Fourier transform spectrometry in the ultraviolet-visible region


Material Information

Fourier transform spectrometry in the ultraviolet-visible region
Physical Description:
iv, 71 leaves : ill. ; 28 cm.
Glick, Mark R., 1962-
Publication Date:


Subjects / Keywords:
Fourier transform spectroscopy   ( lcsh )
Mass spectrometry   ( lcsh )
Interferometers   ( lcsh )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph. D.)--University of Florida, 1989.
Includes bibliographical references (leaves 68-70).
Statement of Responsibility:
by Mark R. Glick.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001479618
notis - AGZ1576
oclc - 21060280
System ID:

Full Text









ABSTRACT . . ... . .. iii


Historical Overview . . .1
Fourier Transformation . . 4
Advantages . . . 9
Use in the Ultraviolet-Visible Region ... ... 17


Introduction . . .. .. .. .20
Experimental . . ... 21
Results and Discussion . . ... 23


Introduction . ... . 28
Experimental . . . 29
Results and Discussion . . ... 33


Introduction . . .. 44
Experimental . . ... 47
Results and Discussion ... . .... 51

CONCLUSIONS . . ... . 63



Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



Mark R. Glick

May 1989

Chairman: James D. Winefordner
Major Department: Chemistry

An investigation of the analytical merits of Fourier transform

spectrometry, at short wavelengths, was made using a Michelson

interferometer capable of operation in the ultraviolet-visible region

(UV-Vis). The multiplex effect on signal-to-noise ratio was examined.

Advantages and disadvantages of Fourier transform spectrometry in the

UV-Vis relative to conventional spectrometry are discussed.

A review of the use of the Michelson interferometer for spectroscopic

purposes and for Fourier transform spectrometry is presented. The

interferometers currently available for operation in the UV-Vis are


Molecular absorption measurements were made using an interferometer and

a diffraction grating. This approach did not have significant advantages

over conventional methods and the multiplex disadvantage degraded signal-

to-noise ratio. Detection limits are given for some polycyclic aromatic

compounds (PAC).

Molecular fluorescence measurements of PAC frozen in a Shpol'skii

solvent at 77K were performed by Fourier transform spectrometry. Emission

spectra were acquired by using a monochromator to select the excitation

radiation and a cutoff filter after the interferometer. Excitation

spectra were acquired by passing broad-band, ultraviolet radiation through

the Michelson interferometer onto the frozen sample. The detection limits

from the emission spectra were sufficiently poor to prohibit analytical

use. Detection limits obtained from the excitation spectra were

comparable to conventional methods, in some cases even better, due to the

use of broad-band excitation.

A Fourier transform spectrometer for continuum-source, atomic

absorption measurements was constructed and evaluated. Degradation of

signal-to-noise ratio due to the multiplex effect was reduced by using a

grating for dispersion of the radiation before the interferometer.

Continuum radiation passed through a flame containing the analyte. A 5 nm

window of radiation, centered around the absorption lines of interest,

was collected and passed through the interferometer onto a photomultiplier

tube. Detection limits using the interferometer were poorer than

conventional measurements, but a few advantages were realized. Complete

absorption line profiles were acquired, including measurements of the true

background in the region of absorption. Calibration curves were extended

by using the line profile. Wavenumber accuracy was high and spectral

resolving power could be easily modified.


Historical Overview

Fourier Transform Spectrometry uses the interferometer that was

designed in 1881 by A. A. Michelson.1 He first constructed the

interferometer for purposes other than spectroscopic, but before the end

of the nineteenth century he had devoted much time to spectroscopic

interferometry. Michelson's use of the interferometer, and the

rediscovery of the interferometer in the 1950s, are responsible for the

widespread use of Fourier transform spectrometry.

Michelson's first experiment with his interferometer was to test for

the existence of the ether-the stationary, luminiferous medium through

which light waves were thought to travel. He sent, at exactly the same

time, a beam of light a fixed distance in one direction, and a second beam

of light the same distance at a right angle. The two beams were reflected

by mirrors and combined at the starting point.

Michelson reasoned that if the two beams returned at the same time,

then no interference pattern would be observed. The stationary ether

could not exist. However, if one of the beams travelled with the flow of

the earth through the ether, and the other perpendicular to the flow, then

the beams would return at slightly different times. An interference

pattern would be observed and existence of the luminiferous ether would

be demonstrated.


With great care to stabilize the interferometer, Michelson made several

attempts to observe the delay that the ether would produce in one beam of

light. Those famous experiments yielded a null result-no interference

pattern was ever observed-and the existence of the ether was refuted.

By 1888 Michelson was using the interferometer for spectroscopic

purposes. Long before numerical algorithms were available to decipher the

information contained in the complex interferograms, Michelson was using

interference patterns (visibility curves) to observe hyperfine structure

in spectral lines. He proposed the use of atomic radiation as a standard

of length when he noticed the narrow spectral width of a red cadmium line,

and he was the first to propose the existence of the Balmer doublet of

hydrogen. Each was a result of visually inspecting the interference

patterns in his interferometer.

After Michelson, and before the advent of computers, interferometry and

the use of the Michelson interferometer were rare in spectroscopy. The

direct methods of obtaining high-resolution, spectral information with a

dispersive spectrometer replaced the disagreeable method of

interferometry. Any advantages of interferometry that would be manifested

by the Fourier transformation were hidden in its complexity.

Michelson realized the limitations of visual inspection and intuition.

Simple spectral features could be obtained from visibility curves, but

retrieving information in the presence of complex light sources was almost

impossible. To simplify the visibility curves, he used a predispersing

element, a prism spectroscope, to allow only a portion of the radiation

to enter the interferometer. "This is necessary because the spectra of

most substances consist of numerous lines. It is usually better to


separate the various radiations before they enter the interferometer."2

This trick would be rediscovered in the 1980s.3

Michelson built an analog computer to help decode his visibility

curves. Recognizing that the observed interference pattern is the sum of

many harmonics, he built a mechanical device for adding up to 80 harmonic

oscillations and displaying the resultant harmonic waveform. By choosing

the harmonics he created synthetic visibility curves to compare with those

experimentally obtained. Compared to Fourier transform spectrometry,

which extracts spectral information from the interference pattern, the

harmonic analyzer worked backward; the interference pattern was fabricated

from an artificial spectrum. Michelson realized, however, the power of

his interferometer to resolve fine spectral features.

Not until the 1950s were the combined capabilities of interferometry

and Fourier transformation realized. Fourier transformation was shown to

be a suitable, if elaborate, method of decoding the spectral information

contained in the interference pattern. In 1951 the Fourier transform was

first applied to data from a Michelson interferometer.4 The increasing

application of computers soon made the calculations routine. Only one

more feature was needed to make interferometry attractive to


In the 1950s, several researchers pointed out the advantages of

interferometry over conventional spectrometry. Jacquinot noticed that

because an interferometer has a circular geometry, and a conventional

spectrometer uses slits, there is a light throughput advantage.5 Fellgett

demonstrated the multiplex effect, an advantage that arises in

interferometry because the detector views the entire signal during the


recording time. These advantages, and several others, revived interest

in the technique. They are described in more detail in a section below.

Fourier transform spectrometry was shown to be more than a spectroscopic

curiosity; it actually worked better than conventional spectrometry in

some cases.

Fourier Transformation

A simplified diagram of a Michelson interferometer is shown in Figure

1. Most interferometers used in Fourier transform spectrometry are simply

better instrumental implementations of the first Michelson interferometer.

The phase difference between the two beams of radiation which creates

the interference is introduced by separating the incident radiation and

by using unequal pathlengths. The amplitude division of the radiation is

accomplished by dividing the incident radiation with a semi-transparent

beamsplitter. One beam travels a fixed pathlength to a stationary mirror

and is reflected. The second beam travels a variable pathlength to a

movable mirror and returns. The two beams recombine at the beamsplitter,

with constructive and destructive interference depending on the introduced

phase delay.

To illustrate the interference effect, monochromatic radiation

simplifies the explanation. When the path difference (retardation)

between the two legs of the interferometer is zero or a multiple of A,

where A is the wavelength of the incident radiation, the recombining beams

are exactly in phase and interfere constructively. The intensity of the

radiation at the detector is the sum of the intensity of the two beams.

All of the radiation goes to the detector and none returns to the source.












When the retardation is a multiple of IA, the beams will be exactly

1800 out of phase and will interfere destructively. The intensity of

radiation at the detector will be zero. All of the radiation returns to

the source.

If the movable mirror in the interferometer is moved at a constant rate

away from the zero retardation point, the recombining beams will undergo

periodic destructive and constructive interference. For a monochromatic

source of radiation the signal at the detector (the interferogram) will

vary sinusoidally. For polychromatic radiation the detector signal will

depend on the individual sinusoidal variations. The signal will be a

composite of the component interferograms.

The intensity at the detector, I(x), can be expressed as a function of

the retardation, x (cm):

I(x) = I(a) [1 + cos 2xrx]

where I(a) is the intensity of the source radiation, and a is the

frequency of the radiation (cm-1). The frequency of the sinusoidal signal

at the detector depends on the frequency of the incident radiation, a, and

on the rate of change in the retardation, x. The intensity of the signal

depends on the intensity of the radiation.

This simple relationship is complicated by nonidealities in practice.

The beamsplitter may not be exactly 50% reflecting and 50% transmitting,

and its efficiency often varies with the frequency of the radiation. An

experimentally determined factor that depends on a is often used for

correction of the acquired interferograms. Other imperfections that may


arise with a dependence on radiation frequency are variations in detector

response and instrumental response.

In the above equation, I(x) is defined as the cosine Fourier transform

of I(a). The Fourier theorem shows that the reverse is also true. I(a)

is the cosine Fourier transform of I(x). The Fourier transform of the

interferogram signal obtained at the detector, I(x), is the intensity of

the radiation, I(a).

In essence, the spectral information of I(a), the intensity of

radiation at a particular frequency, is encoded in the sinusoidal signal

at the detector. In this way it is an indirect method when compared to

conventional spectrometry which measures I(a) explicitly. Michelson

decoded the observed interferograms by experienced inspection but was

unable to work with complicated interferograms. In Fourier transform

spectrometry this information is decoded by numerical means, by the

Fourier transform.

A thorough explanation of the application of the Fourier transform

exists.6 A simple discussion is sufficient for an understanding of the

operation of Fourier transform spectrometry in the ultraviolet-visible.

To apply the Fourier transform, each point in the interferogram is

multiplied by a cosine function of a given frequency and of unit

intensity. ,The frequency of the cosine function is varied during the

deciphering process. If the interferogram does not contain a component

at that frequency, then the result will be zero. If the frequency of the

function corresponds to a component harmonic of the interferogram, then

the result will be a cosine wave, the magnitude of which is equal to the

intensity of the radiation at that frequency.

The relationship between the two domains, between the spectral

information and the detector signal, can be expressed mathematically.

When polychromatic radiation is passed through the interferometer, the

signal at the detector is the sum of the individual sinusoidal signals:

I(x) I(a) cos2rax 6a + constant

The useful feature of the Fourier transform is the reciprocal


I(a) = I(x) cos2wox Sa + constant.

This expression relates the intensity of a spectral component to the

interferogram at the detector. The Fourier transform converts from one

domain (retardation, x) to a reciprocal domain (frequency, a) which is

more practical. If x is measured in cm, then a is in cm-.

In practice the measurement is more complicated. The relationship

between I(a) and I(x) implies that spectral information can be obtained

over the entire range from -- to +- cm1 (really from 0 to +-o) with

infinitely-high resolving power. This would require a retardation of C

cm, with an infinite number of measurements. The restriction of moving

the mirror a finite distance, and of collecting a manageable number of

data points, limits the resolution and spectral range.

In comparison to conventional spectrometry, spectral information is

obtained in a roundabout way. Spectral information contained in the

incident radiation is encoded by the Michelson interferometer. The

constructive and destructive interference pattern which appears at the

detector contains all of the spectral information in a single signal.

That signal is a composite of many sinusoidal signals which are varying

individually. The Fourier transform decodes the spectral information by

transferring the signal from one domain to another. In Fourier transform

spectrometry the transformation is from the displacement domain to the

spectral domain.

The calculations to perform the Fourier transform on the interferogram

are sufficiently complex and numerous to require the computing capability

of a mini-computer. To reduce the calculation time to under a minute,

algorithms have been devised.7 Today, the actual Fourier transformation

is often a transparent operation to the spectroscopist.


Michelson used the interferometer at an early date to predict hyperfine

splitting of atomic emission lines before any method of observation

existed, but interferometry remained a complicated and indirect method for

obtaining spectral information. At the beginning of the twentieth

century, high-resolution spectrometers were available, capable of

measuring highly resolved spectra directly. Not until the 1950s, when

advantages of interferometric techniques were first demonstrated, was the

Michelson interferometer and Fourier transformation recognized as a

powerful spectroscopic tool.


P. B. Fellgett, in 1951, pointed out in his thesis the signal-to-noise

ratio advantage that arises from the multiplex nature of interferometry

compared to dispersive spectrometry.4 It is now called the multiplex

advantage or Fellgett's advantage, and it appears because the

interferometer allows the entire spectral band to be viewed by the

detector for the duration of the measurement period. This signal-to-noise

advantage is great enough to warrant the use of interferometry in cases

where it applies.

Figure 2 compares dispersive spectral acquisition with multiplex

acquisition. A dispersive technique divides the spectrum into N

resolution elements, determined by the resolving power of the

spectrometer. If the total time of spectral acquisition is T, then each

resolution element is sequentially observed for a time T/N in a dispersive

technique. A multiplex technique, such as Fourier transform spectrometry,

observes each resolution element for the entire measurement period, T.

The multiplex technique, then, views each element for a factor N longer

than a dispersive technique.

For detector-noise limited systems, the signal for each element is

proportional to the time of spectral integration: T/N for the dispersive

case and T for the multiplex case. The noise for each element is

proportional to the square root of the time of integration: (T/N)" for

dispersive and T% for multiplex. The ratio of the signal to the noise is

(T/N)' for dispersive and T% for multiplex. For equal spectral acquisition

time, T, the multiplex technique should have a signal-to-noise ratio

advantage of N4.








Fellgett's advantage appears only when the detector is the limiting

source of noise, when the noise is not proportional to the intensity of

the radiation. This restriction is fulfilled in the infrared region,

where detectors are thermal or photo-conductive devices and inherently

noisy compared to the radiation source. In this region Fellgett's

advantage is fully realized, and for this reason, Fourier transform

spectrometry in the infrared is popular.

In regions of shorter wavelength, in the ultraviolet-visible region,

the detectors are photoemissive devices and are not the limiting sources

of noise. In these regions the spectroscopic systems are usually limited

by the source noise.

For source shot noise, the noise is proportional to the square root of

the signal. A multiplex technique in the ultraviolet-visible region, with

a signal N times larger than a dispersive technique, will also be N" times

more noisy, exactly cancelling Fellgett's signal-to-noise advantage of N0.

For source fluctuations, the noise is directly proportional to the

signal. A multiplex technique in the ultraviolet-visible region, with a

signal N times larger, will be N times more noisy, resulting in a net

signal-to-noise disadvantage of N-.

In practice, the multiplex effect results in a disadvantage in the

ultraviolet-visible region that depends on the nature of the signal. For

dense spectra, when the analyte signal is weak compared to the non-analyte

signal, the signal-to-noise ratio can be small. The noise from the non-

analyte signal is distributed to the weak analyte signal. For sparse

spectra, the multiplex effect is expected to degrade the signal-to-noise

ratio to a lesser extent.


The multiplex effect on the signal-to-noise ratio has been discussed

in the literature.8 The signal-to-noise ratio for a multiplex technique

is defined with regard to the limiting noise. In detector noise limited

cases (as in the infrared) the expression is given by

h Ra T JT Ra
(S/N)m = -
R, T 2 JRn

where Ra is the analyte count rate (s-') and Rn is the detector dark count

rate. The expression for a dispersive technique is given by

Ra T/N
(S/N)d Ra -
Rn T/N ;Rn N

The ratio of the two is an expression for the multiplex effect when the

system is limited by detector noise:

(S/N), JN
A =---

(S/N)d 2

The factor of two appears because half the source radiation in an

interferometer is returned to the source. A multiplex advantage appears

when N is greater than 4, or when the number of elements viewed in the

dispersive technique is greater than 4. Unfortunately, detector noise

limited systems seldom occur outside the infrared.


For the case where the noise in the above expressions is determined by

the photon shot noise in the background, then a slight disadvantage

(A < 1) appears. For fluctuation noise in the background then A can be

significantly less than 1, a serious disadvantage. In the ultraviolet-

visible region the multiplex effect will depend on which noise source is


Fellgett's advantage, or any net multiplex disadvantage that results

in interferometry, cannot be the driving force for using Fourier transform

spectrometry in the ultraviolet-visible region. However, there are other

advantages when interferometry is compared to conventional spectroscopic


One of those advantages was demonstrated by P. Jacquinot in 1954.5

Jacquinot's advantage refers to the greater throughput (or light

transmission) of an interferometer compared to a dispersive spectrometer

of equal resolving power.

For a spectrometer or an interferometer, the throughput depends on the

resolution. When an interferometer has resolving power equal to that of

a dispersive spectrometer, the throughput advantage can be one or two

orders of magnitude. The advantage arises from the geometry of the


Resolving power is defined as R = a/6a. In a dispersive spectrometer

R is limited by the width of the dispersive grating and the slit width.

In an interferometer R is determined by the area of the mirrors and

detector, and the aperture area. The circular geometry of the

interferometer entrance aperture results in a throughput advantage for the

same resolving power.


Throughput, 9, is defined as the product of the solid angle of the

incident beam, 0 (sr), and the area of the beam that is viewed by the

detector, A (cm2):

6 A 0, cm2 sr

The maximum throughput of a Michelson interferometer is given by

9I 21 AI ---, cm sr

where AI is the area of the mirrors. The throughput of a grating

spectrometer is given by

h AG Aa
eG cm2 sr
f o2

where AG is the area of the grating, h is the slit height, and f is the

focal length.

Jacquinot's advantage is the ratio:

I8 2r AI a / Omax 2w AI f a2

eG h AG Aa / f a2 h AG Cmax

The limitation of the dispersive spectrometer is apparent from the

above equation. To achieve the same resolution as the interferometer,

narrow slits must be used. AG is small compared to AI.


Jacquinot's advantage appears in all regions, whether the system is

detector noise limited or source noise limited. However, in the

ultraviolet-visible region, Jacquinot's advantage may be completely

overcome by a multiplex disadvantage.

Connes' advantage refers to the accuracy and precision with which

Fourier transform spectrometry can measure spectral frequency. An

internal monochromatic source of radiation can be used as a reference.

Precise measurement of the calibration frequency can be performed as long

as the reference source is not moved. Frequency accuracy can also be very

high, and external calibration against a source of known frequency is

possible for extreme accuracy.

The most important feature of a Michelson interferometer, when

considering operation in the ultraviolet-visible region, is the high

resolving power. The resolution of a Michelson interferometer varies with

the movement of the mirror, R 2La, where L is the path difference. The

resolving power of a Michelson interferometer can approach R = 106, which

can only be achieved by the best, dispersive instrument. The resolving

power can also be easily adjusted to suit the application, by varying the

length of mirror travel.

The analytical applications that fully utilize the high resolving power

of Fourier transform spectrometry will realize an advantage of the

Michelson interferometer in the ultraviolet-visible region. The

simultaneous wide-range spectral acquisition that is possible may be an

advantage, but this multiplex effect will also result in a signal-to-noise



Use in the Ultraviolet-Visible Region

The researchers who in the 1950s discovered the advantages of

interferometry were most interested in operation in the ultraviolet-

visible region. Michelson certainly was limited to the visible region.

However, few modern interferometers have been designed for operation at

short wavelengths.

Almost all of the Michelson interferometers used today in Fourier

transform spectrometry are designed for operation in the infrared region.

Fellgett's multiplex advantage is the driving force. Although Jacquinot's

throughput advantage does appear at shorter wavelengths, the multiplex

effect can cause a signal-to-noise disadvantage when measurements are not

detector noise limited. The few interferometers that do operate in the

ultraviolet-visible region are used mainly for physical rather than

analytical studies.

A Michelson interferometer designed for the ultraviolet-visible region

is conceptually no different from the commercial, infrared

interferometers. Appropriate optics must be selected for different

spectral regions, but the basic design is the same. Technically, however,

the construction of an ultraviolet-visible interferometer is difficult.

Tolerances must be reduced as the wavelength decreases. Stability and

optical flatness to ;A must be maintained throughout the scan.

One of the first modern interferometers for use in the ultraviolet-

visible region was built by Horlick and co-workers.9-12 To achieve the

necessary stability the moving mirror was driven on an air-bearing

suspension system, which prohibited its use in the vacuum-ultraviolet.

The stability of the drive mechanism was satisfactory for maintaining

alignment to 200 nm with modest resolution.


Horlick demonstrated the use of Fourier transform spectrometry in the

ultraviolet region by measuring atomic emission from an inductively-

coupled plasma and from hollow-cathode lamps. The signal-to-noise

degradation that resulted from the multiplex effect was significant, and

detection limits were one or two orders of magnitude poorer than with a

conventional spectrometer. Attempts were made to reduce the spectral

bandpass of the radiation entering the interferometer to reduce the

multiplex effect.

Three research interferometers with high resolving power have been

constructed for the ultraviolet-visible region. One was a stepped-scan

interferometer in Orsay, France, that used cats-eye reflectors13, and the

other a 1 m interferometer at Kitt Peak.14-16 The Kitt Peak interferometer

also uses cats-eye reflectors to maintain optical alignment by minimizing

tilt. High-resolution measurements of atomic emission and inductively-

coupled plasma emission were performed with the systems. Signal-to-noise

is poorer than with a conventional system, but other advantages of Fourier

transform spectrometry made the measurements worthwhile, especially for

physical studies at high resolution. A third interferometer with even

higher resolving power (5 m) has been constructed at Los Alamos National

Laboratory using the same instrumental design.

Two Michelson interferometers that work in the ultraviolet-visible

region are commercially available, and one of them (Bomem) was used in the

current study. Chelsea Instruments makes a Fourier transform spectrometer

that was designed by A. P. Thorne and is capable of operation to 180 nm.17

In principle it is very similar to the interferometer made by Bomem, Inc.,

except that the Chelsea instrument uses cats-eye reflectors for stability.


The Bomem instrument uses a method of dynamic alignment. By continuously

monitoring the interferogram during the scan it adjusts the mirrors to

maintain alignment.

Both interferometers avoid undersampling of the interferogram

aliasingg) by dividing the fringes of the 632.8 nm HeNe laser. The

frequency of the laser used for reference is 15798 cm-1. To accurately

sample an interferogram the upper limit on the spectrum is 15798 cm-1, or

632.8 nm. To go beyond this limit the interferogram is sampled 8 times

for each HeNe fringe, enabling a spectral range to 63192 cm-1. The data

set generated by this method is enormous and is reduced by real-time


The Bomem interferometer that was used in the current study is capable

of operation from 222 nm in the ultraviolet to 2 mm in the far-infrared.

The interferometer is of conventional design, with flat mirrors at the two

legs of the interferometer. The moving mirror is driven by a steel belt

attached to a direct torque motor. During the acquisition of the

interferogram the alignment is dynamically maintained by monitoring the

HeNe interferogram from three paths through the interferometer. Servo

mechanisms adjust the tilt of the mirrors during the scan.



Fourier transform spectrometry in the ultraviolet-visible region,

because it is source shot noise limited, can have a signal-to-noise ratio

disadvantage compared to dispersive spectrometry, especially with dense

spectra.8 At the expense of poorer signal-to-noise ratio, Fourier

transform spectrometry can be satisfactory for high resolution

measurements.18'19 However, low-resolution spectroscopic studies, such as

molecular absorption measurements, are not expected to be performed with

advantages over conventional methods. The broad, dense spectra with high

intensity throughout a wide spectral range should show a significant

signal-to-noise ratio degradation compared to dispersive spectrometry.

To investigate these expectations, a Michelson interferometer was used to

obtain low-resolution, molecular absorption spectra of polycyclic aromatic

compounds (PAC).

The molecular absorption bands of PAC are in the ultraviolet region,

and any source radiation outside of these bands that reaches the detector

will contribute to the noise of the system. Filters can sometimes be

found to correspond to the region of interest, but this method of limiting

the radiation is not versatile. In this work, a plane grating was used

to select the lamp radiation that entered the interferometer, as

previously suggested.3'14 The spectral window of radiation that passed


through the sample cell and onto the detector could be selected to match

the absorption band of the sample.


Figure 3 shows a schematic diagram of the experimental setup used for

this study. It was similar to a conventional spectrophotometer in most

respects. The Michelson interferometer replaced the spectrometer.

A 150 W xenon arc lamp (ILC Technologies, Sunnyvale, CA) was used as the

excitation source. This lamp has high radiant output in the ultraviolet-

visible region and uses an integral parabolic reflector for collecting and

collimating the radiation.

Collimated radiation from the source was diffracted by a plane-ruled

grating with 600 lines/mm (SLM-Aminco, Urbana, IL). An iris diaphragm

passed only a portion of the dispersed radiation (approximately 70 nm

FWHM). A quartz lens focused the radiation on a second aperture at the

entrance port of the interferometer (DA3.02, Bomem Inc., Vanier, Quebec).

The use of the grating added some versatility to the system. The grating

could be rotated to select the excitation region of interest, while

excluding any radiation that would simply add noise by the multiplex


The absorption cells were 1 cm quartz cuvettes. The cuvettes were

mounted in the sample compartment of the interferometer, with an aperture

to block the HeNe laser radiation that is used for alignment in the

interferometer. An elliptical reflector (Bomem) collected the radiation

transmitted through the cell and focused it onto an end-on photomultiplier

tube (R647, Hamamatsu). No filters were used.

r 0


- i 0--
- Q 0 I

I C)

^ .* ) / _________M ______________ ^f
? I ------ ------ ----I--I"
M <. / <"
a ^^ x ^ ^
*^ / '-1


Each reference spectrum was obtained by adding 1000 interferograms

with an instrumental resolution of 64 cm-1. The sample absorption spectra

were obtained by coaddition of 100 interferograms. The scan rate of the

interferometer mirror was 0.15 cm/s. At this rate, the time of spectrum

acquisition was approximately 5 min.

Results and Discussion

The absorption spectrum of benzo(a)pyrene is shown in Fig. 4. For the

complete acquisition of the absorption spectrum, two measurements were

made using two different windows of excitation radiation. The separate

measurements were then linked by the computer. In Fig. 4, the arrow

points to the point at which the spectrum was connected.

One of the disadvantages of this approach was apparent. Much of the

multiplex effect was destroyed by using a predispersing element to select

a 70 nm window. The signal-to-noise degradation is reduced, but the

information that can be obtained in one pass was limited.

The profiles of the excitation radiation used for the measurements are

shown above the benzo(a)pyrene absorption spectrum. The relative

intensity of the two windows is indicated by the magnitude of the curves.

The source profiles were obtained by using solvent blanks and by rotating

the grating until the window of radiation was centered on the region of

interest. They were used as the reference spectra for the absorption

measurements. Calibration of the grating position was not necessary,

because the interferometer provided an absolute wavenumber reference.

Several analytical figures of merit were calculated for the absorption

measurements of the PAC and are listed in Table 1. The limit of detection



/1 /



( I

*4 -

.--" o


*0 0(

7 "" .,-4

0 r.

o 0
i I P-
Mw g

0 Cx

m C C 3 4
.^ ^ ^ 4j

.^ --^ n) n
<. C X
^'' -. *'- a)O


was calculated as the sample concentration having an absorbance equal to

three times the standard deviation of the blank (abl). One hundred

interferograms were coadded for each blank measurement, and abl was

calculated with 16 blanks. For a conventional spectrophotometer, ab1 may

vary from 10-3 to 10-5 absorbance units, depending upon the particular

instrumental parameters.20 Assuming abl 10-3 and the molar absorption

coefficients given in Table 1, the detection limits for benzo(a)pyrene and

benzo(ghi)perylene expected with a conventional instrument would be 30 ppb

and 15 ppb. These detection limits are an order of magnitude better than

those obtained by Fourier transform spectrometry.

Since the present system should be shot noise limited,22 the signal-

to-noise ratio is expected to be directly proportional to the square root

of the source intensity, and to the square root of the integration time

(or number of spectral coadditions).22'23 Flicker noise should remain

localized to the generating spectral region.22,24

The degradation in signal-to-noise ratio with source intensity decrease

is evident in Fig. 4. The spectrum becomes significantly noisier beyond

36000 cm-1, where the source intensity is low. An increase in source

intensity should result in improved detection limits. This could be

accomplished by using a source with a higher ultraviolet output, such as

a 300-1000 W xenon arc lamp. Better collection optics between the source

and interferometer could also improve the radiation intensity at the

sample. Replacing the grating with a quartz prism would increase the

intensity, because all diffraction orders could be collected. However,

the effect of increased throughput would be partially compensated by the

multiplex effect.


Fourier Transform, Molecular
Absorption Measurements

Peak LODa Average Measured Literature21
A (nm) (ng/mL) %RSDb e (L/ mol cm) e (L/ mol cm)

Benzo(a)pyrene 383.7 500 8.4 26,569 26,500

Benzo(ghi)perylene 288.2 800 3.3 42,600 43,400

Benzo(ghi)perylene 299.4 300 2.8 54,200 59,700

a100 coadditions

bn 3


To check the dependence on integration time, signal-to-noise ratio

measurements taken with coadditions of 100 and 1000 were compared to the

signal-to-noise ratio found for a single scan. The signal-to-noise ratio

was calculated as the mean absorbance signal over a given spectral range,

divided by the standard deviation in the absorbance values across that

range. When the coadditions were increased by a factor of 100, the

signal-to-noise ratio was found to increase by a factor of 10.3 (= 1100).

When the number of coadditions was increased to 1000, a factor of 19.5

increase in signal-to-noise ratio was observed (< J1000). This indicates

that some long-term noise, such as source fluctuations, were contributing

at long integration times. A factor of two can be gained in the signal-

to-noise ratio if the number of coadditions for each sample is increased

from 100 to 1000. The time of acquisition increased by a factor of ten.

Using a grating or prism to select a band of excitation radiation has

advantages. The multiplex disadvantage is minimized by a reduction of the

extraneous radiation at the detector. Filters do not have to be obtained

for each region of interest, and choosing the region can be automated.

The relatively low dispersion of the grating used here resulted in a broad

excitation band (70 nm), and a higher-dispersion grating could be used to

give smaller bands of exciting radiation. For those studies in which a

smaller excitation bandwidth is more suitable, such as continuum-source

atomic absorption,17 the use of a grating with higher dispersion would be

appropriate. Reducing the bandwidth of the exciting radiation at the

detector would improve the signal-to-noise ratio.



Fourier transform spectrometry has greatest advantage when the signal

measurements are limited by the detector noise, as they are in the

infrared region. The multiplex advantage which appears in the infrared

is usually not realized when measurements are made in the ultraviolet-

visible region, where the system is limited by the photon shot noise. The

shot noise in the total signal is uniformly distributed throughout the

ultraviolet-visible spectrum and negates Fellgett's multiplex advantage.25

Where the spectral component of interest is weak relative to the mean

spectral intensity, there can even be a multiplex disadvantage.8,26 Dense

spectra, such as those in absorption spectroscopy, have a lower signal-

to-noise ratio in regions of low intensity. Source flicker noise, which

is dependent on the frequency and not uniformly transferred over the

spectrum, can further degrade signal-to-noise ratio and spectral

resolution.22 Only in a few cases will there be a signal-to-noise

advantage in the ultraviolet-visible region. When a spectrum is sparse,

such as an emission spectrum, and the line of interest is more intense

than the mean spectral intensity, then there will be a signal-to-noise

ratio advantage.27

To minimize the effect of shot noise on an ultraviolet-visible

spectrum, the detector can be blinded to portions of the spectrum. If

only the line of interest reaches the detector, then there can be no

multiplex disadvantage. One way of minimizing unwanted radiation reaching

the detector is to place a low-resolution monochromator before the

interferometer to allow only a small window to be viewed.3 The signal-

to-noise ratio is improved, but much of the multichannel detection

capability is lost. An alternative approach is to limit the spectral

bandpass after the interferometer and before the detector.

Low-resolution molecular spectroscopic studies with large spectral

windows are expected to have a multiplex disadvantage. Shot noise from

the dense spectra will be distributed over the entire region, increasing

detection limits and obscuring weak spectral features. In this study, a

Michelson interferometer was used to acquire molecular fluorescence

excitation and emission spectra with low resolution. The spectra of

coronene and other PAC in frozen Shpol'skii solvents were obtained by

limiting the spectral bandpass of radiation reaching the detector.

Optical filtering was necessary for satisfactory results. The effects of

the multiplex technique on the quantitative determination of these

compounds will be discussed.


A schematic diagram of the fluorescence emission system is shown in

Fig. 5. The source was a 150 W compact xenon-arc lamp (ILC Technologies,

Sunnyvale, CA), operated at 14 A. Radiation from the lamp was focused on

the entrance slit (1 mm) of a small monochromator (H-10, Jobin-Yvon), to

select a band of ultraviolet radiation for excitation. The excitation

radiation was incident on a quartz Dewar flask (SLM-Aminco, Urbana, IL)

filled with liquid nitrogen. Samples were contained in quartz tubes and






were frozen by quickly immersing them in the liquid nitrogen. A 3-in

focal length quartz lens focused the fluorescence emission on the 1 cm

aperture at the entrance port of the interferometer (DA3.02, Bomem Inc.,

Vanier, Quebec).

An appropriate, long-pass, cutoff filter was used at the exit port of

the interferometer to eliminate scattered excitation radiation. A

photomultiplier tube (R647, Hamamatsu, Bridgewater, NJ) detected the


For the acquisition of fluorescence excitation spectra, the

experimental setup was changed to place the sample after the

interferometer. A schematic diagram of the system is shown in Fig. 6.

The source was a 150 W compact xenon-arc lamp. For all measurements, a

water filter was used to remove infrared radiation, and two Schott filter

glasses (UG11) were used to remove the visible radiation. A 2-in focal

length quartz lens focused the ultraviolet radiation on the 1 cm aperture

at the entrance port of the interferometer.

The quartz Dewar flask, filled with liquid nitrogen, was placed at the

focal point in the sample compartment of the interferometer. A quartz

tube containing about 1 mL of the sample was quickly immersed in the

liquid N2. A 1:1 image of the tube was focused onto the entrance aperture

of the cooled photomultiplier tube (9789QB, Thorn EMI, Fairfield, NJ),

operated at -250C. For coronene (Fluka, Ronkonkoma, NY) and anthanthrene

(Aldrich, Milwaukee, WI), a 440 nm interference filter (S0l-440) with a

10 nm bandpass and a long pass cut-off filter (LG 420) were used to allow

only a particular fluorescence emission band to reach the detector. For

benzo(a)pyrene and benzo(ghi)perylene (Aldrich), different interference

a e

o ^o a..

\ o I
rCDr r4

a .- I
c a


0 0
d w 93 -


I co


filters (10 nm FWHM) were used having transmission maxima at 420 nm and

450 nm, respectively.

All filters were obtained from Corion Corp. All solutions were

prepared in UV-grade heptane or hexane (Burdick and Jackson, Muskegon,


The interferogram signals from the photomultiplier tubes were amplified

with a transimpedance amplifier (Model Al, Thorn EMI Gencom Inc., New

York, NY), and then filtered and amplified with a pre-amp (Model 113,

Princeton Applied Research, Princeton, NJ). The interferogram was viewed

on an oscilloscope and sent to the interferometer's analog-to-digital

convertor for processing.

Each spectrum was obtained by coadding 3000 interferograms with an

instrumental resolution of 20 cm-1. The scan rate of the mirror drive was

0.15 cm/s.

Results and Discussion

Figure 7 is an emission spectrum of perylene acquired with the Fourier

transform setup shown in Fig. 5. The spectrum shows the characteristic

narrowing that occurs in frozen Shpol'skii solvents. The half-width of

the narrowest peaks is approximately 0.5 nm. For the acquisition of the

perylene emission spectrum, the monochromator was set to allow a band of

radiation at 400 nm to strike the sample. A 420 nm cutoff filter (50%

transmission at 420 nm) was used to minimize the scattered excitation

radiation reaching the photomultiplier tube.

The greatest drawback in this experimental design is the need to change

the cutoff filter every time the excitation radiation changes. For other

& ITSlU1a I a aIlo a i


PAC with different excitation spectra the monochromator must be set to the

excitation maximum and the cutoff filter must be changed.

The signal-to-noise ratio for the emission spectra was poor due to the

multiplex effect from all of the radiation striking the photomultiplier

tube at one time. The emission spectrum in Fig. 7 was for a sample of 200

ppm perylene.

Realizing that the system was shot noise limited, and that multiplexing

the radiation actually resulted in a reduction of signal-to-noise ratio,

the second system shown in Fig. 6 was chosen for further evaluation. By

placing the sample after the interferometer, excitation spectra could be

acquired. With this setup there was no need to change the cut-off filter

for different PAC because the same portion of the fluorescence emission

was collected.

Since photon shot noise was the limiting source of noise in this study,

attempts were made to reduce the bandwidth of radiation reaching the

photomultiplier tube. The selection of the Schott filter glasses was made

to reduce the excitation light to the region below 400 nm. Stray light

below 400 nm which would reach the detector would not only contribute to

the noise on the excitation spectrum, but would also appear as signal.

Rejection of the stray light was achieved by using a 420 nm long pass cut-

off filter at the detector.

A broad excitation bandpass was chosen to take advantage of the

multichannel capability of Fourier transform spectrometry, although it was

limited to the radiation passed by the Schott glass. The excitation

spectrum was obtained by passing the radiation through the interferometer

to the frozen sample and then measuring the fluorescence. The spectra


obtained for the four PAC are shown in Fig. 8. Only coronene exhibits the

line-narrowing expected from the Shpol'skii effect. One explanation for

the broad excitation spectra observed for benzo(a)pyrene,

benzo(ghi)perylene, and anthanthrene may be found from symmetry

considerations. The coronene molecule is highly symmetrical and therefore

may exist in the frozen, n-alkane, crystalline matrix in only a few

orientations. Having lower symmetry, the other three molecules

undoubtedly occupy a larger number of sites in the matrix. Since the

samples are illuminated with a broad band of ultraviolet radiation, all

of the sites are equally excited, and the resulting spectra are broad.

This broadening was not observed in fluorescence emission spectra where

narrow excitation bands were employed.

Analytical figures of merit were determined for coronene frozen in

the Shpol'skii matrix (Table 2). The calibration for coronene yielded a

response with a linear dynamic range covering more than three orders of

magnitude, with a log-log slope of 0.972. The detection limit was

calculated as the concentration corresponding to a signal equal to three

times the standard deviation of the blank. The standard deviation of the

blank was taken as one-fifth of the peak-to-peak noise in the blank

spectrum at the wavenumber of interest. The detection limits reported

from the literature were determined using wavelength-dispersive

fluorescence spectrometers.28,29

The detection limits for the Fourier transform measurements are better

than those from dispersive instruments. Several explanations can be given

for the increased sensitivity. For the Fourier transform measurements,

a higher source intensity reaches the sample, since no dispersive device



25000. 28010. 31020. 34030. CM-I

Wa v e n umb er

Figure 8. Excitation spectra at 77K. Top to bottom:
benzo(a)pyrene, benzo(ghi)perylene,
anthanthrene, coronene in hexane.


Analytical Figures of Merit for Coronene
Frozen in Shpol'skii Solvent.

Wavelength Excitation Detection Linear Precision
Discrimination Maximum Limit Dynamic
Technique Range

cm-1 ng/mL Orders of %RSD
(Temperature) (nm) (ng) Magnitude

FT 29486 2 3.2 12.6
(77 K) (339.14) (0.4)

Grating 29400 100 3.7
Spectrometer (340) (3)
(77 K)28

Grating 29700 20 3.0 7.6
Spectrograph (337) (1)
(15 K)29


is employed. A wider fluorescence bandwidth (10 nm) is allowed to reach

the detector through the interference filter. In addition, many

coadditions were made at each point on the calibration curve.

Figure 9 demonstrates the ability of this experiment to obtain

directly the spectrum of the excitation source. With the cut-off and

interference filters in place at the detector, the quasi-linear spectrum

of coronene was observed. The interference filter had the effect of

reducing the photon shot noise since the detector viewed less radiation.

It also blocked significant phosphorescence emission of coronene which

would have skewed the spectrum because of its long lifetime. The source

spectrum was obtained by removing the cut-off and interference filters at

the detector.

Figure 9 also shows a limitation with this instrumental set-up.

Since the UG11 interference filter was used on the light source, only

those analytes with excitation peaks in the region between 26,000 and

32,000 cm-i could be reliably detected. This limitation could be

eliminated by placing a low-resolution diffraction grating between the

source and the interferometer. This allowed the source spectrum to be

matched to the excitation bands.

No signal-to-noise multiplex advantage can be cited as motivation for

molecular studies of this kind, but other potential advantages have been

suggested. The multichannel capability of Fourier transform spectrometry

can only be utilized by allowing a large window of radiation to pass

through the interferometer. For absorption and fluorescence emission

measurements, an increase in the shot noise would result. In this study,

excitation spectra were obtained by measuring the fluorescence emission.





* 0

C 0


c 0


0o 0

A I!sua 0 u I a A I D 1 a

By using a 10 nm bandpass interference filter, which limited the radiation

reaching the detector, the shot noise was reduced, but the multichannel

capability was retained because all of the excitation band was allowed to

pass through the interferometer.

The high speed of spectrum acquisition using Fourier transform

spectrometry has also been suggested as an advantage of using an

interferometer in the ultraviolet-visible region. Even with the limited

spectral window used in this study, defined by the interference filter,

the signal-to-noise ratio was sufficiently poor that coaddition of

interferograms for several minutes was necessary. The acquisition of

dense spectra such as fluorescence excitation spectra has a multiplex

disadvantage which requires long integration times.

Stray light is only a minor problem, since only modulated light which

passes through the interferometer is interpreted as signal. This reduces

the need for stringent protection of the detector from room light,

although stray light may contribute to the shot noise. Computer

manipulation of spectra is one inherent advantage to collecting spectra

digitally, and the ease of obtaining a source spectrum makes it possible

to acquire corrected excitation spectra.

The photon shot noise that limits Fourier transform studies is not as

significant in dispersive spectrometry because it is largest where the

signal is the highest. In dispersive spectrometry, the shot noise can be

reduced significantly by electronic low-pass filtering. However, in

Fourier transform spectrometry, the discriminate filtering of higher

frequencies is limited by the sampling rate of the interferogram

acquisition. The sampling rate is determined by the speed of the moving


mirror. The Bomem interferometer uses HeNe laser fringes for reference

and takes 8 samples/fringe to acquire the interferogram. With a mirror

movement of 1.0 cm/s, the sampling frequency is 253 kHz, which prevents

high-frequency filtering below 500 kHz. At a mirror velocity of 0.01

cm/s, the sampling frequency is 2.53 kHz, allowing a 5 kHz high-frequency

rolloff. The reduction in noise is a factor of 10, but the analysis time

increases by a factor of 100. Furthermore, at such low scan rates, the

ability of the interferometer to keep the mirrors aligned is degraded.

The use of Fourier transform spectrometry for the acquisition of low-

resolution, spectrally-dense molecular spectra has disadvantages. The

multiplex effect of always viewing the total photon shot noise spreads the

noise throughout the spectrum, resulting in a lower signal-to-noise ratio

compared to a dispersive system. The multiplex disadvantage is serious

in regions of low intensity. To overcome this effect, long integration

times are essential, effectively negating any speed advantage of Fourier

transform spectrometry that might have been possible.

The analytical figures of merit determined for coronene compare well

with those which have been obtained with conventional instrumentation, but

since the noise at the wavenumber of interest can be affected by radiation

in other regions of the spectrum, the detection limits of this technique

can be expected to degrade with increased sample complexity. Non-analyte

signal reaching the detector will increase the noise. A sample

pretreatment step such as solvent extraction or chromatographic separation

would be required in such cases.

An alternative arrangement, with the sample before the interferometer

and the fluorescence collected at the entrance port, showed poorer


results, since fluorescence emission at all wavelengths reached the

detector simultaneously. The increased photon shot noise in this case

resulted in greater noise. This problem could be eliminated at the

expense of losing multichannel information by placing a monochromator at

the exit port of the interferometer, thereby reducing the spectral window

at the detector from approximately 100 nm to 10 nm or less. In this case,

a slew scan technique to obtain the entire emission spectrum could be

employed as was done by Horlick et a124 and suggested by Winefordner et

al.8 An advantage of this arrangement would be the added selectivity

resulting from the wavelength tunability at the source. A broad

excitation band would not be required and more conventional Shpol'skii

spectra should be observed.



Atomic absorption spectrometers are commercially available only with

atomic line sources. A requirement of these instruments is a separate

source for each element-a drawback that complicates multi-elemental

analysis. This approach increases the cost of the system and demands

optical alignment each time the lamp is changed. To avoid these

complications, atomic absorption with a continuum source has been


Even in the first description of the atomic absorption method in 1955,

the use of continuum sources was suggested.30 An intense continuum source

in atomic absorption would mean that one lamp could be used for the

determination of all elements. Alignment problems and cost would be

reduced. However, to achieve high sensitivities, continuum sources were

abandoned for spectral line sources.

The sensitivity of an atomic absorption method depends on the effective

spectral bandwidth of the excitation radiation relative to the absorption

line.31'32 For conventional line source atomic absorption, the effective

spectral bandwidth is determined by the width of the atomic emission line

in the source. When continuum sources are used, the monochromator

determines the effective spectral bandwidth. To achieve the sensitivity

of line source atomic absorption when a continuum source is used,


effective spectral bandwidth of the monochromator must approach the

bandwidth of the atomic absorption line. Much work on the development of

continuum source atomic absorption dwells on the achievement of this

spectral resolution.

Several approaches have been taken. High resolution spectrometers and

high resolution interferometers have been used to provide the narrow

spectral bandwidth required in continuum source atomic absorption. To

compete with conventional atomic absorption methods, the newer methods

must have comparable detection limits and simple instrumental


High resolution Fabry-Perot interferometers have been used with

continuum sources.33'34 Effective spectral bandwidths approaching those of

line source methods were achieved, but significant disadvantages appeared.

The use of Fabry-Perot interferometers is more complicated, and optical

alignment is critical. To take advantage of the wide spectral coverage

of continuum sources, the interferometer must be capable of operation over

a wide range. Fabry-Perot interferometers, however, have limited use

because of the lack of highly reflective materials that cover a wide

spectral range. In addition, medium resolution monochromators are often

needed to allow only a portion of the spectral range to enter the


High resolution, dispersive spectrometers have been used to provide the

required resolving power. Echelle-grating spectrometers are the most

practical and have been used in many investigations.35-37 Effective

spectral bandwidths equal to line source atomic absorption can be

achieved, and simultaneous, multi-element determinations are possible.


The majority of current research in continuum-source atomic absorption is

based on echelle-grating spectrometer systems.

Other approaches have been taken to achieve high resolution. Spectral

line modulation38 and sample modulation39'40 have been used with medium

resolution monochromators. Resonance monochromators have also been used

to decrease the effective spectral bandwidth.1-43

Of all the methods, the use of echelle-grating spectrometers has been

the most successful and is the subject of several reviews. 44-46 Detection

limits for more than thirty metals are usually a factor of 2 poorer than

those of line-source atomic absorption. Alternative approaches to using

a continuum source must be compared to this method.

The Michelson interferometer has been suggested as a component in an

atomic absorption spectrometer with a continuum source.8,17 Fourier

transform spectrometry has the advantages of high spectral resolution,

wavelength accuracy, and high throughput. Sensitivities of an approach

using a Michelson interferometer should be comparable to those of other

continuum-source methods. The wide spectral coverage of the

interferometer should yield complete absorption spectra.

The disadvantages of Fourier transform spectrometry in the ultraviolet-

visible region have already been documented here and in the

literature.8,26,27 Because the system is photon shot noise limited in that

region, the multiplex advantage is not realized. Dense spectra, in which

radiation over a wide range is present, should even show a multiplex


This investigation was to determine whether the advantages of Fourier

transform spectrometry in continuum-source atomic absorption will


compensate for the expected disadvantages of Fourier transform

spectrometry in the ultraviolet-visible region. A commercial Michelson

interferometer was used for atomic absorption measurements in a flame.

A predispersing element was used to select a spectral window around the

atomic line of interest. Multi-element determinations were performed for

elements within the spectral window. Detection limits for several

elements are given. Background correction is possible because absorption

spectra are acquired. A single source can be used for the determination

of many elements, and linear calibration curves can be extended by using

the absorption profile.


Two experimental setups were used in this investigation. In the atomic

absorption spectrometer shown in Fig. 10, the flame was placed after the

interferometer. Radiation from the continuum source was focused on an

iris diaphragm, and then collimated by a second quartz lens. The

collimated radiation was dispersed by a plane-ruled grating, and a third

lens collected a portion of the radiation and focused it on the entrance

port of the interferometer.

Collimated radiation from the exit port of the interferometer was

passed through the flame directly onto the photomultiplier tube, as shown

in Fig. 10, or was directed with quartz prisms to the photomultiplier tube

4 ft away from the flame.

In the second setup, the flame was placed before the interferometer.

A diagram of the atomic absorption spectrometer is shown in Fig. 11.

Radiation from the continuum source was passed through the flame without




Z r
o a



any spectral dispersion. A lens was used to focus the radiation from the

lamp on an iris diaphragm, before collimation by a second lens. A second

iris diaphragm allowed only a portion of the collimated radiation from the

less turbulent region of the lamp image to pass through the flame.48 After

the flame and before the interferometer, a grating was used to disperse

the radiation. A quartz lens focused a portion of the dispersed radiation

onto the entrance port of the interferometer.

The Michelson interferometer (DA3.02, Bomem, Vanier, Quebec) in these

setups was commercially available and used without modification. A

photomultiplier tube (R647, Hamamatsu, Bridgewater, NJ) was used for

detection of the interferogram. The source of radiation was an

unfiltered, 300 W xenon arc lamp (Cermax, ILC Technologies, Sunnyvale,

CA). Predispersion of the radiation was accomplished by a 2400 gr/mm,

plane-ruled grating, blazed at 300 nm (SLM-Aminco, Urbana, IL). The

grating could be rotated to select the radiation window of interest, which

was focused onto the entrance aperture of the interferometer. The

spectral halfwidth of the source radiation entering the interferometer was

approximately 5 nm.

The fuel-lean air/acetylene flame was produced by a 10-cm slot burner

(Perkin Elmer, Norwalk, CN). Collimated white light, stopped to 1 cm

dia., passed through the flame to the grating. In both designs, an

aperture was used to reduce the collection of flame emission.

Each absorption measurement was made by recording a reference spectrum

with 1.00 cm-1 resolution, unless otherwise indicated. One hundred

interferograms were added for both reference and absorption spectra. The

scan rate of the interferometer was 0.15 cm/s, for an average time of


spectrum acquisition of 10 min. Even at concentrations as high as 10

mg/mL, no analyte emission could be detected with the optical

configuration shown in Fig. 11.

Results and Discussion

A grating was used for predispersion of the radiation entering the

interferometer to limit the window of radiation striking the detector.

In the photon shot noise limited region, predispersion resulted in an

increase in the signal-to-noise ratio, because the photon flux at the

detector is reduced. This limited the spectral region that could be used

for multielement analysis to 5 nm. The entire absorption spectrum in that

window could be obtained by the interferometer, but unlike echelle-grating

systems, multiple lines at discontinuous spectral regions cannot be

simultaneously measured.

Gratings with poorer dispersion could be used to obtain a larger

spectral window, which would permit the acquisition of absorption spectra

over an even wider range. The larger spectral window, however, would also

increase the radiation at the detector and degrade the signal-to-noise


To demonstrate the multiplex disadvantage, the effect of photon flux

at the detector on the signal-to-noise ratio of sodium absorption lines

was investigated. The Bomem interferometer uses an internal HeNe laser

for alignment, which strikes the photomultiplier tube. To attain a

spectrum with high signal-to-noise ratio, the HeNe laser radiation was

spatially blocked at the exit port of the interferometer. In Fig. 12,

Case I shows the relative intensity of the HeNe laser radiation that could

o o

X. sKualUi aA I UIl a








not be blocked, in comparison to the excitation radiation from the

continuum source. For Case II, the internal laser radiation was not

blocked, and an external HeNe laser was also used to increase the photon

flux striking the detector. In Case II of Fig. 12, the laser radiation

has a peak intensity almost 200 times that of Case I. The excitation

radiation was not changed.

The effect on signal-to-noise ratio of the absorption spectrum of the

sodium doublet is shown in Fig. 13. The poorest signal to noise is

obtained in Case II, when the HeNe laser radiation is much more intense

than the excitation radiation. The noise from the HeNe laser radiation

was distributed over the analytical lines. Unfortunately, the radiation

from the internal laser cannot be blocked entirely, and inevitably it

reaches the photomultiplier tube. In the shot noise limited region, this

contributed to a multiplex disadvantage. A solar-blind photomultiplier

tube would avoid this particular problem.

Detection limits for several elements with lines in the ultraviolet and

visible region are shown in Table 3. The detection limits are at least

an order of magnitude poorer than those which have been obtained by an

echelle-grating spectrometer and continuum source. The same trend of

poorer detection limits as the analytical line moves to shorter

wavelengths that is observed in other continuum source AAS methods was

observed here.

To determine the cause of the poorer detection limits, the sensitivity

and noise of the system were investigated. The sensitivity is related to

the effective spectral bandwidth and in this system is determined by the

mirror movement of the interferometer. Table 4 lists elements frequently

0. 2


I 0


Wavenumber (cm-1)

Figure 13.

Comparision of signal to noise
in multiplex experiment.

17008. 0


Detection Limits for Fourier Transform, Atomic
Absorption with a Continuum Source

Element Wavelength (nm) Detection limit (pg/mL)
FT Dispersive46

Ag 328.068 0.2 0.007

Cu 324.754 0.1 0.01

Mn 279.482 0.2 0.01

Na 588.995 0.02 0.003


Atomic Absorption Lines and Widths

ELEMENT A (nm) AA (pm)

Antimony 217.6 1.1
Arsenic 193.7 1.1
Beryllium 234.9 3.4
Cadmium 228.8 1.2
Calcium 422.7 4.1
Chromium 359.4 2.9
Cobalt 242.5 1.6
Copper 324.7 2.4
Gold 242.8 1.1
Iron 248.3 1.7
Lead 217.0 0.9
Lithium 670.8 17.
Magnesium 285.2 2.8
Manganese 279.5 2.0
Mercury 253.6 1.2
Nickel 232.0 1.5
Platinum 265.9 1.2
Potassium 766.5 11.
Selenium 196.0 1.0
Silver 328.1 2.0
Sodium 588.9 8.2
Tellurium 214.2 1.0
Thallium 377.6 2.2
Zinc 213.9 1.3


determined by continuum source atomic absorption and the half-width of the

most prominent lines. The highest sensitivity will be achieved when the

lines are fully resolved. Fig. 14 shows the effect of instrumental

resolution on the measured absorbance at the 327.396 nm line of copper.

As the spectral bandwidth decreases, the observed peak absorbance

increases, until the bandwidth is less than that of the line. Below an

instrumental resolution of 0.03 A the absorbance does not change.

All absorbance measurements made for the determination of the detection

limits in Table 3 were made with an instrumental resolution of 1.00 cm-1,

at the copper line this corresponds to 11 pm. Although this resolution

did not result in maximum absorption, it was chosen as a compromise. Much

longer spectrum acquisition times would have resulted if better resolution

was selected. An instrumental resolution of 3 pm would yield a higher

absorbance for copper, but the time of acquisition would become


One advantage that is realized to some extent is the multichannel

capability. Atomic absorption spectra can be obtained over the profile

of the line, as long as the line falls within the selected window of

radiation entering the interferometer. This would allow the use of the

profile of the absorption line for diagnostic purposes, background

correction, and for extension of the calibration curve. The capability

of extending the linear range of the calibration curve has been

demonstrated with continuum source AAS.49,50 The automatic acquisition of

the absorption spectrum that is possible with this system also allowed

this type of extension. Figure 15 shows the extension of the linear

portion of a calibration curve for sodium. Absorption measurements were



*r-I 0



I j i I I I I I I
0 0

Od .g

o o


*q 0

0 0
aou sqjo sqy




a s


r l .,-4

-= Cd

0 0


S- io

o 8

0 0


o o t'E


taken from the spectrum at the peak of the line profile and on the edge

of the line profile.

The limited multielement capability is demonstrated by the absorption

spectra of Fig. 16 and Fig. 17. Four absorption peaks corresponding to

100 pg/mL of three elements, Cu, In, and Ag, are shown in Fig. 16. At

shorter wavelengths, four absorption peaks corresponding to 100 pg/mL of

two elements, Mn and Mg, are shown in Fig. 17. The analytical use as a

simultaneous multielement method is limited, because only those elements

which happen to have absorption lines within the 5 nm spectral window will

appear in the spectrum. The use of a grating with poorer dispersion would

allow the simultaneous determination of more elements, but signal-to-noise

ratio would decrease due to the increased total light flux reaching the

photomultiplier tube.

Fourier transform atomic absorption spectrometry with a continuum

source is limited as an analytical technique since the spectral range,

sensitivity, detection power, and acquisition time are interdependent and

limited. Nevertheless, the technique may have limited use in specialized

analytical applications, especially where ease of background correction

and line identification are important, and where several selected elements

must be measured simultaneously.






S a


0 0

0 4-

40 E-
0 o

0 P
O 441O



aoupqjo sqv










: I

n e

0 0

0 4)



The analytical use of Fourier transform spectrometry in the

ultraviolet-visible region has been investigated. Five spectrometric

systems using a Michelson interferometer were constructed and evaluated.

Molecular absorption spectra were acquired by using a dispersion grating

before the interferometer to select a window of radiation. Pseudo-line,

fluorescence excitation and emission spectra of polycyclic aromatic

hydrocarbons frozen in a Shpol'skii solvent were obtained. The Michelson

interferometer replaced the spectrometer in a conventional setup. Atomic

absorption spectra were acquired by using a continuum source for

excitation and a dispersive grating before the interferometer.

Detection limits for the absorption study were predictably poorer than

in a conventional system. A preliminary investigation using a continuum

source without any spectral discrimination before the interferometer

showed that the signal-to-noise ratio of the measurement was degraded by

the multiplex effect. Attempting to reduce this effect, a predispersing

grating was used before the interferometer, allowing only 70 nm of

radiation. This improved the signal-to-noise ratio, but coaddition of

interferograms was still necessary for satisfactory results. The greatest

advantage of Fourier transform spectrometry in the ultraviolet-visible

region, the high resolving power, is not used in a molecular study of this

type. The greatest disadvantage, the degradation of signal-to-noise ratio


by the multiplex effect, makes this spectrometric system of limited

analytical use.

Shpol'skii narrowed fluorescence emission spectra were acquired by

placing the Dewar flask before the interferometer, using a low-resolution

monochromator for selecting the excitation radiation. A long-pass cutoff

filter was necessary at the detector to remove scattered excitation

radiation. One of the drawbacks of this configuration was the need to

change the cutoff filter for each change of excitation. This

disadvantage, along with the poor signal-to-noise ratio that resulted from

the multiplex effect of all of the emission radiation striking the

detector, prompted a more thorough investigation of an alternative setup

for the acquisition of fluorescence excitation spectra.

The main difference between the excitation setup and the emission setup

is that for the acquisition of excitation spectra, the filters do not need

to be changed. By using a Schott glass filter that passes radiation in

the ultraviolet and covers the excitation spectra of most polycyclic

aromatic hydrocarbons, one filter can be used for all determinations. A

cutoff filter and interference filter were used at the detector to

eliminate scattered radiation and to allow only a small fraction of the

fluorescence to be detected. The effect is that the detector viewed only

the fluorescence, the multiplex disadvantage was not prohibitive, and

complete excitation spectra could be acquired.

Coaddition of spectra was still necessary to improve the signal-to-

noise ratio, but detection limits were an order of magnitude better than

those of a conventional system. The use of continuum source excitation

also contributed to the excellent sensitivity. The spectrometric system


may be of use analytically because it can provide the high resolving power

needed for some molecular studies. Unfortunately, by using continuum

source excitation, the Shpol'skii effect was only observed for one sample,

coronene, because other, less symmetrical molecules occupied many

different sites in the frozen matrix. The continuum source excited all

of the sites. For other molecular studies, e.g. gas-phase molecules, that

require high resolution in the ultraviolet-visible, Fourier transform

spectrometry may be of use. The multiplex disadvantage could be


For the acquisition of atomic absorption spectra, two experimental

configurations were used, both using a dispersive grating before the

interferometer to reduce the multiplex disadvantage. Placing the flame

after the interferometer prevented the background emission from modulation

by the interferometer, but also located it closer to the photomultiplier

tube. The bright flame emission was difficult to block at the detector,

and the interferogram was an AC signal on top of a noisy, DC offset. To

simplify the setup, the flame was placed before the interferometer for

further investigation.

Although any emission reaching the interferometer would be modulated

and appear as signal, by using collimated excitation radiation, the

emission could be optically disregarded. The grating allowed only a 5 nm

window of radiation to pass through the interferometer, which reduced the

amount of information that could be obtained at once, but also reduced the

multiplex disadvantage. The grating could be rotated to select the

excitation window, and multi-element determinations could be performed if

several atomic lines were within the same window.


Although limited analytically, the atomic absorption spectrometer

constructed with the Michelson interferometer has a few advantages. The

high resolution capability of the interferometer can be utilized for

diagnostic purposes. The acquisition of complete absorption spectra can

be used for true background correction, as well as an investigation of

atomic line profiles. Calibration curves can be extended by using the

entire line profile for their construction.

The Michelson interferometer is limited in its usefulness to routine

analytical use in the ultraviolet-visible region. The multiplex effect

results in a significant signal-to-noise degradation in many cases,

especially where the spectral feature is weak and surrounded by other

strong spectral signals. The throughput advantage that is cited in the

infrared appears in the ultraviolet-visible region, but is often offset

by the multiplex disadvantage.

Three other advantages, compared to dispersive spectrometers, could be

of some analytical use. The wavenumber accuracy and precision of the

interferometer permits the confident identification of spectral features

in a complex spectrum. This advantage has already been realized in

Fourier transform atomic emission studies. A second advantage that could

be of use is the simultaneous coverage of a wide spectral range with high

resolution, even better than a diode array. However, much of this

advantage must be sacrificed for improvement of the signal-to-noise ratio.

The multiplex disadvantage arises because of this feature. In these

studies a predispersing element was often used to limit the spectral range

of radiation entering the interferometer.


The third advantage is most important-the high resolving power of an

interferometer. Resolution that is easily achieved by an interferometer,

even in the ultraviolet-visible region, is possible only by the biggest

conventional spectrometers. Only those analytical studies that require

the high resolving power of the Michelson interferometer will use Fourier

transform spectrometry in the ultraviolet-visible region.


1. A. A. Michelson, Studies in Optics, Phoenix Edition (University of
Chicago Press, Chicago, 1962).

2. A. A. Michelson, Light Waves and their Uses (University of Chicago
Press, Chicago, 1903).

3. E. A. Stubley and G. Horlick, Appl. Spectrosc. 39, 811 (1985).

4. P. B. Fellgett, Thesis (University of Cambridge, 1951).

5. P. Jacquinot, J. Opt. Soc. Amer. 44, 761 (1954).

6. P. R. Griffiths and J. A. de Haseth, Fourier Transform Infrared
Spectrometry (Wiley-Interscience, New York, 1986).

7. J. W. Cooley and J. W. Tukey, Math. Comput. 19, 297 (1965).

8. J. D. Winefordner, R. Avni, T. L. Chester, J. J. Fitzgerald, L. P.
Hart, D.J. Johnson, and F. W. Plankey, Spectrochim. Acta 31B, 1 (1976).

9. G. Horlick and W. K. Yuen, Anal. Chem. 47, 755A (1975).

10.G. Horlick, E. G. Codding, and S. T. Leung, Appl. Spectrosc. 29, 48

11.E. A. Stubley and G. Horlick, Appl. Spectrosc. 39, 800 (1985).

12.G. Horlick and W. K. Yuen, Appl. Spectrosc. 32, 38 (1978).

13.P. Luc and S. Gerstenkorn, Appl. Opt. 17, 1327 (1978).

14.L. M. Faires, Spectrochim. Acta 40B, 1473 (1985).

15.L. M. Faires, B. A. Palmer, R. Engleman, Jr., and T. M. Niemczyk, SPIE
380, 396 (1980).

16.L. M. Faires, B. A. Palmer, R. Engleman, Jr., and T. M. Niemczyk,
Spectrochim. Acta 39B, 810 (1984).

16.A. Thorne, J. Anal. Atomic Spectrom. 2, 227 (1987).

17.B. D. Anderson, T. Yu, P. J. McKeown, and M. V. Johnston, Appl.
Spectrosc. 42, 1121 (1988).


18.B. T. Jones, M. R. Glick, B. W. Smith, and J. D. Winefordner,
Spectrochim. Acta 44A (1988).

19.J. D. Ingle and S. R. Crouch, Spectrochemical Analysis (Prentice-Hall
Inc., Englewood Cliffs, N.J., 1988).

20.E. Voigtman and J. D. Winefordner, Appl. Spectrosc. 41, 1182 (1987).

21.J. M. Harnly, Anal. Chem. 58, 933A (1986).

22.E. A. Stubley and G. Horlick, Appl. Spectrosc. 39, 805 (1985).

23.L. Mertz, Transformations in Optics (Wiley, New York, 1965).

24.T. Hirschfeld, Appl. Spectrosc. 30, 68 (1976).

25.A. S. Filler, J. Opt. Soc. Amer. 63, 589 (1973).

26.E. L. Inman, A. Jurgensen, and J. D. Winefordner, Analyst 107, 538

27.B. T. Jones and J. D. Winefordner, Anal. Chem. 60, 412 (1988).

28.A. Walsh, Spectrochim. Acta 7, 108 (1955).

29.J. D. Winefordner, Appl. Spectrosc. 27, 109 (1963).

30.V. A. Fassel and V. G. Mossotti, Anal. Chem. 35, 252 (1963).

31.C. Veillon and P. Merchant Jr., Appl. Spectrosc. 27, 361 (1973).

32.G. J. Nitis, V. Svoboda, and J. D. Winefordner, Spectrochim. Acta 27B,
345 (1972).

33.P. N. Keliher and C. C. Wohlers, Anal. Chem. 46, 682 (1974).

35.P. N. Keliher and C. C. Wohlers, Anal. Chem. 48, 140 (1976).

36.A. T. Zander, T. C. O'Haver, and P. N. Keliher, Anal. Chem. 48, 1166

37.R. L. Cochran and G. M. Hieftje, Anal. Chem. 50, 791 (1978).

38.M. Marinkovic and T. J. Vickers, Anal. Chem. 42, 1613 (1970).

39.V. G. Mossotti, F. N. Abercrombie, and J. A. Eakin, Appl. Spectrosc.
25, 331 (1971).

40.E. F. Palermo and S. R. Crouch, Anal. Chem. 45, 1594 (1973).

41.M. B. Blackburn and J. D. Winefordner, Can. J. Spectrosc. 27, 137


42.P. L. Larkins, B. Radziuk, and J. C. van Loon, Spectrochim. Acta 38B,
473 (1983).

43.T. C. O'Haver and J. D. Messman, Prog. Anal. Spectrosc. 9, 483 (1986).

44.J. M. Harnly, Anal. Chem. 58, 933A (1986).

45.T. C. O'Haver, Analyst 109, 211 (1984).

46.M. R. Glick, B. T. Jones, B. W. Smith, and J. D. Winefordner, Appl.
Spectrosc. 43, 1101 (1989).

47.R. L. Cochran and G. M. Hieftje, Anal. Chem. 49, 2040 (1977).

48.T. C. O'Haver, J. M. Harnly, J. Marshall, J. Carroll, D. Littlejohn,
and J. M. Ottaway, Analyst 110, 451 (1985).

49.J. M. Harnly and T. C. O'Haver, Anal. Chem. 53, 1291 (1981).


Mark Glick was born in 1962. He attended Goshen College, Goshen,

Indiana. He entered graduate school in 1985.

I certify that I have read this study and that in my opinion it
conforms to acceptable standards of scholarly presentation and is fully
adequate, in scope and quality, as a dissertation for the degree of Doctor
of Philosophy.

James D. Winefordner, Chair
Graduate Research Professor of

I certify that I have read this study and that in my opinion it
conforms to acceptable standards of scholarly presentation and is fully
adequate, in scope and quality, as a dissertation for the degree of Doctor
of Philosophy.

J hn Dorsey
Asb ~Aate Professor of Chemistr

I certify that I have read this study and that in my opinion it
conforms to acceptable standards of scholarly presentation and is fully
adequate, in scope and quality, as a dissertation for the degree of Doctor
of Philosophy.

Ge hard M. Schmid
Associate Professor of Chemistry

I certify that I have read this study and that in my opinion it
conforms to acceptable standards of scholarly presentation and is fully
adequate, in scope and quality, as a dissertation for the degree of Doctor
of Philosophy.

Anna F. Brajter-Toth
Associate Professor of Chemistry

I certify that I have read this study and that in my opinion it
conforms to acceptable standards of scholarly presentation and is fully
adequate, in scope and quality, as a dissertation for the degree of Doctor
of Philosophy.

Charles M. Allen
Professor of Biochemistry and
Molecular Biology

This dissertation was submitted to the Graduate Faculty of the
Department of Chemistry in the College of Liberal Arts and Sciences and
was accepted as partial fulfillment of the requirements for the degree of
Doctor of Philosophy.

May 1989
Dean, Graduate School


3 1262 08553 6679