Noise in FT-IR spectral data processing

Journal of Molecular Structure, 175 (1988) 329-334

329

Elsevier Science Publishers B.V., Amsterdam - - Printed in The Netherlands

NOISE IN FT-IR SPECTRAL DATA PROCESSING

M.Gil, N.Iza and J.Morcillo. Dept. of Physical Chemistry Univ. Complutense.

(Spectroscopy),

28040-Madrid.

Fac. of Chemistry.

SPAIN.

ABSTRACT. A series of exploratory works on resolution enhancement, precision in wavenumber, integrated band intensity measures, and noise level control were performed on FT-IR spectra using standard mathematical techniques of data processing. In this communication, the infrared (IR) absorption band corresponding to the ~8 mode of dichloromethane, CH2C12, was studied in benzene solution. Fourier self-deconvolution of the CH2CI 2 ~ band was carried out using 8 standard software supplied for the purpose. Second and fourth derivative spectra were obtained with the "Nicolet" software parameter "DR1". Self-deconvolution in Fourier space and derivative techniques were used to decrease the band full-width at half-height (FWHH) and achieve an apparent band resolution enhancement. The total area under the self-deconvoluted band is not exactly the same as under the original band. Under optimum self-deconvolution conditions, the integrated area increases by 6.9 % as compared with the out-of-phane C-H bendin~ W(CH 2) original band. However, a linear relationship between the integrated area of the self-deconvoluted band and FWHH of the original Lorentzian component was observed. The applicability and potential advantages of the self-deconvolution and derivation methods in spectral data processing are strongly limited by the noise level or signal-to-noise ratio (SNR) of the original IR spectra.

INTRODUCTION. The introduction of interferometric techniques and strong improvements in performance,

sophistication and degree of digitization of infrared

(IR)

instruments have permitted a wider application of Fourier self-deconvolution m d mathematical derivative procedures

(refs. 1-5). Both these resolution

enhancement methods have been compared in some recent studies by Fourier transform infrared

(FT-IR) spectroscopy

(refs. 6-8). Kauppinen et al. have

described the principles and feasibilities of Fourier sef-deconvolution and derivatives in IR spectra (refs,

9-11), and the effect of these spectral data

processing methods on resultant noise (ref. 12). More recently, generalized approaches to the subject have been published

further

(refs. 1,13-14).

The purpose of this paper is to investigate: resolution enhancement grade; reproducibility of band position (in cm -I) and integrated band intensity; degradation of signal-to-noise ratio enhacement procedures,

0022-2860/88/$03.50

(SNR),

and

in order to apply both resolution

self deconvolution in the

© 1988 Elsevier Science Publishers B.V.

Fourier space and mathematical

330 derivatives,

to the

~8 IR absorption band (ref.15) of dichloromethane,

CH2C12,

in benzene solution.

EXPERIMENTAL A 0.1491M

dichloromethane

(Merck, Uvasol)

was examined using a KBr fixed-pathlength recorded at room temperature Globar source,

on a Nicolet 60-SX spectrometer

a DTGS detector,

(interferograms)

(Merck, Uvasol)

equipped with a

and a Ge/KBr beamsplitter. 50 scans -i at 1 cm nominal resolution

were coadded and signal-averaged

for both solution and solvent samples. zero filled,

solution in benzene

cell (385 ~m). FT-IR spectra were

Interferograms

were phase corrected,

and apodized by the standard Happ-Genzel

transformation.

Digital substractions

of solvent and any water vapor present

in the spectra were readily performed using a simple original procedure "macro-program" satisfactory

designed for the purpose.

cancellation

obtained from results in

of

Thus, a FT-IR spectrum of CH2CI 2 with

of benzene absorptions

and water vapour bands was

suitable assays with different backgrounds

a color raster scan

one

function prior to Fourier

by examining the

display.

Fourier self-deconvolution of the CH2CI 2 out-of-plane C-H bending band, -i at 1266.5 cm , was carried out with a standard software package (ref.

W(CH2),

16) provided by Nicolet,

and based on the algorithm of Kauppinen et al. (ref.9)

of the National Research Council of Canada procedure

is

(FWHH) of the Lorentzian designated enhancement

This deconvolution

lineshape function.

line used for the self-deconvolution,

The bandwith

2 o (ref.9),

"VFO" parameter by Nicolet, was ranged from 0.i to 5.0. The resol~ic~ factor or degree of resolution enhancement,

DIAGRAM

BASIC PARAMETERS OF THE w(.CH2)

K= 2a/A~i/2

(A~I/2 being

1

ORI6INAL BAND .

IA A?

I

~

(N.R.C.C.)

performed using the Lorentzian

)P ( c m -~' )

"MAX =

1266.56

AMAx

0.9653

=

A = EMAX

168.2

A-~

5.27

=

N (RMS) =

o,oqq %

(C = O , l q 9 1 M ~ B = 385 . M ) .

(MOL - 1 L

cN-1

cM-1).

(±0.02 CM-1).

(&-~ /2) H = 2 . 3 7 CM-1 ~ ( A ~ 12) L = 2.90 CM-1. (H AND L • "HIGH" AND "LOW" FREQUENCIES, RESPECTIVELY).

})M=x

BAND SKEWNESS PARAMETER. ~ = ( ( ( a ~ /2) L / (a-~ /2) H} - 1} .IO0 = NOISE ,

cM- 1 .

s

SIGNAL-TO-NOISE RATIO •

22.q % • SNR =

2270 •

331

0

i! ¢mL. S

A _

GINAL BAND ,

1

,VF1 = 0.,5

!

VFO=3.2

,

VF,=15

,

,

Q

-~m I'

=~1

'

VFO=3.51

.VFi:~'6

,'l¢~ZS "I~'B6 I~'q7

"r

]

VF,=,.O

1~08 I~IZS

1:~86

WRVENUMBERS

l]'q?

WRVENUMBERS

I

I l~'OBI~Z5

VFO=3.2'

VFi=,I~

t~B6

1'2q7

' I;'OBI3ZS

1:~lB6

WAVENUMBERS

Fig.l.-Spectralnoisevariati~with different self-deconvolution degrees. w(CH2) band. (B-H) Self-deconvoluted bands.

the

FWHH of the original self-deconvoluted band)

12q7

i~OB

WAVENUMBERS

(A) Original

(refs. 9,12), designated "VFI'~

was ranged from 0.25 to i0.0. First, second and fourth derivatives of the w(CH2) band were obtained with the Nicolet software parameter "DR1" (soft-key), applied once, twice or four times, respectively. measurements

Also, a standard subroutine for performing band area

(Simpson's Rule) was used without baseline correction.

RESULTS AND DISCUSSION Diagram 1 presents the values measured for principal parameters of the w(CH2) band. Tables 1-3 show the results of self-deconvolution in Fourier space and derivatives on the w(CH2) band. The optimum deeonvolution for this band (VFO=3.2; VFI=I.80.

See Table i) produces 9100% increase in maximum absorption intensity.

The corresponding decrease in band FWHH or increase in spectral resolution enhancement was 1.8 times that of the original band. The spectral noise is increased

~ 5.5 times in the self-deconvoluted band. In general,

a

strong

degradation effect in SNR with the application of self-deconvolution can be appreciated

(Table 1 and Figure i). Except in infra-deconvolution cases, the

maximum position of the W(CH2) band,

VMax, remained invariable

(at ±0.01 cm -I)

in self-deconvoluted bands. Integrated area values of the studied band, at three different wavenumber ranges of integration,

vary linearly with the "VFO" self-deconvolution parameter

(Fig.2). On the contrary,

variation in the resolution enhancement factor, "VF~'(~

382

TABLE 1 SPECTRALDATAOF w(CH2) BANDFOR DIFFERENT GRADESOF FOURIERSELF-DECONVOLUT VF0

VFI

(2°)

(K)

c

v MAX (CM- I )

C

0.i 0,25 0.5 0.5 0.5 0,5 0,5 1,0 2.0 2.0 2,0 2,0 3,2 3.2 3.2 3,2 3,5 3.5 3,5 4.0 4.0 5,0 5.0 5.0

5,0 5,0 0,25 0.5 1,0 5,0 10.0 0,25 0,25 0,5 1,0 1,5 1,0 1,5 1.75 1,8 i, 0 i. 5 1.75 i. 0 2,0 0,25 0.5 I, 0

B~MAX

ADECONV.

=

-MAX

"'MAxADECONV '

AAI~x

SNR

xlO3 i

1266,56

0.9653

1266.56 1266,56 1266,56 1266,56 1266,56 1266,56 1266,56 1266.56 1266,10 1266,57 1266.57 1266.56 1266,56 1266,56 1266.57 1266,57 1266,56 1266,57 . 1266,57 . 1266,09 1266.09 1266.56

0.9846 1,0150 0.9887 1.0469 1,0644 1.0703 1,0705 0.9067 0.7135 1.0584 1.3559 1.4713 1,4910 1,8312 1.9576 1.9808 i. 5085 i. 9156 . 1. 5260 . 0,4028 0,7905 i. 5281

. .

0c

2270 1230 1420 3960(?) 2260 1550

+19.3 +49,3 +23.3 +81,6 +99.1 +105,0 +105,2 -58,6(??) -251,8(??) +93,1 +390,6 +506,0 +525.7 +865.9 +992,3 +i015,5 +543.2 +950.3 . +560.7 . -562,5(??) -174,8(??) +562,8

1450 1430 2130 5200(?) 2720(?) 1430 910 1750 880 510 500 2150 870

2780(?) 250 1910 2350( ? )

"0VER-DECONVOLUTEDSPECTRA.

-AMAx

CORIGINALBAND. does not produce f

.

.

.

noticeable

changes

in

.

integrated

L

band area

Under optimum

self-deconvolution, increases ¢m-I

(I) (See Table 2).

conditions

for w(CH 2) bard

the integrated

by 6.9% as compared

original

band.

foreseen

in theory

with the

This discrepancy, (ref.9),

area

not

may be

cm_T

7.5

explained • 1

1

1

t

2 3 VFO (2r)

(1325-1210) cm -1" I

4

[~

~ig.2.-Integrated area ( I ) of s e l f -deconvoluted w(CH^) bands VS. VFO(2o) parameter used in self-deconvolution.

noise

by the existence

of variable

in the self-deconvoluted

Application

second and fourth, decreases

bands.

of even derivatives, to the w(CH 2) band

their FWHH by a factor of 2.8

and 4.3, respectively

(Table 3). Thus,

second and fourth derivatives

greatly

333 improve decays

resolution

enhancement

by a factor

of mathematical

compared

of 3.7 and 13.7,

derivative

to self-deconvolution,

respectively.

procedures

although

Nevertheless,

is advantageous

the application

to resolve

inherently

TABLE 2 INTEGRATED INTENSITIES OF w(CH2) BANDFOR DIFFERENTGRADESOF FOURIERSELFDECONVOLOTION,

VFO

VF1

(2°)

(K)

Av: EcONV'

Ic

(*3Av~ ) (CM-I)

(CM-I)

Ic

Ic

N(MS)

(*6A,~ ) (1325-1210) (CM-I)

(%)

(CM-1)

5.27D

7,295

7,711

7.805

0.044

0,I

5.0

5,17

7,313

7,721

7,813

0.081

0.25

5,0

5.01

7.336

7.731

7,814

0,070

0.5

0,25

5.32(??)

7,373

?,747

7.816

0,025(~

0,5

0,5

4,90

7.374

7,747

7,815

0.044

0.5

1,0

4,79

7,374

7,747

7.815

0.065

0,5

5,0

4.75

7.374

7.747

7.815

0,070

0.5

i0.0

4.75

7,374

7,747

7.815

0.070

D

D

1.0

0.25

6.41(??)

7.447

7,778

7.819

0,047

2,0

0.25

9.62(??)

7,593

7,842

7,825

0.019(?)

2.0

0,5

5.72(??)

7.593

7,845

7.825

0.037(~

2.0

1,0

3,97

7.602

7,843

7.825

0,070

2.0

1.5

3.50

7.602

7.842

7,825

0.111

3,2

1.0

4.19

7,'791

7.921

7.834

0,057

3.2

1.5

3,15

7,791

7.924

7,833

0.113

3,2

1.75

2.86

7,796

7.925

7.833

0,242

3.2

1,8

2,81

7.791

7.922

7,832

0,250

3,5

1,0

4,32

7.844

7,944

7.836

0,047 0,115

3,5

1.5

3,15

7,844

7.948

7,835

4.0

1.0

4.57

7,923

7,976

7,839

0.036(9

5.0

0.25

20,37(??)

8.069 E

7.982 E

7,851 E

0.407

5.0

0,5

10,21(??)

8.033

8,046

7,845

0,052

5.0

1,0

8.090E

8,043E

7.845E

0.043(?)

5.13

CINTEGRATEDBANDAREA.DORIGINALBAND. EpRESENCEOF "FooTs"OR SIDE LODES.

TABLE 5 SPECTRAL PARAMETERSOF w(CH2) ORIGINAL AN DERIVATIVE (FIRST. SECONDAND FOURTH) ABSORPTION BANDS.

DERIVATIVE - - 0 ORDER (0RI61NAL BAND)

VNAX OR "NIR AMAX

(CM- I )

|C(cM-I)

SNR

(CM- I )

OR AMIN

1266.56

0.9653

5,27

7,805

0,4425

2.89

-0,0136

370

]ST ORDER 2ND ORDER 4TH ORDER

A~

(1325-1210)

2270

1266.56

0,0871

1,89

-0.0155

110

1266,56

0,0490

1,22 D

-0,0144

30

CINTEGRATED BAND AREA. DNEAROF INSTRUMENTALRESOLUTION.

SNR

334 broad absorption Finally,

bands into distinct peaks

second derivative application

(refs. 1,5,7).

to the W(CH2) self-deconvoluted

band

under optimum conditions reduces the FWHH by a factor of 1.7. This combined resolution

enhancement procedure

a decrease

in original bandwidth by a factor of - 3 with a measured total noise

level (in RMS) of N=0.728%.

(self-deconvolution

+ 2nd derivative)

causes

This noise level is smaller than that corresponding

to the second derivative of the original band smaller than that corresponding to the second derivative of the original band remaining for derivatives

the maximum

for the (self-deconvolution

(Table 3). Moreover,

(or minimum)band

+ 2nd derivative)

as well as

position stayed constant

case.

Other aspects referring to the subject here discussed will be published shortly.

REFERENCES 1

2 3 4 5 6 7 8 9 i0 ii 12 13 14 15 16

H.H.Mantsch, H.L. Casal and R.N. Jones, in R.J.H.Clark and R.E.Hester (Editors), Spectroscopy of Biological Systems, Wiley, New York, 1986. Chap.l. And references quoted therein. W.F.Maddams and W.L. Mead, Spectrochim. Acta, 38A(1982)437-444. F.M.Wasacz, J.M. 0linger and R.J. Jakobsen, Biochemistry, 26(1987)1464-1470 W~J. Yang, P.R. Griffiths, D.M.Byler and H.Susi, Appl. Spectrosc.,39(1985), 282-28?. H.Susi and D.M.Byler, Biochem, Biophys. Res. Commun., 115(1983)391-397. W.K. Surewicz, M.A. Moscarello and H,H.Mantsch, Biochemistry, 26(1987)3881-3886; J.Biol. Chem, 262 (1987) 8598-8602. C.Chapados, J.B~liveau, M.Trudel and C.Levesque, Appl. Spectrosc., 40(1986) 773-782. J.M.Olinger, D.M. Hill, R.J. Jakobsen and R.S. Brody, Biochim. Biophis. Acta, 869(1986) 89-98. J.K.Kauppinen, D.J.Moffatt, H.H.Mantsch and D.G.Cameron, Appl.Spectrosc., 35 (1981)271-276. J.K.Kauppinen, D.J.Moffatt, H.H.Mantsch and D.G.Cameron, Anal. Chem., 53(19@1) 1454-1457. J.K.Kauppinen, D.J.Moffatt, H.H. Mantsch and D.G.Cameron, Appl. 0pt.,21(1982) 1866-1872. J.K.Kauppinen, D.J.Moffatt, D.G.Cameron and H.H. Mantsch, Appl. 0pt.,20(1981) 1866-1879. D.G. Cameron and D.J. Moffatt, Appl. Spectrosc., 41(1987) 539-544. W.I. Friesen and K.H.Michaelian, Appl. Spectrosc., 39(1985)484-490 K.Tanabe, Spectrochim. Acta, 30A(1974)1891-1900. D.A.C. Compton, FT-IR Spectral Lines (Nicolet Instr. Corp.), 5(1983)4-7.