Upload
m-s-rihawy
View
220
Download
0
Embed Size (px)
Citation preview
8/3/2019 LOD and Other Error Sources
1/17
Sample Preparation
Page 1 2009 Bruker Biosciences Commercial in Confidence
,Powders and Solids
Sources of Error
02/11/2009Bruker Confidential1
n
X-ray Fluorescence
Overview
General introduction in sample preparation for XRF
Sample Preparation for Liquids, Powders and Solids Introduction
Penetration- and Information Depth
Grain- or Particle Size Effect (Powders)
Crystallographic- or Mineralogical Effect (Powders)
Sample Preparation Techniques
Sources of Error Introduction
Sample Preparation Errors
Instrument Errors
02/11/2009Bruker Confidential2
Definitions
Sample PreparationIntroduction
Parameters of interest in Sample Preparation:
Physical form (liquid, powder, rock, bulk, ...)
Sample size (does it fit in the sample cup)
Ease of sampling: representative
Quantity of sample
Time per sample
Cost per sample
Reproducibility of preparation
02/11/2009Bruker Confidential3
Sample PreparationIntroduction
XRF is an analysis by comparison
Accuracy of the analysis will be dependent on:
Quality of standards: primary or secondary standards
Quality of calibration: use of the right correction model
Calibration maintenance: re-calibration against drift
Instrument reproducibility: stable spectrometer?
Sample taking: representative for the bulk?
Sample preparation: is the method reproducible?
no sample preparation reproducibility... no measurement reproducibility
02/11/2009Bruker Confidential4
8/3/2019 LOD and Other Error Sources
2/17
Sample Preparation
Page 2 2009 Bruker Biosciences Commercial in Confidence
Sample PreparationIntroduction
Sample preparation is the largest source of error
Many forms of samples exist, 3 broad categories of specimen
preparation can be distinguished:
directly measurable after one simple treatment
significant pre-treatment, e.g. heterogeneous samples
special treatment, e.g. radioactive samples
Types of samples:
Bulk solids
Powders
Li uids
02/11/2009Bruker Confidential5
Gases
Sample PreparationIntroduction
Methods of specimen preparation:
Form TreatmentBulk SolidsHomogeneousHeterogeneous
PowdersHomogeneousHeterogeneous
L i u ids
Grind to give flat surfaceDissolve or react to give a solution or ahomogeneous melt
Grind and press into a pelletGrind and fuse with a flux (e.g. borax)
02/11/2009Bruker Confidential6
Homogeneous (concentrated)Homogeneous (diluted)Heterogeneous
GasesAirborne Dusts
Analyse directly or dilutePreconcentrationFilter to remove solids
Aspirate through a filter to remove thesolids
Sample PreparationPenetration- and Information Depth
Penetration Depth
Information or Critical Depth
02/11/2009Bruker Confidential7
high energy photons
Surface treatment is moreimportant for heavy matrices !
Light matrices are more criticalwhen it comes to informationdepths
Sample PreparationPenetration- and Information Depth
Line Energy Graphite Glass Iron Lead
Cd KA1 23,17 keV 14,46 cm 8,20 mm 0,70 mm 77,30 m
Mo KA1 17,48 6,06 3,60 0,31 36,70
Cu KA1 8,05 5,51 mm 0,38 36,40 m 20,00
Ni KA1 7,48 4,39 0,31 29,80 16,60Fe KA1 6,40 2,72 0,20 * 164,00 11,10
Cr KA1 5,41 1,62 0,12 104,00 7,23
S KA1 2,31 116,00 m 14,80 m 10,10 4,83
Mg KA1 1,25 20,00 7,08 1,92 1,13
F KA1 0,68 3,70 1,71 0,36 0,26
N KA1 0,39 0,83 1,11 0,08 0,07
02/11/2009Bruker Confidential8
0,001 m = 1 nm = 10
Atomradius: 0,5 - 3
C KA1 0,28 * 13,60 0,42 0,03 0,03
B KA1 0,18 4,19 0,13 0,01 0,01
8/3/2019 LOD and Other Error Sources
3/17
Sample Preparation
Page 3 2009 Bruker Biosciences Commercial in Confidence
Sample PreparationSample taking
Bulk10x kg
10 - 1000 g
Sampling
Pulverising,Pelletising orFusing
02/11/2009Bruker Confidential9
1 - 10 g
Specimen
X- ray Spectrometer Precise and accurate result
Sample Preparation
Definition of Accuracy and Precision
High AccuracyHigh Precision
High AccuracyLow Precision
02/11/2009Bruker Confidential10
Low AccuracyLow Precision
Low AccuracyHigh Precision
Sample Preparation for
Liquids, Powders and Solids
02/11/2009Bruker Confidential11
Sample PreparationPowders
Sampling
Grinding or pulverizing reducing the grain size and homogenising grinding vessel: WC, ZrO2, Stainless Steel
o hardness and application (contamination possibility)o cross contamination after cleaning, pre-grinding
grinding time grinding or binding aid
sieve diameter
Direct measurement in a liquid cup or...
02/11/2009Bruker Confidential12
binding aid pressure and pressure time
8/3/2019 LOD and Other Error Sources
4/17
Sample Preparation
Page 4 2009 Bruker Biosciences Commercial in Confidence
Sample preparation simple recipes, not rocket science
Pressed PelletsUseweighedamount of
Addgrindingtablets
Mill ingrindingvessel
sample
Finishedpowder
Pour intoSample
GetPressed
02/11/2009Bruker Confidential13
material Press Pellet
Effect of binder on analyte line intensity
Sample PreparationPowders
Sample PreparationPowders
minimum quantity is generally 5g
little quantities: small spot analysis,
r r
pressure as pure pellet, on a support
layer of boric acid, in steel rings or in
aluminum cups
binder if necessary
(contamination with binder elements!)
Mowiol (components: C, H and O)
boric acid (components: B, H and O)
wax (components: C and H)
r ind in a ids
02/11/2009Bruker Confidential15
Sample PreparationPowdersGrain- or particle size effect
analysed layerSiO2
Al2O3
02/11/2009Bruker Confidential16
ISi= f (NpSiO2)Volume
IAl = f (NpAl2O3)Volume
Analysed layer = homogeneous distribution
8/3/2019 LOD and Other Error Sources
5/17
Sample Preparation
Page 5 2009 Bruker Biosciences Commercial in Confidence
Sample PreparationPowders
Crystallographic- or mineralogical effect
SiO2
AlxSiyOz
pressed powdersloose powders
analysed layer
02/11/2009Bruker Confidential17
ISi= f ((NpSiO2)+(NpAlAlxSiyOz ))Volume
IAl= f (NpAlxSiyOz)Volume
SSiO2(kcps/%)SAlxSiyOz
Sample PreparationPowders as Pressed pellets
02/11/2009Bruker Confidential18
Quelle: Socachim 2006
Influence of particle size and pressure on analyte intensity
Sample PreparationPowders
Sample Preparation
Fusion technique
8/3/2019 LOD and Other Error Sources
6/17
Sample Preparation
Page 6 2009 Bruker Biosciences Commercial in Confidence
Sample preparation simple recipes, not rocket scienceFusions
Pour flux meltingagent into mortar
Weigh sampleand flux
Add sample toflux
Mix sample andflux material
02/11/2009Bruker Confidential21
Pour materialinto crucible
Heat crucible inmelting furnace
Casting mold withfinished glass bead
Sample PreparationFused beads
better reproducibility, accuracy
matrix and particle size effects areeliminated
Hom eneity is im roved
Flat surface;
Relatively fast (depends on)
Sample PreparationFused beads
What is fusion?
Chemical reaction based on kovalent binding betweenboron groups and oxides. The result is a glass bead.
Quelle: Socachim2006
Sample PreparationFused beads
Quelle: Socachim 2006
8/3/2019 LOD and Other Error Sources
7/17
Sample Preparation
Page 7 2009 Bruker Biosciences Commercial in Confidence
Sample PreparationFused beads
Quelle: Socachim 2006
Sample PreparationFused beads
Quelle: Socachim 2006
Sample PreparationFused beads
What is fusion?
Chemical reaction based on kovalent binding betweenboron groups and oxides. The result is a glass bead.
Quelle: Socachim2006
Sample PreparationXRF-analysis offused beads
minimum quantity in general: 1g
modern fluxes
lithiumtetraborate Li2B4O7
lithiummetaborate LiBO2 fusion in electrical, induction or gas
02/11/2009Bruker Confidential28
mixtures of Li2B4O7 and LiBO2
o eventually some LiBr can be added as
a non-wetting agent
platinum ware necessary
(crucibles and moulds)
, -automated
8/3/2019 LOD and Other Error Sources
8/17
Sample Preparation
Page 8 2009 Bruker Biosciences Commercial in Confidence
Sample PreparationFused beads
Quelle: Socachim 2006
Sample PreparationFused beads
What is fusion?
Chemical reaction based on kovalent binding betweenboron groups and oxides. The result is a glass bead.
Quelle: Socachim2006
Important parameters
Sample PreparationXRF-analysis offused beads
Particle size (< 100m)
Sample to flux ratio (Dilution varies to 1/30)
Type of flux (LiT,LiM, both)
Temperature ( 975C 1275C)
Time (3-10, 20 Minutes) depends on.
Furnace or fusion device
Oxidizing agent (Nitrates, Iodates)
Additives like Halides (NaI, LiBr, LiI)
Crucible and dish material (preheating is important)
Cooling and solidification
Additives
Sample PreparationXRF-analysis offused beads
Efficiency of fusion and casting
o LiF, B2O3, Li 2CO3
Non wetting
o Iodides, Bromine's,Perjodides, (Chlorides) of Li, NH4,
Heavy absorbers
o La, Ce, Ba oxides
Internal standards
8/3/2019 LOD and Other Error Sources
9/17
Sample Preparation
Page 9 2009 Bruker Biosciences Commercial in Confidence
Cooling and Solidification
Sample PreparationXRF-analysis offused beads
0C
Complete crystallization
LiMBNaTB
Crystallization Curve
Time
White Opaque Bead
Bead will crack as a result of stress
Cooling OK
Sample PreparationSolution for sample effects in Powders
Elimination of grain size- and mineralogical effects:
Grain SizeEffect
MineralogicalEffect
Grain sizereduction
YES NO
Pressed Pellet YES NO
02/11/2009Bruker Confidential34
Fusion YES YES
Sample Preparation
Preparation in Liquid cups
Sample preparation simple recipes, not rocket scienceLiquids
Cover bottom of liquidcup with transparentfilm
Check for holesPipette sampleamount to definedweight
02/11/2009Bruker Confidential36
8/3/2019 LOD and Other Error Sources
10/17
Sample Preparation
Page 10 2009 Bruker Biosciences Commercial in Confidence
Sample PreparationLiquids
Water, oil, fuel, solvents, slurries
He-device necessary
In principle liquids are homogeneous
Depending on sampling, measurement in liquid cups
or on filters (liquids with particles: filtering and
measuring filter and filtrate)
filters with hydrofobic ring
Dilution leads to intensit loss
02/11/2009Bruker Confidential37
Sample PreparationLiquids
element range: from Na to U
minimum quantity in general: 5g
selection of foil
resistance
(chemical attack by sample
material, embrittlement by X-
ray exposure)
"as thick/stable as necessary"
02/11/2009Bruker Confidential38
transm ttance absorption
LLD for light elements
"as thin/transparent as
possible"
Sample Preparation: LiquidsDegradation Resistance of Thin-Film Substances
Chemical Classification Mylar Polycarbonate Polypropylene Polyimide(Kapton)
Prolene Ultra-Polyester
Acids, dilute or weak G G E N G G
Acids, concentrated G G E N E G
Alcohols, aliphatic N G E G E NAldehydes U F E E E U
Alkalies, concentrated N N E E E N
Esters N N G G G N
Ethers F N N U N F
Hydrocarbans, aliphatic G N G G G G
Hydrocarbons, aromatic F N F F F F
02/11/2009Bruker Confidential39
Hydrocarbons, halogenated F N N F N F
Ketones N N G G G N
Oxidizing agents F N F N F F
E = Excellent, G = Good, F = Fair, N = Not Recommended, U = Unknown
Sample Preparation: LiquidsTransmittance Curves for Various Thin-FilmSubstances
02/11/2009Bruker Confidential40
8/3/2019 LOD and Other Error Sources
11/17
Sample Preparation
Page 11 2009 Bruker Biosciences Commercial in Confidence
Sample PreparationSolids
Types of solid samples:
Metals & Alloys: Steel, Aluminum, Zinc,
Plastics: PE, PP,
Rocks/Large pieces of metals (Fe-alloys)
Odd shaped Solids
The aim is obtaining a surface that can easily be measuredwith XRF
02/11/2009Bruker Confidential41
Sample PreparationSolids
Metals & Alloys
. . Resurfacing: depends on the type of material
Grinding: refer to powders
Turning: for soft materials: e.g. Pb and Al
Polishing: for hard materials: e.g. steel Important parameters:
Hardness of material
Grit size
02/11/2009Bruker Confidential42
Cross contamination
Sample preparation simple recipes, not rocket scienceMetals
Clamp metal cut Check sample heightat molding cutter
Mill Check surface
02/11/2009Bruker Confidential43
Sample PreparationSolids
Plastics as hot pressed tablet or as grains in a liquid cup
Rocks/Large pieces of metals (Fe-alloys) Reduce size and treat as powder
Odd shaped solids Dissolve in acid or make as a glass bead
Small s ot calibration: direct anal sis
02/11/2009Bruker Confidential44
Standardless Analysis
8/3/2019 LOD and Other Error Sources
12/17
Sample Preparation
Page 12 2009 Bruker Biosciences Commercial in Confidence
Sample PreparationInfluence of bad sample surface
R( k c ps) = f ( 1 / d 2)
Non flat samples
Concave (Distance, Shadow effect)
Convex (Distance, Shadow effect)
Surface finish (Shadow effect)
= part of spectrum not
being irradiated,
especially for lighter
elements
Sample PreparationConclusions
Sampling and sample preparation depend on many
parameters
Important physical effects have to be taken into account for
the specimen preparation
Powder- and liquid samples can be prepared in a relatively
simple way
Different preparation techniques exist to prepare specimens
46
Sample preparation is always compromise: costs, time tospend, analytical tasks and available lab stuff
Sample PreparationThe one commandment
Whatever sample preparation method you
choose, verify its reproducibility before you
start preparing the calibration standardsSources of Error
02/11/2009Bruker Confidential48
S l P i
8/3/2019 LOD and Other Error Sources
13/17
Sample Preparation
Page 13 2009 Bruker Biosciences Commercial in Confidence
Sources of ErrorIntroduction
Possible sources of error:
Sample taking
Samp e preparation
Instrument precision
Counting Statistical Error (CSE)
Chemical data of standards
Inaccuracies during calibration
===> Total error in the end result of the analysis of the unknownsamples
02/11/2009Bruker Confidential49
Sources of ErrorTypes of error
Random Errors: non-systematic fluctuations Equipment errors
o Generator and/or X-ray tube instability
Operator errorso Bad sample preparation
o Count ing s ta t i s t i cs (time dependant)
Systematic Errors: a consistent deviation Equipment errors
o Errors in calibration (model, corrections)
02/11/2009Bruker Confidential50
o Dead time losses Sample errors: Absorption, enhancement , particle effects, chemical state
Sources of ErrorTypes of error
i
(Rp)i = Si C(%)i + (Rb)i
p i
X
Si Random Error
Systematic Error
02/11/2009Bruker Confidential51
(Rb)i
Concentration (%)i
Sources of ErrorSensitivity
The Sensitivity Si of the spectrometer for an element i
= the number of detected X-ray photons N (counts)
divided by the measurement time t
divided by the concentration Ci of the element i
( kcps / % or kcps / ppm )
02/11/2009Bruker Confidential52
i
i
Ct
NS
=
S l P ti
8/3/2019 LOD and Other Error Sources
14/17
Sample Preparation
Page 14 2009 Bruker Biosciences Commercial in Confidence
Sources of ErrorThe Counting Statistical Error (CSE)
r = counts per second (cps)
CSE
N
NCSE
=
=
%100(%)
%100(%)
r =S . C
02/11/2009Bruker Confidential53
trCSE
=
%100(%)
Longer counting time,better relative CSE
Sources of ErrorThe Counting Statistical Error
For N = 100
N +/- 1 * SQRT (N) 1 68.3 % 90 - 110
N +/- 2 * SQRT (N) 2 95.5 % 80 - 120
N +/- 3 * SQRT (N) 3 99.7 % 70 - 130
02/11/2009Bruker Confidential54
Sources of ErrorThe Counting Statistical Error
The precision is limited by the CSE, mostly expressed as %CSE(relative value)
This means for 3-sigma:
= 100 SQRT (N) = 10 %CSE = 30 %
= 1000 SQRT (N) = 30 %CSE = 10 %
= 10 000 SQRT (N) = 100 %CSE = 3 %
NCSE
%100(%) =
For 1-sigma
02/11/2009Bruker Confidential55
= 100 000 SQRT (N) = 300 %CSE = 1 %
= 1000 000 SQRT (N) = 1000 %CSE = 0.3 %
= 10 000 000 SQRT (N) = 3000 %CSE = 0.1 %
Sources of ErrorThe Counting Statistical Error Sensitivity
N = S . t . C A high precision measurement
Counting statistical error (%)
u
a low statistical error
a high number of counts
therefore you need
an instrument with high sensitivity
or
long measurement times
3
%1003 =N
N
02/11/2009Bruker Confidential56
%1003
=
=
CtS
N
Sample Preparation
8/3/2019 LOD and Other Error Sources
15/17
Sample Preparation
Page 15 2009 Bruker Biosciences Commercial in Confidence
Sources of ErrorMaximum Count Rate
The maximum count rate is limited by the detection system:
--
20 - 50 kcps
for the whole spectrum
02/11/2009Bruker Confidential57
Sources of ErrorMaximum Count Rate
Example 1:
major element
e.g. Cu in brass
0.1% counting statistical error
10 000 000 counts
Minimum measurement time given by maximum of detector countrate:
ED-XRF: 200 - 500 s
02/11/2009Bruker Confidential58
Sources of ErrorMaximum Count Rate
Example 2:
1 % Ni in steel
1 % counting statistical error
100 000 counts
the concentration of Fe is close to 100%
the intensity of Fe is about 100 times higher than the intensity of Ni
02/11/2009Bruker Confidential59
Sources of ErrorMaximum Count Rate
ED-XRF with Si(Li)-detector
maximum total signal 20 - 50 kcps
maximum Ni signal 200 - 500 cps
minimum measurement time 200 to 500 s
02/11/2009Bruker Confidential60
Sample Preparation
8/3/2019 LOD and Other Error Sources
16/17
Sample Preparation
Page 16 2009 Bruker Biosciences Commercial in Confidence
Sources of ErrorThe Lower Limit of Detection (LLD)
e ana y ca per ormance ortrace elements is determined bythe peak-to-background ratio
typically the background in XRFis primary radiation scattered bythe sample
02/11/2009Bruker Confidential61
Sources of ErrorThe Lower Limit of Detection
the peak-to-background ratio is
equal in both diagrams
the measurement shown bottom left
is clearly better because the nett
signal is better separated from the
background signal
What parameter can
describe this difference ?
02/11/2009Bruker Confidential62
Sources of ErrorThe Lower Limit of Detection
Lower Limit of Detection (LLD)e ne as:
the concentration giving
a net signal equal to
3 times the background noise
background noise
= CSE of the background signal
02/11/2009Bruker Confidential63
tcpsBKG = )(
Sources of ErrorThe Lower Limit of Detection
Lower Limit of Detection in ppm
The LLD is typically defined for t = 100 s (S in cps/ppm)
tcpsBKG = )(3
tLLDCSsignalNett = )(_
02/11/2009Bruker Confidential64
tSppmLLD )( =
Sample Preparation
8/3/2019 LOD and Other Error Sources
17/17
Sample Preparation
Page 17 2009 Bruker Biosciences Commercial in Confidence
Sources of ErrorConclusions
Two main types of error: random and systematic
Equipment, sample and operator are the largest sources of error
Sensitivity and measurement time are very important parameters in
the counting statistic
Good counting statistics means reaching low limits of detection
Many parameters have to be taken into account in order to obtain a
good analytical result
02/11/2009Bruker Confidential65