Group Title: Laser-excited atomic fluorescence in a graphite furnace /
Title: Laser-excited atomic fluorescence in a graphite furnace
Full Citation
Permanent Link:
 Material Information
Title: Laser-excited atomic fluorescence in a graphite furnace
Physical Description: vi, 115 leaves : ill. ; 28 cm.
Language: English
Creator: Goforth, Douglas Edmon, 1960-
Publication Date: 1986
Copyright Date: 1986
Subject: Fluorescence spectroscopy   ( lcsh )
Atomic spectra   ( lcsh )
Graphite   ( lcsh )
Chemistry thesis Ph. D
Dissertations, Academic -- Chemistry -- UF
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
Thesis: Thesis (Ph. D.)--University of Florida, 1986.
Bibliography: Bibliography: leaves 110-114.
General Note: Typescript.
General Note: Vita.
Statement of Responsibility: by Douglas Edmon Goforth.
 Record Information
Bibliographic ID: UF00099331
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: alephbibnum - 000989388
oclc - 17689590
notis - AEW6250


This item has the following downloads:

PDF ( 4 MBs ) ( PDF )

Full Text








I would like to express my sincerest gratitude to Dr. James D.

Winefordner. His guidance and insight have been extremely beneficial

to me. I would like to acknowledge Dr. Benjamin Smith and Dr. Edward

Voigtman for their help with my research. I would also like to thank

my family for their continual support during my education. Lastly, I

would like to thank my friends, who have been an inspiration to me.




ABSTRACT........................... .... ..... .................. .. v


1 INTRODUCTION..............................................1

Theoretical Considerations.................................4
Types of Fluorescence Transitions.....................4
Fluorescence Expressions.............................7
Formation of the Atomic Vapor.............................10
Review of Literature..................................... 13


Laser System.........................................15
Furnace System......................................20
Detection System.....................................27
Procedure.................. ......................... 28
Analysis of Graphite Cuvettes.............................28
Analysis of Graphite Coatings.............................29
Importance of Graphite Coatings......................29
Coating Procedure...................................32
Results and Discussion....................................36
Choice of Graphite Cuvette...........................36
Comparison of Graphite Coatings......................37
Comparison with Previous Studies......................45

3 AN ENCLOSED FURNACE SYSTEM................................51

Experimental....... ............................. 51
Enclosed Furnace System..............................51
Detection Electronics...............................54
Standard Reference Materials.........................56
Procedure...................................... .... 57
Benefits of Computer System...............................58
Results and Discussion....................................61
Comparison of Atmospheres............................61


SRM Results .........................................72
Comparison With Previous Studies.....................72


Experimental.......... ....... .... ..... ...... ........... 76
Results and Discussion.................................... 81

5 THE COPPER VAPOR LASER AS A PUMP LASER....................87

Comparison of a Nitrogen Laser and a CVL as a Pump
Experimental.............................................. 88
Copper Vapor Laser System............................ 88
Procedure.. ..........................................92
Results and Discussion..................................93
Problems Associated With the Fiber Optic..............93
Comparison of Results.................................96

6 CONCLUSIONS..............................................107

Summary ................................................. 107
Future Work..............................................108

REFERENCES............................................... .. 110

BIOGRAPHICAL SKETCH...............................................115

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




August, 1985

Chairman: James D. Winefordner
Major Department: Chemistry

Laser-excited atomic fluorescence in a graphite furnace was

evaluated while varying different aspects of the graphite furnace and

the laser system. The fluorescence system consisted of a nitrogen-

pumped dye laser, a graphite cuvette, a monochromator-photomultiplier

system, a boxcar average, and a readout system (chart recorder or


A plain graphite cup, a graphite rod, and a slotted graphite cup

were tried as the graphite cuvette. In order to protect the graphite

cuvette and also to enhance atomization, a pyrolytic coating, a

tantalum carbide layer, a tantalum foil liner, and a calcium matrix

which forms a carbide layer were compared as coatings for the graphite

cuvette. The plain graphite cup with the pyrolytic coating was found

to give the best results for volatile elements. A graphite tube

furnace was evaluated and found to give improved detection limits over

the plain graphite cup for nonvolatile elements.

An argon atmosphere, a hydrogen-argon atmosphere, and a low

pressure atmosphere were compared as the environment surrounding the

graphite furnace. Both the argon atmosphere and the hydrogen-argon

atmosphere were found to give very good results.

A copper vapor laser was compared to the nitrogen laser as the

pump laser. The high repetition rate (6 kHz) of the copper vapor

laser makes it very attractive for use with the transient signal

generated by the graphite furnace. However, the copper vapor laser

gave mixed results due to inefficient frequency-doubling and also

because of losses associated with using a fiber optic for the laser


Overall, most of the detection limits were in the picogram to

subpicogram range with an analytically useful range from 3 to 7 orders

of magnitude.


Atomic fluorescence spectrometry (AFS) is one of the most

sensitive analytical methods for trace analysis. Atomic fluorescence

is based on the photon excitation of atoms to an excited state which

subsequently undergo radiational deexcitation over a short period of

time (order of nanoseconds). The frequencies as well as the

intensities of the light emitted by electronically excited atoms are

determined by the electronic configurations of the atoms in both their

ground and excited states. Therefore, the spectrum of every atom is

made up of discrete lines which reflect the electronic structure of

that atom.

The fluorescence signal is linearly proportional to the source

irradiance until saturation is achieved. Some of the criteria (1)

important for choosing an excitation source are high irradiance over

the absorption line, good short and long term stability, simplicity of

operation, a usefulness for all spectral lines, and a freedom from

stray light. Several types of conventional sources have given

satisfactory results (2,5): metal vapor discharge lamps, hollow

cathode lamps, electrodeless discharge lamps, and xenon arc lamps.

Both metal vapor discharge lamps and hollow cathode lamps are

relatively low in intensity compared to other sources and therefore

have not been used recently.

Prior to lasers, electrodeless discharge lamps were the most used

AFS sources (4,5). These sources are tedious to produce and no

reliable production and operation methods are available for all

elements. The use of these sources has dropped off considerably since

the advent of lasers. The xenon arc continuum lamp was used in the

early stages of AFS and more recently the Eimac xenon arc lamp was

used in several multi-element AFS systems (6,7).

The pulsed, tunable dye laser (8-10) has evolved as the problem

solving tool for atomic fluorescence analysis. The dye laser has the

following unique features: 1) tunable over the visible wavelength

region and with frequency-doubling, covers the region down to 220 nm;

2) extremely high peak powers; 3) coherence, both spatial as well as

temporal, thereby leading to high power densities (small spot size)

and narrow linewidths monochromaticityy); and 4) pulsed with a low

duty cycle so that the maximum benefit of signal-to-noise ratio should

be obtained for background noise-limited systems by gated operation.

In almost all cases, the detection system has consisted of a

photomultiplier tube and single channel detection. With pulsed laser

systems, the most widely used data processing system has been a

photomultiplier tube wired for a fast response and a boxcar average

(11). The boxcar average has a variable gate that can be operated by

a reference signal so that information is collected only during the

laser firing. In this manner, the background noise is only measured

during the short "on" time of the detector.

Laser-excited fluorescence has been applied primarily to

atmospheric pressure flames (12-14), electrothermal atomizers (15-25),

and inductively coupled plasmas (ICP) (22,26,27). The results for

flames and plasmas have not shown a significant improvement over other

analytical techniques (11,28). Flames are limited by scatter caused

by unvaporized particles in the atomizer when there is a large excess

of matrix (Mie Scattering) (28). Also, native flame species (OH, CH,

CN, C2) or molecular species formed in the combustion process can give

rise to a fluorescence background (29). The ICP does not have the

scattering or background fluorescence of the flame and has been used

in combination with pulsed laser excitation (22) to achieve very low

detection limits for refractory elements. However, electrothermal

atomizers provide the ultimate sensitivity in analytical laser

fluorescence work. For a graphite cuvette, there are increased atomic

concentrations because the atomic vapor is maintained in a smaller

volume than a flame or ICP, and the graphite cell is essentially a

static cell as opposed to a dynamic system like a flame. The graphite

furnace also has decreased quenching of radiationally excited atoms.

Under ideal conditions, the limiting noise for atomic fluorescence in

a furnace should be shot noise from the detector, which is much

smaller than the Mie scatter and the flame background flicker noise.

One added attraction for the electrothermal atomizer is that solid

sampling can be achieved (30).

Theoretical Considerations

Types of Fluorescence Transitions

Various types of atomic fluorescence transitions have been used

for analytical studies. Figure 1-1 shows the types of fluorescence

transitions. There are basically five types of atomic fluorescence

transitions (31). Resonance fluorescence occurs when the same lower

and upper levels are involved in the excitation and deexcitation

processes. Direct line fluorescence occurs when the same upper level

is involved in the excitation and deexcitation processes; however,

different lower levels are used. In stepwise line fluorescence,

different upper levels are involved in the radiational excitation and

deexcitation processes. Sensitized fluorescence occurs when one

species is excited (called the donor) and transfers excitation energy

to an atom of the same or another species (called the acceptor),

either of which deexcites radiationally. Lastly, multiphoton

fluorescence results when two or more photons excite an atomic species

which then radiationally deexcites.

The fluorescence is termed Stokes if the excitation energy is

greater than the fluorescence energy; if the fluorescence energy is

greater than the excitation energy, the process is called anti-

Stokes. The fluorescence is said to be "excited," if the radiational

excitation and fluorescence processes involve only excited states.

Thermally assisted fluorescence occurs if the excitation process

involves a collisional excitation following the radiational


Figure 1-1. Types of Atomic Fluorescence: a) resonance; b) excited
state resonance; c) Stokes direct-line; d) excited state
Stokes direct-line; e) anti-Stokes direct-line; f)
excited state anti-Stokes direct-line; g) Stokes stepwise
line; h) excited state Stokes stepwise line; i) anti-
Stokes stepwise line; j) excited state anti-Stokes
stepwise line; k) thermally assisted Stokes or anti-
Stokes stepwise line (depending upon whether the absorbed
radiation has shorter or longer wavelengths,
respectively, than the fluorescence radiation); 1)
excited state thermally assisted Stokes or anti-Stokes
stepwise line (depending upon whether the absorbed
radiation has shorter or longer wavelengths,
respectively, than the fluorescence radiation); m)
sensitized (D = donor, D* = excited donor, A = acceptor,
A* = excited acceptor, hvE = exciting radiation, hvF =
fluorescence radiation); n) two-photon excitation via a
virtual level; o) two-photon excitation via a real level.

a b

m *
c d e f

g h i

j k 1

D + h D

D A A +D

A A h F
n F

n 0

Fluorescence Expressions

The treatment of the intensity expressions (1) assumes that there

is only a two level system, i.e., ground state, 1, and first excited

state, 2. The basic fluorescence radiance expression is given by

BF )= () 21E 0 kvdv (1)


1 = path length in direction of detection system, m

4w = number of steradians in a sphere, sr

Y21 = fluorescence power (quantum) efficiency, W fluoresced/W


Ev12 = spectral irradiance of exciting radiation at absorption

line, Wm-2Hz-1

kdv = integrated absorption coefficient over absorption line,


h12 g1n2 -1
Skdv = n( )2[1 - ], m Hz (2)


n1 = concentration of analyte atoms, m-3

hv12 = energy of the exciting photon, J

c = speed of light, ms-

B12 = Einstein coefficient of induced absorption, m3j- s-Hz

1,g92 = statistical weights of states 1 and 2, respectively,


ni,n2 = concentration of states 1 and 2, respectively, m-3,

(n1+n2 = nT, the total concentration of atoms in all


The factor in brackets accounts for the absorption decrease caused by

stimulated emission.

Assuming a steady state approach, where the excitation rate

equals the deexcitation rate

B 12Ev12)n B21Ev12 (3)
(k12 + 12 1)n1 = (k21 + A21 c V2)n2


k12,k21 = excitation and deexcitation nonradiational (collision)

rate constants, s-1

A21 = Einstein coefficient of spontaneous emission, s-1

B21 = Einstein coefficient of induced emission, m3Js- Hz

B12 = Einstein coefficient of induced absorption, m3Js-1Hz

n1,n2 = concentrations of electronic states 1 and 2, m-3

c = speed of light, ms-

The fluorescence quantum efficiency, Y21, is defined as

S21 (4)
21 A21 + k21

and A21 is related to B21 and B12 by

87hV 12 hv3 12 g1
A = ( )B = ( )()B (5)
21 3 21 3 g 12
c c

where h is the Planck constant. Combining these expressions, BF


B hv* E* I?
BF (4 )Y21Ev12[n( 2)B12E 2 (6)E
v12 v12


E* 21 (7)
v12 B21Y21

The saturation spectral irradiance, Eslv2 is the source irradiance

which results in a fluorescence radiance 50% of the maximum possible


Es E* ( ) (8)
v12 v12 + g 2

Making substitutions and also substituting for nI in terms of nT, BF

is given by

hv B
BF= ()21E12nT (12[ 12 (9)
F= 4 Tr21 12'T( 0 E 02
1 +-

From these results, several conclusions can be made:

(i) BF is linear with nT as long as the optical density is low.

(ii) BF is linearly dependent upon the source irradiance and the

fluorescence quantum efficiency as long as E12 << ES12

1 hv12
BF = 21Ev12T- c 12 (0)

which is the case for excitation by conventional sources.

(iii) BF is independent of the source irradiance and the

fluorescence quantum efficiency if E 12 > ES'12. There is saturation

of the upper state and BF is the maximum possible value for a given nT

h1 V
BF = ()Y21 Es12 nT 12 1

or by substituting for ESv12

BF = ()hvl2A21nT( 1 ) (12)

Therefore, the fluorescence is determined by the well known parameters

A21' g1' g2' hv12' and by nT. At high concentrations, the exact

expression for BF becomes very complex. There is also a leveling of

the signal due to self-absorption.

Formation of the Atomic Vapor

The purpose of the graphite furnace is to atomize the analyte

salt. Various works (32-38) have attempted to determine the

atomization mechanism for many different elements. These works were

carried out using atomic absorption spectrometry. The measurements

are made more difficult because the possibility of high thermal

gradients, the rapid rates of rise of temperatures, and the transient

signals by these atomizers (39,40) may not allow local thermal


Sturgeon et al. (37) and Sturgeon and Chakrabarti (38) proposed

that there are four potential routes to the atomic species for an

analyte in the graphite furnace

MO / + C + M + CO + M (13)
s/1 s g g

+ M2g + 2M (13a)

MO + M + 0 (14)
s g g

MO + MO + M + 0 (15)
s g g g

MX + MX + M + X (16)
s g g g

where M = metal, 0 = oxygen, C = carbon, and X = halide. If the

sample is in the form of a nitrate or sulfate, the metal oxide usually

results on heating. Therefore, the majority of precursors to gas

phase atoms are metal oxides or metal halides. The appearance

temperature (temperature of the furnace at which the signal begins)

was used as an indication of Ea, the bond dissociation energies of

gaseous species. By determining the precursors to the analyte atoms,

they could predict the atomization mechanism. Other works (41-43)

have suggested that the thermal dissociation of the carbide is an

important mechanism for the refractory elements

MCs/ Mg + Cs (17)

High dissociation energies of carbides are responsible for the high

atomization temperatures required for such elements as Mo and V.

The analyte concentration (3,44) within furnace atomizers depends

upon the sample characteristics, the rate of heating of the furnace,

the thickness of the sample on the furnace walls, and the environment

within and type of furnace. Although there are mathematical models

(38,45,46) describing the time-dependent characteristics of the atom

population within a furnace, they do not attempt to take into account

any experimental parameters. Holcombe and Rayson (47) used a Monte

Carlo simulation to model the release of Ag atoms from the surface of

the graphite. The data supported the proposed mechanisms (48) for the

vaporization of Ag in a graphite furnace.

Losses of the atomic vapor (47,49) include diffusion out of the

ends of the (tube) furnace, diffusion out of the sample injection

port, and losses through the graphite walls. A pyrolytic coating on

the graphite (50-52) is utilized most often to minimize losses of

analyte through the graphite walls. Some chemical interference

include the formation of volatile halides (53,54) and oxides (55) with

the analyte atoms. Matrix modifiers have been used to minimize the

losses of halides and addition of hydrogen to the sheath gas has

reduced interference from oxides.

Review of Literature

The first analytical use of atomic fluorescence was in the 1960s

by Winefordner and Vickers (56) using a flame as the atomizer and a

metal vapor discharge lamp as the source. Since that time, atomic

fluorescence has utilized the sources and atomizers previously

discussed. General reviews on atomic fluorescence are given by

Winefordner et al. (3), Browner (57), Sychra et al. (2), Wineforder

(1,58,59) and Omenetto and Winefordner (60). Omenetto and Winefordner

(61) have given a review of applications for laser fluorescence, and

Omenetto and Human (11) have given some general considerations for

laser fluorescence.

Most of the graphite cells for atomic absorption have been based

on the L'vov furnace (62). Massmann (63) first demonstrated the

usefulness of graphite cells for atomic fluorescence spectrometry.

His cell consisted of a cup shaped graphite cuvette in which the

source radiation enters through the top and the fluorescence is viewed

through a slit cut into its wall. West and Williams (64) constructed

a carbon filament (1-2 mm diameter) within a chamber with quartz

windows. These early works as well as other furnace designs (65,66)

used conventional light sources. A review of fluorescence

electrothermal atomizers is given by Winefordner (67), and general

reviews on electrothermal atomizers are given by Kirkbright (68) and

Wineforder and Vickers (69).

More recently, many workers (15-25) have utilized lasers as the

excitation sources for furnace atomic fluorescence. The pulsed dye

laser is the most versatile laser source for atomic fluorescence. In

the work of Bolshov et al. (19), the fluorescence of atoms above a

graphite cup furnace was excited by a Nd:YAG-pumped dye laser. The

fluorescence was sent to a broadband amplifier and fed into a gated

voltmeter (boxcar averager. The best ever detection limits were

obtained for Pb, Fe, Na, Pt, Ir, Eu, Cu, Ag, Co, and Mn (most notably

2 fg of Pb).

Using a similar system but with a nitrogen-pumped dye laser,

Tilch et al. (23) achieved a detection limit of 5 fg of lead. Human

et al. (22) also achieved a 5 fg detection limit for Pb with an

excimer-pumped dye laser. Using a graphite boat, Hohimer and Hargis

(17) found a detection limit of 2 pg for cesium.

Slightly varied systems include using tungsten spiral atomizers

(24) or graphite tubes (25). Miziolek and Willis (20) studied lead

using double-resonance fluorescence spectroscopy. Greenlees et al.

(70) recorded bursts of six or more fluorescence photons in about 1 s

when a single Na or Ba atom passed through their continuous wave laser

beam in a vacuum. In a similar experiment, Pan et al. (71) measured

sodium atoms and their motion through the laser beam in 200 torr of

helium gas.


Much of the work utilizing laser-excited atomic fluorescence in a

furnace has been carried out in an argon atmosphere. Amos et al. (65)

have also found that having a hydrogen flame surrounding the graphite

burns up any entrained oxygen, thus giving a much more reducing

atmosphere for the atomization process. This section describes a

furnace system that was open to the atmosphere, but with an argon

sheath and a hydrogen flame gas. The main focus of this section was

to determine the best graphite cuvette design and also to see if

varying the atomizer surface could improve detection limits.

The experimental system described below generally applies to this

entire work. Modifications that took place will be described as they



The general experimental system used for all experiments is shown

in Figure 2-1 and the components are described in Table 2-1. The

basic setup can be divided into three parts: the laser system, the

furnace system, and the detection system.

Laser System

The nitrogen-pumped dye laser (Model DL-II dye laser, Molectron;

UV-24 nitrogen laser, Molectron, Palo Alto, CA) was operated at 20 Hz

Figure 2-1. Experimental System for Laser-Excited Atomic Fluorescence
in a Graphite Furnace.


Table 2-1
List of Equipment for Fluorescence System

Equipment Manufacturer

Model DL-II Dye Laser Molectron Corp., 928 East Meadow Dr.,
Palo Alto, CA 94303

UV-24 Nitrogen Laser Molectron Corp., 928 East Meadow Dr.,
Palo Alto, CA 94303

Furnace and Controlling Laboratory Built

Model SCR 20-250 Power Electronics Measurements, 405 Essex
Supply for Furnace Rd., Neptune, NJ 07753

Model EU-700 Scanning Heath,a 530 Main St., Acton, MA 01720

Nebulizer Perkin-Elmer Corp., Main Ave. (MS-12),
Norwalk, CT 08846-258

R1414 Photomultiplier Tube Haamaatsu, 420 South Ave., CN420,
Middlesex, NJ 06856

Model 280 High Voltage Princeton Applied Research, P.O. Box
Power Supply 2565, Princeton, NJ 08540

Model 162 Boxcar Averager Princeton Applied Research, P.O. Box
with a 164 plug-in 2565, Princeton, NJ 08540

Model 562-126 Frequency INRAD, 181 Legrand Ave., Northvale, NJ
Doubling Crystals 07647

Model D-5000 Chart Recorder Houston Instruments, 8500 Cameron Rd.,
Austin TX 78753

Photodiode Trigger Laboratory Built

aNow GCA/McPherson.

for the experiments with the hydrogen-argon (H2-Ar) atmosphere. The

laser had a spectral line width of 0.015 nm (full width half maximum,

FWHM), a pulse width of 5 ns, and a typical pulse energy of 0.5-0.8 mJ

in the fundamental region (360-1100 nm). The dye laser was equipped

with a frequency-doubling system which allowed frequency-doubling in

the range of 220-360 nm and a pulse energy of 5-40 lJ. Since

replacement crystals were difficult to obtain from Molectron, during

the course of this research this system was replaced by an external

frequency-doubling system (Model 562-126, INRAD, Northvale, NJ) which

gave similar output characteristics.

The laser beam was allowed to diverge over a 2 m distance to the

atomizer. The fundamental beam had a diameter of about 2 mm at the

atomizer; the frequency-doubled beam had a diameter of 6 mm at the

atomizer due to optics associated with the frequency-doubling

system. The radio frequency interference from the nitrogen laser was

reduced to a negligible level by enclosing the laser in a well

grounded brass screen Faraday cage.

A laboratory built flowing dye system was used for dye

circulation. The flowing dye system provided several advantages over

the static cell system. Higher repetition rates (e.g., 30 Hz) could

be used because of the large volume of dye and also because the dye

moved very rapidly through the dye cuvette. In the static system,

only 2 mL of dye were confined in the dye cuvette; here, the higher

repetition rates caused thermal distortion of the dye and a reduction

in beam quality. Also, because of the larger volume (50 mL) of dye in

the flowing dye system, the dye would last much longer before it

required changing.

When changing dyes, the flowing dye system was flushed at least

five times with solvent. If the solvent for the new dye was

different, the new solvent was flushed through the system several

times. Finally, the new dye was rinsed through the system once before

filling the system with new dye.

Table 2-2 lists the dyes (Exciton, Dayton, OH) used during the

course of this research. The dyes all had sufficient lifetimes (~2

hours of operating time) that there was no significant degradation in

energy per pulse while running a calibration curve. The dye laser was

turned on only during the atomization cycle of the furnace, thus

extending the lifetime of the dye even more.

Furnace System

The furnace system (Figure 2-2) was similar to that of Wittman

(21) and consisted of a cylindrical phenolic block with water cooled

copper blocks on either side to support the graphite electrodes and

provide electrical contact with the SCR power supply (SCR 20-250, 5kW,

Electronics Measurements, Neptune, NJ). The gas flow control system

consisted of one solenoid valve for the sheath and flame diluant gases

(argon) and one solenoid valve for the flame gas (H2). Since the

flame gas was only used during the atomization stage, the solenoid

switch was very convenient for controlling the gas. The flame gas and

the flame diluant gas were fed through a capillary burner below the

center of the graphite cuvette. Concentric rings surrounding the

capillary burner allowed the sheath gas to cover the entire length of

Table 2-2
Dyes Used in N2-Pumped Dye Laser

Freq. Dbl.
Dye Concentration Solvent WL Range WL Range
(M) (nm) (nm)

PBBO 2.3X103 p-dioxane 388-417

BBQ 1.0X103 p-dioxane 365-407

Coumarin 540A 1.0X102 ethanol 515-583 258-291

Rhodamine 590 5.0X10-3 ethanol 570-630 285-315

Rhodamine 610 3.7X10-3 ethanol 593-646 297-323

Rhodamine 640 5.7X10-3 ethanol 620-673 310-336

Oxazine 720 1.0X10-3 ethanol 658-723 329-362

Figure 2-2. Configuration of Furnace for the Hydrogen-Argon Atmosphere.


:* H 0 IN





the graphite cuvette, thus protecting the cuvette from the

atmosphere. As the graphite cuvette heated up during the atomization

cycle, the hydrogen would ignite, burning up any entrained oxygen.

The gas flow rates were ca. 3 L/min of argon sheath gas, 2 L/min of

argon flame diluant, and 1 L/min of the hydrogen flame gas. Each gas

had its own rotameter that was calibrated with a mass flow meter

(ALK-50K, Hastings, Hampton, VA). A plain graphite cup was used as

the cuvette for all of the reported results.

The heating rate and the final temperature of the graphite were

controlled by the voltage setting on the power supply and also by a

timing circuit. Figure 2-3 shows a simplified diagram of the timing

circuit used. The power supply was programmed so that the output was

linear with the resistance in the controlling circuit (72), with 100 n

corresponding to full power. The length of heating time was

controlled by a timer chip (555, Texas Instruments, Dallas, TX) which

controlled a relay (W171DIP-12, Magnecraft, Chicago, IL). The

normally closed (NC) relay pins were in parallel with a 100 n

resistor. This parallel circuit was in series with a 0-25 Q variable

resistor. The circuit was then connected to two programming inputs on

the power supply.

With the switch of the relay in its normally closed position,

there was no significant resistance in the controlling circuit. The

drying cycle was controlled manually by the 0-25 0 variable

resistor. After drying, a pushbutton was used to initiate the

atomization cycle. When the 555 timer was started, it sent a signal

activating the relay. The normally closed switch was opened, causing

Figure 2-3. Furnace Power Supply Controlling Circuit. NO = Normally open, NC = Normally
closed, 555 Timer = LM 555, Relay = Magnecraft W171 DIP-12, RT = 0-5 MQ, CT =
10 hF.


0-25 fl


the 100 0 resistance in the controlling wires of the power supply, and

thus initiating the set maximum output power on the power supply. The

amount of time the system was activated was controlled by the timing

capacitor, CT (10 oF), and the timing resistor, RT (0-5 MO). A

variable resistor was used for RT so that the atomization time could

be varied between 0 and -5 s. By setting the maximum output voltage

on the power supply and also setting the atomization time with RT and

CT, the graphite cuvette could be reproducibly heated up to any

specified temperature. The temperature of the graphite cuvette was

measured with an optical pyrometer (Model 87C, Pyro, Bergenfield, NJ).

Detection System

A lens (1.5 in dia., 2 in F.L.) and aperture system was used to

project a 1:1 image of the fluorescence into the monochromator (0.35 m

Heath monochromator, stray light--0.1%). The fluorescence was

collected at 900 from the laser beam. A tent of black felt placed

around the furnace and monochromator helped to reduce stray light.

The fluorescence was detected with a photomultiplier tube (Model

R1414, Hamamatsu, Middlesex, NJ) mounted within a well-shielded

laboratory constructed housing. The photocurrent pulse was stretched

slightly by a 1200 0 load resistor (FWHM 100 ns) and connected

directly to a boxcar average (Model 162 with a 164 plug-in, Princeton

Applied Research, Princeton, NJ) operated with a 5 ns gate. The

boxcar was triggered by a photodiode which received a fraction of the

nitrogen laser output. The boxcar output was displayed on a chart

recorder; peak height measurements were made.


The stock solutions were prepared in accordance with the

directions of Smith and Parsons (73). The standards were prepared

from serial dilution of the stock solutions.

Solution samples of 5 pl were deposited using an Eppendorf

micropipette. The analyses were carried out with a drying stage for

20 s at 1000C. During the atomization stage, the temperature was set

for the specific element up to 27000C. When necessary, neutral

density filters were placed between the furnace and the monochromator

to avoid saturating the detection system. The monochromator slit

width was varied between 300 pm and 1500 pm.

After optimization of the experimental setup and conditions,

analytical calibration curves were obtained for each element. The

limit of detection, LOD, was defined as the concentration of analyte

producing a signal which was 3X the standard deviation of the blank.

Analysis of Graphite Cuvettes

The atomizer required a special design for atomic fluorescence

since the fluorescence was viewed at 900 to the laser beam. A

graphite cup (19,23), a Massmann cup (63), a graphite tube with holes

cut in the sides (25), and a graphite rod (21,64) are several designs

that have been tried. The simplicity of the graphite rod makes it

attractive for use with volatile elements. On the other hand, the

graphite cup offers an improvement over the graphite rod because the

sample is in a semi-enclosed environment which gives more of a

"furnace" effect. However, the atoms still have to emerge from the

furnace into the cooler atmosphere before being excited by the

laser. As the atoms vaporize out of the hot furnace they have more of

a chance to react with interferents, such as 02. In principle, the

best atomizers are those which contain the atoms in a hot environment

while they are being excited by the laser. The graphite tube with

holes cut in the sides (25) and the Massmann cup (63) are both

examples of this type.

Before any data were taken, a preliminary investigation was

carried out to determine the best graphite cuvette design. In Figure

2-4, the designs evaluated are shown. For the plain cup and the rod,

the laser beam was directed over the top of the cuvette, and the

fluorescence was viewed at 900. For the slotted cup, the laser was

directed into the cup from above and the fluorescence viewed through

the slots on the side of the graphite cup. The graphite cups were

held between two spring-loaded graphite electrodes. These electrodes

firmly held the cups and helped to give good electrical contacts.

Analysis of Graphite Coatings

Importance of Graphite Coatings

Electrically heated graphite atomizers have found widespread use

in atomic absorption and atomic fluorescence. Unfortunately, in many

cases, full advantage cannot be taken of the high temperatures. The

tubes rapidly deteriorate and a frequent use of standards is necessary

due to a steadily changing response.

Several studies in graphite furnace atomic absorption have

utilized coatings to inhibit reactions between the analyte and the

Figure 2-4. Graphite Cuvette Designs.


0.10" ID

0.1" & 0.16"






graphite. A pyrolytic coating has been the most popular (50-52).

Some alternate coatings include a tantalum treated graphite tube (74),

a molybdenum treated graphite tube (75), tantalum foil liners (76-78),

metal atomizers (24), and a calcium matrix in the standards (79).

In this study, a pyrolytic coating (Pyro), a tantalum carbide

coating (Ta), a calcium matrix which formed a carbide layer (Ca), and

a tantalum foil liner (Foil) were compared. A pyrolytic coating has

the beneficial properties of low permeability to gases, low porosity,

and a resistance to oxidation. The tantalum carbide layer also

prevents interaction between the analyte and the graphite. The

calcium matrix preferentially forms a calcium carbide layer which

protects the graphite much the same as the Ta carbide layer. When

using tantalum foil, there is a reduction in diffusional losses of the

analyte through the graphite wall. There is a prevention of

involatile carbide formation. Finally, analyte compound is reduced by

the tantalum to the analyte element.

Coating Procedure

The graphite was pyrolytically coated in a quartz and graphite

chamber (Figure 2-5). The "coating" gas used for this process

consisted of 90% argon and 10% methane (P-10 gas). The graphite was

heated to 25000C for 10 min while the "coating" gas was flowing

through the chamber at about 13 L/min. The "coating" gas continued to

flow during a 5 min cooling off period.

The tantalum carbide treatment was the same as that of Zatka

(74). The 6% tantalum soaking solution was prepared by weighing 3 g

of tantalum metal into a 100 mL PTFE beaker, adding 10 mL of 1/1,

Figure 2-5. Pyrolysis Chamber.

13 L/MIN


v/v hydrofluoric acid, 3 g of oxalic acid dihydrate, and 0.5 mL of 30%

hydrogen peroxide. The solution was carefully heated to dissolve the

metal. More peroxide was added if the reaction became too slow. When

dissolution was complete, 4 g of oxalic acid and approximately 30 mL

of water were added. The acid was dissolved, and the solution diluted

to 50 mL. The solution was stored in a plastic bottle.

The graphite cuvettes were then immersed in the 6% tantalum

soaking solution in plastic vials. The vials were placed in a

desiccator and then evacuated by a water pump for 20-30 s.

Atmospheric pressure was then restored in the desiccator, the graphite

cuvettes were removed from the soaking solution and dried first in air

(30 min) and then at 1050C (1 h). Each cuvette was then placed in the

furnace system and heated gradually (30 s) to 10000C and then for a

few seconds to 25000C. The treatment was repeated again but the tubes

were soaked for only 10 s under reduced pressure.

The third method consisted of lining the inner surface of the

graphite cup with 0.025 mm thick tantalum foil. The liner was shaped

around a smaller metal rod and then allowed to expand in the cup. The

cup was then taken through several heating cycles which caused the

liner to expand so that there was no clearance between the liner and

the cup.

The fourth method consisted of making the standards in a 1000 ppm

Ca (as the nitrate) matrix (79). The samples were measured in a

pyrolytically coated graphite cup. An ashing step was included to

decompose the calcium nitrate to calcium oxide. A temperature of

6000C was used for this step.

Results and Discussion

Choice of Graphite Cuvette

From the preliminary study concerning graphite cuvettes, the

plain graphite cup was found to give the best results. The graphite

rod worked well with volatile elements, but with the less volatile

elements, atomization was inefficient; the sample was vaporized off of

the graphite without benefit of a semi-enclosed furnace, which

provides higher temperatures inside the furnace. Also, the graphite

rod could only hold 2 pL samples and had a tendency to crack at high


The slots on the slotted cup were cut in opposite sides of the

cup. These slots allowed the fluorescence to be viewed through one

slot while the opposite slot reduced the emission from the hot

graphite which was directly viewed by the spectrometer. In this way,

atoms were still in the hot environment of the cup when they were

excited by the laser. Even with the opposite slot for reducing the

emission from the luminous graphite, the measured emission still

swamped fluorescence signals in the visible region. Extra apertures

placed between the atomizer and the monochromator were of little

help. Enlarging the width of the slots on the graphite did not help

appreciably either. Another problem was the scatter caused by the

laser hitting the bottom of the cup. If the excitation and

fluorescence lines were close together, the background scatter was

very large. Therefore, the plain graphite cup was used; the best

results were obtained even though there was a chance for interference

because of cooling of atoms as they exited the graphite cup.

Comparison of Graphite Coatings

Table 2-3 lists the limits of detection (LODs) obtained for the

different graphite coatings. Figures 2-6, 2-7, and 2-8 show the log

signal versus log concentration calibration curves for Al, Mn, and Cu,

respectively. The bending over of the curves at high concentrations

was due to self-absorption. The pyrolytic coating gave the best

overall results for the elements that were measured. These results

were expected since the pyrolytic coating has been shown before (44)

to give the best overall results.

The calcium matrix detection limits were all slightly worse than

for the pyrolytically coated graphite. The reason for this was that

the Ca matrix resulted in a much larger scatter signal even with the

ashing step, than the other methods. Any carbide layer that may have

formed did not help to enhance the signal.

Except for Al, the Ta foil gave worse detection limits than the

pyrolytic coating and also produced calibration curves that were

nonlinear. Possible reasons for this were that the foil caused more

scatter than the graphite, giving a large background signal. Also,

the foil liner seemed to degrade much faster than the other

coatings. As the condition of the Ta foil worsened, the efficiency of

atomization became poorer. This condition could have accounted for

the low slope of the Mn calibration curve and also the relatively

short analytically useful range (AUR) for Al, when using the foil

liner. Another possible reason for nonlinearity of the calibration

curve was that the foil is known to adsorb oxygen very easily (77).

The adsorbed oxygen occupied active sites at which atomization can

Table 2-3
Limits of Detection (pg) by Laser-Excited Atomic Fluorescence
for Various Coatings of a Graphite Furnace (5 IL aliquots)

Coating Ala Mn Cu

Pyrolytic 5.X102 7.X100 7.X100

Ta Foil 3.X102 1.X102 6.X101

Ta Carbide 1.X102 1.X101 5.X101

Ca Matrix 7.X102 3.X101 4.X101

aLimit of detection is defined as 3 G/m, where 0 = standard deviation
of the blank and m = slope of calibration curve.

Figure 2-6. Log Intensity vs. Log Mass for Al Utilizing Different Coatings. Slopes: Pyro =
0.98, Ca = 1.03, Ta = 1.06, Foil = 0.95. Typical RSD for each point is 10%.



4.0- Co


M / /O
c2.0- // /

0.0 2.0 4.0 6.0 8.0

Figure 2-7. Log Intensity vs. Log Mass for Mn Utilizing Different Coatings. Slopes: Pyro =
0.98, Ca = 0.99, Ta = 1.00, Foil = 0.80. Typical RSD for each point is 10%.

-' 4.0



2.0 4.0 6.0



Figure 2-8. Log Intensity vs. Log Mass for Cu Utilizing Different Coatings. Slopes: Pyro =
0.91, Ca = 0.95, Ta = 0.95, Foil = 1.30. Typical RSD for each point is 10%.


4.0- Ta

c 2.0

0.0 2.0 4.0 6.0 8.0

take place. The oxygen would be removed after several heating cycles,

but the lowest concentration may have given uncharacteristically low

responses before the oxygen was removed. The atomization mechanism

for each of these elements (37,78) would determine whether any of

these possible interferents would actually cause any problems.

The Ta carbide coating worked much better than the Ta foil

liner. The detection limits were not only improved, but also the AUR

and the linearity of the calibration curves were improved. The

improved detection limits for Al also suggest that this coating may

work well with some of the more refractory elements. For refractory

elements, the atomization process has to compete favorably with

carbide formation. The Ta carbide layer precludes subsequent carbide

formation, thereby enhancing the atomization efficiency.

The pyrolytically coated and the carbide coated cups gave

analytically useful results for about 70 atomizations. On the other

hand, the Ta foil lined cups would last only about 50 atomizations.

All three cups gave a reproducibility of about 10%.

Comparison with Previous Studies

In Table 2-4, LODs obtained in this work are compared with other

literature values for the same technique and also furnace atomic

absorption spectrometry. Figure 2-9 shows the log-log calibration

curves in the H2-Ar atmosphere. The LODs for Sn, Pt, and In are

similar or improved over previous works. The limiting noise in these

cases was the photomultiplier dark current. The blank scatter was

very small due to the relatively large separation in the excitation

Table 2-4
Limits of Detection (pg) by Laser-Excited Atomic Fluorescence
in a Graphite Furnace (5 uL aliquots)

Element Wavelength (nm) LAFSa GFAASc
Exc. Fl. This Study Lit.b

In 410.2 451.1 5.X10-2 1.X10-1 NRd

Sn 286.3 317.5 2.X10-1 5.X102 2.X101

Pb 283.3 405.8 2.X10-1 2.X10-3 5.X100

Mn 279.8 280.1 7.X100 2.X10-1 4.X10-1

Cu 324.8 327.4 7.X100 2.X10-1 2.X100

Pt 265.9 270.2 6.X101 1.X102 2.X101

Al 394.4 396.2 5.X102 NRd 1.X100

aLaser-excited atomic f
detection is defined a
blank and m = slope of

luorescence with a graphite
s 3 a/m, where a = standard
calibration curve.

furnace. Limit of
deviation of the

bData for In and Sn from (23); Pb, Mn, Cu, and Pt from (19).

CData from (80).

NR = No Report.

Figure 2-9. Log Intensity vs. Log Mass When Using the Hydrogen-Argon Atmosphere. Slopes:
Al = 1.04, Mn = 0.98, Cu = 0.91, Pt = 0.93, Sn = 1.00, In = 1.10, Li = 0.90, Pb
= 0.80. Typical RSD for each point is 10%.


0.0 2.0 4.0 6.0






and fluorescence wavelengths. The improvements for Sn and In were

also due to the hydrogen flame. Tin has previously been shown (55) to

give improved detection limits when hydrogen was added to the argon

sheath gas for atomic absorption spectrometry. Indium probably has

improved detection limits due to its high volatility. Certainly, the

cool hydrogen flame is hot enough to atomize In; keeping the In atoms

in this atmosphere should minimize losses due to side reactions.

Copper and Mn have about an order of magnitude worse LODs than by

previous fluorescence spectrometric studies. The poorer LODs were due

to the close proximity of the excitation and fluorescence

wavelengths. There was an increase in the magnitude of the scatter,

and there was also an increase in the variability of scatter. The

limiting noise here was a result of particles coming off of the

graphite which flowed into the laser beam and scattered light into the

monochromator. Although the pyrolytic coating improved detection

limits and helped to protect the graphite surface, further

improvements may have been found if the graphite had been

pyrolytically coated by a commercial company; this would have insured

the reproducibility of the coating from cup to cup. Another possible

reason for the better LODs of Bolshov et al. (19) is that they used a

Nd:YAG pump laser. This laser is capable of giving about an order of

magnitude increased energy per pulse in the frequency-doubled range of

the dye laser, when compared to the nitrogen laser as a pump source.

The increased pulse energy should give an accordingly larger signal

until saturation is achieved (1).

Lithium was the only element in which resonance fluorescence was

used. The poor detection limit when compared to atomic absorption

spectrometry is indicative of the scatter problems associated with

resonance fluorescence. Lithium does not have a useful pair of

excitation-emission lines that can be used for nonresonance


The poor detection limit for Pb was due to a large amount of

contamination, both from the graphite cuvette and from the distilled-

deionized water and reagents used to make up the standards. The low

slope of the log-log calibration curve was due to a poor background

correction. Even with the contamination, the detection limit was an

improvement over the detection limit for atomic absorption


The aluminum LOD was two orders of magnitude worse than that for

atomic absorption spectrometry. Part of the reason is due to the

closeness of the excitation and fluorescence wavelengths as previously

discussed. Another problem with Al is that it has a low volatility.

The atomic absorption detection limits were carried out in a tube

furnace, which provided much better atomization conditions, because of

the semi-enclosed design. The fluorescence detection limits could be

improved if the atomization was carried out in an enclosed furnace.

In general, the hydrogen flame acted only to provide a reducing

atmosphere for the atomization process. Only in the case of In, which

was very volatile, did the temperature of the hydrogen flame possibly

help to atomize the sample.


The graphite furnace is generally surrounded by an argon

atmosphere. The inert argon atmosphere keeps out contaminants from

the air and also prevents the graphite from burning. Low pressure

fluorescence (70,71) has also been utilized and it has resulted in the

detection of individual atoms as they passed through a continuous wave

laser beam by using a photomultiplier-photon counter system. This

section describes an enclosed furnace system that was compatible with

either an argon (Ar) atmosphere or a low pressure (LP) atmosphere.

The detection electronics were also changed at this time. The

chart recorder was replaced by a personal computer (PC) as the readout

for the fluorescence system. The benefits of this system will be



Enclosed Furnace System

The laboratory built furnace (Figure 3-1) consisted of a phenolic

base plate and a glass dome, which enclosed the copper electrodes and

the graphite cup. The two copper electrodes were connected to the

leads of the power supply by steel bolts that were fed through holes

in the phenolic. A brass plate with an array of small holes was press

fit into the center of the phenolic base. This plate allowed

Figure 3-1. Configuration of Furnace for Argon and Low Pressure

I 0 vacuum
LA /l--/ /_ pump

Graphite cup

Groove Electrode
with O-

argon to flow into the enclosure. Besides the benefits of the inert

atmosphere, the argon also swept the atoms up into the laser beam.

Two steel 1/4" tubes were also press fit through the phenolic to let

in cooling water for the electrodes. A groove was cut into the

phenolic surrounding the electrodes and water lines. Using vacuum

grease, an o-ring in the groove formed a sufficient seal with the flat

edge of the glass dome to allow a pressure down to 25 torr. Sealant

was used around all of the fixtures (brass plate, electrodes, bolts,

steel tubes) to keep air from entering the chamber. The glass top had

three 1" quartz windows attached with epoxy. One window allowed the

laser beam to enter while the opposite window allowed the laser beam

to leave the enclosure. The exit window was used to minimize air

entrainment and scatter from the glass dome (the curvature of the

glass dome caused considerable scatter). There was also a quartz

window at 900 to these windows so that the fluorescence could pass

through to the monochromator. There was a small outlet in the glass

so that the Ar could escape; a Cajon connector was used to connect

this outlet with a vacuum pump (Model 1400, General Electric, Skokie,

IL) when the LP atmosphere was used. In this system, the laser was

tuned using a nearby flame-monochromator system and the monochromator

used with the furnace was tuned using a hollow cathode lamp.

Detection Electronics

The chart recorder used previously was replaced with a personal

computer (Model 5150, IBM PC, Armonk, NY). Table 3-1 lists the

equipment utilized. In order to make this change, a boxcar average

that was compatible with the computer system had to be used. The

Table 3-1
Equipment List of Modified Electronics System
Using the Enclosed Furnace System

Equipment Manufacturer

SR 250 Boxcar Averager Stanford Research Systems, 460
California Ave., Palo Alto, CA 94306

SR 225 Analog Processor Stanford Research Systems, 460
California Ave., Palo Alto, CA 94306

SR 245 Computer Interface Stanford Research Systems, 460
California Ave., Palo Alto, CA 94306

Model 4163 Preamplifier Evans Associates, P.O. Box 5055,
Berkeley, CA 94705

Model 5150 Personal Computer IBM Corporation, Old Orchard Rd.,
Armonk, NY 10504

boxcar average (SR 250, Stanford Research Systems), analog processor

(SR 225, Stanford Research Systems), and computer interface (SR 245,

Stanford Research Systems) were designed to be compatible with the

IBM PC. A preamplifier (Model 4163, Evans Associates, Berkeley, CA)

was used just before the boxcar average to provide gain and to act as

a filter.

The data acquisition program (SR 265, Stanford Research Systems)

controlled both data acquisition and data reduction. The computer

scan was started just prior to the atomization cycle and was stopped

(manually) after the signal was completed. The computer allocated a

position in memory (bin) every time the system received a trigger

pulse (one trigger pulse for each laser shot). The system was

synchronized by connecting the "Busy Output" of the boxcar average to

the "Sync" on the computer interface. The "Busy Output" provided a

TTL synchronizing pulse when the unit was triggered. In this manner,

a real time output was given by the computer. The integrals of the

peaks were used as the measurement method.

Standard Reference Materials

In order to demonstrate the feasibility of this technique, the

standard reference materials (SRMs) wheat flour (SRM 1567), spinach

(SRM 1570), and steel (SRM 364) were determined for Cu and Mn.

All three samples were placed in a desiccator for 48 hours prior

to weighing out any sample. After drying, the samples were weighed

out above a given minimum sample size for each particular sample. The

specified sample size was to insure homogeneity.

The dissolution technique (S. Hanomura, private communication,

1986) for the wheat flour and spinach consisted of placing each sample

in a round-bottomed (RB) flask. Nineteen milliliters of purified

nitric acid and 1 mL of sulfuric acid were added to the flask. The

distilling apparatus included a nitric acid preserver, which consisted

of a sidearm fitted with a stopcock. Gentle heating evaporated the

nitric acid up into the condensing apparatus. As the nitric acid

vapor recondensed, it would collect in the nitric acid preserver.

When all of the nitric acid had evaporated out of the RB-flask, the

heat would be turned off for 1 min, and then the nitric acid would be

allowed to slowly drain from the nitric acid preserver back into the

RB flask. The entire process would then be repeated. After several

hours, the solution became clear. At this point, the dissolution was

complete. The solution was transferred to a volumetric flask and then

diluted up to 50 mL. The steel sample was put into solution by

placing it in 30 mL of nitric acid along with 2 mL of sulfuric acid.

The solutions were diluted up to 250 mL. Further dilutions were

carried out as needed to bring the concentrations to a workable level.


For the Ar atmosphere the procedure was the same as that for the

H2-Ar atmosphere. The argon flow rate was 5 L/min. For the LP

atmosphere, the sample was dried at atmospheric pressure while argon

was flowing through the enclosure. The argon was turned off and the

enclosure was pumped down to 25 torr. The sample was then atomized as

before. The SRM samples were determined in the Ar atmosphere. For

these samples, a charring step was included with temperatures of 5000C

for Mn and 600C for Cu.

Benefits of Computer System

The major advantage of using the computer for data reduction was

that real time outputs were obtained. The data reduction routines

allowed smoothing and background subtraction before measuring the

integrals of the peaks. L'vov (44) emphasized the importance of

correlating the integrated absorbance reading (rather than peak

absorbance) with concentration. A matrix can cause the peak to become

shorter and broader, causing erroneous measurements for the peak

height method but leaving the integral the same. Also, refractory

elements, such as Mo and V, tend to have severe tailing of the peak.

An improved signal would result by using the integral method.

Figure 3-2 shows a printout of one of the peaks for In in an Ar

atmosphere. If several more features were added, the computer could

be used for determining the mechanistic pathway for the atomization

(37,38). One feature that was added later was a spike in the printout

that occurred at the beginning of the atomization cycle. In this

manner, one could measure how long after the heating ramp started

before the signal began (atomization time). By also adding in a

temperature measuring system (using a calibrated photodiode) for the

graphite cup, the atomization temperature could also be determined.

These values could then be used to indicate the mechanistic pathway

for atomization. The temperature feedback system would also help in

assuring the reproducible heating of the furnace. The temperature

Figure 3-2. Printout of In Peak Atomized in the Argon Atmosphere.
1 bin a 33 ms. Note: The time shown prior to the peak
is arbitrary and is not indicative of the atomization

.2 1







; i i

feedback system was not added due to time constraints and also because

the research was moving in a different direction.

Results and Discussion

Comparison of Atmospheres

Figures 3-3 and 3-4 show printouts of the peak shapes for the

H2-Ar and LP atmospheres, respectively. There was a slight difference

in the peak shape of the H2-Ar atmosphere and the other two

atmospheres. The ignition of the hydrogen may have caused the

atomization to go to completion much faster (from the beginning of the

peak to the maximum peak height). Suzuki et al. (81) suggested that

the atomization mechanism may be altered in the presence of

hydrogen. More information, such as the atomization temperature,

would be needed to confirm this fact.

The results with H2-Ar and LP atmospheres are compared in Table

3-2. The log-log calibration curves for the Ar and LP atmospheres are

shown in Figures 3-5 and 3-6, respectively. In general, the LODs for

the low pressure atmosphere were poorer than for the other two

atmospheres. The original hope in carrying out this low pressure work

was that the quantum efficiency for the fluorescence would be

drastically improved because there would be no interferents to react

with the atoms. Unfortunately, any improvement in quantum efficiency

that may have occurred was more than nullified by the increased

diffusion rates of the atoms. At low pressures, the atoms have high

diffusion rates because there are few collisions with the inert gas.

Since the atoms were in the path of the laser beam a shorter period of

Figure 3-3. Printout of In Peak Atomized in the H2-Ar Atmosphere.
1 bin a 33 ms. Note: The time shown prior to the peak
is arbitrary and is not indicative of the atomization



0 i



o -. qJ



S 78 9 11 13B
5a ?R _8 119 !38


Figure 3-4. Printout of In Pel
Atmosphere. 1 bit
to the peak is ar
atomization time.

ak Atomized in the Low Pressure
n a 33 ms. Note: The time shown prior
bitrary and is not indicative of the

S 38 5 B 7


96 18 13 150
90 110 130 150'




__ I

.. .. . .. . . .. . . .

Table 3-2
Comparison of Hydrogen-Argon (H2-Ar), Argon (Ar), and
Low Pressure (LP) Atmospheres (5 pL aliquots)

Limits of Detectiona AURb Slopec

H2-Ar Ar LP H2-AR Ar LP H2-Ar Ar LP

Cu 7.X100 2.X100 2.X102 4.5 5 3 0.91 1.05 0.95

Mn 7.X100 1.X100 7.X100 3 4 3 0.98 0.98 1.05

Pt 6.X101 1.X100 2.X101 3 3.5 >5 0.93 1.05 0.90

Sn 2.X10-1 1.X101 >6 5 1.00 0.95 -

In 5.X10-2 3.X10-1 7.X10-1 6 6 5 1.10 0.91 1.50

Li 4.X102 4.X102 4.X103 3 3 2.5 0.90 0.92 0.62

a(pg), 3 0/m, where 0 = standard
of calibration curve.

deviation of the blank and m = slope

bAnalytically Useful Range (orders of magnitude).

CSlope of log signal vs. log mass calibration curve.

Figure 3-5. Log Intensity vs. Log Mass for Elements Measured in an Ar Atmosphere. Slopes:
Cu = 1.05, Mn = 0.98, Pt = 1.05, Sn = 0.95, In = 0.91, Li = 0.92. Typical RSD
for each point is 10%.



"m .^- Mn Sn
6o.0- Cu




-2.0 0.0 2.0 4.0 8.0 8.0

Figure 3-6. Log Intensity vs. Log Mass for Elements Measured in a Low Pressure Atmosphere.
Slopes: Cu = 0.95, Mn = 1.05, Pt = 0.90, In = 1.50, Li = 0.62. Typical RSD for
each point is 10%.



8.0 +

U) Pt
c-I Mn
6.0 Cu

4.0 L



0.0 +.
-2.0 0.0 2.0 4.0 6.0 8.0

time and because of the relatively slow repetition rate of the laser,

there was a decrease in signal.

An anomalous result occurred for Sn which gave no signal under

low pressure except at 5000 ng, the highest amount used. This

probably occurred due to the alteration of the mechanism of

atomization; Rayson and Holcombe (55) suggested that oxygen attenuated

the Sn signal by reacting with the Sn when measurements are at

atmospheric pressure. However, at this time, there is not a good

explanation for the loss in signal at low pressures.

The slopes of the log-log calibration curves for In and Li were

also not unity, i.e., signal is a cx when x=1 is considered a linear

calibration. The low slope (0.62) for Li at LP is probably due to

poor background correction caused by scatter. The low pressure

apparently magnified this effect.

The reason for the extraordinarily high slope for In (1.50) at LP

is not well understood. Sturgeon and Chakrabarti (38) have shown that

the atomization time can change with pressure. However, at a

particular pressure, the atomization time as well as the transport of

the atoms into the laser beam should remain constant. There is no

apparent reason why the signal should increase so disproportionately

with analyte concentration.

The LODs in the H2-Ar atmosphere and the Ar atmosphere were

similar. The detection limits for Cu, Mn, and Pt were improved in the

Ar atmosphere. Although the hydrogen flame provided a reducing

atmosphere, the entire system was still open to the air.

Interferents, such as 02, may have gotten through the argon sheath and

the hydrogen flame to react with the atoms, causing losses. The

enclosed Ar atmosphere was much easier to control and much more

reproducible. The reasons for the excellent detection limits for Sn

and In in the H2-Ar atmosphere have previously been discussed.

SRM Results

The results from the SRM samples are shown in Table 3-3. The

results were good considering the SRM samples were determined with

aqueous standards. One of the benefits of atomic fluorescence over

atomic absorption is that in many cases accurate results can be

obtained without having to make standards in a matched matrix.

Comparison With Previous Studies

In Table 3-4, LODs obtained in this work are compared with other

literature values for the same technique and also furnace atomic

absorption spectrometry. The LODs for Sn and Pt in the H2-Ar and Ar

atmospheres, respectively, are considerably improved over the previous

results, while the In LODs are slightly improved.

These results show that both the H2-Ar and Ar atmospheres work

very well. The best results occurred for elements with a relatively

large separation in their excitation and fluorescence wavelengths.

Also, the more volatile elements tended to give better results. The

Al results suggest that an enclosed furnace was needed to improve the

atomization efficiency of nonvolatile elements.

Table 3-3
SRM Standards (ppm, except where noted)

Mn Cu

Sample NBS# Expt. Given Expt. Given

Wheat Flour 1567 7.30.8 8.5+0.5 2.5+0.1 2.00.3

Spinach 1570 128.9 1656 1712 122

Steel 364 0.23% 0.25% 0.29% 0.24%

Table 3-4
Limits of Detection (pg) by Laser-Excited Atomic Fluorescence
in a Graphite Furnace (5 UL aliquots)

Element Wavelength (nm) LAFSa GFAASc

Exc. Fl. H2-Ar Ar LP Lit.b

Cu 324.8 327.4 7.X100 2.X100 2.X102 2.X100 2.X100

Mn 279.8 280.1 7.X100 1.X100 7.X100 2.X10-1 4.X10"1

Pt 265.9 270.2 6.X101 1.X100 2.X101 1.X102 2.X101

Sn 286.3 317.5 2.X10- 1.X101 5.X102 2.X101

In 303.9 325.6 5.X10-2 3.X10-1 7.X10-1 1.X10-1 NRd

Li 670.8 670.8 4.X102 4.X102 4.X103 MRd 3.X10

Pb 283.3 405.8 2.X10-1 2.X10-3 5.X100

Al 594.4 396.2 5.X102 NRe 1.X100

aLaser-excited atomic fluorescence with a graphite furnace. Limit of
detection is defined as 3 a/m, where a = standard deviation of the
blank and m = slope of calibration curves.

bData for In and Sn from (23); Pb, Mn, Cu, and Pt from (19).

CData from (80).

dNR = No Report.


In most of the atomic fluorescence works utilizing a graphite

furnace, an open furnace (15-24), such as a graphite cup, has been

used. A tube furnace has been utilized most often in atomic absorption

spectrometry. The semi-enclosed design of the tube provides better

conditions for atomization than does the graphite cup. The difficulty

with using a tube furnace for atomic fluorescence is viewing the

fluorescence at 900. Dittrich and Stark (25) observed fluorescence

using holes in the sides of a graphite tube to send the laser beam

through and then viewing the fluorescence through the end of the tube.

However, they only reported improved sensitivities (slope of calibration

curve) and did not discuss any other figures of merit.

Because of the poorer atomization conditions of the graphite cup,

many of the refractory elements have not been studied by atomic

fluorescence using a graphite furnace. The refractory elements

generally require very high temperatures for atomization due to the

formation of involatile carbides with the graphite (41-43). In

Chapter 2, poor results were obtained for Al when using the graphite

cup. In order to improve on those results, better atomization

conditions were needed. This section describes a tube furnace to be

used with laser-excited atomic fluorescence.


Figure 4-1 shows the layout of the tube furnace system. The

laser beam passed through a hole (1/4") in the mirror and then excited

the atoms in the graphite tube. The fluorescence was collected back

in the same direction as the laser beam. The fluorescence was

reflected off of the mirror and then a 1:1 image of the fluorescence

was focused by a lens (2 in dia., 3 in F.L.) into the monochromator.

The tube furnace was placed in the enclosed Ar atmosphere; there was a

glass dome over the tube furnace and both the laser beam and the

fluorescence had to pass through a quartz window. The quartz window

was between the tube furnace and the mirror. One change made to this

quartz window was that the glass arm holding the window was cut at a

450 angle upward. Consequently, the small (~5%) laser reflection from

the window was directed upward rather than back towards the mirror

where it might be directed into the monochromator, causing a large

scatter signal. The graphite tube (Figure 4-2) was held between two

spring-loaded graphite electrodes. The graphite tube was not

pyrolytically coated due to excessive heat from the bulky tube furnace

during pyrolysis; the pyrolysis chamber could not withstand this

heat. Two power supplies (TCR 34, Electronics Measurements, Neptune,

NJ) were wired in parallel to provide rapid heating of the graphite

tube. The same controlling circuit (Chapter 2) was used for

controlling these power supplies. The laser system, the Ar furnace

system, and the detection electronics were the same as described in

Chapter 3.

Figure 4-1. Optical Layout of Tube Furnace System for Laser-Excited Atomic Fluorescence.





Figure 4-2. A Graphite Tube Furnace for Laser-Excited Atomic Fluorescence.

I o


I 50
;< ---- 0.50" --

The mirror and graphite tube system were lined up by placing the

mirror in such a way that the laser beam would pass through the hole

in the mirror and also the graphite tube. The monochromator was set

in the visible wavelength region and the output from the

photomultiplier tube was connected directly to a chart recorder. The

graphite tube was then heated for a long period of time while the

mirror and lens were aligned to give the largest signal from the

graphite emission. This alignment insured that the center of the

graphite tube was focused onto the monochromator slit.

Results and Discussion

Table 4-1 lists the results for the tube furnace. Figure 4-3

shows the calibration curves for the elements measured. In comparing

these results with the previous graphite cup results, the detection

limit for Cu was poorer and for Al was better. Also, it should be

noted Mo and V were not even measurable in the graphite cup, but were

detectable in the tube furnace, which shows an improvement for the


The Cu results were slightly worse in the tube furnace than for

the cup furnace because Cu is relatively volatile and therefore

atomizes easily in the graphite cup. When using the graphite tube,

the amount of fluorescence collected was significantly less than with

the cup furnace. In the tube furnace there were light losses when

fluorescence escaped through the hole in the mirror which allowed the

laser to pass through. Also, the graphite tube limited the solid

angle of light exiting the tube. For these reasons, the tube furnace

Table 4-1
Limits of Detection (pg) for
the Tube Furnace (5 VL aliquots)

Element Wavelength (nm) This Worka GFAAS
Exc. Fl. Tube Cup

Al 394.4 396.2 1.X102 5.X102b 1.X100

Cu 324.8 327.4 8.X100 2.X100 2.X100

Mo 313.3 317.0 1.X102 NSC 1.X102

V 385.6 411.2 2.X105 NSc 2.X101

aLimit of detection is defined as 3 G/m, where 0 = standard deviation
of the blank and m = slope of calibration curve.

bHydrogen-argon atmosphere.

cNS = No Signal.

Figure 4-3. Log Intensity vs. Log Mass for the Graphite Tube Furnace. Slopes: Cu = 0.90,
Mo = 0.80, Al = 0.55, V = 0.95. Typical RSD for each point is 10%.


s.o0 Cu


nc x



c 2.0


0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

did not work as well as the plain graphite cup for the more volatile


The improved results for Al, Mo, and V showed that the graphite

tube was more effective in atomizing these elements than the graphite

cup. The semi-enclosed furnace was necessary for atomizing these

refractory elements (Mo, V).

The results were still significantly worse than the values for

atomic absorption, especially for V. The results from this work could

have been greatly improved if a pyrolytic coating for the graphite had

been used. The pyrolytic coating helps to minimize carbide

formation. Also, the limiting noise on the blank was laser scatter

into the monochromator from particles ejected from the graphite

surface. The pyrolytic coating would have minimized most of this

scatter. Another problem was the laser scatter caused by the front

quartz window. Even though most of the scatter was deflected upward,

a small portion would still be measured by the photomultiplier tube.

In most cases, at least a "2" (orders of magnitude) neutral density

filter had to be used in front of the monochromator so that the

detection electronics would not be overloaded. By placing the front

quartz window at the Brewster angle (82), such scatter would have been


The other major problem was light losses through the hole drilled

in the mirror. A 1/8" hole was originally used but alignment was much

more difficult, and there was more scatter as the laser beam passed

through the hole. This scatter was worse when a frequency-doubled

beam was used since this beam had a much larger diameter.


Even with the problems associated with viewing the fluorescence

at 900, better LODs were obtained for nonvolatile elements using the

tube furnace. The more volatile elements worked better with the plain

graphite cup.


Laser-excited atomic fluorescence with a graphite furnace has

generally employed either a nitrogen laser (17,21,23,25), an excimer

laser (22), or a Nd:YAG laser (16,18-20) as the pump laser. All of

these lasers have very high energies/pulse (e.g., 10 mJ/pulse for a N2

laser) and broad tuning ranges when used with a dye laser (360-900 nm

in the fundamental and 220-360 nm in the frequency-doubled range).

For the most part, they can only be operated on the order of 50 Hz or

less. In most applications, this repetition rate is quite adequate.

When using a graphite furnace though, the atomic vapor emerging from

the furnace is in the path of the laser for only about 2 s at most.

Because the pulse widths of these lasers are on the order of 10 ns and

because of the slow repetition rate, many of the atoms escape without

ever being excited by the laser. One possible solution to this

problem is to use a copper vapor laser (CVL), which operates at 6 kHz.

The CVL was developed in 1966 (83) and is still somewhat of a

novelty among scientists. The early lasers achieved only 20 mW

average power and a lifetime of a few hours. Now, the CVL operates at

average output powers of 10-40 W and lifetimes of several hundred

hours (84), before reloading of Cu is necessary. Some of the possible

applications (85) of this laser are underwater research, photography

and holography, semiconductor research, and fingerprint detection.

Comparison of a Nitrogen Laser and a CVL as a Pump Laser

Table 5-1 compares some important specifications for dye lasers

when pumped either by a nitrogen laser or a CVL. The big advantage to

using the CVL with a furnace is that this laser will produce 12,000

laser pulses while the analyte atomic vapor is present in the furnace

(~2 s). With this many laser pulses, every atom should be excited

many times. The CVL also had a longer pulse width, yielding a better

possibility of exciting all of the atoms.

The major disadvantage of the CVL was the limited tuning range of

the dye laser. The wavelengths of excitation for the CVL were 510 nm

and 578 nm; therefore, the fundamental dye range was above these

wavelengths. Even with frequency doubling, there was still a "hole"

in the tuning range. There have only been a limited number of works

(86-88) that have dealt with utilizing different dyes to extend the

tuning range of the laser. This tuning range could easily be extended

to longer wavelengths (e.g., 900 nm), but there would still be a

"hole" in the tuning range.


Copper Vapor Laser System

The laser system consisted of a CVL (Model 251, Plasma Kinetics,

Pleasanton, CA) and a dye laser (DL 13, Molectron, Palo Alto, CA).

The CVL was operated at 6 kHz. The only major differences between

Table 5-1
Comparison of Nitrogen and
Copper Vapor-Pumped Dye Lasers

Dye Laser Characteristics Nitrogen Laser Pump CVL Pump

Fundamental Tuning Range 360-950 nm 525-700 nma

Frequency Doubled Range 217-360 nm 260-350 nmb

Pulse Width 5 ns 24-30 ns

Repetition Rate 0-50 Hz 6 kHz (nominal)

Pulse Energy 1 mJ 0.3 mJ

Peak Power 200 kW 15 kW

published data to date.

bAssumed values from fundamental tuning range.

this dye laser and the nitrogen-pumped dye laser were that the turning

mirrors for the CVL-dye laser were coated to reflect the 510 nm and

578 nm wavelengths (rather than 337 nm) and the dye pump (Model

DL-363, Molectron) had a much higher flow rate. The high flow rate

was necessary to prevent heating of the dye caused by the high

repetition rate of the CVL. Table 5-2 lists the dyes used for the CVL

system. When dyes were used that lased below 578 nm, a dichroic

mirror was used to block the 578 nm wavelength from the CVL.

The external frequency-doubling system (Autotracker II, INRAD,

Northvale, NJ) was capable of tracking the fundamental wavelength so

that the frequency-doubled power would stay optimized. A focusing

lens (1 in dia., 4 in F.L.) was placed just prior to the frequency-

doubling system. While the autotracking accessory was not necessary

for data collection, it was very convenient for scanning the laser

when trying to locate the excitation wavelength in the flame.

Because the CVL-dye laser system was at the opposite end of the

laboratory from the furnace system, a 24-foot fiber optic cable

(quartz, 600 um core, Quartz and Silice, Plainfield, NJ) was used to

send the dye laser beam down to the furnace system. A focusing lens

(1.5 in dia., 3 in F.L.) was used to focus the beam into the fiber

optic. On the furnace end, a focusing lens (1.5 in dia., 2 in F.L.)

was used to project a nearly 1:1 image of the laser beam (as it exited

the fiber optic) over the furnace. In practice, the lens was adjusted

so that the laser beam was the same diameter as the graphite cup. An

aperture was also used to reduce the stray light.

Table 5-2
Dye List for Copper Vapor-Pumped Dye Laser

Freq. Dbl.
Dye Concentration Solvent WL Range WL Range
(M) (nm) (nm)

Oxazine 720 6.6X10-4 methanol 655-700 328-350

Rhodamine 6G + 8.8X10-4
Kiton Red 620 2.1X10-4 methanol 575-614 288-307

Kiton Red 620 1.7X10-3 methanol 588-639 294-319

Because of the limitation in the data transfer rate (1.4 kHz) to

the computer (89), the computer was only triggered at one tenth the

rate as compared to the boxcar average and the laser. A decade

counter (74LS90, Texas Instruments, Dallas, TX) was utilized to divide

the frequency of the "Busy Output" (synchronization pulse) from the

boxcar average by ten. This lower frequency (600 Hz) was then used

to trigger the computer. All data points were collected at the 6 kHz

rate and averaged by the boxcar average. The averaged output from

the boxcar average was sent to the computer at a rate of 600 Hz. In

this manner, no data were lost.

The nitrogen-dye laser system was the same as that described in

Chapter 2.


The elements In, Li, and Na were each determined with both laser

systems. The In was measured using the H2-Ar atmosphere (but with the

computer system as the detection electronics). The Li and Na were

determined in the Ar atmosphere. All three elements were determined

in the plain pyrolytic graphite cup. The CVL system was tuned to the

particular elemental line by moving the end of the fiber optic over to

the flame system used for tuning the nitrogen-dye laser system. A

blade connected to a rotary solenoid motor was used to block the laser

beam when loading samples into the graphite cup.

In order to excite Li at 671 nm, a dye had to be used with its

maximum power near that wavelength. To help choose a dye, the dyes

used for the Ar+ laser (514 nm) and also the frequency-doubled output

of the ND:YAG laser (532 nm) were investigated. Oxazine 720 appeared

promising; therefore, a 1X10- M solution was prepared and found to

lase in the required wavelength region. The power at 671 nm was

monitored while the dye was diluted; maximum power was achieved at a

concentration of 6.6X10-4 M.

Results and Discussion

Figure 5-1 shows the lasing curve for the oxazine 720. The

maximum output power was 0.76 W, which was approximately half of the

power measured for rhodamine 60 at its peak. This power ratio was

similar to power ratios obtained from other pump lasers (90). These

results indicated that the oxazine 720 dye worked very well.

Problems Associated With the Fiber Optic

The main reason for having to use the fiber optic was because of

space restrictions. Before using the fiber optic, the beam was

directed down the laboratory using mirrors. Unfortunately, five

mirrors had to be used to direct the beam around and above other

equipment in the laboratory; also, the beam diverged to a diameter of

3 in by the time it arrived at the furnace. The beam divergence could

have probably been dealt with using collimating lenses, but the main

problem was the five mirrors. Not only were there light losses at

each mirror, but lining up the laser each day became tedious and time

consuming. There was also the safety problem of sending the laser

beam down the length of the laboratory, even though safety shielding

was put in place. Because of all of these reasons, the fiber optic

approach was chosen as the only alternative.

Figure 5-1. Lasing Curve for Oxazine 720 (6.6X10-4 M).

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs