A~om~lies, Artifacts and Common Errors in Using VIbratIonal Spectroscopy Techniques

Mid-infrared Spectroscopy: Anomalies, Artifacts and Common Errors John M. Chalmers University of Nottingham, Nottingham, UK

1

INTRODUC~ION ". .

. .

Anomaly (rnegulanty, dev1ation from rule"),l artifact/artefact ("a thing made by human workmanship"),l and error ("deviation from the right way, blunder, mistake,

~rong-doing"): 1 three nouns (and their dictionary definiti?ns).that when applied to an infrared (IR) (or indeed any v1brational)spectrumor spectralfeatureimmediately imply that the data are imperfect. that the data were derived in an imperfect way, or that the data are impure. There are many sourcesthat give rise to such observationsand comment. Many are derived from poor experimentalpractice (e.g. poor grinding of the analyte in an a1kali-halid~disk preparation),somefrom inappropriatesampling,somefrom contamination (e.g. silicone greasefrom vacuum apparatus), others from instrumental limitations (e.g. too Iowan angle of incidence in an attenuatedtotal reflection (ATR) measurement),some from the environment (e.g. intrusion of absorptionsdue to atmosphericwater vapor and carbon dioxide), while othersclassified similarly merely reflect the fact that th~ measurementwas made at an extreme (e.g.' photoacousticsaturation).Somemay merely be a nuisance andnot detractfrom the measurementpurpose,while others are disastrousand can lead to erroneousconclusions. ?nderstanding, appreciatingand recognizing anomalies, artifacts and common errors is vital if data are to be interpreted and quantified correctly. The intention in this article is to present examples of common imperfections observedin mid-infrared spectraand to discussconcisely their .ori~in, u~lizing, wherever possible, example spectra to highl1ght 1ssues.It cannot hope to cover all such

occurrences,but will hopefully instil precautionand caution in experimentalistsand researchersnew to the technique.

2

ATMOSPHERIC

INTRUSION

Perhaps the most readily recognized imperfections in any IR spectrum, whether, for example, it is a transmission, emission, or photoacoustic measurement, are those absorption bands arising from the presenceof atmosphericmoleculesin the path of the IR beambetween the source and the detector. (This, of course, excludes circumstanceswhere they are required for detection or calibration.) A ro-vibrational mid-infrared transmission spectrum depicting these features is shown in Figure 1. It comprises absorption bands due to both water vapor ~d c.arbon.dioxide. Their spectral contrast.and relative Intens1ty will depend on the spectral resolution of the measurement,the relative concentrationof the two gases, and,in someexperiments,perhapstheir partial pressureand te.m~rature. (~cluding this imperfection may seemoverly trIV1alto ~xpenencedspectroscopists;however,I have seen changesIn the relative intensity of the CO2 bands near 2350cm-l reported in a conferenceproceedingsabstract as being indicati:e of a change in the level of cure in a polymer compos1te!)These bands due to water vapor and carbon dioxide when seenin a Fourier transform infrared (FT-IR) .spe~trum.may be observedas either with positive or ~egative Intens1ty,dependingon the relative extents to ~hich they were present in the sample and background sIng!e-beams~ctra of an Fr-IR spectrummeasurement. SInce, there 1Sno change of dipole moment associated with the stretching frequenciesof homonucleardiatomics, then neither nitrogen nor oxygen give rise to IR absorption

I

2328 Anomalies,Artifacts and CommonErrors in Using Vibrational SpectroscopyTechniques -

Common interference

100

§ 80

nomenclatures fringes, channel

for such a spectral fringes/fringing,

feature are and channel

:~ !;;

60

spectra (see Glossary to the Handbook): The name used often reflects the favorite of a fringes" particular example, the term "interference is community. most commoFor

a

40

usedto describethe effect when observedin thin film po

~ ~

20

mer spectra, they originateand as aofconsequence of sample beingwhere thin, non-scattering uniform thickn

H2O

0 4000

3500

3000

2500

2000

1500

"Channel spectra" useddescriptorfor . is the frequently .

0

Wavenumber / cm-1

~

Figure 1. transmission.spectrum showing abs°fP.tio~bands characteristic of atmosphenc watervaporandcarbondioXIde.The 4 cm-l resolutionspectrumwasgenerated by ratioingFr -IR open single-beam spectrarecordedwith thesamplecompartment of the spectrometer unpurgedagainstthatwith the compartment purged with dry nitrogen. . .. bands. Their presence, ~ong with other e~v~onmenta1 phenomena,such .asCOSmiC rays ?r !amp emiSSiOnS, may however be sometimesdetectedwithin Ramanspectra(see Anomalies and Artifacts in Raman Spectroscopy).-

appearance in high-resolutIon spectra, where .. gas-phase fr all 1. . are instrument related, ansmg om par e Ism WIthin interferometer beam-splitter/compensator assembly.

. . . ever their source, they anse ~om. an mterferencepatt generatedbetween the recombmationof two coherent beamsthat have travelled a different pathlength. The effect is perhaps most easily illustrated in practice of determining the pathlength of an empty transmissioncell, as shown in Figure 2. When the opti path difference between the two beams equals A./2 destructive interferencewill occur, where A.represents wavelength. At optical path differences equal to inte multiples of a wavelength, then the two beams will

3

SUPERIMPOSITION SINUSOillAL

OF A

in phase and constructive interference will occur, i.e.

intensitieswill be additive. The fringe separation,freque of the waveform, for two differing pathlengths Is

WAVEFORM

. . . The appearanceo~ a smusoldal waveform overl~ymg .an IR ~pectrum .IS yet another commo~ cause ~f Imperfection that IS generally well recogmzed, and IS usually most noticeable in paseline regions, where there are no absorption bands. The frequency and amplitude of the sinusoidal wave will depend on its origin, and its intrusivenessmay depend also on the spectral resolution. Under some circumstances,the amplitude may change with wavenumbermost likely increasing with decreasing wavenumber(increasingwavelength).

shown in Figure 2. Figure 3 illustrates the waveform might be observed from a slightly wedged source the interferencepattern. Figure 4 shows two examples fringing observedin IR transmissionspectrarecordedfr thin polymer films.

4 STRAY LIGHT For stray light, the Glossaryto this Handbookcontains two definitions following: "Radiation that does not foIl r

100

~

80 c 0 "00

.!!l (/) E

a

60

..

40

I~

IR

.

.

1

I

cm-

VVl

AAAAf\M

~... 20 0 4000

Empty cell

3500

3000

2500

2000

1500

1000

500

Wavenumber/cm-1 2. Overlaid 4cm-l resolution, transmissionIR spectrarecorded from empty (air gap) transmissionIR liquid cells. The p s of the air gaps were: 12.5~, low-frequency sinusoidal wave; lOO~, high-frequency sinusoidal wave. A schematicof processproducing interferencefringes in the IR spectrumrecorded from a thin, empty transmissionliquid cell is shown in the ins

~~I~\;

I

ct.

~';"f -:;;i Y(

;,~ ,I4}

" Mid-infrared Spectroscopy: . - - Anomalies,Artifacts and CommonErrors 2329 J

-~

from the likes of a bubble in a liquid in a cell or a polymer film in a transmissionmeasurement. IR radiation that has travelled the conventional path and reached the detector without having passedthrough the sample will adversely affect the spectral contrast of any measurement.It will probably have seriousdeleterious consequenceson any quantitative determinations made from the spectrum.It causesloss of spectralcontrast, and can affect bands of differing absorptivities to different extents, Unhindered stray light, that is radiation that

,~

WavenumberI cmFigure 3. Tr~smission spectrum r~co~dedfrom an, empty

bypassesthe sample and effectively passesstraight from the sourceto the detector,will shift the zero position of the

ii !:

wedged nolDlnal 50 I1In pathlength liquid cell. Loosemng the retaining screws at one end of the cell assembly produced the non-parallelismof the cavity.

transmission scale on the measured spectrum.

100

§ 80 :~ 60 ffi

~ ~

40 20

'"

0 4000

3500

3000

2500

2000

1500

1000

500

1

'.

100

100

§ 'w

80

'E

U)

c ,-0

U) .!!!

60

~

40

E U)

U) c ~

I-

~so

~

~ 20

}

0

.

4000 3500 3000 2500 2000 1500 1000 500

O.

(a)

4000

3500

3000

2500

2000

1500

1000

88,56 100

c ,2

80

'E

U) U)

0 ,-

U) c

U)

(\I

,!!! 60 ffi

c I-~ ~

Wavenumber/cm-1

500

(a)

c

~

~

40

4.75 4000

20

(b) (b)

0 4000

3500

3000

2500

2000

1500

Wavenumber Icm-1

1000

500'

100 c ,~80

iilterference fringes recorded from polymer filIIis: (a) 20 I1In

'~

thickness

~

/

(b) 2.5111n thickness

poly(ethylene

terephthalate)(PET). the usual path through a spectrometer and consequently

~ I~

appears in a spectrum at a wavenumber different from its true wavenumber. The term is also used to refer to

radiation that passesaround the sampleinstead of through

3500

3000

2500

2000

1500

1000

500

Wavenumber/cm-1

Figure 4. TransmissionIR spectra, 4 cm-l resolution, showing polypropylene;

60 40 20 0 4000 3500 3000 2500 2000 1500 1000 500

(c)

Wave~umber/cm-1

it, and consequently is not modifiedby the sampleand

Figure5. IR transmission spectra, 4cm-l resolution,recorded

seriously affects absolute and relative intensities." In this section we will be considering the second classification , . ,. .error. It will be broadened,however, ., for .the measurement

from I, I ,2,2-tetracWoroethane in a 12111npathlength liquid cell, The spectrumplotted as full ordinate scale expansion shown as (b) was recordedfrom a specimenthat had a few small bubblesin it; the li qw' d compIete1y fill ed the ce11caVlty when the spectrum '

to mclude a discussion of examples where the radiation has passed through the sample, but the sample was not continuous. The straylight phenomenon in this case arises

shown as (a) was recorded.The spectrum shown as (c) is from the samerecording as for (b) but plotted on an ordinate scale of 100-0% transmission,

~-~"

J;

'

2330 Anomalies,Artifacts and CommonErrors in Using Vibrationa Figure 5 shows tWo transmission spectra recorded from the solvent 1,1,2,2-tetrachloroethanecontained in a 12-11Inpathlength IR transmission cell. Neglecting the differencesas a consequenceof differing contributionsfrom atmospheric water vapor and carbon dioxide, these two spectra look essentially identical, and each could readily be used to characterizethe solvent. The lower spectrum was output to a recorder using an automated software commaQdthat setthe minimum and maximum transmission values to full scale. This is not an uncommon practice.

but thi layer these Figure this. occurr The that e

The ordinatevaluesof Figure5(a) extendfrom 100to 0% transmission, while thoseof Figure5(b) only extendfrom

c: 80 .,9;

88.56 to 4.75% transmission.While the maximum value is

.~ 60

of no significantconsequence here..the minimumis! The"

§ 40

100

spectrumof Figure 5(b) was generatedfrom a samplethat ~ had a few small bubbles in it, that is the solvent did not ~ 20 fully fill the capillary spacebetween the windows of the transmissioncell, such that some stray light fell onto the ~ooo detector.Figure 5(c) showsthe spectrumof the sampleused (a) to generatethe spectrumof Figure 5(b), but this time with the ordinatepresentedas a full transmissionscale.The shift ,. 100 in the true zero is now clearly apparent,and the disastrous" d.l 1 . b d . .. bec consequences

on re ative

an

IntensIties

orne rea I y

observedin Figure6, wherethe absorbance ordinatescale;

c: 80 .9: II)

60

,.~

equivalentsof Figure 5(a) and 5(b) are compared.A similar( catastropheis shown in Figure 7(a) and 7(b), from a pair

~

of transmissionspectr,.~generated from the s~e solvent,

~ 20

;,::;;'~

~

40

..

0

2.5 aJ

4000

2.0

(b)

u

~c:

1.5

2.0

g 1.0 .c ~

aJ 1.5

g

0.5 0.0

1400

1200

1000

800

600

400

0.1

II)

00 .

-e 1.0 0 ~ 0.5,

1.4

0.0

1.2

~ .c

(c)

0.8

Figure .

(; 0.6 :fJ ~ 0.4

non, r betwee record

0.2 0.0

specim in the 1400

(b)

1200 1000 800 Wavenumber/cm-1

600

400

Figure 6. The IR spectra shown as Figure 5(a) and 5(c) but plotted as absorbancespectraover the range 1500-400cm-l as (a) and (b), respectively.

,-,._-

[ :::~::J .

0.3 0.2

(Ij

(a)

~ 1.0

1000

~5~~ the b large i line s to (b).

:: .'~

950

/

900

~ ::

Mid-infrared Spectroscopy:Anomalies,Artifacts and CommonErrors " 60

2331

a few small bubbles that were generatedby molding the film sampleat too high a temperature.The sampleusedfor Figure 8(a) preparedunder optimal molding conditionswas

§ 50 40 o~ 30

continuousand containedno visual bubbles.The deleterious consequencefor quantitative measurementsfrom such a non-continuoussample is illustrated in the comparisonof the two spectramadein Figure 9. While undoubtedlysome of the differencesin relative band intensities betweenthe

°i?;

ffi

~ 20 ~ 10 0 4000

3500

3000

(a)

2500

2000

1500

1000

spectra occur as a result of the differing thermal histories of the two specimens,the bubbles in the film sample are also responsiblefor someof the variation.

500

Wavenumber/cm-1 60

5

CONTAMINATION

To many readers,contamination may seem too obvious a form of spectrumimpurity to include as a section in this article, but sourcesthat are obvious to many are still novel to some, and indeed have been overlooked in research publications by experts (although no referenceswill be 0 quoted for these!). In the context here, "contamination" 4000 3000 2500 2000 1500 1000 500 refers to materials that give rise to absorption bands (b) Wavenumber/cm-1 that may be misinterpreted as being attributable to the Figure 8. IR transmission spectra,4cm-l resolution,recorded sample in its expected form. The examples here cannot from two compression moldedfilm samplesof polypropylene. be exhaustive; they will vary widely among different Spectrum(a) was recordedfrom a .specimen ,;,,~thout bubbles, application areas. The three selected as illustrations are whereas(b) was recordedfrom ~specllllencontainIng a few small from my experienceand encounters.They will hopefully bubbles.Seetext for moredetalls. . .ti. h 1 . ill di . tin f I kin all

e

,.

ms

a

SCIp

eo

00

g

at

a

spectrum

mI

.

y

WI

th

some lateral interrogation as opposed to solely focusing on the problem in hand or anticipated/predictedresult (as clearly happenedwith the publications alluded to above).

!

"

01050

p

950

900 W

850 b

avenum

/ er

800

700

-1 cm

Figure 9. Overlaid absorbance spectra over . ~e range 1050- 700cm-l generated from the two tranSmiSSIOn spectra.

shownin Figure8. The dashedline is for the continuousfilm, andthe solid line is for the film thatcontainedbubbleso

were recorded from samples of a polypropylene film, each of a thickness of about 0.15mm, and both serve as excellentfingelprint spectraof the material. The spectrum of Figure 8(b) extends to 0% T, since the thickness of polypropylene through which the IR radiation passedwas such that it provided extinction for bands with high absorptivities. However, while no IR radiation bypassed the sample,it was recorded from a sample that contained

c

Residual solvent in a sample cast from solution is an obvious source of spectral impurity, and attributed absorptionbands are usually readily observable.However, some samples may deposit in a form that contains the solvent as "solvent of crystallization". For more on this and similar spectralmanifestations,the readeris referredto the article on polymorphism in Volume 5 of this handbook (Polymorpbs, Solvates and . Hydrates). .. Mer

a

few

preparatIons,

most

expenmentalists

have

grown accustomed to th e lffipro . babili.ty 0f ever m aki ng a KEr disk preparation that is entirely free from absorbed .

water and observmg the consequent broad absorption vOH at ca. 3400cm-l and maybe the 8HOH band near 1640cm-l. However, to the unknowing there is potentially a nastierdangerlurking in commercialsuppliesof the alkali halide..This lies in contaminationby the nitrate ion, NO3-. Unless one purchases"spectroscopic grade" KEr, one is very likely to find the materialcontaminatedwith low levels of KNO3. NO3- gives rise to a narrow band with very high absorptivity near l380cm-l; it may be confused,for instance,with the deformation mode of a -CH3 group that might be presentin the analyte of a KEr disk preparation.

2332

Anomalies, Artifacts and Common Errors in Using Vibrational Spectroscopy Techniques -

100 c: .2 w

80

.E w

60

w

c: tIS 40

c:

.-0 .-~

\

~

tIS '-

I-

~ ~

90

E

~ 20 0

NO3-

4000 3500 3000 2500 2000 1500 1000 Wavenumber/cm-1

80 4000 500

(a)

Figure 10. JR transmissionspectrum of a KBr disk preparation made using "non-spectroscopic grade" KBr; the absorbance peak near 1380cm-1 due to nitrate is clearly evident. The backgroundspectrumfor this FT-JR measurementwas madeusing spectroscopic-gradeKBr. (Some interference fringes from the disk are evident to?, more strongly to.wards~owerwavenum~er.

§ 80 .~ -

Other weak absorptionbandsare assocIatedWIththeanalyte beIng examined.) , l1le specbruunshovvn in Figure 10 vvas generated recently (vvhile this handbook vvas in preparation) as a consequence

.E

60

~

40

~

20

~

(b)

0 4000 3500 3000 2500 2000 1500 1000 500 Wavenumber/cm-1

of an attempt by a laboratory technician in an academic institution to find an alternative supply of KBr in a teaching -laboratory. .1. . th e lorm l' S.1li cone 01.1 and Sllcone grease rn 0f

Figure 12. JR, 4cm-1 resolution, transmissionspectrarecorded from: (a) a deposit on a KBr plate generatedby touching one surface.of the plate v:ith a finger while wearing PVC gloves that were slightly wet WIth chloroform; (b) a reference sample phthalate plasticizer. .' of a

poly(dimethyl siloxane) (PDMS) are potentially major sources of contamination, and anyone not involved directly in characterizing silicone materials should be very suspicious of any spectrum that has a sharp band at 1260cm-l. It is a distinctive feature of PDMS, and its presence or otherwise may be ascertained by reference to its. other characteristic absorption features shovvn in Figure 11.

invasive contamination from poly(vinyl chloride) (PVC) products, such as PVC piping and protective PVC gloves (see Figure 12(a)). An example of a phthalate plasticizer spectrum is shovvn in Figure 12(b), the vC=O at about 1725 cm-l and accompanying vveak sharp doublet near 1600 cm- f are usually the give-avvay features.

PDMS is potentially a common source of contamination for those using vacuum apparatus and desiccators. I hav~ seen _bandsof PDMS assigned to other species in both a

6

Pill thesis and a research publication in a highly respected spectroscopic journal. In a similar context, "phthalate plasticizer" (a di-alkyl phthalate) is a common source of 100 .§ 80 .13

60

~ c: ~ 40

';

20

-

0 4000 3500 3000 2500 2000 1500 1000 500 Wavenumber/cm-1 Figure 11. JR,4 cm-1 resolution,transmissionspectrumrecorded from a capillary layer of silicone vacuum grease.The spectrumis that of PDMS.

ANOMALOUS

DISPERSION

A key intrinsic property of a sample to mid-infrared measurements, particularly those made by a reflection technique, is its refractive index. l1le refractive index, n, absorption index, k, are interrelated through the Fresnel laws of reflection and the Kramers-Kronig (K-K) relations~p; see, for example, External Reflection Spectroscopy by Claybourn in this handbook. As one scans through an absorption band, then the refractive index of the sample changes from the average value at positions either of the band, vvhere the sample does not absorb IR radia"'

tion. For organicmaterials,this changeis usually as a lovveringof the samplerefractiveindex to the

vvavenumber side of the absorption band maximum, returning to the average value at the absorption band center, increasing to the lovver-vvavenumber side, before returning again to the average value. This is knovvn as anomalous dis. persion and is illustrated in Figure 13. This dispersion in a

Mid-infrared Spectroscopy:Anomalies,Artifacts and CommonErrors 2333 /

18 E as ~ -

c: as

0

()

"-

"in

Q

.c

~(!)

.

c..

12

a: ~ 6 Wavenumber

/cm-1

Figure 13. Representative spectra superimposed R, absorption index, k, and refractive index, n. ,

..

..

sample s refractive

mdex,

which

mcreases

of reflectanc~,

... w;1th mcreasmg

absorp~yity, can feature in many reflection measurements of ~d-infrare~ spectra, and ~s a consequence it is important that Its effect IS fully apprecIated. An important consequence of anomalous dispersion is the Christiansen effect that may be observed readily in the spectrum recorded from a non-optimal specimen preparation of a powder sample as a mull or an alkali-halide disk (see next section). As examples in this section, we will discuss mid-infrared measurements made by the transflection and internal reflection spectroscopy techniques. I In a transflection experiment, the measurement is usually made from ~thin (typically 0.5!o a few micrometers thickness, t) continuous, non-scattermg sample, often a polymer film d ' . d .. ,

epOSIte

.

on

a

non-IR

absorbmg

..

reflective

substrate,

such as a polished metal surface. The radIation IS mcident

0 2000

1600

1200

800

400

(a)

c:

~ ~

£

1

~ x"

.g "~ I :§.

~

~

2000

1800

1600

(b) Figure

coatIng .

1400

1200

1000

800

600

Wavenumber/cm-1 14.

(a)

Transflection

on a bakin gay. tr

IR

spectrunl

recorded

(b) R efl ec n" on IR

from

a

PES

spec tru m recor d ed

from the surfaceof a thick molding of PES, and the subsequently derived K-K algorithm absorptionindex spectrum.

on the sampleat near nornlal (low angles)of incidence,e, and the detectedradiation is a compositeof radiation that hasbeenspecularlyreflectedfrom the front surfaceand that which has passedthrough the sample and been reflected from the reflective substrate.The latter contribution is usually dominant,and the recordedspectrumresemblesclosely that of a transmissionspectrumof a thicknessequivalentto 2t/ cose', where e' represents the angle of refraction within die samplelayer. However, in regions of high absorptivity, that is for strongly absorbingbands,then the specularcomponentmay becomepronounced,particularly if the sample thicknessis low. Figure 14 shows a transflectionspectrum recorded directly from a non-stick coating on a baking tray. The significant effects, that is of apparentband multiplicity and inversion, on the strongerabsorptionbandsis clearly evident, when this spectrumis comparedwith the reflection and absorption index spectra in the lower part of Figure 14. These spectra,for comparison,were a pure front-surfacereflection spectrumrecordedfrom a thick sampIe of polymer similar to that used for coating the baking tray and the absorption index spectrumgeneratedvia the

(The polymer being examined was a poly(aryl ether sulfone), PES.) More subtle imperfections may be observed in some spectrathat are recordedusing the internal reflection technique. As an example, two spectrarecordedfrom the surface of a polyester film are shown in Figure 15. The spectra appearto differ slightly. The internal reflection spectrum shown as Figure 15(a) was recordedusing a parallelipiped KRS-5 prism with an incidence angle of 60°; the internal reflection spectrum shown as Figure 15(b) was again recordedusing a parallelipiped KRS-5 prism, but this time using an 'incidence angle of 45°. Some of the differences in relative band intensities might be accountedfor by gradients or anisotropy in surface-layermorphology because of the differing surface layer thickness sampled in these two measurements.However, in the context of anomalous dispersion,one should note in particular, for example, the increasedasymmetryto the lower wavenumberside of the absorption bands at about 1725cm-1 and 1250cm-1 in

K-K transfornl of this front-surface reflection spectrum.

Figure 15(b) compared with those in Figure 15(a). These

r ";

2334

Anomalies, Artifacts and Common Errors in Using Vibrational Spectroscopy Techniques! -~

,

2.0 Q)

u

c: co

Q) u

-e 0 cn

c: co .c

.c

'-

g

~

.c ~

O. (a)

1600 1400 1200 1000 Wavenumber/cm-1

3.3

0 1800

(a)

~

1.200

c: co .c

0 cn

.c ~

Q) u c: co .c

0.0 1600 1400 1200 1000 (b)

are artifacts, which might be misinterpreted; for instance, one might have assigned the perceived shoulder at about 1700crn-l in Figure I5(b) to' potentially indicate a high level of surface carboxylate end groups, since the film being examined was PET. (However, intuitively, if this were true then the concentration would be greater in the spectrum of Figure 15a, where a shallower surface layer is probed.) These artifacts arise from the fact that around the positions of the maxima of these bands with high absorptivities, then the condition for ATR becomes violated. That is, since the angle of incidence at these positions is no longer greater than the critical angle, then external rather than total internal reflection occurs at the internal reflection element/sample boundary. (The critical angle, fJc,is given by sinfJc = n2/nl, where nl is the refractive index of the internal reflection element, in this case KRS-5 with n 1 = 2.37, and n2 is the refractive index of the sample. If we assume n2 = 1.5 for the sample in regions where it does not absorb, then fJc ~ 39°. If n2 rises to ca. 1.705, then fJcbecomes about 46°, that is, it is now above the angle of incidence used to record the spectrum shown as Figure I5b.) For a similar comparison on a sample with bands of generally lower absorptivities, that is a polypropylene film, then no such gross distortions

Jc~

~ ~

Wavenumber/cm-1

Figure 15. IR internal reflection spectra,4cm-l resolution, plot as absorbancespectrarecordedusing KRS-5 parallelepipedreflection elementswith both sidescoveredwith PET film. The multiple internal reflection elements were: (a) a 50mm x 3 mm element with an incidenceangle of 60°; (b) a 30 mm x 3 mm elementwith an incidence angle of 45°.

I

,

0

Mid-infrared Spectroscopy: Anomalies, Artifacts and Common Errors

Q)

g ~

2335

1.4 1.2 1.0 0.8

~ 0.6

.c
90~; (b) 75~ < d < 9O1!m; (c) 10~ < d < 75~; (d) d < 10~. [Reproducedfrom 'Diffuse Reflectance Measurementsby Infrared Fourier Transfolm Spectrometry',Fuller M.P. and Griffiths P.R.,Anal. Chern.,1978, 50, 13, 1906,by kind pennission of the American Chemical"Society, @ 1978.]

In addition to particle size, packing density and compaction pressure are two more important factors that can have significant effects on diffuse reflection spectra. Figure 27 compares two diffuse reflection spectra recorded

from a powder sampleof 1% w/w ball-milled KNO3 mixed with ball-milled KBr at two compaction pressures?Also h . diffu fl . d d fr s own IS a

;~c

..~~

se re echon spectrum reCOf e

Q)

-§ ~ 2000 1800 16 (b)

t !!l 'c

:;,

~

c:

~ th

~

~ 2000 1800 16

om a

neat sample of the ball-milled KNO3 at one of the compaction pressures; note the "missing" band of KNO3 at ca. 1380cm-lin this spectrum due to the influence of front surface reflection.

8 SATURATION IN PHOTOACOUSnC

(c)

Figure 26. IR spectraof transmissionmeasuremen (b) a Kube~-Munk plo pure ball-mIlled DPS; (c) reflection spectrum from ball-milled DPS diluted. Ibbett (1988)2by kind pe

MEASUREMENTS The prime first-order relationship that defines the relative intensities of bands within a magnitude photoacoustic Fourier transform infrared (PAlFT-IR) measurementis the ratio of the optical decay length (optical absorptiondepth) to the thermal wave decay length (or thermal diffusion depth). The optical absorption depth is the reciprocal of the linear absorption coefficient of a band, while the thermal diffusion depth is proportional to the thermal diffusivity. The relative magnitudeof thesetwo parameters governs the observed relative band intensities within a

PAlFT-IR spectrum. F is increasing potential increasing absorptivity a deviation from line absorptivity until at hi samplethickness)full s increasein signal is obs or absorptivity.For an the effects of saturatio modulation frequency, investigatedbecomes

..r j..

'"

--~._~

1

Wavenumber/crrJ-

400

-cj~; (

~ii i~~'" "",,-:~ill\

\:--1~'\1 ~'

Mid-infrared Spectroscopy:Anomalies,Artifacts and CommonErrors 2339

t

t

In

...

t

In

In

.2

"~

"~

~:J

:J ::£

:J ::£

I::

~

~ ~ Q)

.0

~ -

I -

-'

I

A I

apertureof such a dimension.The higher values r~present situations where a mid-infrared spectrum essentially free of diffraction-induced artifacts may surely be recorded. (For a low-divergence, high-brightness IR source, such as that emanating from a synchrotron, the lateral spatial res~lu.tionis often perceived as better, since most of the radiation may be focused to a spot of diameter about 101Jffi.Synchrotron-sourcedIR FT-IR microscopy spectra

I::

Ll ~ -a;

~ ~ -a;

.0

.0

~

~

-

4000

400 4000 400 4000 400 Wavenumber/cm-1 (a) (b) (c) Figure 27. Diffuse reflection spectra of ball-milled KNO3. (a) and (b) 10J,0 w/w ball-milled KNO3 dispersedin ball-milled KBr at compactionpressuresof 17.5kPaand~520kPa,respectively; (c) neat ball-milled KNO3 at compaction pressureof ~520kPa. (a) and (b) are on the same ordinate scale; the ordinate scale' expansionof (c) is 1/25 that of the scale for (a) and (b). [(a, b) Reproducedfrom lbbett (1988f by kind permissionof the author. (c} was also reproducedby kind permission of Roger lbbett.] ~

.

.

of. these-effects and others are discussedfully elsewherem this handbook(seePhotoacoustic Spectroscopy).

IR microscope design and practice has been addressed by Messerschmidt,3and by Sommer in Mid-infrared Transmission Microspectroscopy in this handbook.)The

The last decade of the twentieth century saw FT-IR microscopy become a technique of widespreadusageand maturity. Instrumentationis now a commonfeaturein many industrial, biomedical and forensic analytical laboratories, cove~g diverse application areas, many of which are featuredwithin the pagesof this handbook.The technique"is also an jrnportant tool for many academicresearchstudies. Over this decade there has, however, been considerable discussionand debateabout the ultimate in lateral spatial resolution that may be achieved in an FT-IR microscopy transmission measurement. a first approximation, the spatial resolution of a FT-IRTo microscope is approximately

source and focusing optics provide a range of incidence angles, and the sample and its support, if there is one, will likely induce significantrefraction, maybesomediffuse transmission.In truth, there is probably no absoluteanswer to the questionpf what is the ultimate spatial resolution in practice. It will be, at least, both sam~le and wavelength dependent. It is the opinion of this author that for a conventional Ff -IR microscopy set-up, it will likely be realized practically as being in the region of 8-10 Iim. ,As a practical illustration, Figure:28 represents an exampleof a well-publicized industrial applicationof FT-IR microscopy, namely that of characterizingthe layers in a polymer laminate, by examining in transmission sequentiallayers of a cross-sectionmicrotomed from the sample. "The recorded from both the(PET) SO11mlayers wide polypropylenespectra and 101Jffi wide poly~ster are good

equal to the wavelength of the radiation. As a result, the one-dimensionalvalues commonly quoted for the lateral spatialresolution of an FT-IR microscopehave varied from about SlJffithrough to 10IJffior even lSIJffi. It will depend on the wavelengthsbeing examined;the spatial resolution will be better at shorterwavelengths. In the mid-infrared, using a narrow band cooled mercw;y cadmium telluride (MCT) detector, the spectral range scannedis normally about 4000-6S0cm-l. This represents a wavelengthrangeof about2.S-1S.SIJffi.1na c~nventional Ff-IR microscopy system using an incandescentsource, such as a Globar or similar, then the lower values commonly pronouncedfor spatial resolution purport to the fact that it is possibleto record an IR spectrumthrough an

matches against appropriatestandardreferencespectra of the two materials. However, the spectrum recorded from the approximately81Jffiwide polyurethane~adhesive)layer shows absorption features associatedwith both the surrounding layers, which intrude more with increasingwavelength (decreasingwavenumber). A similar examination is shown in Figure 29(a), this time from a polyurethane adhesive layer between two PET films. The spectrum recordedof the adhesivelayer from a free-standingcrosssection clearly shows many absorption featuresassociated with PET. For comparison,an Ff-IR microscopy spectrum recordedfrom the adhesivelayer of the cross-sectionafter it had beensquashedin a diamond window compressioncell is shown also in Figure 29(b). The compressionincreased

9

!

have been reported recorded through a 31Jffi x 31Jffi square aperture.) The key for the spectrum interrogator is how pure is the recorded spectrum, is it really only from material within the sample area defined by the delimiting aperture, or is there any intrusion of spectral features from surrounding areas? Radiation will certainly pass through an aperture significantly narrower than its. wavelength but . . . .' . ' It. will suffer consIderable dIffraction. (DIffraCtion theory, diffraction patterns and the diffraction limit, particularly for point sources, are covered extensively in many physics textbooks, and the situation with respect to optimizing

DIFFRACTION MICROSCOPY

IN FT -IR MEASUREMENTS

~

~

I

2340 Anomalies,Artifacts and CommonErrors in Using Vibrational SpectroscopyTechniques

t

80 c .-0

0) (.)

c

U) .!!2

ItS

-e

E

~

0

.'8

40

~

«

t

I-

~

'0)

0

(.)

c

4000

3500

3000

(a)

2500

2000

1500

1000

Wavenumber/cm-1

J5

g

.!:J

« 80

4

§ '00

(a)

U)

40

.~

t

ItS

0)

~

g

~

/

J5 ... 0 U)

0

.!:J

4000 3500 3000 2500 2000 1500 1000 Wavenumber/cm-1

(b)

«

t 0) (.) c

c

0

ItS .!:J

80

... 0 U) .!:J

.-

U) .!!2

«

E U)

c

~ 40

4

l-~

.

(b)

0 4000 3500 3000 2500 2000 1500 1000 (c) Wavenumber/cm-1 IR mIcroscopy . . . . f three conF 19ure . . 28. 1::"T' r 1. tranSmISSIOn .. spectra 0 fr I. secunve layers of a thin transverse section cut (mlcrotomed) ( ) 50 'd I 1 1 I . om a mu n ayer po ymer ammate: a ~m WI e po ypropylene layer; (b) ~8 ~m wide layer showing the likely presence. of a polyurethane(adhesive)layer plus intrusion of spectral features from the surroundinglayers; (c) 10~m wide polyesterlayer. [ReprOducedfrom ChalmersI.M., Croot L., EavesI.G., Everall N., Gaskin ~.F., Lw:ns~onI., Moore N., Spectrosc.Int. J:, 8, 13 (1990) by kind permISSIOn of IOS Press,Amsterdam.]

.

the wIdth of the softer polyurethane layer to an extent ..that . a "pure" spectrum could be recorded that was charactenstIc

of the adhesIve. 10

DETECTOR BEER'S

NON-LINEARITY

-

LAW

It is generally accepted that in the mid-infrared region room temperaturedetectors,such as the DTGS (deuterated

I

3000

2500

2000

Wavenumber/cm-1

Figure 29. (a) Fr-IR nricroscopyab~orbancespectrafrom transmission measurementsof adjacent layers in a thin cross-section (nricrotomed) from a multilayer film, the composition of which wasPET/adhesive/PET~(b) Fr-IR nricroscopy absorbance tra from transnrissionmeasureJIlents of adjacentlayers in the thin cross-sec . (a) n.on (mIcro . tomed)fr om a muIn' I ayer film used m di d . d . 11 . - -" b h d. 1t h' au een squas e m a amon wm ow compressIonce . . . trigl~~me sulfate), are presently more linear than the mo~ sensItIve cooled detectors, such as MCT. Also, to obtaIn the best linearity, it is not advisable in quantitative analysis to perfonn measurements on bands of absorbance greater than 0.7 (equivalent to a peak maximum of about 20% transmission), and certainly not greater than 1.0. This rule 0f thurnb ISp . arti.cuarYImpo I 1 . rtant 1.f trioanguarap 1 00IZ . atI.on IS .ty (non-ph0tornetrioc accuracy) . bemg . appl 1 ' ed . N on-li nean m . If as a curved cal1.brauon .: . detectorresponsemaDl .f;estsrtse

graph (deviation from a Beer's law plot). The problem becomesmore acute as the band being measuredbecomes narrower. Software of many FT-IR spectrometer will display and plot absorbancescalesto valuesof4 or 8, and sometimeseven higher; thus, it is worth recalling that absorbancevalues of 3 and 4 correspondto only 0.1% and 0.01% transmission,respectively. Thus the signal-to-noise

Mid-infrared Spectroscopy:Anomalies,Artifacts and CommonErrors ratio may be very low near the center of strong absorption bandsand the photometric accuracyof many contemporary Fr -IR spectrometersis only :f:O.l% T. More detailed discussionon the effects of photometric accuracyon Beer's law plots can be found in Beer's Law by Griffiths in this handbook. For very weak bands,the quantitative accuracyis given by a number of parameters,including the noise level on the spectral baseline, the baseline flatness, the level of impurities in the sample, and how well atmosphericH2O and CO2 have been compensated.

3550cm-l. The presenceof the band at higher wavenumber implied entrapmentof isolated (non-hydrogen-bonded) water moleculeswithin the polymer matrix. Th~ second example from Spectra-Structure Correlations: Polymer Spectra covers changes between the common polymorphic forms of polybutene. A polybutene film preparedfrom the melt will likely exist in the type n crystalline modification, but with time it reverts to the higher-density type I form. This change,which may take place over a several days, is accompaniedby steady but si~ficant changes in the absorption pattern in the fingerprint region of the spectrum(see Figure 30) (also, see Figure 18 of Spectra-Structure Correlations: Polymer

11 POLYMORPmC AND

Spectra).

MORPHOLOGICAL METASTABLE AND ENVIRONMENTffIME STATES

The third example selected shows how different thermal histories may affect the state of a dispersedmaterial.

CHANGES,

Figure 31 showsIR spectraover the range l700-1400cm-l recorded from calcium stearate (CaSt) in three different sampling pres~n~tions.4The absorbancespectra~enera~ed from a tranSmISSIOn measurementof caSt as a dispersIon in a KCl disk and the Kubleka-Munk plot of powdered

DEPENDENT

Many solid materialsexhibit polymorphism,and someform changesare accompaniedby very significant differences between their associatedvibrational spectra. Some states may also ~e me~stable, revertin~ in time to a mor~ stable form, agaIn leading to changes10 the IR specfi"umof the material. Other equilibrium states may be altered by the methodof samplepresentation,containmentor preparation. Clearly, many suchchangesmay be invoked deliberatelyor )

1.4 1.2

\ ~g1.0 g 0.8

mentionedin this section all occurredat room temperature using conventionalmethodsof sampleanalysis.It-is hoped that they serve to illustrate some, perhaps unanticipated, consequences of analysinga sampleby IR spectroscopy.In all casesthe IR spectrumchangedwith time. Polymorphism and its effects on vibrational spectra, par-

.

~ 0.6 0.4

anticipated through variations in experimental parameters, such as temperature or pressure, but the few ex~ples ,

2341

0.2 0.0

1300 1200 1100 1000 900 800 700 600 500 400

(a)

Wavenumber/cm-1 1.4 1.2

~ 1:0 ~

ticularly compoundsof interestto the pharmaceuticalindus-

"§ 0.8

try, are discussedin detail in Polymorphs, Solvates and Hydrates in Volume 5 of this handbook.Also, examples of spectral changesresulting from different crystal forms of organic polymers may be found in the Volume 3 article SPectra-Structure Correlations: Polymer Spectra. Two examplesrelevant to this part of our discussion may be. d. S t St tu C I ti P I S ., J.oun m pec ra- ruc re orre a ons: 0 ymer pec. .. tra asFIgures 12 and 18 and theIr assocIatedtexts. The first concernsdesorptionof water from a polymer film while it is mounted in the IR beam in the samplecompartmentof

~

850 800 750 Wavenumbericm-1 .. 1. Figure 30. (a) Nme overlaid JR, 4cm- resolutIon,spectra . recorded from a polybutene film over a penod of about 3 days, duringwhichtimeit revertedfrom a typeII to a typeI crystalline form. (b) Scale-expanded plot over the range950-750cm-l of theplotsshownin (a), showingtheincreasewith time of absorp-

an IR spectrometer .

tins~t 923cm-l, 848cm('- and 8l6cm-l, ~ a~sorption d~rease WIth time at 9O3cm-l, and an apparent shift m band maxImum

(see Figure

12 of. Spectra-Structure . .

.

Correlations: Polymer Spectra). The peculianty here IS that the wat~r that was in equilibrium with the polymer film gave rise to vOH stretching bands at 3650cm-l and

~-"'~~~

0.6 0.4

0.2950 (b)

900

with time from 763 cm-l to 758 cm-l. The time intervals between when a spectrumwas recordedafter the initial spectrumwas taken were 11,20,62, 135,345,512,2660 and 4196 min.

.~~

"~-

~

I 2342 Anomalies,Artifacts and CommonErrors in Using Vibrational SpectroscopyTechniques I ~ [t,

metastablefonn of caSt, since in time the CaSt reverted slowly to the more stable crystalline fonn, with a consequent the bands and 1540cm-l. Theshift finalofexample in to this1577cm-1 section shows morphological

~

I

changesinduced into a sample of PEEK while contained under pressurein a diamond window compressioncell. The

~ .;

analysisconcernedexamining by FT-m microscopy a visible small defect area in a PEEK molding. The defect was isolatedfrom the ~olding and compressedbetweenthe windows of a diamondcompressioncell to facilitate analysisby

§

~ ~

\

.g

FT-m microscopy.Theresultantspectrumshowedreadily

~

.

that the defect areahad,a different morphology to the bulk material of the molding, in that it was amorphous.However, over about 2 days the contained sample underwent pressure-inducedcrystallization, as evidencedby changes in its spectrum(Figure 32). Its s~ctrum then looked very similar to that recordedfrom the crystalline bulk material, such that the spectrally observedphysical difference information betweenthe defect and thIebulk material was lost! M~y materials may undergo morphological or state changes,suchas pressure-,temperature-or solvent-induced crystallization, with time or as a consequenceof differing histbries or containmentenvironments.Most of thesephys-

~1

~ g

~

(c) 1700

'

1600

'

1500

.

1400

Wavenumber/cm-1

ical changeswill be reflected by changesin the materials' IR spectrumthat may be considerable.The few examples F19ure . a) . in powderedKCI; (b) absorbance spectrumof caSt stabilized here emphasIzethe f~ct that an m s~~ of ~ sample PVC film; (c) absorbance spectrumof caSt in a KCI disk representsthe fingerpnnt of.that samplem Its phYSICal state preparation.[Reproducedby permissionfrom J.M. Chalmers, and environmentthe time th~ spectrumwas recorded,and M.W. Mackenzieand H.A. Willis, Appl. Spectrosc.,38, 763 not just its perceivedchemical structure. (1984).] 31 (

.

Diffu

fl . f caS di ed se re ectlon spectrum 0 t spers

PVC containing CaSt. as a ,stabilizer, d~spersedin powderedKCI generatedfrom a diffuse reflection measurement. both exhibit CaSt bands with maxima at 1577cm-1 and 1540cm-l. However, these bands appear at 1601cm-l and 1562cm-1 in a transmission measurementrecord~ from a sample of the stabilized PVC powder prepar~ as compressionmolded (hot pressed)film from the melt. The bandsat 1601cm-l and 1562cm-l are attributed to a

12

INVERTED

IN EMISSION

SPECTRA Mid-infrared emission studies are mostly limited to examining thin specimensor a thin layer/film on a weakly absorbingsubstrate.Typical of thesemight be an adsorbed specieson a catalyst surface or a coating on a beverage

t , 1400

BANDS

I.II~~

I

, , , " 1200 1000 Wavenumber/cm-1

, 1400

, , , " 1200 1000 Wavenumber/cm-1

, 1400

, , , I 1200 1000 Wavenumber/cm-1

-2 days in diamond window compression cell

Figure 32. IR absorbance spectrarecordedfrom an Ff -IR microscopytransmission measurement of a defectisolatedfrom a PEEK molding. The defect was flattened and containedunder pressurein a diamond window compressioncell.

~~--""'-

~~~,

Mid-infrared Spectroscopy: Anomalies,Artifacts and CommonErrors 2343 can, that is regimes where it is possible to generate ~ spectrum that is characteristic of the fingerprint pattern

+

associatedwith th~" analyte.Although the useof IR emission spectroscopyas an analytical tool has become sparser with the arrival of Fr -IR sampling techniques such as PA and diffuse reflection, in an article such as this, it is

.§ (/)

and applications of IR emission spectroscopy together

(a)

.!Q

~

worth notinga few oddities.The theory,samplehandling

2000 1800 1600 1400 1200 1000 800 600 400

including some discussion of spectral distortions may be found in Infrared Emission Spectroscopyby Mink in this handbook. In optically opaque samples it is possible to observe apparentlyinverted bands in their emission spectra.These distortionsarisefrom re-absorptionby a colder surfacelayer

+ I:: 'ffi

'E W

of radiationoriginatingfrom within the bulk of a sample.

2000 1800 1600 1400 1200 1000 800 600 400

Mink (seeInfrared Emission Spectro~copy)hasdiscussed and illustrated (seeFigures 7 and 8 in Infrared Emission Spectroscopy)somecharacteristicsof self-absorption.Also discussedfor organicmaterialswere the so-called"reduced emission" phenom~na(see, for example, Sheppard5),at wavenumbers where significant front surface reflectibn occurs. The wavenumberdependentabsorptance(a) of a sample . . IS given as: a

= 1-

(t

+ r)

(1).

where t and r represent the transmittance and reflectance of the sample respectivel

. ' y: . . The eqUIvalentexpressIon for elDlSSIonspectrometry, 'through Kirchhoff's law, in temlS of a sample'semissivity (6) is: 6 = 1 - (t + r) (2)

Wavenumber/cm-1

(b)

Wavenumber/cm-1

+ I::

.~ .~ W

",...,.

.

2000 1800 1600 1400 1200 1000 800 600 400 (c) Wavenumber/cm-1 . .. Figure 33. IR, 4cm-l resoluuon, ellllSSIonspectra recorded at 348 K from a s.eriesof PET films: (a) 2.5 ~ thick; (b) 9 ~ thick; (c) l00~m thick. [Reproduced from J.M. Chalmers and M.W. Mackenzie, 'Solid Sampling Techniques', in M.W. Mackenzie

(ed.), "Advancesin Applied FourierTransformInfraredSpectroscopy",J. WIley & Sons,Chichester,105-188(1988).Copyright 1988. @JohnWiley & SonsLimited. Reproducedwith permission.]

Rememberingthat there is dispersion.in the refractive index as one traversesan absorption band (see preceding section), then there is therefore an equivalencein the case of emittance. Fig~e 33 shows an ~xample of "reduced emission" observed in the emission spectra of a series of PET films of differing thickness.5 B'arid splitting is clearly observed on the vC=O band at ca. 1725cm-;I in Figure 33(a). Loss of spectral contrast is clearly seen in the emission spectrum of Figure 33(b), as the more intensebands approachsaturation(seeInfrared Emission Spectroscopy). The minima observed at 1725cm-l and 1265cm-lin Figure 33(c), may be explained by the effect of "reduced emission", while the minima in the vicinity of 1900cm-I and 600cm-I are genuine regions of low

instance not as a consequenceof selective reflection. In these cases,filler material may act as a relatively efficient blackbody emitter, that is, as a conventionalsourcefor the surrounding polymer matrix, which being cooler absorbs at its characteristicfrequencies.Examplesof suchan effect may be seenin the spectraof Figure 34. The featureswithin the emission spectrum recorded will depend not only on the sampletemperature,but also on the filler level and its dispersion,the)film thickness,the thennal balancebetween the filler and the polymer, and the wavelengthdependence of the Planck function. (More detailed discussionof this and further examples may be found in Chalmers and Mackenzie.6)

emissivity.6 In the caseof ,organicpolymers,°.pticalopacitymay be

13

the consequenceof a high level of a filler material, such as carbon black, rather than a~ a result of absorptionbecause of an excessiyesample thickness. Spectra recorded from such samplesmay also show inverted spectra,but in this

EXTERNAL

REFLECTION

-

MIXED-MODE SPECTRA In recent years, increased analytical use has been made of external reflection as a convenient sampling technique,

!

[I 2344 Anomalies, Artifactsand CommonErrors in UsingVibrationalSpectroscopy Techniques

,,---~)""

~

!

1

)

"r,-f"\~~Jh#j\a

"", "

1

u

,

c::

~-~.. ,W~f'ij fu~

i :.-.~ (c) ~,

§ t

:~

"'"' v:

3500

~

,

3000

2500

'"u

2000

1500

1000

Wavenumber/cm-1 Figure 35. FT-IR microscopy, external reflection spectrarecorded from a series of PEEK samples:(a) 3 mrn thick; (b) 125~m thick; (c) 10IJ.Inthick.

t

sample (a 10!lID thick film) is dominated by absorp~on features and resembles thatisrecorded from an a transflectIon measurement, but on which superimposed interference

c::-.

.~

fringe pattern. This is becausethe dominant featuresarise from radiation that hasbeentransmittedthrough the sample

'E

W , 1600

1400

1200 ~1000

and then been reflected back from ~e films' lower surface, but attenuated by the characteristic. absorption bands

Wavenumber/cm

of the sample. At high wavenumber (short wavelength),

Figure 34. IR, 4cm-1 resolution,emissionspectrarecordedat 3"48K from a. seriesof 20% c,arbon-black-filled. PET films: (a) 150~m thick; (b) 1~1J.Inthick; (c) 51J.Inthick. Spectra have beenoffset for clanty. [Repr.oduced from J.M. Chalmers and M.W. Mackenzie,'Solid SamplingTechniques',in M.W. Mackenzie(ed.), "Advancesin Applied Fourier Transform InfraredSpectroscopy", J. Wiley & Sons,Chichester,105-188 (1988). Copyright 1988. @John Wiley & Sons Limited. Reproduced with permission.]

the middle spectrum,from a film of 125~m thickness,has similar characterto the spectrumrecordedfrom the thinnest film but with a different fringe frequency.However,at low ' ,wavenumber the spectrum more resemblesa true: ~ontsurface reflection spectrum,particularly in the viclmty of bandswith high absorptivities.Clearly, neither of the spectra recorded from the two thinner samplesis in a form appropriateto a K-K transform approach.

particularly in combinatio~ with FT-IR microscopy measuretnents.The benefit has been derived from the ready

14 SCALE EXPANSION: DIGITIZATION

application of the K-K transformation of Fresnel reflection spectra to the optical constants of the sample, and thereby

extracting the analytically useful absorption index spectrum. Successfulapplication of this approachdependson the purity of the front-surfacereflection spectrum.The sampIe must be optically flat, homogeneous,and optically thick. If the sampleform does not meet theserequirements,then a "mixed-mode" spectrummay be recordedthat is useless for such a treatment.As an example,Figure 35 shows IR spectrarecordedfrom three samplesof the samepolymer, each of a different thickness, from a front-surface reflection experimentalarrangement.Neglecting the featuresdue to atmosphericwater vapor and CO2,the spectrumfrom the thickest sample(3 mm thickness)is what is expectedfrom a front-surfacereflectiorfmeasurementand appropriatefor K-K treatment.The spectrumrecorded from the thinnest

STEPS AND PLOT SOFTWARE CHARACTERISTICS The number of data points in a mid-infrared recorded interferogramrelatesto the resolution at which the desired spectrum is to be analyzed. The point density in the transformed spectrum may possibly be increasedby an integer multiplication of the number of recorded data points, depecnding on the level of zero-filling chosen.When the spectrum is plotted with high abscissaexpansion,the spectrum may appear disjointed, being composed as a series of points connected by straight lines. Some plot software algorithms further increasethe point density (not the spectral resolution though) through interpolation, by, for example,fitting polynomials through successiveblocks of data points. Like zero-filling, spline interpolation serves

mi;;::

~:i;;:t., ,

primarily as an aestheticfunction implementedto improve band contour; the spectralresolution is implicit and defined by thenon-zero-filled interferogram.

aperture card sample mount. In Figure 36(a), the singlebeam background,used was that of the open beam. The mismatch in beam dimensions between the two ratioed single-beamspectra,producedthe sloping backgroundand

15, SINGLE-BEAM BACKGROUND CHOICE

To compensate for the .vignetting causedby the sample

overallreducedtransmission scalefor thesamplespectrum.

..

The choice of single-beam background for measuring a sample's FT-IR spectrum should be that which most closely representsthe conditions under which the sample's single-beam spectrum is recorded. For example, if an

card mount, the single-beam background spectrum for Figure 36(b) was recorded through an empty aperture sample mount located as near as possible in the same position, in the samplecompartmentat the beam focus, as for that used to record the sample single-beamspectrum. The improvementin spectralintegrity is self-evident.

accessorysuch as an internal reflection unit is used, then the backgroundspectrumshouldbe taken through the same unit in the same alignment in the sample compartment fitted

16

with the internal reflection element to be used, but without

the sample.This consideration;of matching as closely as possiblethe throughputand light pathsunder which the two single-beamspectra are recorded, will minimize artifacts and distortions in the sample's spectrum. As a trivial exampleof minimizing such distortions, the spectraof Figure 36 may be compared.The pair of spectra were 'recorded from a polysulfone film of about 10~ thickness. The film was mounted in a 25 mm x 11mm 50 ..~ 40 .~ 30

~ ~ 20 0 '4000 3500 3000 2500 2000 ~;oo 1000 500 (a) Wavenumber/cm 80 § 60 .~ .-

equivalent spectra,e.g. Kubelka-Munk, absorption index or photoacoustic intensIty plots. It is not uncommon to observe within an appropriately factored difference spec-trum features that may indicate a band shift, a band narrowing, a b~d increaseorlan increaseor decreasein band intensity. ~t high ordinate scaleexpansions,similar effects can sometimesarise as artifacts, and not be associatedwith real changes. For instance, assuming there are no time-

l

'.

from and then replacedinto the samplecompartmentbeam betweenthe recordings. 500

Figure 36. IR, 4cm-l resolution, transmissionsrectra recorded from a polysulfone film. The film was mounted In the 25 mm x 11mm apertureof a samplecard mount. The single-beambackground spectrumused for (a) was the unrestricted spectrometer open beam, while that for (b) was an open beam but with an empty card mount sited in the sampling position.

17 INSTRUMENT

-

Up to this point, the discussion in this article has concen... . trated mostly on ~omalles, artIfac~s and errors arISl~g out of sample preparation and presentation and spectral display. This has been deliberate, since in mid~infrared measurements, these are the most common sources of malpractice

,.

\.

-'~-":

,,",,~i_-"~'~~::4"

~

.

~ 4000 3500 3000 2500 2000 1500 1000 Wavenumber/cm-1

Many artifacts or distortions may be introduced into spectra as a consequenceof data treatmento~ manipulation by software algorithms. Some, s~ch as phase correction considerationsin high-resolution spectra,are beyond the scope of this article. Others, such as decreasingsignal-to-noise ratio with increasinglevels of deivatization,loss of spectral contrast and detail with smoothing,or introduction of side lobes in Fourier self-deconvolution,have beendiscussedin the appropriatearticles within this handbook. Spectralsubtraction(differencespectroscopY),is a manipulation that is both extremely valuable and seriously open to abusein the interpretationof its results.It should always

spectra generated from consecutively recorded spectra from the same sample, but where the sample was merely removed

20

0

TREATMENT

dependentchanges,then such artifacts can appearin a difference spectrumgeneratedby subtractingtwo absorbance

~ 40 ~

(b)

AND

be undertaken between two absorbance or two absorbance-

~ 10

~

DATA MANIPULATION

I

, 2346 Anomalies,Artifacts and CommonErrors in Using Vibrational SpectroscopyTechniques ,

-

~ssociatedwith measurementsmade on condensedphases at spectralresolutionsof 1 cm-l or lower. Instrumentalartifacts, such as vignetting and aliasing, leading to wavenumber shifts and spectral folding respectively, are mentioned in ano.therarticle (seeAnomalies and Artifacts in Raman Spectroscopy) in considerationsof anomaliesappropriate to Fourier transform (Fr)-Raman spectra..These and other performanceand design criteria, such as mirror alignments, and roll, tilt, yaw and shear elimination or compensation during interferometer mirror travel, become much more pertinent to more speci3!ized applications, such as highresolution spectroscopy,and will not be discussedhere. For most commercial Ff spectro~etersdesignedtoday for operating in the mid-infrared region, they are of no apparent consequenceto the user, except for all but the most sensitive measurements,and of no real concern for applications by conventional sampling methods undertaken at low or moderat~spectralresolutions. A classified list of 50 categoriesof potential ordinate error in Ff spectroscopyhas been published with commentsspectrophotometer by Birch and Clarke: The classifications are: nonideal properties, interaction of non-ideal instrumentand sampleproperties,effectsof non-idealproperties of the sample, and deficiencies.of, the measurement procedures.More recently, these authors have published a prelimifiary appraisal of the interreflection errors in F.f spectroscopy.8Discussions on component specifications and design and performancerequirementsfor Ff-IR spectrometers may be found in many other publications.9-11 Some such as the Zachor-Aaronson disto_rtion,related to non-constantmirror scan velocity, becomemore needy of considerationin problematic environments,such as process installations}2 A detailed at"Hirschfeld the problems involved in quantitative Ff early -IR bylook Tomas was published in 1979.13In mid-infrared dispersivespectrometry,the inher. ent polarization in 'gratingswas an important consideration.

'

have been shared will hope~lly make newcomersto the field more aware'of the pot~ntial hazardsassociatedwith poor experimental practice, short-cutting well-established methodologiesor attempting to circumvent (violate!) the laws of optical physics. The empha~isin this article has beendeliberatelybiasedtowards samplepresentationtechniques,sincein generalpractice,this is the areamost prone to misconception,misunderstanding,misuseand error. This is perhaps in contrast to the article on anomalies and artifacts in Ramanspectroscopy(seeAnomalies and Amfacts in Raman Spectroscopy), for which the tendency out of necessityis more towardsthe instrumentationand its parameters.However, I would recommendthat anyonenew to the field of vibrational spectrosc~pyread both theseartic1es,and also the article covering ariomaliesin near-infrared spectroscopy (see Anom.alies in Near-infrared Spectroscopy). Each contains differing insights to the problems faced in recording high-quality ~pe:ctrafree from artifacts and distortions,and collectively iliey should-providea good basefor developing successfulapproachesto both qualitative and quantitative vibrational . spectroscopypractice.

ABBREVIATIONS AND ACRONYMS CaSt DPS' K-K PA/Fr-IR PDMS PEEl( PES' PET PVC

Calcium Stearate Dipheny1Sulfone Kramers-Kronig PhotoacousticFourier Transform Infrared Po1y(dimethy1~i10xane) Po1y(ary1ether eilier ketone) Po1y(ary1ether sulfone) Po1y(ethy1ene terephthalate) Po1y(viny1 chloride) "

REFERENCES

However, the effect is significantly weaker in Ff-IR spectrometers (see Mid-infrared Spectroscopy of the CondensedPhaseand Griffithset al.ll and Hirschfe1d13). And, while the reflectance and transmittanceof b~am-sp1itters may be 'quite different for the two (horizontal and vertical) polarizations, the efficiency is usually quite close to unity .

, 1. B.A. Macdonald(ed.),'Chambers TwentiethCenturyDictionary', W & R Chambers Ltd., Edinburgh(1979). 2. R.N. Ibbett,PhD thesis,Universityof EastAng1ia,Norwich (1988). 3; R.O.Messerschnridt, 'Minimizing OpticalNonlinearitiesin InfraredMicrospeotroscopy', in "PracticalGuideto Infrared Microspectroscopy", ed.H.J. Humecki,MarcelDekker,Inc.,

is CLOSINGREMARKS

New York, 3-39 (1995). 4. J.M..Chalmers, M.W. Mackenzie and H.A. Willis, Appl.

.

Spectrosc., 38(6),763 (1984). As

one

progresses

through

composin

an

article

g

like

thi

5. s,

?ne tends to recall (~nfortunate1y!) exampies of many m~tancesone has expenencedor observedwhere nuisances and.nuancesha:e oc~urre~ ~at resulted in spectral distortions and artifacts m mId-infrared spectra. Those that

N.

Sheppard,

Measurem~nt

'The of

.use Infrared

of

F°u.ne~ EmISSIOn

Transform Spectra',

In

for.

the

"AnalytIcal

Applicationsof FTlRto ,MolecularandBiologicalSystems".. ed.J.R.Durig,D. Reidel,Dordrecht,125-140(1980). 6. J.M. ChalmersandM.W. Mackenzie,'Solid SamplingTechniques',in "Advancesin AppliedFourierTransformInfrared

.

.

~ethods

. .' ,

c

.

. Mid-infrared.Spectroscopy: Anomalies,Artifacts and CommonErrors 2347 -

11 y so "" : ,

Spectroscopy", ed. M.W. Mackenzie, I. Wiley & Sons, Chich~ster,105-188 ~1988). 7. I.R. BIrch and F.I.I. Clarke, Spectrosc.Hur., 7(4),16 (1995). 8 I . . . (1999). .R. BIrch and F.I.I. Clarke, Anal. Chlm. Acta, 380, 369

11. P.R. Griffiths and I.A. de Hasetli, 'Fourier Transform Spectrometry',~. Wile~ ~ Sons,Inc., New Yor~ (1986).. 12. D.W. Vidrine, 'Mid-infrared Spectroscopym,Chelnlcal Process Analysis', in "Spectroscopyin ProcessAnalysis", ed. I ..M Chalmers, Sheffi eId A cadelnlc . Press, SheffieId, 96- 138

9. R. Geick, Top. Curren(Chem., 58, 75 (1975).

(2000). 13. T. Hirschfeld,- 'Quantitative FT-IR: A Detailed Look at the

10; S.F. Iohnston, 'Fourier Transform Infrared Spectrometry.A ConstantlyEvolving Technology', Ellis Horwood, Chichester (1991).

Problems Involved', in "Fourier T~sform Infrared Spectroscopy", eds I.R. Ferraro and L.J..Basile, Academic Press, New York, 193-242, Vol. ff (1979).

~

\

/

\

(t!

.

t~

')

,

:

t, i

\

,

i

"

\

,