2 + SiO2CuKaGround or unground?particle statisticsMicroabsorption (FeS)ideally switch to CoKaFluorescence (FeS)high backgroundmonochromator, energy-discriminating detector, switch to CoKaPreferential orientation (Mg(OH)2)Extinction? (SiO2)Micronize!!!!All of these problems are reduced by micronizing to sub-micron particle/crystallite size
54 Problematic sample: Rietveld analysis LiMn1.4Ti0.1Ni0.5O4 (lithium battery cathode material)Mn fluoresces with both CuKa and CoKa !Use a monochromator or energy discriminating detectorGood peak-to-background, but...Fluorescence is still there even if you can’t see itVery high absorption impacts particle statistics (X-rays only penetrate a few 10s of microns)Solution by changing tube?CrKa 2.29Å (unusual, high air scatter/attenuation and limits lower d-spacings attainable)FeKa 1.94Å (very unusual and low power tubes)MoKa 0.71Å (unusual and beta-filter artefacts visible)
55 LiMn1.4Ti0.1Ni0.5O4 Cu Co Mo Cr P/B = 9.4 P/B = 4.5 P/B = 84 P/B = 87 A primary monochromator would get rid of this high angle tailP/B = 84CrP/B = 87(P/B = 54 without air-scatter sink to reach angles >100)
56 Variable counting time (VCT) The physics of XRD dictate that intensities drop with angleMost of the information (reflections) is at higher anglesCan regain much of the information by counting for longer at higher anglesConstant Counting TimeI ~ LP * thermal vibration * f2Variable Counting TimeBoehmite (Madsen, 1992)
57 VCT data - quantitative analysis Also possible to improve detection limits in quant analysis by counting for longer where minor phases expectedFixed count timeVariable count time (normalized)Example from presentation by Lachlan Cranswick
58 VCT data - structure refinement Extract more structural details if reflections still visible at high anglesUsing a PSD split pattern into sectionscan also increase step size with angle as well to save some time…Jadarite structure with thermal ellipsoids
59 Phase-IDPhase-ID usually undertaken using vendor-supplied software with the ICDD Database (PDF2 or PDF4)The database is not free so budget accordinglyPDF4 requires yearly renewal but has more featuresPDF2 good enough for search-match and OK for 10 yearsA free database called the Crystallographic Open Database (COD) exists but there is no quality checking – user beware…The Powder Diffraction File uses XRD ‘fingerprints’ – if they haven’t been deposited they won’t show upDatabase entries are allocated a ‘quality mark’ but occasionally the newer ones are actually worse!Experimental quality marks ‘*’ > ‘I’ > ‘A’ > ‘N’ > ‘D’Calculated from ICSD, etc ‘C’Background subtraction recommended before search-match if it is high but don’t bother with Ka2 stripping, etc
60 Phase-ID Improve your odds in the search-match make a sensible guess as to the likely elementsdoes your sample really have plutonium in it?!if you have elemental analysis results then use thembut consider possibility of amorphous phasesSearch-match in EVA on a sample of zircon
61 Be sensible… Use common/chemical sense don’t believe results just because the computer tells youeven oxygen has entries in the PDF2!Where software supports it ‘residue’ searches can be very helpful in identifying minor phases
62 Don’t be led astray…Minor peaks - make sure they aren’t Kb or tungsten linesvendor software can often identify these (e.g. EVA below)WLaCrKaCrKb
63 No luck – what next? Do you have a large systematic error in the data? your diffractometer alignment should be checked regularly with a standardmodern search-match software can cope with a reasonable error but it has limitsLook for possible analogues which may appear in the PDF2LaCoO3 similar to LaNiO3 with slightly different lattice parametersanalogues may have significantly different relative intensitieshowever: LiMnO2 (Pmmn) completely different from LiCrO2 (R-3m)LaCoO3, R-3ca = 5.449, c = ÅLiMnO2LaNiO3, R-3ca = 5.456, c = ÅLiCrO2
64 Getting desperate yet? Put the sample under optical microscope does it seem to have the number of phases you expect?If it contains Fe or Co try a magnet!Possible contaminationmortar and pestle not cleanmaterial from micronizer grinding elements (newer corundum elements not as good as the older ones – use agate)Last possibility to consider….maybe you have found a new phasethen the fun really starts!
65 Conclusions…Use the appropriate sample mounting technique for the sample and the data requirementsGraininess, microabsorption and preferential orientation are all related to particle and crystallite sizeDo yourself a big favour by micronizing your sample if possible!Preferential orientation can be corrected during analysis but the others can’t……Assumptions of the Brindley correction never met in real lifePoor application of Brindley correction worse than no correction
66 Yet more conclusions….There are times when the newest diffractometer (PSD, etc) isn’t the best one for the jobfluorescence can be your #1 enemy!secondary optics can be your friend No such thing as the perfect configuration for everyoneVCT data can help in a number of waysimprove the detection limit for minor phasessignificantly improve the quality of a structure refinementIf you don’t remember anything else remember this..think about your samplesa one size fits all approach doesn’t work!
67 Acknowledgements Ian Madsen (CSIRO) I couldn’t improve on his explanation of microabsorption so I used it!Responsible for the CPD QPA round robin sample 4 which still give people nightmaresMati Raudsepp (UBC) for spray drying the mica sample and the SEM
68 ReferencesG.W. Brindley, “The effect of grain and particle size on X-ray reflections from mixed powders and alloys….”, Philosophical Magazine, 3 (1945),Commission on Powder Diffraction webpagelinks to all the round-robin information, guidelines and papers (freely available)
70 OK so you found a new phase…. Before getting to the refinement step you have to figure out a rough idea of the structureThere are some different steps in the processPeak fitting (most of the time)IndexingSpace group determinationStructure determination
71 Indexing….You need to know whether cubic, monoclinic, etc and what are the lattice parametersThe instrument should be as good as you can get itlab data more difficult than synchrotronIn the lab it may require a special high resolution dataset over a limited range (usually only use the first lines)Accurate peak positions the main goalSoftware packagesTOPAS (LSI and LP Search), Crysfire (Dicvol, Treor, etc), MacMaille, etc
72 Space group determination…. This is where you often need to know a little crystallography….Conventional way to do this is to study systematic absences (International Tables necessary and Chekcell software useful)Maximum likelihood software Extsym can give a list of probable extinction symbols from a Pawley refinement (not same as SG!)TOPAS makes a guess at extinction symbol and often correctSample density very useful (buy a pycnometer!)Can then calculate ‘volume per formula unit’Allows easy exclusion of space groups where the multiplicities are too highWith organics a rough guide of 18Å3 per non-H atom can help
73 Solution…. A number of approaches possible Conventional direct methods (EXPO, SXTL software)Real space methods (DASH, TOPAS, FOX, etc)Charge flipping (Superflip, TOPAS, etc)Real space methods tend to be more reliable with poor resolution dataPowder diffraction data often regarded as ‘poor’ by defaultCharge-flipping a powerful method with higher resolution powder datahas been know to work with iffy data, but not mine…..!
74 Example…. aspirin Aspirin tablets usually very pure Done with capillary transmission but…Highly crystalline organic with reflections to high 2q anglesEasily indexed to a monoclinic cellCharge flipping does need a space group to workBasic structure solves in a matter of minutes…
75 Example... sucrose Even easier to get hold of than aspirin Also highly crystallineSimulated annealing with a z-matrix works with normal data, charge flipping with VCT data
76 Example... wollastoniteAn inorganic reflection example…. but a difficult oneWollastonite needles show severe preferential orientation when top-loadedNormally I would say ‘make a better sample’ but sometimes it will still worksThe basic simulated annealing approach still the same with some tweaksSEM of wollastonite
77 Wollastonite – SA strategy TOPAS input file setupSpace group P-1SiO4 always tetrahedral – safe to use simple z-matrixAnti-bump for Ca-Ca, Ca-O and Si-SiOctahedrally coordinated Ca-OToo many oxygens with SiO4 z-matricesNeed to merge oxygens to get correct unit cell contentsUse “occ_merge O* occ_merge_radius 0.9”
78 Wollastonite – SA strategy TOPAS lets you to use a ‘trick’ to solve badly orientated datafor details read the paper!The structure from simulated annealing matches the literatureLiterature structureRaw structure from SA with 4th order SH – blue atoms are merging oxygens
79 Wollastonite – SA strategy Just a couple of plots to prove the sample was orientated!Simulated powder pattern from the SA structure without PO correctionFit to the data for the raw SA structure with 4th order SH