Vous êtes sur la page 1sur 8

Uses of Derivative Spectroscopy

Application Note

UV-Visible Spectroscopy Anthony J. Owen

Derivative spectroscopy uses first or higher derivatives of absorbance


with respect to wavelength for qualitative analysis and for
quantification. The concept of derivatizing spectral data was first
introduced in the 1950s, when it was shown to have many advantages.
However, the technique received little attention primarily because of the
complexity of generating derivative spectra using early UV-Visible
spectrophotometers. The introduction of microcomputers in the late
1970s made it generally practicable to use mathematical methods to
generate derivative spectra quickly, easily and reproducibly. This
significantly increased the use of the derivative technique.

In this application note we review briefly the mathematics and


generation methods of derivative spectroscopy. We illustrate the
features and applications using computer-generated examples.

Agilent Technologies
Innovating the HP Way
Introduction

If a spectrum is expressed as
absorbance, A, as a function of 1.0 Absorbance 1.0 Absorbance
wavelength, l, the derivative
spectra are:
0.5 0.5

0.0 0.0
Zero order 300 400 500 600 700 300 400 500 600 700
A = f ()
2.0E-04
First order 0.01
dA 0.0E+00
= f ( )
d 0.00 -2.0E-04

-4.0E-04
Second order
-0.01
1st derivative -6.0E-04 2nd derivative
d2A
= f () 300 400 500 600 700 300 400 500 600 700
d2
2.0E-04 3rd derivative 1.0E+06 4rd derivative

1.0E-04
5.0E-07
0.0E+00
0.0E+00
-1.0E-04
-5.0E-07
-2.0E-04

Figure 1 shows a computer 300 400 500 600 700 300 400 500 600 700
simulation of the effects of
derivatization on the appearance Figure 1
Absorbance and derivative spectra of a Gaussian band
of a simple Gaussian absorbance
band. Derivative spectra are
always more complex than zero-
order spectra.

A first-order derivative is the rate The most characteristic feature of A strong negative or positive band
of change of absorbance with a second-order derivative is a with minimum or maximum at the
respect to wavelength. A first- negative band with minimum at same wavelength as lmax of the
order derivative starts and finishes the same wavelength as the absorbance band is characteristic
at zero. It also passes through zero maximum on the zero-order band. of the even-order derivatives.
at the same wavelength as lmax of It also shows two additional
the absorbance band. Either side positive satellite bands either side Note that the number of bands
of this point are positive and of the main band. A fourth-order observed is equal to the derivative
negative bands with maximum and derivative shows a positive band. order plus one.
minimum at the same wavelengths
as the inflection points in the
absorbance band. This bipolar
function is characteristic of all
odd-order derivatives.
Quantification Obtaining derivative spectra

If we assume that the zero-order Derivative spectra can be obtained the amplitude and wavelength
spectrum obeys Beers law, there by optical, electronic, or math- shift of the derivatives varies with
is a similar linear relationship ematical methods. Optical and scan speed, slit width, and resis-
between concentration and electronic techniques were used tance-capacitance gain factor.
amplitude for all orders of deriva- on early UV-Visible spectropho-
tive: tometers but have largely been Mathematical techniques
superseded by mathematical To use mathematical techniques
Zero order techniques. The advantages of the the spectrum is first digitized with
A = b c mathematical techniques are that a sampling interval of Dl. The size
derivative spectra may be easily of Dl depends on the natural
First order calculated and recalculated with bandwidth (NBW) of the bands
different parameters, and smooth- being processed and of the
dA d
= bc ing techniques may be used to bandwidth of the instrument used
d d improve the signal-to-noise ratio. to generate the data. Typically, for
UV-Visible spectra, the NBW is in
nth order Optical and electronic the range 10 to 50 nm. First-
techniques derivative spectra may be calcu-
d A d
n n
= bc The main optical technique is lated simply by taking the differ-
dn dn wavelength modulation, where the ence in absorbance between two
wavelength of incident light is closely spaced wavelengths for all
l wavelength wavelengths :
A absorbance rapidly modulated over a narrow
e extinction coefficient wavelength range by an electrome- ( A+ A )
b sample path length D+ /2 =
c sample concentration chanical device. The first and
second derivatives may be gener-
For single component quantifica- ated using this technique. It is Where the derivative amplitude,
tion the selection of wavelengths popular for dedicated spectropho- Dl, is calculated for a wavelength
for derivative spectra is not as tometer designs used in, for intermediate between the two
simple as for absorbance spectra example, environmental monitor- absorbance wavelengths.
because there are both positive ing. First-derivative spectra may
and negative peaks. For the even- also be generated by a dual For the second-derivative determi-
order derivatives there is a peak wavelength spectrophotometer. nation three closely-spaced
maximum or minimum at the same The derivative spectrum is gener- wavelength values are used:
lmax as the absorbance spectrum ated by scanning with each ( A 2 A + A + )
but for the odd-order derivatives monochromator separated by a D =
2
this wavelength is a zero crossing small constant wavelength differ-
point. ence. Higher-order derivatives can be
calculated from similar expres-
Taking the difference between the First and higher-order derivatives sions.
highest maximum and the lowest can be generated using analog
minimum gives the best signal-to- resistance capacitance devices. This method involves simple linear
noise ratio but may lead to in- These generate the derivative as a interpolation between adjacent
creased sensitivity to interference function of time as the spectrum is wavelengths. A better method is
from other components. scanned at constant speed that proposed by Savitzky and
(dA/dt=S). For the first derivative: Golay. To calculate the derivative
at a particular wavelength, l, a
dA 1 dA
= window of n data points is
d S dt selected and a polynomial is fitted
Higher-order derivatives are using the least squares method:
obtained by using successive
derivators. The electronic method
A = a0 + a1 + + al l
suffers from the disadvantage that
Features and applications
An advantage of this method is For clarity the points made in the For example, the absorbance
that it can be used to smooth the following discussion are illustrated spectrum of the steroid testoster-
data. If the polynomial order, l, is using computer-generated ex- one has a single, broad, featureless
less than the number of data amples. In the figures, dotted lines band centered at about 330 nm but
points (2n+1) in the window, the show the baseline, dashed lines the second derivative has six quite
polynomial generally cannot go show component spectra, and solid distinctive peaks.
through all data points and thus lines show the analyte spectrum
the least squares fit gives a made up from the component Resolution
smoothed approximation to the spectra. As figure 1 shows, the derivative
original data points. This feature centroid bandwidth of the even-
can be used to counteract the Graphics order derivatives decreases with
degradation of signal-to-noise that As shown in figure 1 there is an increasing order. Relative to the
is inherent in the derivatization increase in the number of bands as zero-order spectrum the derivative
process (see below). The coeffi- higher orders of derivative are centroid bandwidth for a Gaussian
cients a0...al at each wavelength calculated. This increase in the band is observed to decrease to
multiplied by the factorial of the complexity of the derivative spectra 53 %, 41 %, and 34 % of the original
order are the derivative values: a1 can be very useful in qualitative bandwidth in the second, fourth,
is the first derivative, 2xa2 the analysis, either for characterizing and sixth orders respectively.
second derivative, 6xa3 the third materials or for identification.
derivative, and so on. Spectra that are very similar in This feature may be used in
absorbance mode may reveal qualitative analysis to identify the
Savitzky and Golay developed a significant differences in the presence of two analytes with very
very efficient method to perform derivative mode. similar lmax values that are not
the calculations and this is the resolved in the absorbance
basis of the derivatization algo- spectrum. Figure 2 shows a
rithm in most commercial instru- computer simulation.
ments. Other techniques for
calculating derivatives, for ex-
ample, using Fourier Transforms,
are available but are not commer-
1.5 Absorbance
cially popular.

One consequence of these math- 1.0


ematical methods for the calcula-
tion of derivatives is that data
points at the beginning and end of 0.5
the wavelength range are lost. If
three data points are used for the 0.0
process then one data point will be 400 500 600
lost at each end of the range for
each derivative order. If five points
are used then two points will be Absorbance
lost and so on. 5.0E-06

It should be clearly understood 0.0E-06


that, although transformation of a
UV-Visible spectrum to its first or
higher derivative usually yields a -5.0E-06
more complex profile than the
zero-order spectrum (see below),
400 500 600
the intrinsic information content is
not increased. In fact, it is de-
creased by the loss of lower- Figure 2
order data such as constant offset Resolution enhancement
factors.
In absorbance mode, when two However, as noted above the Other background effects that are
Gaussian bands with 40 nm NBW information content of derivative directly proportional to higher
and separated by 30 nm, are added spectra is, in fact, less than the orders of wavelength with the
the result is a single band with a absorbance spectra and it can general form:
maximum midway between the easily be shown that the improve-
two component bands. The two ments in quantitative accuracy are A = a 0 + a1 1 ...+ a n n
components are not resolved. In the result of other effects as
the fourth derivative the presence described below. can be eliminated by using higher
of these two bands is clearly orders of derivative but such
visible with maxima centered Background elimination spectral features with exactly this
close to the lmax of the component A common, unwanted effect in form are very uncommon so this
bands. spectroscopy is baseline shift. This effect has little practical use.
may arise either from instrument
Although the bands have been (lamp or detector instabilities) or
resolved there is no indication of sample handling (cuvette reposi-
whether these arise from two tioning) effects. Because the first
chromophores in a single com- derivative of a constant absor-
pound or in two different com- bance offset is zero, using the first-
pounds. It is often claimed that derivative spectra always elimi-
this increased resolution and the nates such baseline shifts and
increased differentiation between improves the accuracy of quantifi-
spectra in the derivative mode cation. This is illustrated in figure
allows multicomponent analysis 3 where a 0.1 A offset, that would
of mixtures of components with cause a 10 % quantitative error for
similar spectra that cannot be the analyte, is completely elimi-
resolved in the absorbance mode. nated by the first derivative.

1.0
1.0

0.5
0.5

0.0

0.0

300 400 500 600 700 300 400 500 600 700

Figure 3
Background elimination
Discrimination Scattering is inversely propor-
Probably the most important tional to the fourth (Rayleigh,
effect of the derivative process is small particles) or second
that broad bands are suppressed (Tyndall, larger particles) power of
relative to sharp bands and this the wavelength. Because the
suppression increases with relationship is inverse, the use of
increasing derivative order. derivatives will not eliminate the
scattering component from the
This arises from the fact that the spectrum as has been claimed in
amplitude, Dn, of a Gaussian band some publications. However,
in the nth derivative is inversely because the scattering component
proportional to the original resembles a very broad absor-
bandwidth, W, raised to the nth bance, using derivatives discrimi-
degree: nates against it and reduces its
effect on quantification.
1
Dn =
Wn

Thus for two coincident bands of


equal intensity but different
bandwidth in the zero order, the
nth derivative amplitude of the 1.0 Absorbance
sharper band, X, is greater than
that of the broader band, Y, by a
factor that is dependent on the 0.5
relative bandwidth and the
derivative order:
DXn WYn
= 0.0
DYn WXn 400 500 600

Figure 4 shows the effect of taking 0.05 1st derivative


derivatives of two bands, one with
160 nm NBW and one with 40 nm
NBW. In absorbance mode they
0.0
have equal amplitude, in first
derivative the narrower band has
four times greater amplitude and
in the second derivative it has -0.05
sixteen times the amplitude. 400 500 600
0.005
This property is used to improve
2nd derivative
the accuracy of quantification of a
narrow band component in the 0.000
presence of a broad band compo-
nent and to reduce error caused
by scattering. -0.005

Scattering is a common problem in


-0.010
biological analyses resulting from 400 500 600
the measurement of small particu-
lates present in the sample. Figure 4
Discrimination against broad bands
Figure 5 shows an absorbance
band with 40 nm NBW and the
same band in the presence of a 1.0
scattering background. Without
any correction, the amplitude at
500 nm is 1.0920 A instead of 1.0 A
0.5
because of the scattering contribu-
tion. Quantification at this wave-
length results in an error of 9.2 %.
Using the first derivative the 0.0
contribution from the scattering 300 400 500 600 700
component is reduced such that,
using peak maximum to minimum,
the signal in the presence of
scattering is 0.02992 instead of 0.01
0.03024, that is a quantification
error of only -1.1 %. This ability to 0.0
discriminate against scattering
components is widely used in the
analysis of biological fluids that -0.01
contain a high level of particulates,
and in pharmaceutical analyses
300 400 500 600 700
where particulate excipients in
tablets and capsules cause quanti- Figure 5
fication errors. Scatter elimination

Matrix suppression
The analytical problem is often not
simply scattering, baseline shift, or
unwanted broad absorbing Absorbance
components. It is a combination of 1.0
two or more of these that results
in a broad absorbing background
matrix. 0.5

In qualitative analyses,
derivatization often allows the
detection and positive identifica- 0.0
tion of trace levels of a component
in the presence of a strongly
6.0E-05
absorbing matrix. This is illus- Absorbance
trated in Figure 6. A trace, 0.01 A,
4.0E-05
of a 40 nm NBW component with
lmax at 500 nm was added to a
2.0E-05
synthetic matrix. The matrix
comprised offset, second and
fourth-order scatter, and 320 nm 0.0E+00
NBW components with lmax at 300
and 600 nm. In absorbance mode -2.0E-05
the presence of this component is 200 400 600 800
virtually undetectable. In second-
order derivative mode its presence Figure 6
is obvious. Matrix supression
Instrument considerations Signal-to-noise ratio

In quantitative analyses, Virtually all current UV-Visible An unwanted effect of the


derivatization improves the spectrophotometers generate derivatization process is that the
accuracy of quantification in the derivative spectra by mathematical signal-to-noise ratio decreases as
presence of interference caused by means so instrument consider- higher orders of derivatives are
a broad absorbing component, ations for generation of derivative used. This follows from the
matrix, or scattering. Thus in the spectra by optical and electronic discrimination effect and the fact
example given above, quantifica- techniques are not discussed. that noise always contains the
tion of the analyte in the absor- Instrument requirements for sharpest features in the spectrum.
bance mode without any correc- derivative spectroscopy are, in Thus, if the spectral data used in
tion results in an error of nearly general, similar to those for the derivative calculation is at
5000 % (absorbance of 0.502 A conventional absorbance spectros- 2 nm intervals, the noise has a
instead of 0.01 A). Using the copy but wavelength reproducibil- 2 nm bandwidth. If the analyte
baseline-to-valley signal of the ity and signal-to-noise are of band has a bandwidth of 20 nm
second-order derivative the error increased importance. then the signal-to-noise ratio of the
is -2.1% (2.37 x 10-5 instead of first derivative is ten times worse
2.42 x 10-5 A/l2). The increased resolution of than the zero-order spectrum. The
derivative spectra puts increased decrease in signal-to-noise ratio
An example of discrimination demands on the wavelength can be reduced by using the
against a broad absorbing matrix reproducibility of the spectropho- smoothing properties of the
is the quantification of caffeine in tometer. Small wavelength errors Savitzky-Golay polynomial
soft drinks. Soft drinks generally can result in much larger signal smoothing technique but great
contain a mixture of natural and errors in the derivative mode than care must be taken as too high a
synthetic products with added in the absorbance mode. degree of smoothing distorts the
colorants, resulting in a broad derivative spectrum.
featureless absorbance over a The negative effect of
wide wavelength range. In absor- derivatization on signal-to-noise Alternative techniques, such as
bance mode, quantification of also puts increased demands on using a reference wavelength or
caffeine is inaccurate because of low noise characteristics of the full spectrum multicomponent
the matrix effect but good accu- spectrophotometer. It is an analysis with a scattering spec-
racy can often be achieved using advantage in this case, if the trum as standard, may often be
the second-derivative spectra. spectrophotometer can scan and used to achieve the same analyti-
average multiple spectra before cal goals but without the reduced
derivatization to improve further signal-to-noise penalty.
the signal-to-noise ratio.

For the derivatization process it is


important to be able to control the
Anthony J. Owen is product
degree of smoothing that is
manager at Agilent Technologies,
applied in order to adapt to
Waldbronn, Germany.
differing analytical problems. In
the case of the Savitzky-Golay
method this means being able to For the latest information and services visit our
world wide web site:
vary the order of polynomial and http://www.agilent.com/chem
the number of data points used.

Copyright 1995 Agilent Technologies


All Rights Reserved. Reproduction, adaptation
or translation without prior written permission
is prohibited, except as allowed under the
copyright laws.

Publication Number 5963-3940E

Agilent Technologies
Innovating the HP Way

Vous aimerez peut-être aussi