Raman spectroscopyof silicate glasses and melts
in geological systemsStéphanie Rossano & Bjorn Mysen
Lava flow (Hawaï)
Melting of crystallinematerials
1100°C
≈ 30°C
Fast cooling (quench)to avoid
crystallization
Glass/melt science : the scientific issues
Geological interest Information on the Earth historyFormer historyMelt differentiationDynamic behaviourEarth evolution
Build / constrain / complete geochemical models
Stromboli (july 1999) Cré
dit p
hoto
@ S
. Ros
sano
Moldavite (15 Ma) Cré
dit p
hoto
@ S
. Ros
sano
viscosity, colouration,nucleation, … can bemodulated by a change :• in composition• in thermal treatment
Link composition / thermal historyto physico-chemical properties
Glass/melt science : the scientific issues
Technological interest
Crédit photo @ S. Rossano
Glass/melt science : the scientific issues
Cultural heritage interest
Understand alteration processesForesee glass durability
PhD J. Sterpenich (1998)
condensation
Meteoric water
Les-Noës-Près-Troyes (Aube)
Flow of asbestos« lava »
Flow of nuclear waste« lava »
Glass/melt science : the scientific issues
Environmental interest
Slow coolingRapid cooling
Liquidstable
Glassout of equilibrium
Supercooledliquidmetastable
Crystalstable
mMelting temperatureGlass transition
temperature
Glass vs crystal ?
Glass vs crystal ?
Crystalline CaFeSi2O6Fe, Ca, Si, O
Glass of hedenbergitecompositionFe, Ca, O, Si
No periodicityNo more special atomic planes / directions
Increase of « disorder »
Qn species
Q4
Q2
QnTetrahedralentities centredon Si4+ or Al3+
Number of bridgingoxygens
Distribution of environments
Na2O-4SiO2 (classical MD)
Na, Si, O
Bridging O
Nonbridging O
Glass structure
Q2
Chargecompensationneeded for the
NBO
• competition betweencations• Distribution ofcationic environments
FeCaOSiNaAl
Tektite (Classical MD snapshot)
Complex compositions
FeCaOSiNaAl
Tektite (Classical MD snapshot)
Probing glass structureLocal order
Selective techniques 57Fe Mössbauer,
XAS,NMR…
Raman spectroscopy
• Species• Connectivity of the
glass network
1st order Raman spectrum +Al3+-O + P4+-O + Fe3+-O…
Qn
vibrationsVolatile vibrations CO2,CO, CH4, H2, H2O, N, S…
200 420028001300600
Alkaline earth - OAlkaline - OO - transition metal
800
Regions of interest
Network connectivity
Pre-treatment steps
Modified from Malfait & Halter (2008)
• Comparison between a seriesof spectra
• Extraction of quantitativeinformation
Pre-treatment steps
Baseline determination / substraction
• Spectrometer dependent
Modified from Malfait & Halter (2008) 200 400 600 800 1000 1200
K2O-MgO-2SiO
2
Raman shift, cm-1
• Spectrometer dependent• Fluorescence signal
Pre-treatment steps
Baseline determination / substraction
Modified from Malfait & Halter (2008)
• Spectrometer dependent• Fluorescence signal• Glass composition
Mercier et al. (2009)
Pre-treatment steps
Baseline determination / substraction
Modified from Malfait & Halter (2008)
Pre-treatment stepsT / ν correction
Modified from Malfait & Halter (2008) Hass, 1970
€
Icor = Imeasω
(ω0 −ω)4 (1− e
−hvω /kT )
Corrections necessary at high T or low frequency region
Pre-treatment steps
Daniel et al. 1995
CaAl2Si2O8 @2009 K
(1) Exp. spectrum
(2) Measured Thermalemission (laser off)
(1) - (2)
€
Icor = Imeasω
(ω0 −ω)4 (1− e
−hvω /kT )
Daniel et al. 1995
CaAl2Si2O8
Liq.
SupercooledLiq.
Glass
-
T
T0
Na2TiSi2O7
Reynard & Webb 1998
The Long formula does notcorrect for T dependence ofsd order Raman scattering
€
I (2) (ω) =Icorr (T )− Icorr (T0 )
n T − n T0
€
Icorr (T )− I(2) (ω)
€
Icorr (T )
€
I (2) (ω)
Daniel et al. 1995
Minimize misinterpretations
Changes observed are due toconfigurational changes oranharmonic effects
Pre-treatment stepsNormalization
Modified from Malfait & Halter (2008)
Comparison of relativeintensities among spectra of asame family
• in a region with no Raman signal• to the most intense point• to the integrated Raman signal
Comparison between studiesdifficult
Tsujimura et al., 2004
Fingerprinting
Identification of Ramanbands due to dissolved
species
1st order Raman spectrum +Al3+-O + P4+-O + Fe3+-O…
Qn
vibrationsVolatile vibrations CO2,
CO, CH4, H2, H2O, N, S…
200 420028001300600
Alkaline earth - OAlkaline - O
O - transition metal
800
Isolated SO42- S6
2- Isolated SO42-
Quantification (external standard)
Lenoir et al., 2009
• Magmatic systems• industrial glasses : refining agent
Without refining agent2 Na2SO4 2Na2O + O2 + 2 SO2 With refining agent
Calibration lines are composition dependent
Sulfur At 1450°CA bubble of 0.1 mm140h = elevation of 1m
Lenoir et al., 2009
Application to melts
Fingerprinting
• Identification of pigments• Classification of samples
Tournié et al., 2011
Fingerprinting
Brawer & White (1977)
Comparison between glass and crystal
200
400
600
800
1000
0 200 400 600 800 1000 1200
verre CaMgSi2O6cristal CaMgSi2O6
cm-1
glass & crystal of CaMgSi2O6 composition
Statistical analysis
• Decomposition intoGaussian components• Least squares refinement• A posteriori assignment
Na2O-MgO-8SiO2
Modified from Trcera et al. (2011)
Polymerization
Na2O-MgO-xSiO2 with 2 < x < 10
Trcera et al. (2011)
Please : verify that theenvironment around atoms donot change
Trcera (personal com.)
1290 1300 1310 1320 1330 1340 1350
NMS2 norm.NMS3NMS4NMS6NMS8NMS10
Rel
ativ
e in
tens
ity
(arb
. uni
ts)
energy (eV)
Mg K-edge
Guide to interpret the trends Polymerization
Controversies : • Shape of the individual bands• Number of components• Assignments
Quantification of Qn species
29Si NMR spectroscopy
Not easy for meltsFrantz & Mysen (1995)
Statistical analysis
• Principal components analysis
• Linear combinations of partialspectra
• Optimization procedure
No need of external calibration melts
Necessity of a large number ofspectra in the same glass family
Number of independentcontributions
Internal standard for Qn
quantification
Statistical analysisExtraction of individual signals
assigned to Qn species
Malfait et al., JNCS 353 p4029 (2007) Zotov (1999)
Toplis & Reynard (2000)
melt dynamics0.9Na2O-0.1Al2O3-2SiO2+ 6mol% P2O5
Na3PO4 Na4P2O7
Reaction extent : A910-950/A800-1350
Phase separation
Conclusions and perspectives
• Raman has been extensivelyused to try to get insight intoglass connectivity
• Simple experiments
• Difficulty to analyze /interpret the spectra
GLASS MELT
Zotov, 1999 ; Spiekermann, 2012
• Quite easy to perform ascompared to other techniques(such as NMR for ex.)
• Access to the dynamics /kinetics of melt transformation
• Necessity to correct for sd
order Raman effect
Calculation could help to assign the peaks