Laser excited atomic flourescence spectrometry as a powerful tool for analytical applications and spectroscopic studies


Material Information

Laser excited atomic flourescence spectrometry as a powerful tool for analytical applications and spectroscopic studies
Physical Description:
vii, 175 leaves : ill. ; 29 cm.
Gornushkin, Igor, 1958-
Publication Date:


Subjects / Keywords:
Trace elements -- Analysis   ( lcsh )
Fluorescence spectroscopy   ( lcsh )
Laser spectroscopy   ( lcsh )
Chemistry thesis, Ph. D   ( lcsh )
Dissertations, Academic -- Chemistry -- UF   ( lcsh )
bibliography   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph. D.)--University of Florida, 1997.
Includes bibliographical references (leaves 167-174).
Statement of Responsibility:
By Igor B. Gornushkin.
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 - 028020912
oclc - 38268360
System ID:

This item is only available as the following downloads:

Full Text








I would like to express my deep respect and gratitude to Jim Winefordner who

invited me to join his group and whose guidance, support, and friendly attitude accompanied

me throughout this work.

I want to thank Ben Smith whose office door is always open for everybody who is

looking for advice and whose truly encyclopedic knowledge helps to promote all the

projects going in the laboratory.

I would like to acknowledge Michael Bolshov for his friendship, help, and support

since my early days in analytical spectroscopy. I further extend my gratitude to Nicolo

Omenetto for many good ideas and much good advice which greatly supported nearly all the

projects I have done.

I want to address my great appreciation to all members of JDW group, past and

present, who became my real friends and with whom I was glad to share work and leisure.

My special thanks go to Martin Clara and Jason Kim who helped me in promoting laser

plasma projects.

I heartily thank my wife Lena; I can barely express how much her love, support and

confidence mean for me. I also thank my mother and sister whose love and concern I always



AKNOWLEDGEMENTS................................................................................ ....ii

ABSTRACT................... .................................................................................. vi


ANALYTICAL CHEMISTRY: BASIC INFORMATION................................

Fundam entals....................... ....................................
Instrumentation............................... .......... .........................7
Analytical Performance of LEAFS.................................................. 1
Analysis of Real Objects.............................. ........................ 14


Introduction................................... ................................................... 16
Furnace Design........................ ........... ........................20
Experimental Set-Up............................. .... ........................22


O bjectives........................................... .............................................. 26
Theoretical Model........................ ...... ........................28
Experimental Procedure.......................... ... ........................39
Results................. ........................................................................ 40
C opper...................................................................................... 44
Silver............................ ................................ 49

C onclusions......... ........................................................... ........................ 61


Introduction.............................. ... .. ... .... ........ ............. 64
Experim ent...................................... ...............................................65
R esults................................................ ................................... ............ 69
Spatial distribution........................................ ...... .............. 69
Saturation ofthe silver transition.................................................. 75
Analysis of Sea Water............................. ...... ......................78
Analysis of Solid Reference Materials................................................88
Conclusions...................... ......................................................... 92


Introduction.......................... ........ .............................93
Evaluation of Limits of Detection........................ ................... ..............98

INERT ATMOSPHERE........................ .......... ........................105

O bjectives........................ .. ....................................................105
Experim ent............................................................................................. 106
Determination of Lead in Copper, Brass, Steel, and Zinc.....................12
Lifetime of the Metastable 6p2'D Level of Lead................................... 120
Theoretical................................ .......................120
Conclusions...................... ....................... ......................... 125


O bjectives......................................... ................................................. 127
E xperim ent................................. ........................ ......................128
Determination of Cobalt in Soil, Steel, and Graphite.................................131
Conclusions........................ .... ........ ..................... 141

PRODUCED LEAD AND TIN PLASMAS................................. ............ 142

Introduction................... .............................................................142
Experim ent....................................................................................................145
Analysis of Resonance Shadow Images................................................... 147
Shock Wave Propagation................................................ .......................156
Conclusions................... ........................................................ 161

9 CONCLUSIONS AND FUTURE WORK....................................................162


A ACRONYMS USED IN TEXT.......................... .......................165


R EFER EN C E S.............................................................. ..................................................167

BIOGRAPHICAL SKETCH...........................................................................................175

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



Igor B. Gomushkin

August, 1997

Chairman: Prof. James D. Winefordner
Major Department: Chemistry

Laser-excited atomic fluorescence spectrometry (LEAFS) with a novel diffusive tube

electrothermal atomizer (ETA) has been used for the study of atomization and diffusion

processes and for the direct trace analysis of complex matrices.

A novel ETA was a graphite tube sealed by two graphite electrodes. A sample was

introduced into the tube and the furnace assembly was heated. The vaporized sample

diffused through the hot graphite and the atomic fraction of the vapor was excited by a

tunable dye laser above the tube.

Temporal behavior of atomic fluorescence of Cu, Ag, and Ni atoms, diffused through

the furnace tube, was studied at different temperatures; the values for activation energies

and diffusion coefficients were derived on the basis of the diffusion/vaporization kinetic


The femtogram/nanogram concentrations of silver were determined in coastal

Atlantic water and soil samples. Use of the new ETA resulted in significant reduction of

matrix interference, ultra-low limits of detection, good accuracy and precision.

LEAFS coupled with laser ablation (LA) was studied in terms of its analytical and

spectroscopic potential. Low concentrations of lead (0.15 ppm 750 ppm) in metallic

matrices (copper, brass, steel, and zinc) were measured in a low pressure argon atmosphere.

No matrix effect was observed, providing a universal calibration curve for all samples. A

limit of detection of 22 ppb (0.5 fg) was achieved. Also, the lifetime of the metastable

6p21D level of lead was measured and found to be in good agreement with the literature data.

A simple open-air LA-LEAFS system was used for the determination of cobalt in

solid matrices (graphite, soil, and steel). The fluorescence of cobalt was excited from a level

which was already populated in the ablation plasma and was monitored at the Stokes-shifted

wavelength. Detection limits in the ppb to ppm range and linearity over four orders of

magnitude were achieved.

The resonance shadowgraph technique has been developed for time-resolved imaging

of laser-produced plasmas. The shadowgraphs were obtained by igniting the plasma on the

lead or tin surface and by illuminating the plasma by a laser tuned in resonance with a strong

atomic transition. UV-photodecomposition of lead and tin clusters was visualized. The

evolution of the plasmas was studied at different pressures of argon. A shock wave

produced by the laser ablation was monitored and its speed was measured.




The method of atomic fluorescence spectrometry (AFS), which is based on selective

absorption and following emission of light by atoms and molecules, has been known since

the 1930s. Unfortunately, it was not widely used in an analytical practice until the 1970s due

to the absence of powerful light sources and, consequently, due to a very low sensitivity as

compared to other routinely used methods of spectral analysis (atomic absorption

spectrometry (AAS), atomic emission in flames, arc and spark discharges, inductively

coupled plasma (ICP), etc.). However, when powerful lasers, especially tunable dye lasers,

became available, the method of laser excited atomic fluorescence spectrometry (LEAFS)

was found to be one of the most sensitive and selective methods of atomic spectral analysis,

allowing, in some cases, the determination of single atoms or molecules in the analytical

volume [1,2].

The physical principle of LEAFS is based on the behavior of the atomic system in

a strong laser field [3-7]. Two types of laser-induced fluorescence are usually used in

experiments: resonance fluorescence, when the fluorescence light is monitored at the same

wavelength as that of the laser excitation (two-level atomic scheme, Figure la), and the

shifted or nonresonant fluorescence, when the fluorescence is collected at longer wavelength


relative to the excitation wavelength (three-level atomic scheme, Figure Ib). The latter is

more favorable, allowing the elimination of strong laser scattering by using a simple spectral

filter (a monochromator or an interference filter with a narrow bandpass).

As an example, consider a more general three-level atomic fluorescence. The rate

equations which describe a time behavior of the levels populations N, (i=1,2,3) are given by


dN1 N
-- =-NpvB12N,2(A2, +PB21+R2)+-3
dt 31
S-N1pB-N2 2(Ap21 21+R21+A23R23) (1),

dN3 N.
d =N2(A23+R23) 3
dt T'31

where p, is the spectral energy density (J cm" Hz'), A21 and A23 are the Einstein coefficients

for spontaneous emission (s-i), R21 and R23 are the probabilities of non-radiative (quenching)

transitions (s'-), B12 and B2, are the Einstein coefficients for stimulated emission (cm3 J-1 s-

Hz), and T3, is the lifetime of the level 3 (s).

The solution of differential equations (1) is usually expressed in terms of the

population of the upper excited state (N2) from which fluorescence photons are emitted.

There are several factors which affect the final form of the solution and the magnitude of the

fluorescence signal measured in the experiment; these factors include T3,, the lifetime of the

1 I Ji 1 I iAA *L

a) b)

Figure 1. Diagrams of atomic levels commonly used in laser-induced atomic fluorescence
experiments; a) two-level resonance fluorescence; b) three-level shifted fluorescence.
A2, and A23 are the Einstein coefficients for spontaneous emission from the level 2 to the
levels 1 and 3, respectively; B12 and B2,, are the Einstein coefficients for stimulated emission
between the levels 1 and 2 caused by laser radiation with the spectral energy density p,; R21
and R23 are the probabilities of quenching transitions from the level 2 to the levels 1 and 3,
respectively; and 3,, is the lifetime of the level 3 decaying to the level 1.

intermediate level 3, and T,, the time constant of the electronic data acquisition system. It

is interesting to analyze the solution for the case of the metastable level 3 (t31-~) and for the

case of the fast decaying level 3 (T31-0). If the time constant To is smaller than the width of

the laser pulse, T<,r,, then the fluorescence signal, expressed in the number of fluorescence

photons, is proportional to the population of level 2: NphN2"" (amplitude registration). If

the time constant r, is greater than the duration of the laser pulse, then the fluorescence signal

is proportional to the integral Nph-o'N2(t)dt (integral registration). Let us consider these

possibilities sequentially.

Large T (metastable level 3). small T. (amplitude registration)

The approximate solution of(1) in its extreme point, i.e. dN2(t)/dt = 0, is given by

N NmaxN PBr12
NphN2 N z-- (2),

where No is the total number of atoms; A21'A21+R21; A23A23+R23; and e=l+g,/g2.

At some critical laser spectral energy density, saturation of the atomic transition can

occur, where the fluorescence from the upper level does not depend any more upon the laser

energy density but only upon fundamental constants: statistical weights and Einstein

coefficients. It is seen from (2), that N2" reaches the saturation value at

A21A23 A21 A2
p>>) B12 (3),
eB12 B12+B21

where the relationship Bl2g1=B21g2 between Einstein coefficients for stimulated emission has

been used. Hence, the saturation value for N2 becomes

satr 2
N2 aNO -- (4),

that yields N2"-1/2No when g,=g2.

Large ,1. large (integral registration)

The number of fluorescence photons is now expressed by

Nphf A23N2(t)dt (5).

The solution of (5) is strongly dependent upon laser parameters (a spectral energy density,

a pulse width), as well as upon parameters of an atomic system (transition probabilities,

quenching constants). In first approximation, for a typical dye laser system, this solution is

given by

N vB2A23Z
^ph'o- (6).
A21+A23 Pv12e

At moderate laser intensities, the fluorescence increases linearly with the laser spectral

energy density p,

Nph-No23T I V (7)

approaching the saturation value at high p,

S-Nh o A(8).

This solution has a limitation A23T1/e 1 which follows from the numerical evaluation of the

integral (5) (not shown here). Taking into consideration this condition, equation (8)

acquires a clear physical sense: when level 3 is metastable, the atoms will undergo

successive transitions 1-2-3 only once within the duration Tr of the laser pulse. Therefore,

in the presence of quenching (R230), only a fraction of the total number of atoms No will

contribute to the fluorescence at frequency v23.

Fast decay of the level 3 (TZ0)

In this case, the three-level atomic scheme converges to the two-level scheme,

providing N3(t)=0. The solution of the system (1) may be exactly the same as that expressed

by Eqns 6 through 8 if the parameter A23-/e in (8) (the saturation mode) is much greater than

1, so that Nph>>N. This means that in a strong laser field, atoms undergo transitions between

levels 1 and 2 without being retained in level 3 and emit fluorescence photons many times

during the laser excitation pulse.

Finite lifetime 3r,

Again, in the first approximation, the solution of(l) is given by equation (6) and the

limiting cases (7) and (8) are possible depending on the relationship between the parameter

pB,12 and the values of A21 and A23.

As one can see, in all cases considered, the situation can be realized in practice when

the number of fluorescence photons emitted is nearly the same order of magnitude as or

greater than the total number of atoms in the analytical volume. In terms of sensitivity, this

provides a significant advantage of the method of laser-induced atomic fluorescence over all

other methods where thermal sources of light are used (Figure 2). In the case of a thermal

light source, the fluorescence signal, proportional to the population of a particular excited

level, n,, is determined by an exponential Boltzmann factor which reflects the distribution

of atoms over all thermally excited energy levels. Typically, this factor is much less than

unity providing that only a small fraction of light source atoms can contribute to the

fluorescence signal. In contrast, for laser-induced fluorescence, the fluorescence signal is

proportional to the population of only one selectively excited level and is determined by the

ratio of statistical weights of levels which are involved into the radiative transition. This ratio

can be close to unity providing that nearly all atoms present in the zone of interaction with

the laser contribute to the fluorescence.


The traditional laser atomic fluorescence spectrometer consists of three main

components: an excitation source (a laser), an atomic reservoir (a graphite furnace, flame,

ICP-plasma, etc.), and a system of data acquisition and processing.

The most popular source of the fluorescence excitation is a tunable dye laser which

is usually pumped by an Nd:YAG laser [7], a nitrogen laser [8], or an excimer laser [9].

Recently, tunable diode lasers have also become widely used due to their low price,

compactness and convenience in operating [10].

The system of atomization of samples is probably the weakest part of any laser

atomic fluorescence spectrometer. It is impossible to design an ideal atom reservoir which

could provide an efficiency of atomization close to unity and a freedom from spectral and

chemical interference. Therefore, a large variety of different atomizers are used in LEAFS

depending on purposes pursued by researchers. Very popular atomizers are graphite furnaces

or electrothermal atomizers (ETA) of various shape and size [7-12], especially those

designed for AAS analysis. Some typical constructions of graphite furnaces used in LEAFS-

ETA experiments are shown in Figure 3. Also, different types of flames (air-acetylene, air-

hydrogen, etc.) [13-15], glow discharges in hollow cathode lamps [16-17], ICP-plasmas [18-

19], and laser-induced plasmas [20-21] are used as atomic reservoirs.

A typical system for the detection of fluorescence consists of a collection optics, a

spectral filter (usually, a monochromator or an interference filter) which transmits light only

at the fluorescence wavelength, a photomultiplier tube (PMT), a gated integrator which

opens the acquisition system for a short time (approximately equal to the width of the laser

pulse), and a computer supplied with a software for a statistical data routine. An amplitude

or an area of the fluorescence pulse, induced by a pulsed laser in a vapor of free atoms, is

measured within the integrator time-gate. The sum of fluorescence signals over the time of

atomization is usually taken as an analytical signal.

1. Thermal (flame, ICP, ...)

n = no g/Z exp[-E,/kT]

Boltzman factor, exp [-E/kT]

X,nm 2000K 4000K 6000K
200 10-" 10-' 10-'
500 10- 10' 10-'
700 10-5 10- 10'

2. Laser Induced Fluorescence
a) Two-level scheme b) Three-level scheme

n-, =g/(g,+g,)n


g,=g2 g--9g2

0.5 0.1

Figure 2. Comparison of a thermal light source with laser-induced fluorescence

Sample vapors

Laser beam /

Graphite furnace


Graphite furnace

Pierced mirror

Figure 3. Graphite atomizers popular in LEAFS: a) an open type atomizer--a graphite
cup; b) a closed type atomizer--a graphite tube


Analytical Performance of LEAFS

Traditionally, the potential of a new analytical method in comparison with other

methods is determined by using calibration plots constructed on the basis of aqueous

standards and by measuring limits of detection (LOD) of elements by the linear extrapolation

of these plots to the level of triple standard deviation of the background signal.

LEAFS-ETA analysis of aqueous samples was carried out in references 3 and 7-12.

The authors used a similar sampling routine: a small sample aliquot (10 lL-50 tL) was

dosed into a graphite furnace, then dried and atomized at temperatures between 1200 C and

30000C depending on the thermal-chemical properties of the element to be determined. An

atmosphere of inert gas at normal or reduced pressure was used, or the sample was atomized

in a vacuum, as in Bolshov et al. [7]. Dougherty and Prely [9] used LEAFS-ETA with a

Zeeman background correction to obtain lower LODs and to improve an accuracy of the

analysis relative to LEAFS-ETA without background correction. The atomizer (a graphite

cup) was placed into a strong magnetic field modulated at a frequency of 60 Hz that caused

atomic energy levels to be periodically split into Zeeman sub-levels. A pulsed dye laser was

tuned in resonance with an atomic transition and operated at a repetition rate of 80 Hz. The

fluorescence was excited by every second laser pulse at times when the field was off. When

the field was on, no excitation occurred, and the background signal was measured. Cobalt

was determined in aqueous standard solutions at the level of 0.3 pg (LOD) with a relative

standard deviation of 0.13. Without the background correction, the LOD and the RSD were

0.7 pg and 0.3, respectively. Similar results for the determination of cobalt in aqueous

samples are reported in the literature [7, 11-12]; LODs in the range of 0.2 pg 0.3 pg were

achieved and RSDs ranged between 0.2 and 0.6.

Winefordner and Goforth [8] studied the analytical performance of LEAFS-ETA by

using graphite furnaces of different shapes and size. Four elements (Al, Cu, Mo, and V)

were determined in aqueous solutions. The use of a closed type atomizer, a graphite tube,

yielded LODs of 100 pg, 8 pg, 100 pg, and 2-10 pg for Al, Cu, Mo, and V, respectively.

The use of an open atomizer, a graphite cup, yielded LODs of 500 pg and 2 pg for Al and Cu,

whereas Mo and V were not detected at all. The conclusion was made that graphite tube

furnaces are advantageous over graphite cup atomizers, especially, for analysis of refractory

materials. The same authors also studied how different furnace coatings (pyrolytic coating,

a tantalum and a tantalum carbide foil alignment) and different inert atmospheres (argon,

argon-hydrogen) affect the results of the LEAFS-ETA analysis [9]. The best LODs for six

elements (Cu, Mn, Pt, Sn, In, and Li) in the range of 0.3 pg (In) 400 pg (Li) were obtained

by using a pyrolytically coated graphite tube furnace and an argon buffer gas at atmospheric


A high repetition rate (-10 kHz) copper vapor laser in a combination with ETA [12]

or with a hollow cathode glow discharge [16] was used as a source for the fluorescence

excitation. Such high repetition rate laser systems allow a significant increase in the probing

efficiency as compared to commonly used low repetition rate (-10 Hz 100 Hz) lasers. The

limits of detection of 0.3 pg, 3 pg, 1 pg, and 0.03 pg were obtained [12] for Co, Ni, Fe, and

Pb, respectively, and LODs as low as 0.015 pg for Pb and 2 pg for Ir were reported in

Bolshov et al. [22].

The determination of 14 elements (Al, B, Ba, Ga, Mo, Pb, Si, Sn, Ti, TI, V, Y, Zr,

and U) was carried out by Human with coworkers [18] with the use of a ICP-LEAFS

combination. The fluorescence could be excited either from atomic or from ionic energy

levels that substantially enlarged a choice in finding the most convenient and free from

spectral and chemical interference transitions. The detection limits obtained were in the

range of 0.4 ng/mL 20 ng/mL, lower than average LODs achievable by the ICP-emission

method. Theoretical aspects of the ICP-LEAFS combination are given in detail in Omenetto

et al. [19]. Similar LODs for 19 elements (Ag, Al, Ba, Bi, Ca, Cd, Co, Cr, Cu, Fe, Ga, In,

Li, Mg, Mn, Mo, Na, Ni, Pb, Sr, Ti, TI, and U) in the range of 0.08 ng/mL 100 ng/mL were

obtained by Weeks et al. [14] by using laser-induced fluorescence in flames.

LEAFS in an air-hydrogen flame with the use of a diode laser was demonstrated by

Barber and colleagues [15]. The LOD of 0.2 ng/mL was obtained for Rb. A special cell

containing rubidium vapor was used for the correction of the diode laser wavelength. An

interesting application of LEAFS in flame was also reported by Simeosson et al. [23] who

used photodissociation lasers for the determination of In and TI in aqueous solutions.

Photodissociation lasers are simple low pressure cells containing volatile halides of metals

and pumped by an excimer laser. They are line sources with fixed wavelengths which are

inherently tuned to atomic transitions of the same elements as those confined in active laser

media. LODs obtained for TI and In were 10 ng/mL and 80 ng/mL, respectively. The

authors concluded that a fluorescence spectrometer on the basis of simple and cheap

photodissociation lasers has a good potential for the sensitive determination of elements

which form volatile halides.

Analysis of Real Objects

The utility of LEAFS as an analytical method has been demonstrated in numerous

applications of LEAFS to the analysis of real objects: environmental, industrial, or


The direct determination of antimony in environmental and biological samples by

LEAFS-ETA was carried out by Enger et al. [24]. An intensified charge coupled device

(ICCD) was used for continuous monitoring of a large spectral region. This allowed for the

control and for the correction of various interfering background signals which appeared

when samples of complex composition were analyzed. Measurements of the Sb content in

drinking water, river water, marine sediments, human blood serum, and whole blood were

performed in the concentration range from several tg/mL to low pg/mL level with the

accuracy less than 10 % RSD. The LEAFS-ETA technique was also successfully used for

the direct determination of lead [25] and silver [26] in seawater (-1 ng/L-10 ng/L), iridium

in industrial solutions (60 pg/mL 1900 pg/mL) [22], lead in blood (-50 ng/mL 500

ng/mL) [27], metals (Cu, Fe, Mn, Pb, Sn, TI) in air (-1 fg/m3- 1 ng/m3) [28], etc.

Recently, the combination of LEAFS with laser ablation (LA) has become a popular

method for the analysis of solids. In this method, very little or no sample preparation is

needed that significantly reduces the chance for contamination due to handling. On the other

hand, an extremely sensitive LEAFS method allows the determination of elements in laser

plasmas at the part-per -billion (ppb) level of concentration. Quentmeier and colleagues [29]

carried out the determination of four elements (Si, Mn, Mg, and Cr) in steel, copper, and

aluminum. The plasma was induced by a Nd:YAG laser on the surface of solid targets, and

the fluorescence of two elements was excited simultaneously by two dye lasers. The internal

standardization was proposed to account for matrix interference and to obtain a universal

calibration plot independent of matrix. Concentrations at a low part-per-million (ppm) range

were detected. The LEAFS-LA method was also recently used for the determination of

cobalt (low ppm level) in soil, steel, and graphite [30] and for the determination of lead in

metallic matrices in the concentration range of 20 ppb 700 ppm [31].



One of the major problems, inherent in all analytical techniques using

ETAs, is how to account for the various interference which can exist when analyzing a

particular matrix. These interference are primarily due to an imperfection of the atomic

cell design and usually include i) spectral interferences--black body radiation from the hot

furnace, laser scattering from sample soot particles ejected into an analytical zone,

molecular fluorescence from matrix concomitants; and ii) chemical interferences-solid

phase reactions within the sample, sample or sample vapor interaction with the atomizer

material, interaction between the analyte vapor and matrix gases within the furnace

atmosphere (generation, decomposition, and transformation of various compounds) [32].

Many efforts have been made to improve the construction of the graphite furnace

ETA and to optimize it for the analysis of the largest possible variety of real materials.

In most early research on the LEAFS-ETA technique, open furnaces, such a graphite cup

or a graphite rod, were used [10,33-34]. Since about 1985, commercially available

graphite tube ETAs have been reported to be a much better choice for LEAFS-ETA than

open graphite atomizers, because a graphite tube atomizer suffers less from diffusional

losses and vapor-phase interference. In addition, better detection limits for the


nonvolatile elements have been obtained [8-9,35-36]. Such graphite tube ETAs are widely

used in atomic absorption spectrometry (AAS); past studies have indicated that there is no

difference between the techniques of absorption and fluorescence in terms of the

optimization of analytical methods because of the similarity of physical-chemical processes

when the same furnace is used for both [37]. This leads to the conclusion that such

electrothermal atomizers are equally well suited to both AAS and LEAFS.

During the last two decades, new type graphite furnaces, the furnaces with graphite

filters, have been proposed for the analysis of liquids and solids by AAS [38-41]. The

design of the furnaces was based on the separation of the atomization and analytical zones

by means of a porous graphite filter which became transparent to vapors of some metals

at high temperature. For example, the atomizer "capsule-in-flame" [38,39] has been

applied for the analysis of powdered samples by AAS. The atomizer was a graphite cylinder

(capsule) squeezed between two cylindrical graphite washers. The washers had holes to

allow passage of a light beam along the capsule close to its surface. The air-acetylene burner

was placed beneath the furnace. Powdered sample was introduced into the capsule cavity.

The capsule was heated by an electric current and the sample vapor was released into the

flame which was used to provide sufficiently high temperature in the analytical zone above

the capsule. Such a furnace design allowed the elimination of soot particles which could be

ejected into the analytical zone, the elimination of uncontrollable sample losses (that usually

occur in open-cavity atomizers under pulse vaporization conditions) and the use of larger

amounts of sample with respect to commercial ETA. It was also found that "capsule-in-

flame" atomizer provides a relative freedom from chemical and spectral interference.

Schmidt and Falk [40] carried out the direct AAS-determination of Ag, Cu, and Ni

in biological and vegetation samples using a specially designed "ring chamber tube" with

a sample chamber separated from the absorption volume. This "ring chamber tube" could

be directly inserted into a commercial atomizer. These authors demonstrated the principal

advantages of vapor filtration, such as suppression of background and interference, and

determined several elements in solid materials and found good agreement with certified


Several other types of closed-cavity atomizers for AAS have recently been reviewed

by Katskov and coauthors [41] (see Figure 4). Also, a new design of a furnace with a

graphite filter has been proposed. The furnace was a pyrolytic graphite tube with a spool-

shaped insert made from porous graphite. The sample was injected into the ring cavity

between the tube and the insert. After the pulse vaporization of the sample, the atomic vapor

along with gaseous matrix components diffused through the porous insert into the analytical

zone in the center of the graphite tube. In the course of the determination of Cd, Pb and Bi

in the presence of excess matrices (NaC1 or CuCI2), Katskov et al. [41] discovered that

molecular vapor entering the analytical zone was delayed relative to the atomic vapor.

Further investigation [42] showed that this delay could be attributed to the formation of

molecular intercalation compounds of different stoichiometry which were implanted into the

graphite structure between crystalline layers. Enthalpies of formation were obtained on the

basis of the kinetic analysis of different exponential portions of the signal which contributed

to the total shape of the signal at different temperatures.





1 cm

Figure 4. Atomizers with the sample vapor filtration for AAS; a) furnace with a ring cavity;
b) capsule-in-flame; c) capsule-in-furnace; d) ring chamber tube. Shaded areas represent
sampling volumes; arrows represent light source beams.

Therefore, one can conclude that the use of a diffusive media in an ETA, made of

a porous graphite (or, perhaps, other porous material), provides a high-temperature

separation of matrix components and leads to minimization of spectral and chemical


Furnace Design

The design of the furnace is shown in Figure 5. The graphite tube (1), which served

as the porous filter, was made from a carbon electrode of density 1.65 g/cm' (Type F, ultra

pure, Ultra Carbon Corporation, Bay City, Michigan, USA) used for spectral emission

analysis. The length and the external diameter of the tube were 22 mm and 6 mm,

respectively. The diameter of the inner channel was 3.5 mm in the central part of the tube

and was enlarged to 4.6 mm toward the tube ends at a distance of 3 mm from each end.

The graphite tube was held between two spring-loaded graphite electrodes (3,4) of

4.5 mm in diameter with rounded tips. The electrodes firmly held the tube providing a good

seal of the inner tube channel and good electric contact between the tube and the electrodes.

One of the electrodes (3) had an insert in the form of a smaller diameter (3 mm) graphite rod

(2) which had a hollow for a sample. This electrode could slide back and forth in a stainless

steel mount (5). To load a sample, the electrode was pulled out to the position where the rod

hollow was between the mount and the tube. The maximum acceptable volume for a single

dosing was about 50 tL.

The furnace assembly was constructed from a commercial carbon rod atomizer

CRA-90 (Varian Techtron, Springvale, Australia). The atomizer, initially designed for AAS,

32 1

Laser beam

W -- --1--t-i--t -- -.

SPower supply g

Figure 5. Furnace with graphite filter: 1, porous graphite tube; 2, graphite rod with a
sampling boat; 3-4, electrodes; 5, steel mount; 6, insulator; 7, shim chimney for the laminar
argon flow supply;8, stainless steel platform.

was modified to suit the purposes of the LEAFS experiment. Two rectangular windows

were cut in the stainless steel electrode mounts to allow passage of a laser beam along the

graphite tube furnace close to its surface. The CRA-90 control unit was also modified to

provide an atomization time up to 10 min, instead of a maximum of 5 s initially specified.

The furnace assembly was inclosed in a gas-tight chamber filled with argon at one

atmosphere pressure during the atomic fluorescence measurements.

Experimental Set-Up

The LEAFS-ETA instrumentation with a diffusive tube furnace is shown in Figure

6. The components of the experimental set-up are listed in Table 1. The 308 nm line of

a xenon-chloride excimer laser was used to pump a Cresyl Violet 670 dye laser at a

repetition rate of 25 Hz. Stray light from the excimer laser incident on a photodiode was

used to trigger a boxcar integrator. The dye laser output was frequency doubled by a

KDP crystal to produce a UV excitation wavelength which was directed through a quartz

window into an atomizer chamber and was used to excite atoms generated above the

diffusive tube furnace. Tuning the laser to the wavelength of the atomic transition of

an element of interest was carried out with a second spectrometric system consisting of an

air-acetylene flame in which the element stock solution was nebulized, a monochromator,

a photomultiplier tube (PMT), and an oscilloscope with a 300 MHz bandwidth. The

fluorescence signal from atoms in the flame was visually monitored on the oscilloscope


The furnace (Fig.5) was enclosed within an atomizer chamber which has been

designed to be compatible with an argon atmosphere at controlled pressure. The

Excimer laser

Dye laser

Figure 6. Schematic of the experimental set-up

Table 1. Components of the experimental system

Component/Model Manufacturer

Excimer laser/LPX100

Dye laser/EPD-330

Frequency doubler/KDP



Boxcar integrator/SR265



Infrared thermometer/OS3709



Lambda Physic, Gottingen, Germany

Lumonics, Ottawa, Ontario, Canada

Interactive Radiation, Northvale, NJ, USA

CVI Laser, Albuquerque, NM, USA

Hamamatsu Photonics K.K., Toyooka Vill, Japan

Stanford Research Systems, Palo Alto, CA, USA

Stanford Research Systems, Palo Alto, CA, USA

Varian Techtron, Springvale, Australia

OMEGA Engineering, Stamford, CT, USA

CLUB American Technologies, INC., Fremont,

Metagraphics Software Corp., Copyright 1988 by
Stanford Research Systems, Palo Alto, Ca, USA


chamber contained two parts: the steel platform, on which the furnace assembly was

mounted and a cubic aluminum top with quartz windows on each face. A laminar argon

flow was supplied from beneath the furnace. The distance between the laser beam and

the furnace surface was 4 mm. The fluorescence emitted by excited atoms was detected

at 900 to the direction of the laser beam. To collect the fluorescence from the entire

excitation pathlength (-2 cm) and, hence, to increase the collection efficiency, the

monochromator was rotated on its side (90o) so that the entrance slit was parallel to the

laser beam. A 0.24 m focal length monochromator with an effective aperture of f/3.9,

blazed at 250 nm, was used. Two biconvex lenses (51 mm diameter, 76 mm focal length)

projected a 1:1 image of the analytical volume on the monochromator slit. The optical

system was aligned by using emission from an electrically heated tungsten filament,

positioned on the excitation axis. The thin tantalum foil diaphragm was placed in front of

the furnace at a distance of 10 mm to block thermal blackbody radiation emitted by the hot

furnace in the horizontal direction. The blackbody radiation in the vertical direction was

detected by a photodiode, calibrated by an infrared thermometer, to monitor the

temperature of the furnace surface.

The detection system included a PMT, a voltage amplifier with a gain of ten, and

a boxcar integrator with a gate width of 20 ns. To keep the PMT in the linear range, the

fluorescence signal was attenuated by a set of colored glass filters which were calibrated

at the excitation and fluorescence wavelengths. The data were acquired and manipulated

with a personal computer which recorded the fluorescence signal as a function of

atomization time.




In spite of extremely high sensitivity of LEAFS, it has one serious limitation.

LEAFS is essentially a single-element technique. This is due to the nature of the excitation

source, a laser: it is impossible to obtain highly intense light in a large spectral range but

only within a narrow wavelength window (-0.01 nm) selected by a laser cavity from a

spectral band (-20 nm) of a laser dye. The high cost of good and reliable laser systems (dye

laser pumped by an excimer or Nd:YAG lasers) makes their application unpractical for a

determination of only one element in time. Therefore, up to this date, LEAFS instruments

are not commercialized and are not used in routine analysis.

However, there are some ways to overcome this drawback. Tunable dye lasers which

are usually used in LEAFS can be tuned to different wavelengths within the generation

range of the dye. If several elements have transitions within the same spectral range then

they can be sequentially excited by the laser. Unfortunately, it takes some time to tune the

laser to the chosen wavelength (hardware limitation). This time ranges from seconds for

lasers with diffraction gratings to milliseconds for lasers with opto-acoustic wavelength

selectors. Nevertheless, if a construction and variable parameters (temperature, ramp rate)

of the atomizer are chosen in such a way to allow the sequential release of atoms into the

analytical zone (a zone where free atoms interact with laser light), similar to

chromatography, then the time between the appearance of two elements in the analytical

zone could be made long enough to provide tuning the laser from one wavelength to

another. This would increase the performance of LEAFS-ETA and, instead of the

determination of only one element in one atomization cycle, one could sequentially

determine several elements in the same sample and the same atomization run. To do this,

knowing the diffusion coefficients and their temperature dependence is essential.

Therefore, evaluation of diffusion coefficients of elements in a graphite atomizer was

a part of this work. The study presented in this chapter attempts to provide some new data

on processes involved in the diffusive graphite tube ETA by using a kinetic approach to

examine the diffusion of three elements (Ag, Cu, and Ni) through hot graphite. The elements

copper, silver and nickel were chosen for the diffusion study for the following reasons. First,

copper and silver have low melting points (1083 oC and 962 oC, respectively) and the

formation of atoms in the analytical zone can be easily measured over a wide temperature

range. Second, copper and silver do not form stable carbides. Third, the atomization of

these three elements in graphite ETAs has already been extensively studied, and the

abundance of experimental data which have been accumulated in the literature is useful for

comparison with our results. Finally, all three elements have strong atomic transitions within

the spectral range of the dye laser available for our experiment.

Theoretical Model

The problem of diffusion, complicated by vaporization and chemical interaction, may

be solved by considering the generalized diffusion equation [43,44]

dn a2n
D- kn (9)
dt ax2

where n is the number density of atoms in the substrate (cm-3), t (s) the time, x (cm) the

coordinate along the diffusion path, D (cm2/s) the diffusion coefficient and k (s-)the first-

order reaction rate constant. Unfortunately, this equation can be solved analytically only for

a very few special cases. Therefore, the use of the semi-empirical models, which will be

described below, is often the only way to solve the problem.

When solid state diffusion, vaporization, or chemical reactions leading to the

formation of free atoms in the analytical zone are considered as successive phenomena, the

kinetics of the process are determined by the stage with the largest rate constant. The rate

equation for the change in amount of analyte atoms, N, in the analytical zone can be written

in the form of a differential equation [45]:

dN(t) dN dN (10)
dt dt dt

where the two terms on the righthand side express the rate of supply and the rate of removal

of the atoms. The number of atoms in the analytical zone can also be expressed in terms of

an integral equation [45-46] which is more convenient for practical use:

N(t) = fS(t')R(t-t)dt' (11)

where S(t) = (dN/dt),, is the supply rate of the atoms, and R(t) is the normalized response

of the system (R, = 1) to the supply of an infinitely rapid pulse of atoms. If the removal

function R(t) is very rapid in comparison with the supply function, then the convolution

integral in (11) can be transformed into the form:

N(t) = S(t)fR()dt S(t)R (12)

where TR is the equivalent time constant of the removal function. Thus, the temporal

behavior ofN(t) (or the shape of the absorption or the fluorescence signal) reflects the time

dependence of the supply function. In many publications on AAS [45-50], it has been shown

that the supply function can be adequately described by first-order kinetics. Such kinetics

can be attributed to specific processes under chosen experimental conditions: solid or gas

phase reactions, evaporation, thermal decomposition, adsorption, and surface and bulk

diffusion. The same information in terms of rate constants can be obtained by analyzing

either the initial or the decay portions of the analytical signal. It was shown [47], however,

that the latter is more advantageous because it is easier to attain an isothermal environment

for the furnace and to make the results independent of temperature-sensitive parameters such

as the atom residence time in the analytical zone and spectral line characteristics [51].

The main (if not sole) available parameter, related to the atomization mechanism,

which can be obtained from direct measurements of the absorbance or the fluorescence

signals, is an activation energy. This energy can then be compared with tabulated data of

thermodynamic constants to determine which of the possible mechanisms of sample

transformation is more likely. However, the principal difficulty in the study of the

atomization mechanism by AAS or LEAFS is the impossibility of creating experimental

conditions favoring one process over all others. All the processes are interrelated and often

occur simultaneously. This is usually reflected by the shape of the Arrhenius plot (logarithm

of a measured quantity such as rate constant, absorbance, fluorescence, vs inverse

temperature) in the form of breaks and irregularities which have been observed for many

elements [45,48,52].

In some early studies of the mechanism of atom formation in the ETA [48,52], solid-

state diffusion through the porous wall of the furnace was considered as a pure loss process.

This type of loss was regarded as having a minimum influence on the removal function (see

Eqn 11) compared to the diffusion in the gaseous phase. Nevertheless, Smets [48] pointed

out that grain boundary diffusion in porous graphite could strongly affect the atom supply

and the kinetic behavior of the analytical signal. This could result in the temporal shape of

the signal being markedly different from that obtained after evaporation from an ideal,

impermeable surface.

A macrokinetic theory, developed to account for the surface diffusion in graphite

ETAs and to explain the appearance of breaks and inflections in the Arrhenius plots, was first

proposed by L'vov et al. [53] (see Figure 7). The theory described surface diffusion in

Diffusion (D) + Vaporization (V)

i \ .. ..

............... X

Set of
Sample x=

a) Low temperatures: D > V

E =H
a vapor

b) Moderate temperatures: V ~ D

Ea =1/2( Qdi+ Hvapor)

Figure 7. Macrokinetic model of diffusion

porous graphite accompanied by vaporization. The model of a porous body was represented

as a set of equal cylindrical capillaries of radius p. The sample mass was taken to be initially

located at some depth below the surface. Two limiting cases were considered. First, if the

rate of diffusional transport is much larger than that of vaporization (the authors denoted this

as "effusional kinetics"), then the supply function S can be expressed in the form

S = Tp2qk (13),

where q is the number of atoms per unit area and k is a vaporization constant (s-). In the

second case, when vaporization is very fast compared to diffusion ("diffusional kinetics"),

the supply function can be written

S = 2tpVDk (14)

where D is the surface diffusion coefficient (cm2/s). Taking into account the Arrhenius-like

temperature dependence of both k and D,

k koexp( -A ) and D = Doexp -e (15)

where AH is the heat of vaporization, Q the diffusion activation energy, R ,the gas

constant, and ko (s") and Do (cm2/s) preexponential factors, one can obtain the values for

activation energy from (13 -15)

E,, = I and Ea (Q + AI) (16)

for the two cases under consideration. Therefore, the Arrhenius plot will show different

slopes at different temperature regions, only if AH Q. For most of the metals which are

usually determined by AAS or LEAFS, the heat of vaporization AH is larger than Q. The

typical value for Q is -80 kJ/mol (as was shown in reference [54] for 12 elements), and AH

typically lies within the range of 100-400 kJ/mol. In the high temperature region,

atomization is usually driven by diffusional kinetics, i.e., competitive diffusion/vaporization

process, whereas at lower temperatures effusional kinetics are dominant and the heat of

vaporization determines a time-behavior of the supply function. The proposed theory [53]

was confirmed using as an example the atomization of copper in a graphite ETA. The data

obtained, E = AH =317 kJ/mol for the low temperature region (T5 1400 K) and E = 1/2(AH

+ Q) = 203 kJ/mol for the high temperature region (Tt 1400 K) were in excellent agreement

with tabulated values (AH,40 = 317 kJ/mol, Q=78 kJ/mol, 1/2(AH+Q)=198 kJ/mol).

It is worth mentioning in this relation, that the enthalpy AH should be considered in

a broader context than only as a heat of vaporization. It can also be attributed, for example,

to the enthalpy of a specific chemical process or to the bonding energy of a metallic dimer,

depending on which process is rate-determining under the conditions chosen. For example,

the values for the activation energy in the range of -75-95 kJ/mol were reported by Smets

[48] for different physical chemical transformations of copper oxide.

If we extend our considerations to the range of even higher temperatures, where the

condensed phase can barely exist in a porous medium (graphite), then the mechanism of the

formation of atoms in the analytical zone, beyond the diffusive layer, can be considered as

a permeation of an atomic (or molecular) vapor through the membrane. This permeation may

be a simple flow through a capillary but we shall exclude this case from consideration and

deal only with true diffusion processes. Contrary to gaseous flow, which does not show

very pronounced differences for different gases, true diffusion is a highly specific process

which depends on the solubility and mobility of atoms or molecules in the solid [55]. In this

case, the model proposed by L'vov et al. [53], which represents a porous body as a set of

capillaries, is only approximately valid. A traditional approach to the high-temperature

diffusion problem seems to be more relevant.

A simple model of diffusion of an atomic vapor through a membrane-like graphite

cylinder was proposed by Katskov [51] (Figure 8). In the experiment, he used a 2-step ETA

with a separate vaporizer and a pyrolytic graphite furnace. A spool-like porous graphite filter

was tightly inserted into the furnace. After pulsed sample vaporization, vapors were quickly

injected into the furnace between two concentric cylinders, forming the furnace and the filter

walls, and then diffused through the filter partition to the analytical zone. The absorbance

was measured in the center of the furnace. The temporal profile of the signal reflected the

time-dependence of the supply function S(t) which was a flux of analyte atoms through the

surface of the graphite filter. If the rate of change in concentration due to diffusion is

proportional to the concentration in the filter body, one can write



- Concentr=C Sample

High temperatures

E = Qdif

Rate constant: r = 4D[2a21n(a/b)+b2-a2]-1

Arrhenius eqn: D = Doexp[-Q/RT]

Figure 8. Diffusion out of the hollow cylinder

S(t) d -kNF (17)

where NF is the number of atoms in the filter body and the parameter k (s') is the diffusion

rate of atoms in graphite. If the diffusion is not complicated by other physical chemical

processes, the flux of atoms through the surface of the graphite filter is given by the Fick's

diffusion equation:

S(t)-DsVn (18),

where D (cm2/s) is the diffusion coefficient, s (cm2) is the surface area, and V n is the gradient

of the atomic number density n (cm 3) in the substrate material. Under the experimental

conditions, when the appearance time of the analytical signal is at least 10 to 20 times shorter

than the duration of the whole signal at any chosen temperature, the diffusion can be

considered as quasi-stationary, although neither the concentration nor the concentration

gradient are constant. The solution of the steady-state diffusion equation for the hollow

cylinder is well known [56]:

n = A + B logr (19)

where A and B are constants to be determined from the boundary conditions at r=a, r=b (a

and b are inner and outer radii of the cylinder, respectively). If the inner surface r=a is kept

at almost constant concentration n(a), and r=b at zero concentration (the reverse conditions

of Katskov's experiment, but not essentially changing the picture), then

n(r) = n(a) r (20)


Integrating (20), one finds the total number of atoms in the cylinder body

h n(a)
N(t)= 2nrn(r)= b [2a 2lna+b 2-a 2]
b b (21)
a 21n-

Combining (17) and (18), and taking into account the surface area of the particle source

S=2TLb, where L is the length of the furnace, one finally obtains
k =4D [2a 21n b -a2] (22)

The diffusion coefficient D can now be easily determined by measuring the rate constant k

from the decay portion of the analytical signal. The temperature dependence of D is

expressed by the Arrhenius equation:

D = D exp -RT (23)
[ R)

where Do is the preexponential factor and Q is the diffusion activation energy. Measuring

k at different temperatures and plotting log k vs 1/T, one can find the values for Q and D0.

Now, it can be easily seen that for a process driven by pure diffusion, the total

activation energy Ea is equal to the diffusion activation energy Q. The proposed model,

however, does not explain the inflections on the Arrhenius plot and does not show from

which part of the plot the diffusion coefficient can be obtained. The Arrhenius plot for

silver, represented by the author [51], showed a break at a temperature of -1560 K. The

value of Ea of 235 kJ/mol, obtained for the low-temperature region (below 1560 K), closely

coincided with the heat of vaporization of silver, although condensation with a consequent

re-vaporization has not been considered. At higher temperatures (T-1560 K), E =103

kJ/mol and chemical interaction has been assumed.

It seems to be a reasonable suggestion, that even in the 2-step ETA, with independent

pulse vaporization, condensation and re-vaporization of the sample take place, especially,

if the temperature of the preheated furnace is far below the boiling point of the analyte, as

is true for silver at T< 1560 K (B.P. 2485 K). Therefore, the diffusion equation should also

include a term to account for vaporization. On the other hand, the model proposed by L'vov

explains the behavior of the Arrhenius plot quite satisfactorily and gives a value of the

activation energy for the low-temperature range equal to the heat of vaporization. Also, for

the moderate-temperature region, the theoretical predictions of E, coincided with the

observed values within a range of uncertainty -20 % (such a comparison is given in

reference [53] for 12 elements). However, in the range of very high temperatures, where the

condensed phase of the analyte no longer exists and in the absence of chemical interaction,

the diffusion model of Katskov [51] seems to be relevant. Therefore, for some elements,

satisfying these conditions, one can expect the appearance of a portion in the Arrhenius plot

with a slope proportional to the diffusion activation energy Q.

Experimental Procedure

The experimental set-up (see Figure 6) was described in detail in Chapter 2. Key

features of the experimental routine are given here. A sample (a 10 pL aliquot of a 100 ppm

standard solution ofCu, Ag, orNi) was loaded into the sampling hollow of the diffusive tube

furnace (Fig. 5). After the sample was dried in air by a radiant heat projector, the insert was

placed into the center of the graphite tube which was then sealed by the electrodes and

heated. A five-step heating program was applied for all three elements. The program

included a charming step (10 s, 700 OC), a rapid, brief initial atomization step (1 s, 2000-2300

oC), a relaxation step (5 s, 1000-1300 C), an atomization step (10-300 s, 1100-2300 OC), and

a cleaning step (2 s, 2300 oC). The reason for using two extra steps (initial atomization and

relaxation) with respect to the traditional temperature program (drying-charing-atomizing-

cleaning), was to provide a pulse injection of the sample vapor into the furnace cavity, prior

to diffusion through the graphite partition into the analytical zone. Alternately, if partial

condensation takes place at the temperature of the relaxation step, this step serves to create

a uniform distribution (ideally, close to a monolayer). In this case, the second pulse

atomization is facilitated because the furnace is already pre-heated to a significant

temperature, above the melting point of the element under study, and the condensed phase

is distributed as a thin layer on the inner surfaces of the furnace.

To excite the fluorescence of the element under study, the laser beam was directed

along the graphite tube surface so that no gap remained between the beam and the tube

surface. The wavelengths of 324.8 nm/510.6 nm, 328.1 nm/338.3 nm, and 322.2 nm/361.9

nm were used for the fluorescence ecxitation/collection for Cu, Ag, and Ni, respectively.

Fluorescence vs time profiles for each of the elements studied were stored in the

computer by using the program "Stanford" (Megagraphics Software Corp., Stanford

Research systems, Palo Alto CA, USA). At least three parallel measurements were

performed at each temperature. The rate constants (k) were measured from the decay portions

of the fluorescence signal when a constant temperature was established in the furnace

shortly after the signal maximum was reached. Under the assumption of first-order kinetics,

the exponential function y = a + bexp[-kt] was fitted to the decaying part of the signal by

using the same software ("Stanford") utilizing a least-square method. Then the plot of the

logarithm of the rate constant vs the reciprocal of the absolute temperature (the Arrhenius

plot) was drawn and the activation energy, E, was evaluated from the slope of this plot. The

value of the pre-exponential factor ko (and Do- through the Eqns 22,23 ) was obtained from

the intersection of a high temperature fragment of the Arrhenius plot with the log k-axis at

1/T-0 (T-.). The experimental error was propagated from k to D by using Eqn 22.


Atomic-fluorescence vs time profiles were measured within the temperature range

of 1400 K 2600 K. The absolute amount of each analyte, introduced into the furnace as the

nitrate, was 1 pg. This relatively large amount was used to provide quasi-stationary

conditions for diffusion, so that the atomic number density at the inner wall of the porous

graphite cylinder could be almost constant during the measurements. It was possible,

however, that the fluorescence signal from such a large amount of analyte could become non-

linear due to self-absorption of the fluorescence light by the dense atomic vapor in the

analytical zone. To avoid this effect and also to provide for a fast removal of the atomic

vapor from the space beside the furnace (so that the number density of the analyte atoms was

near zero at the outer furnace wall), a rapid upward flow of argon gas at 4 L/min was used.

Additionally, the proportionality between the amount of the sample and the analytical signal.

was checked. The calibration curves, shown in Figure 9, indicate that the fluorescence

response was proportional to the absolute amount of each element over a range of 6 orders

of magnitude, up to the value of 1 pg.

Special care was taken to measure rate constants under isothermal conditions. The

analytical zone, from which the fluorescence radiation was collected, was restricted in size

by a set of diaphragms (3x3 mm in the longitude direction, equal to the laser cross-section,

and 2 mm in the perpendicular direction) to insure the observation of diffusion from only the

isothermal part of the furnace. Temperature measurements showed, that the temperature was

constant only in the central part of the furnace, at a distance of 4-5 mm from the center, and

was 50-150 OC lower near the furnace ends, depending on the magniture of the final

temperature of atomization. The typical distribution of the atomic vapor above the surface

of the graphite filter is shown in Figure 10. These data were obtained when a pure copper

sample was vaporized at a temperature of 1800 OC. It is seen from Figure 10 that the density

of the atomic vapor above the furnace wall can be quite satisfactory approximated by a step-

function with a width of about 5 mm. Therefore, the observation of a 2 mm segment of the











-1 0 1

Figure 9. Calibration plots for Cu, Ag and Ni obtained in the diffusive tube ETA by using
aqueous standards. m is the mass of each analyte (Cu, Ag, or Ni) in the 10 pL sample
aliquot; the solid lines represent a linear regression fit to the experimental points.

6 -5 -4 -3 -2
log m, Pg

I I .. I I



20- .





-15 -10 -5 0 5 10 15 20
Distance from center, ma

Figure 10. Spatial distribution of fluorescence signal and copper atoms
above the graphite tube with a constant atomic vapor flow.

analytical zone allowed us to avoid the small non-uniformity of the temperature along the

furnace wall.


In the experiment, rate constants for the atomization of copper were obtained over

a temperature range of 1550-2600 K. The upper limit of this range was equal to the

maximum temperature which could be obtained with our particular ETA unit. The lower

limit was determined by the uncertainty in the results at temperatures below 1500 K. In this

case, the total time of the atomization exceeded 300 s and signal shapes for consecutive runs

could differ significantly. This resulted in errors of up to 100%, when an exponential fit was

applied to such a slowly decaying portion of the signal. Therefore, the experiment was not

run below a temperature of 1500 K and the results were only considered as satisfactory if

they fell within an uncertainty range of :30 %.

Figure 11 shows the Arrhenius plot for copper which was constructed on the basis

of the two sets of data. The upper part of the plot was constructed from data obtained in a

temperature region of 1550-2600 K by using a LEAFS-ETA technique (circles, connected

by the solid line). The lower part (crosses, connected by the dashed line) represents some

data, extracted from L'vov et al. [53], which have been obtained in the temperature region

of 1250-1600 K by using an AAS-ETA technique and a quasistatic method for the

evaluation of the parameter k. This method was similar to ours and also utilized the decay

portion of the analytical signal. It can be seen, that the plot consists of three linear parts with

two inflection points: at -1770 K and at -1380 K. In the middle part, corresponding to the

temperature region of 1380-1770 K, the two linear plots, obtained from two different


Temperature, K
3000 2500 2000 1500
l I. I I I




-ji.o II I 1-3.5
3 4 5 6 7 8

10000/T [K1]

Figure 11. Arrhenius plot for copper constructed on the basis of the two sets of data:
obtained in this experiment (circles connected by the solid line) and extracted from L'vov
et al. [53] (crosses and stars connected by the dashed lines). The lines represent a linear
regression fit to the experimental points in different temperature regions.


experimental systems, show nearly the same slopes. The complete coincidence of the plot

from L'vov et al. [53] with our data should be considered as accidental because, as was

reported in the same paper, the rate constant k could vary in consecutive runs under the same

experimental conditions, as reflected by the two parallel dashed lines in Fig. 11.

Nevertheless, the slope remained the same for all runs, permitting the extraction of a value

of the activation energy E, which could be attributed to the specific process under


At temperatures below 1770 K, a value for E, equal to 195 kJ/mol was obtained, in

close agreement with the value of 203 kJ/mol reported by L'vovBayunov and Ryabchuk [53]

for the temperature range of 1380-1600 K. L'vov and co-authors attributed this value of

activation energy to the competitive diffusion/vaporization process in a porous graphite

medium (E=1/2(Q+AH) see the theoretical section). A value of the activation energy, also

close to ours and equal to 184 kJ/mol at T>1400 K, has been obtained by Sturgeon,

Chakrabarti and Langford [52]. This value was related to the dissociation energy of the

gaseous copper dimer. Some other data, reported in the literature, are as follows: Fuller [47]

obtained Ea=138 kJ/mol (T>1720 K) and attributed this energy to the heat involved in the

reaction of the reduction of copper oxide by carbon; Katskov and Orlov [57] presented an

Arrhenius plot, from which a value of 154 kJ/mol could be extracted for T,1400 K.

For the region of lower temperatures (below 1400 K, dashed line in Fig.11), which

was not examined in the present experiment, almost all reported values for E. were very

close to the enthalpy of vaporization of copper: AH140=317 kJ/mol [48,52-53].


At high temperatures (above 1770 K), E,=77 kJ/mol was obtained which closely

coincides with the activation energy of 78 kJ/mol for the diffusion process reported in [54].

Another value for the activation energy, which was somewhat close to our result, has been

obtained by Smets [48]: 50 kJ/mol at Ti1800 K. The value obtained for Ea was attributed

to the pure diffusion transport of the copper atomic vapor through the graphite partition,

considering this transport not to be complicated by other concomitant processes (chemical

reactions, sorption-desorption, etc.). Equations (22 and 23) were used to evaluate diffusion

coefficients (D) in the temperature range of 1770-2600 K. Taking into account the geometry

of the furnace, values for D between 3.7-104 cm2/s and 2.0 -10' cm2/s were obtained over the

temperature range studied. The frequency factor was determined from the intercept of the

graph of log( k)=f(l/T) with the logk-axis at 1/T=0 and was equal to 7-102 cm2/s.

To verify the value of the diffusion coefficient obtained by using the tube furnace,

another experiment was also carried out. A small graphite cup (height, 6 mm, inner and an

outer diameters of 3.5 mm and 4.5 mm, respectively, and a depth of 4 mm) was used instead

of the tube furnace. The cup was held between the same spring-loaded electrodes, as in the

previous design, and could be tightly sealed by a graphite stopper. The stopper was made

from the same type of graphite, as the tube furnace, and served as the filter to separate the

analytical zone from a sampling volume. The thickness of the filter varied between 0.8 and

3.3 mm throughout the experiment. After the sample aliquot (10 tL) containing 0.01 jig of

copper was introduced into the cup and dried, the cup was sealed by the stopper and heated

by using a 3-step temperature program: charring (10 s, 900 K), atomization (up to 200 s,

1800 K), and cleaning (2 s, 2600 K). The intermediate steps (the initial atomization and the

short relaxation), used in the experiment with the tube, were excluded because the small

thermal inertia of the cup allowed an instant vaporization of the sample. The absolute

amount of the analyte was also reduced from 1 Ig to 0.01 ig due to the much smaller

volume of the cup compared to the tube. The analytical volume, from which the

fluorescence was collected, was also reduced to -1 mm3 in order to neglect a slight non-

uniformity of the temperature over the filter surface and also to apply the model of steady-

state diffusion through an infinite plane sheet. According to this model [56], the rate of

transfer of the diffusing substance is the same across all sections of the filter and is given by

S = -D- = D -n2 (24),
dx I

where D is the diffusion coefficient, 1 is the thickness of the filter, n is the number density

of the analyte in the filter body, n, and n2 are the number densities in the two surfaces of the

filter. Under the boundary conditions ofn=n0 at x=0 (the filter bottom) and n=0 at x=l (the

filter top) and repeating the considerations (17-21) from the theoretical part, one obtains

k = D or D = kl2 (25),

where k is the rate constant for the diffusion process. The parameter k could also be

obtained from the decay portion of the analytical signal by using an exponential fit.


The temporal behavior of the copper fluorescence signals, obtained at T=1770 K by

using filters of different thickness (0.8 mm, 1.4 mm, and 2.5 mm), are represented in Figure

12. The rate constants, extracted from these signals, were 0.07 s', 0.02 s', and 0.009 s',

which result in values for the diffusion coefficients, via Eqn. 17 of 4.6-104 cm2/s, 3.9-10'

cm2/s, and 5.8-10' cm2/s, which gives an average D=(4.70.7)-10-4m2/s.

The value for D, obtained at the same temperature (1770 K) for a furnace with a

hollow cylinder geometry, was 3.7.10-4 cm/s, which is in a satisfactory agreement with this



Typical shapes of silver fluorescence signals for different atomization temperatures,

together with temperature profiles, measured with photodiode using thermal radiation from

the hot furnace, are displayed in Figure 13. One can see that as the temperature increases

the kinetic of the release of silver atoms changes significantly: the long-tailed signal in

Fig.13c transforms into the sharp, short spike in Fig. 13a. This likely corresponds to the

transition from the dominant vaporization to the dominant diffusion kinetic as will be

discussed below. A first-order exponential function gave a good fit (with a correlation

coefficient close to 1) to the decay portions of the signals.

The Arrhenius plot for silver (Figure 14) was obtained over a temperature range of

1430-2280 K by using a five-step temperature program. Beyond this range, the uncertainty

in measured rate constants exceeded 30 %, and so these data were not considered to be






0.0 -

a b c
0 100 200
Time, s

Figure 12. Fluorescence signal profiles for copper (0.01 pg) obtained at 1770 K by using
a graphhite cup atomizer sealed by the graphite stopper of the different thickness: a) 0.8 mm;
b) 1.4 mm; c) 2.5 mm.




0 50 100 150
Time, s

200 250 300

Figure 13. Fluorescence signal profiles obtained for silver (1 jg) at different temperatures:
a) 2100 K; b) 1700 K; c) 1400 K. Temperature profiles vs time, obtained by using infra-red
blackbody radiation from the furnace, are displayed by the dashed lines.

Temperature, K
2500 2250 2000 1750


4 6
100001T, K"1

Figure 14. Arrhenius plot for silver.









reliable. It is seen that the plot consists of two linear parts. At temperatures below -1750

K, the activation energy, corresponding to the atomization process, was equal to 238 kJ/mol.

This value was in close agreement with that obtained by Katskov [51], 235 kJ/mol (T< 1560

K) and also closely coincided with the heat of vaporization of silver. Smets [48] obtained

a value of 277 kJ/mol at T< 1250 K for vaporization of silver in a pyrolitically coated furnace.

A value of the activation energy of 25112 kJ/mol, also close to ours, was reported by

Katskov and Orlov [57] for a temperature range of 1100-1450 K.

Thus, one can conclude that after the first initial atomization and the relaxation steps

(1900 K, 1 s and 1300 K, 5 s, respectively), the sample vapor has condensed on the inner

surface of the furnace and then re-vaporized during the second atomization step. As was

suggested by Fonseca et al. [58], vaporization of silver occurred preferentially from the

surface of microdroplets. At relatively high masses (1 ig in our experiment) and moderate

temperatures, the condensed phase could exist in the bulk of the graphite membrane during

the total atomization time and be transported through it either by surface diffusion [53] or

by simple capillary action [59]. It is interesting to note, that the inflection point in our plot

corresponds to a higher temperature (1750 K) than the temperatures of inflection reported

in other publications: 1560 K [51], 1450 K [57] and 1250 K [48]. This was probably due to

the larger mass of the analyte (ljg) which we used and the smaller inner volume of our

furnace compared to those used by the above authors, favoring the condensation of the

atomic vapor on the inner surface of the furnace in the form of microdroplets. For example,

Smets [48] used a mass of silver of only 0.2 ng and assumed that prior to the atomization,

the sample was distributed as a surface monolayer. Under such conditions, the condensed

phase could readily disappear at temperatures slightly above the melting point (1250 K, M.P.

1235 K).

At temperatures above 1750 K, a value of the activation energy of 97 kJ/mol has been

obtained. A value of 103 kJ/mol was obtained by Katskov [51] by using a two-step diffusive

graphite furnace with a pulse vaporizer at T> 1560 K. This value was attributed to diffusion,

complicated by chemical interaction of silver with a graphite substrate. Fonseca and

coauthors [58] used a furnace with a graphite platform to study the effect of the sample

concentration, the pretreatment temperature and the roughness of the platform surface on the

magnitude of the activation energy. Varying these parameters, they obtained values for Ea

in the range of 97-147 kJ/mol and also concluded that the atomization process could slightly

deviate from first-order kinetics. Nevertheless, when the thermal pretreatment was carried

out at high temperatures (-570 K), the process was shown to be first-order and the lowest

value for E, (-97 kJ/mol), obtained in this case, was related to the desorption of silver atoms

from the platform surface.

By comparing the Arrhenius plots for copper and for silver (Figs. 11,14) one can see

that, contrary to copper, silver exhibits only one break in the Arrhenius plot. This can be

explained by differences in the degree of interaction that copper and silver show for graphite.

As was pointed out by Guell and Holcombe [60], copper possesses a strong affinity for the

graphite surface, whereas silver exhibits only a weak interaction with graphite [57].

Therefore, at moderate temperatures, when gaseous and condensed phases co-exist, we can

expect that the process of atom release is driven by different mechanisms: by competitive

diffusion/vaporization for copper and by vaporization from microdroplets for silver.

The values for diffusion coefficients within the temperature range of 1750 K-2280

K, estimated by using equations (14)-(15), were between 1.4.10 cm2/s and 6.5-10' cm2/s.

The preexponential factor Do was determined to be 1.1 cm2/s. Compared to the diffusion

coefficients which can be calculated on the basis of eqn. (23) by using the results from

Katskov [51]: E,=103 kJ/mol and Do=29 cm2/s, our values for D were approximately one

order of magnitude lower. This was, probably, a result of differences in types of graphite

used in these two experiments.

To verify the results obtained, a further experiment was carried out. Small graphite

cylinders of length 1 cm and diameter 0.6 cm and density 1.65 g/cm3 were used as diffusive

media. The cylinders were tightly enclosed into an impermeable molybdenum tube of the

same length. Two 20gL-aliquots of the stock solution (1000 ppm of silver) were deposited

on each end of each of the cylinders. After drying, the cylinders were placed into a

temperature-stabilized laboratory furnace (Lindberg, USA) where they were annealed for 2

hours at a temperature of 900 OC under an argon atmosphere. After this time, the graphite

substrates were uniformly saturated by silver. This was checked separately for several

cylindrical specimens by sawing off sections and determining the silver by LEAFS-ETA.


The solution of the diffusion equation for such a finite system is given by Jost[55]

and can be well approximated by

n 8 h2
exp[-t/t] T (26)
no T72 ntD

at initial and boundary conditions of n-no for 0 < x < h at t=0 and n-0 for x=0 and h=0 at

t> 0, where no and n are the initial and current atomic number densities in the cylinder body,

respectively, h is the cylinder length, D is the diffusion coefficient, and x is the coordinate

along the direction of the diffusion transport. Consequently, T and D may be obtained from

the slope of the linear graph of log n/no vs time. The fluorescence signal profiles were

measured at different temperatures in the same manner as in the previous experiments. The

current values of n were determined by subtracting the portions of the analytical signal,

integrated between t-0 and t=ti,t2, etc., from the total area of the signal. A similar study of

the dynamics of the diffusant release is also described by Hensel and coauthors [61].

The plots of log n/no vs time, obtained at four different temperatures, are presented

in Figure 15. The diffusion coefficients, calculated on the basis of the Eqn. (26), were

1.7-103 cm/s, 2.1-10-2 cm/s, 2.8-10-2 cm2/s, and 4.5-10-2cm2/s at temperatures of 1340 K,

1580 K, 1840 K, and 1970 K, respectively. The values for the activation energy and the

preexponential factor, determined from the Arrhenius plot, were 109 kJ/mol and 17 cm2/s.

These parameters were very close to those obtained by Katskov [51]; E,=103 kJ/mol, Do=29

cm2/s, and the diffusion coefficients listed above coincided with those given by Katskov

within an uncertainty of30 %. However, some discrepancy with our previous results


-0.5- A

-1.05- ,

o d

-2.0 -


-2.5- b


-3.0 .
-50 0 50 100 150 200 250 300
t, S

Figure 15. Diffusion out of the graphite cylinder bounded by the impermeable molybdenum
surface, parallel to the axes of the diffusion transport, eqn. (26). a) 1970 K; b) 1840 K; c)
1580 K; d)1340K


(Ea=97 kJ/mol, D0-=.1 cm2/s) was observed. This was not very surprising because the two

diffusive media in our two experiments were different in terms of surface-to-volume ratios:

in the first case this ratio was equal to -10, and in the second, to -2. As has been shown [58-

59], the surface topology can strongly affect the shape of the analytical signal and,

consequently, the values of the activation energy and the diffusion coefficient. Also,

Zherdev and Platonov [62] have pointed out that a combination of factors (the concentration

used, the method's sensitivity, the time of diffusion annealing, etc.), involved in different

experimental methods could distort the picture observed or emphasize only some of the

features of the migration process. For example, analysis of the data on the diffusion of Cs

in reactor graphite of various grades, scattered throughout the literature [62], showed that

values of the diffusion coefficient could vary by as much as 8 orders of magnitude from

author to author, although the scatter in values of the activation energy was relatively small.


The activation energy for diffusion of Ni was found to be 161 kJ/mol over a

temperature range of 1770 K-2530 K. No breaks in the Arrhenius plot were observed

(Figure 16). A single-slope Arrhenius plot for Ni was also obtained by Katskov and Orlov

[57] for temperatures between 1970 K and 2560 K with a value for the activation energy of

1198 kJ/mol. This value was attributed to vaporization of the metal from a microdroplet

covered by a carbon core. According to the authors, the core can form on the microparticle

of the melted metal if it possesses a strong affinity for carbon. The affinities for carbon were

estimated (in %) for several elements and for the elements studied in the present work are:

Ni 2.6, Cu 0.031 and Ag 0.0018. As was shown, no core formed for Ag, some core






T, [K]
2700 2400 2100



-2.5 1 1 I ,-
3.5 4.0 4.5 5.0 5.5 6.0

10000/T, K"

Figure 16. Arrhenius plot for nickel


formed for Cu but destroyed at relatively high temperatures, and a rigid carbonic core formed

on the surface of molten nickel even at high temperatures.

Two values for the activation energy for Ni were reported by Sturgeon and coauthors

[52]: 210 kJ/mol and 416 kJ/mol. The first one was related to dissociation of the gaseous

nickel dimer whereas the second was associated with thermal vaporization of Ni from the

condensed phase. The break in the Arrhenius plot occurred at a temperature of 1690 K

(lower than the minimum temperature used in the present work).

An activation energy of 29917 kJ/mol (2000 K < T < 2440 K) was obtained by

Rojas and Olivares [50], when a dry residue of a nickel standard solution was vaporized from

the surface of a pyrolitic graphite furnace. First-order kinetics for the release of the metal

vapor was proven.

The diffusion coefficients obtained on the basis of the model expressed by Eqns (17-23)

were in the range of 5.6-10-'cm/s 1.5-103cm2/s (1770 K < T < 2530 K) with a value for the

preexponential factor Do of 3.2 cmZ/s. The data for diffusion of Ni in electro-graphite,

reported by Weisweiler and Nageshwar [54], were at least two orders of magnitude lower than

ours. Such a large discrepancy with our results may be explained by a principal difference in

the transport process in these two experiments. Diffusion of the metal into the bulk of a

massive graphite prism [54] was carried out from a molten metal. Prior to the concentration

analysis, the upper layer of the prism was removed, so that the influence of surface effects,

which could take place, was minimized. In our case, when the surface-to-volume ratio was

equal to -10, surface effects, such as sorption on surface irregularities or capillary action [58-

59] could strongly affect the transport process. Also, the relative content of the two phases

(condensed and gaseous) on the surface of the porous diffusive medium is important because

the transport mechanism, is supposed [62], to consist of a superposition of transport across the

volume of the crystallites and vapor diffusion in the pore volume.


A method for the kinetic study of atomization and diffusion of three elements (Cu, Ag

and Ni) in a porous graphite tube atomizer by means of laser excited atomic fluorescence

spectrometry has been presented. Activation energies for vaporization/diffusion of Cu, Ag and

Ni have been obtained for a temperature range of 1430 K 2530 K and found to be in a

satisfactory agreement with values of the activation energy reported by other authors. Diffusion

coefficients for diffusion of the metal vapor through a porous graphite cylinder were estimated

on the basis of the model proposed by Katskov [51] for a high temperature region. The values

of activation energies and diffusion coefficients obtained in the experiment are given in Table

2 and Table 3, respectively.

The results of this study can be of importance both for further understanding of

mechanisms involved in the formation and the transport of atoms in a porous graphite medium

and for analytical applications. The difference in diffusion rates for different atomic vapors in

graphite, obtained in the present work, together with the differential transport rates of

molecular and atomic vapors, observed in previous studies [42], could be the basis for

development of methods of high-temperature separation of constituents of complex matrices

in order to improve the analytical performance of techniques utilizing ETAs (AAS, LEAFS,

etc.), particularly, in terms of freedom from spectral interference.

Table 2. Activation energies (in kJ/mol) of the release of atoms in ETAs: comparison of
the data obtained in the present experiment with the literature

Temperature This experiment Literature

< 1400 K -"- 317 [52-53]

1400 K 1700 K 195 184 [52] 203 [53]

> 1700 K 77 78 [54]

<1750K 238 235 [51]

>1750 K 97 103 [53]

1770 K 2530 K 161 210 [52]

Table 3. Values for pre-exponential factor (Do) and for diffusion coefficients (D)

Temperature Do [cm2/s] D [cmI/s]

1770 K 2600 K 7.10-2 3.710-4 -2.0-103

1750 K 2280 K 1.1 1.410- 6.510-

1770 K 2530 K 3.2 5.6-10 1.5-10-3



In the next experiment, the idea of the transport of the analyte vapor through a

porous graphite filter was utilized together with a new arrangement of the LEAFS-ETA

experiment. Contrary to the widely used practice of exciting atomic fluorescence within

the hot graphite tube and collecting the analytical signal by means of a pierced mirror

placed in front of the tube [8,24,36,63-64], in the present experiment, the fluorescence

was excited directly above the outer furnace surface. This allowed for discrimination

against the strong thermal background radiation produced by the furnace and for

improvement of the fluorescence collection efficiency. The performance of the atomizer

was studied in the course of the determination of silver in sea water. As was previously

reported [40-41], the use of an atomizer with the graphite filter significantly reduced

matrix effects; this is one reason why such a complex matrix as sea water was chosen for

the real analysis. Silver was chosen because it exhibits excellent figures of merit for

determination by LEAFS-ETA [33], possessing a high volatility and a low reactivity.

Also, there are several studies in the literature concerning the determination of Ag in sea

water [65-66]. The determination of Ag in solid standard reference materials (SRM) was

intended primarily for the investigation of the accuracy, feasibility, and versatility of the

LEAFS technique with the diffusive tube atomizer.


Solution samples of 10 iL were deposited into an inner graphite rod hollow (2, Fig.5)

with an Eppendorfmicropipet. When analyzing sea water, a 10 tL injection of sea water and

a 10 gL injection of an aqueous standard were sequentially dispensed to the same spot of the

rod hollow to obtain a calibration plot affected by the presence of the matrix.

A special sampler was designed to analyze solid powdered samples. It consisted of

a narrow graphite cylinder with a stainless steel piston. One end of the cylinder was

sharpened to a cone so that the cylinder tip was smaller than the size of the sampling hollow.

The piston was lodged in a marked position and the sample powder was pressed into the

channel. The powder was then released by pushing the piston down the cylinder channel.

The average dosing mass (usually, in the range of 10-20 mg) of each sample was determined

in advance in series of weighing. Weighing errors were taken into account in the

calculation of concentration.

Two sets of silver standard solutions were used throughout the experiment. One set

was prepared from solid silver nitrate (analytical-reagent grade, Fisher Scientific, Orlando,

FL, USA) by dissolving 0.1575 g of the salt in 100 mL of high-purity water (18 MQ cm)

obtained from a deionizing water system (Milli-Q, Millipore Corp., Molsheim, France).

Working standard solutions were then prepared by sequential dilution of this stock standard

(1000 ppm of Ag) down to a concentration of 0.01 ppb. The second standard solution set


was obtained by a similar stepwise dilution of a 1000 ppm commercial Ag stock solution

(Inorganic Ventures, Inc., Brick, NJ, USA).

Synthetic sea water was prepared from the solid salts of NaCI, Na2SO4, KC1, CaCl2,

and MgCl2 of analytical-reagent grade (all from Fisher Scientific, Orlando, FL, USA)

according to the recipe given in Khoo et al. [67] (Table 4). The certified reference materials

(soils) SRM 2709-2711 (NIST, Washington, DC, USA) were used without preliminary


The dye laser output (656 nm) was frequency doubled by a KDP crystal to produce

the 328 nm excitation wavelength. This was directed through a quartz window into an

atomizer chamber (see Fig. 6) and was used to excite Ag atoms (5s2Sla 5p2p32 transition)

generated above the diffusive tube furnace. The energy level diagram for silver, showing

the transitions used in the experiment, is given in Figure 17.

Prior to initial use for analysis, new diffusive tubes were fired several times at 2500

'C until the signal, detected at the wavelength of silver transition (338 nm), dropped to the

background level. The limiting background noise was primarily due to the laser radiation

scattered from atomizer parts. It was significantly reduced by placing an acetone filter

between the furnace and the monochromator slit and by blocking all possible sources of laser

stray light. The width of the monochromator slit was 2000 jim, providing a high detector

throughput and the best signal-to-noise ratio.

Before sampling, the furnace was fired several times at a temperature of -2500 0C

until the background signal fell to the baseline. Then the sample was deposited into the

Table 4. Major constituents of seawater [67]

NaCI 2.4 %

Na2SO4 0.4%

KC1 0.1%

CaCI2 0.1%

MgCI2 0.5 %

2 2P

1 2s

2p 3
- 1/2

Figure 17. Simplified energy level diagram for silver. The laser excited transition is drawn
with a thick solid line; fluorescence transitions with thin solid lines; and collisionally excited
and deactivated transitions with wavy lines.

sampling hollow inside the furnace and 3-step heating program was applied. The program

included a drying step (40 s, 110 C), a charring step (10 s, 600 C), and an atomization step

(40s, 1500 C, ramp rate 800 K/s). This program was used for all samples, unless otherwise

specified. As the expected silver concentration in the sample increased, additional colored

glass filters were placed between the furnace and the monochromator to avoid saturating the


Analytical calibration curves were obtained by using 10 AL aliquot of standard

solutions which covered a concentration range of 6 orders of magnitude (0.1 tg/L 0.1

mg/L). The analysis of the sea water and soils was carried out within the calibration

procedure to minimize a temporal drift of experimental conditions. The limit of detection

(LOD) was estimated as the concentration of the analyte (silver) producing a signal 3 times

the standard deviation of the blank (pure water).


Spatial distribution

First, the 3-D distribution of silver atomic vapor diffused through the graphite tube

was studied. Figure 18 (a,b) represents the distribution of silver atoms in X- and Y-

directions (lengthwise and transverse with respect to the tube axis). The measurements were

carried out under steady-state atomization conditions, i.e. at a steady state of atomic vapor

flow. The steady state flow was established by placing a small piece of a pure silver wire

into the furnace (continuous atomic source), and by holding the furnace at constant

temperature (1500 OC) during the 4 min interval necessary to run one longitude or one

transverse scan. The scan was done by moving the entire furnace relative to the exciting

7ni mm

I eA

SiJ 'I I I1

1 m
, ,l

SI ,' II', l 'l

-1 O ta5 0 f5 t 10 15 20
ODistance front cnter, mm



J 7mm

1 mm

-is -0 tn -s f to is 20
Ditne from center, mm

Figure 18. Spatial distribution of silver atoms diffused through the graphite tube with a
stationary state of atomic vapor flow. (a) Silver fluorescence signals recorded at discrete
distances (-2.5 mm) from the center of the furnace during one scan along the X -direction
at different heights above the furnace; (b) fluorescence signals recorded in the same manner
in the transverse direction (Y).

laser beam while leaving the collection optics fixed. The fluorescence was measured in a

particular point for 5 s (with a volume of 0.25 x 0.25 x 2 mm3) above the furnace, then the

monochromator entrance slit was blocked for another 5 s while the furnace was moved to the

next point, and so forth.

The distribution in the X- direction (Fig. 18a) has a bell shape that probably reflects

the atomic distribution within the furnace. For non-isothermal atomizers (like the one under

study), the temperature inside the furnace reaches its maximum in the center and decreases

towards the tube ends which contact with the water-cooled electrodes. This causes partial

condensation of the atomic vapor near the tube ends and is responsible for the decrease of

concentration in this part of the furnace. Diffusion of the vapor through the tube wall makes

this effect even more pronounced because the diffusion coefficient is proportional to


The second observation is that the concentration of free silver atoms does not change

significantly with the distance above the tube up to a height of 15 mm. This indicates, first,

(not surprisingly), that under laminar argon flow (1 L/min) the transverse diffusion of the

analyte vapor is much slower than the convection forced by the argon stream, and, second,

that gas-phase reactions in the absence of matrix are insignificant even in a relatively cold

zone far from the furnace. The latter confirms the conclusion given by Chekalin and

Marunkov [64], who studied in detail the mechanism of silver atomization, that "silver is one

of the few elements for which chemical reactions, either gaseous-phase or heterogeneous,

can be neglected under controlled conditions" (p. 1413). This statement, made in reference

to a hot furnace enclosure, can now be extended to the environment near the furnace.

The distribution in the Y-direction (Fig. 18b) is similar to the one along the X-axis

and reflects the geometry of the tube.

The same distribution was also obtained for nonsteady state vapor flow conditions,

when 10 4L of silver stock solution (1 ppm) was used as a sample. The similarity of the

results for both steady state and nonsteady state flows indicates that the equilibrium

distribution of free silver atoms is reached inside the tube prior to any significant diffusion

of vapor through the furnace wall.

Also, relative rates of heating of different parts of the furnace were measured. To

make this possible, a small orifice was drilled in the center of the tube and a telescope,

terminated with a photodiode which detected infra-red blackbody radiation was

sequentially focused at the inner tube wall, the inner rod, and the outer wall through a set of

diaphragms. Figure 19 shows the temperature vs time behavior for the three furnace regions.

As can be seen, for the relatively low final temperatures (1100-1300 OC, Fig.19a,b), heat

dissipation predominates over heat supply due to the high thermal conductivity of graphite,

responsible for the heat transfer from the hot central part of the tube to the water-cooled ends,

and to the cooling of the furnace by the cold argon stream. In particular, the thermal

conductivity mechanism affects the temperature of the radiationally heated inner rod which

contacts directly with the water-cooled electrode. As the final temperature increases (up to

1500-1700 OC, Fig.19c,d), this difference in heating rate becomes less; the inner and outer

walls reach the final temperature almost simultaneously, while the inner rod lags slightly

behind. All measurements were carried out at the atomizer control unit ramp rate specified

as 800 K/s.

0 2[

020 1100 0C 020

015 o"- 0.1

005 00 0 o.Os

0.0 -o.os
1to 20 o30

Time, s

Figure 19. Temperature growth curves for different parts of the furnace: a,b low
temperature range; c,d working temperature range. The dashed line corresponds to the
temperature of the inner wall of the tube; dotted line to the outer wall; and the solid line to
the inner graphite rod with a sampling boat.

Oi35 -------------------

I 20 3 0


This type of differential heating of the sampling zone and the bulk of the furnace is

widely used in AAS carried out in furnaces with a platform [68-69]. These furnaces have

been found to be advantageous with respect to simple graphite tubes without an insert,

providing an enhanced analytical signal due to the substantial reduction in gas-phase

reactions. This occurs because of the temperature delay between the platform and the bulk

of the furnace allows the analyte vapor to enter the inner tube volume at a higher

temperature. Our new furnace can, therefore, be considered as analogous to a furnace with

a platform, prior to the diffusion of the analyte vapor through the porous graphite tube wall.

Additionally, it is interesting to compare open tube atomizers with our sealed graphite

furnace. Akman and Dtner [70], who studied interference mechanisms of NaCl on Zn and

Co using the AAS-ETV technique, showed that there are some losses of the analyte during

the pretreatment (charring) step. This is due to formation (in the condensed phase) of

volatile compounds which can be expelled from the open ends of the furnace before the

atomization step. Also, according to Holcombe [71], thermal gas expansion and or matrix

gas evolution are significant for analyte loss in open furnaces, especially, for volatile metals

and high heating rates.

It is clear, that the closed graphite tube is free from these undesirable effects inherent

to open tube furnaces. Even if a volatile compound of an analyte has been formed during the

charring stage, this compound can be adsorbed on the furnace wall and then secondarily

decomposed and evaporated at the atomization temperature. The presence of the graphite

filter between the sampling and the analytical zones can also result in partitioning of the

matrix gases and the analyte atomic vapor due to differing diffusion rates through the


graphite and to the formation of intercalated compounds implanted into the graphite

crystalline structure [41-42]. This is the most likely reason for the reduction of matrix

interference in the atomizer with a graphite filter.

Saturation of the silver transition

The saturation parameter is an important figure of merit linked directly with other

analytical characteristics, such as accuracy and limit of detection (LOD). Operating under

saturation conditions is favorable because it provides the largest possible analytical signal,

reduces flicker noise caused by fluctuations in the laser irradiance and interference caused

by variations in the quenching environment.

On the other hand, if the laser output far exceeds the optimum value, excessive laser

scattering can significantly increase the background level and result in a poorer LOD. The

latter is particularly important in our case, where stray laser light is a source of the limiting


For these reasons, the saturation parameter was measured and compared with the

theoretically estimated value. The fluorescence saturation curve (Figure 20) was obtained

with the laser operating at its maximum power while recording the time integrated

fluorescence signal, and gradually decreasing the laser irradiance by inserting into the beam

before the atomizer chamber several calibrated colored glass filters to reach the region of

linearity between the fluorescence signal and the laser power. The saturating medium (the

silver atomic vapor) was considered to be optically thin when 10 iL of the 1 ppm

concentration aqueous standard was atomized, and steady state fluorescence radiance

Spectral energy density, p 1013, J s/nm

Figure 20. Saturation curve for 2S,;/ 2P32 silver transition. The dashed horizontal line
indicates 50% of the steady state saturation value.


conditions were assumed during the detection of the signal by the boxcar with a wide gate

(50 ns) [72]. In accordance with the definition given in Omenetto et al. [73], the saturation

parameter, E',, was determined as the laser irradiance producing a steady state value of the

excited state population which is 50% of the steady state saturation value. In terms of

energy, this parameter was found to be equal to 0.5 pJ.

The theoretical value of E', was estimated on the basis of the model proposed by

Omenetto et al. [73]. Under our experimental conditions, silver can be considered as a

sodium-like system (Fig. 17) in which the third level (5p2P32) is radiatively coupled with the

ground state (5s2 S ,) and collisionally with the level directly reached by the laser radiation

(5p2P3a). It was evident from the experimental saturation plot that strong optical transitions

2-1 and 3-1 predominate over the collisional deactivation of levels 2 and 3; therefore,

collisional deactivation has been neglected in the calculation. It was also assumed that rate

constants responsible for the mixing of levels 2 and 3 (AE=921 cm') are balanced, and the

rate constant was derived from the experimental data obtained in reference [74]. As a result,

the saturation spectral irradiance, E',, was calculated to be equal to 510-9 J cm-2 s"' Hz-' or,

in terms of energy, 0.3 pJ, in excellent agreement with the experimental value.

In this way, the laser energy available in our experiment was found to be more than

an order of magnitude in excess of the saturation spectral irriadiance. This made possible

to enlarge the laser cross-section (from 0.1 cm2 to 0.25 cm2 ) and, thus, to excite the

fluorescence in a larger volume of the analyte vapor. In addition, calibrated colored glass


filters were used to further attenuate the laser irradiation to the point where it exactly

satisfied the saturation condition.

Analysis of Sea Water

The determination of trace elements in sea water is pursued with great difficulty

because, first, the high salt content of the sea water matrix often causes analytical

inaccuracies and, second, the concentration of the element of interest is often below or very

close to the detection limit of the most sensitive analytical techniques.

The most common methodologies used are atomic absorption spectrometry (AAS)

[75-77], isotope dilution mass spectrometry (ID-MS) [65-66,78], and inductively coupled

plasma mass spectrometry (ICP-MS) [79-80]. However, none of these methods is sensitive

enough to detect ng/L concentrations without a preliminary separation/preconcentration

procedure which can be very tedious in this matrix.

Recently, the use of the LEAFS technique with a graphite tube atomizer for the direct

[25,81] and semi on-line [37] analysis of sea water and marine sediments has been reported.

Cheam and coauthors [25] used the in situ standard addition procedure for the determination

of lead in sea waters down to femtogram levels. No separation/concentration steps nor

chemical modifier were used. Enger et al. [81] carried out the direct determination of

antimony in environmental and biological samples at pg/mL concentrations. An intensified

charge coupled device (ICCD) was used as a detector providing the possibility to control and

correct for various background signals. Yuzefovski and colleagues [37] developed a method


for the determination of ultratrace amounts of cobalt in sea water by means of LEAFS-ETA

coupled with semi on-line flow injection microcolumn preconcentration. Despite the

adequate results obtained during the analysis of aqueous reference materials, the authors

experienced contamination problems introduced by the preconcentration step.

In the present approach to the direct determination of silver in natural sea water,

synthetic sea water was first used to develop a reliable calibration procedure. The normal

aqueous standards calibration could not be directly used due to matrix interference caused

by the extremely high salinity of sea water which was responsible for an appreciable

depression of the analytical signal, even in atomizers with a graphite filter [41]. Therefore,

two calibration plots (Figure 21a) were created to account for the interference. The first one

was obtained by using aqueous standards, and the second, a 1:1 mixture of the synthetic sea

water and the stock solution of a known concentration. The sample and the standard were

sequentially pipetted into the sampling hollow of the furnace inner rod without premixing

the individual solutions. Such premixing would seem more convenient to avoid the dosing

sequence; however, the formation of a precipitate in the form of insoluble silver chloride, due

to the substitution reaction of silver nitrate (the initial form of silver compound) with the

excess of metallic chlorides contained in the sea water, leads to a highly inhomogeneous

distribution of the precipitate in this hypothetical standard solution. The calibration plots

shown in Fig.21a differ from each other by a factor of 0.23 ( 1.7 on a linear scale). This

ycaIe(Y) A B ncate(X) /^
Paamm Value sd

2 -1 0 2 3 4 5 6
-A 1 :66 0,031

b 2 212 0 0770
S0o0 2 01 2o N12

S 09 0 13 0

R -s 00.04115, N-.9

-z : -- ,-

Log m, pg

Figure 21. Calibration plots for the determination of silver in sea water. Solid lines
represent the calibration on the basis of the stock aqueous solutions; dashed lines on the basis
of 1:1 mixtures of the stock solutions with (a) synthetic sea water, and (b) natural sea water.

-2 -1

S4 5 6


factor, which we denoted as a "depression factor", served to characterize numerically the

matrix interference and served as a correlation coefficient allowing the determination of

silver by using a normal aqueous standardization only. As can be seen, the two plots have

exactly the same slope indicating complete recovery of the analyte.

A similar two-plot calibration procedure was then carefully repeated using natural

ocean water samples. The samples were collected from the Atlantic Ocean near the central

West coast of the Florida peninsula. Not surprisingly, in a series of consecutive experiments

we found that the depression factor ranged from 0.28 to 0.32 (1.9 and 2.1 on a linear scale),

reflecting an increased complexity of the real matrix with respect to the synthetic one. The

log-log calibration plots (Fig.21b) were found to be linear in the range of 1 ng/mL 10

ig/mL, corresponding to 10 pg 0.1 tg. The slightly sub-linear slope of the graphs (0.93)

can probably be attributed to a partial loss of the analytical signal at higher concentrations

due to multiple reflections of fluorescent light from several colored glass filters which were

used to keep the PMT within the linear dynamic range. The relative standard deviation

(RSD) did not exceed 10% over the entire concentration range with a maximum of 9.5% at

the 10 pg/mL level. A slight difference between the same calibration plots (solid lines in

Fig.21a and Fig.21b) is due to the change in experimental conditions for consecutive runs

which were carried out on different days. At concentrations lower than 1 ng/mL,

contamination of deionized water by residual silver caused a curvature on the calibration

plots; the data points corresponding to these concentrations were discarded under the linear


regression data fit. The blank was measured at the laser line detuned from the silver

transition by 0.1 nm. This off-line signal was entirely determined by laser stray light when

the pure water blank solution was atomized, and by non-selective molecular luminescence

and laser scattering when a mixture of deionized water and sea water was used as the blank


The fluorescence signal from the ocean water silver was 30-fold smaller than the last

point made on either of the calibration plots and approximately 4 times higher than the

threefold fluctuation of the background. Special care was taken to avoid procedural

contaminations at such low concentrations. In the course of the experiment, it was found

that laboratory air pollution affects the result of the analysis at concentrations below I

ng/mL. This was due to the adsorption of silver from the air by the graphite tube surface,

activated by multiple firings, during the several seconds after the atomizer chamber top was

removed for consecutive sample dosing. The source of the air contamination was probably

the open flame, located about 1.5 m from the furnace, in which a high concentration silver

solution was occasionally used to set the laser wavelength. A typical signal from a low

concentration aqueous standard is shown in Figure 22a. Peak 1 corresponds to the

fluorescence of silver adsorbed on the furnace surface, and peak 2 to the fluorescence from

the bulk of the sample atomized within the tube and diffused through it with some delay

relative to the surface silver. This delay was caused not by only a spatial separation of the

two sources of silver atoms but also by the differential heating rates of the outer and inner


4 .5 .. .....
S 4.0 a ..... .

3.5 Temperature
n 2.5 1- Fluorescence from silver adsorbed on the furnace surface
) W 2- Fluorescence from silver containing in the sample aliquot
1.5- 2
o i 1 Fluorescence



b .1 : I. .

10 20 30 40
Time, s

Figure 22. Signal profiles vs time. (a) No furnace preheating used: 10-fold attenuated signal
from the aqueous stock solution (10 luL, 1 ppb); (b) Furnace preheating used: I signal from
the stock solution (10-fold attenuation, 10 pL, 1 ppb), II signal from natural sea water (10
gL), II off-line blank signal from sea water (10 uL). Dashed lines show the temperature
of the furnace vs time.


furnace parts (see Fig.19). The two signals could be easily time-resolved without changing

the experimental conditions. However, the uncontrollable, sporadic character of the surface

contamination very often resulted in overlapping of the two peaks, introducing an additional

random error into the result of the analysis. Therefore, a special furnace temperature

program was used over the whole concentration range to separate these peaks completely.

The program included short and rapid (1 s, 800 K/s) heating to the atomization temperature

after the drying and charring steps, short cooling (5 s), and then fast reheating and holding

the furnace under atomization temperature during a time sufficient for the complete

evaporation of the sample (20 s). Figure 22b shows the signal profiles obtained from the

ocean water and the stock solution samples. The dashed line curve corresponds to the

emissive temperature of the furnace measured by the photodiode. One can see that the two

peaks are now completely separated and do not interfere with each other. The loss of the

analyte due to the first, cleaning temperature spike was estimated not to exceed 5% which

was within the instrumental error.

The degradation of the filter with increasing number of firings was controlled by

periodic measurements of the fluorescence from an aqueous standard of 0.1 ig/mL

concentration. This signal remained stable over approximately 50 firings. This number of

firings was taken as a life-time for one filter before it was replaced. A visual examination

of the filter after multiple firings showed an increase in the roughness of the surface and


some defects in the points of electrical contact between the filter and the spring-loaded


The influence of sea water sample volume on the magnitude of the analytical signal

was also studied (Figure 23). As the sea water aliquot was increased, the signal showed a

linear growth up to a volume of 30 uL. After that, a large quantity of steam, probably caused

by the condensation of matrix constituents in the cold zone of the atomizer chamber,

appeared in the analytical volume and masked the fluorescence signal. The chosen sample

volume was 10 uL which provided no on-line preconcentration and a reduced blank signal

relative to that when larger volumes were used. Additionally, using smaller volumes

minimized plugging the graphite filter with excessive matrix material.

Finally, silver concentration in the coastal Atlantic seawater samples were determined

by using the aqueous standards calibration plot. The analytical signal, averaged over ten

measurements, was corrected for the off-line background signal and then multiplied by a

factor of 2 (the depression factor) which introduced a correction for the matrix interference.

The concentration was found to be equal to 146 pg/mL or, in terms of absolute values, 140

60 fg. The results somewhat close to ours for the concentration of silver in different

regions of the Pacific Ocean were obtained by ID-MS following preconcentration using a

dithizone-chloroform extraction method (the concentration range of 0.5 3.2 pg/mL was

reported to be uncertain) [65,82] and by AAS -ETA with a matrix modifier after solvent

extraction and back extraction (5 8 pg/ml, RSD=10%) [66] (Table 5).






Fram \Vau sd
A -23333 24944
B 2 011547
R = 099834
SD= 1.6e2 N=3

10 15 20 25 30 35 40
Sunple volume, uL

Figure 23. The magnitude of the silver fluorescence signal as a function of sea water
sampling volume.


Table 5. Concentration of silver in seawater (pg/ml) obtained in this experiment in
comparison with data from the literature


direct analysis preconcentration by extraction, matrix modifier


(coastal Atlantic) (open Pacific)

14 6 0.5-3.2 5-10


The LODs, obtained by linear extrapolation to 3obla~ levels (blank signals were

measured at the detuned from the silver transition line) were 40 fg (or 4 ppt relative) and 90

fg (or 9 ppt relative) when either pure aqueous or 1:1 mixed sea water standards (which

were considered as real samples) were used for the construction of calibration plots,

respectively. As seen from the literature, the absolute detection limit of 90 fg is among the

lowest detection limits ever obtained for direct sea water analysis of silver.

Analysis of solid reference materials

Three moderately contaminated soil standards (SRM 2709 SRM 2711; San Joaquin

Soil, Montana Soil I and Montana Soil II, respectively) were chosen because they allowed

an excellent opportunity to evaluate the behavior of this atomizer with direct solid sampling

in the presence of a rather complex matrix.

The samples (soils) were highly homogeneous powdered substances with a high

content of silicon (-30%) and organic matrix (-55%). After a calibrated mass of a particular

SRM sample was placed into the sampling hollow, the furnace was heated by using the

following temperature program: drying at 110 oC, 40 s; ashing at 800 oC, 25 s; and

atomizing at 1500 C, 30 s. During the atomization step, silver and silver compounds were

released from the melted matrix, whereas the bulk of the matrix did not evaporate completely

and left a glassy residue bead at the end of the atomization cycle. This bead was

mechanically removed from the sampling hollow, and the furnace was cleaned at a higher

temperature prior to the next sample dosing. An attempt to clean the furnace at very high


temperature (2500 C), without removing the residue, failed. Although the rest of the matrix

was evaporated, the performance of the graphite filter rapidly decreased, probably due to

clogging of the graphite tube pores with matrix constituents. This degradation of the filter

porosity could be deduced from observations of the gradual decrease in the fluorescence

signals from aqueous samples which were injected after the cleaning cycles when the matrix

residues were not removed.

Typical signal profiles obtained from the SRM and the stock solution samples are

shown in Figure 24. It is seen that the two signals are delayed with respect to each other and

the peak corresponding to the solid sample is much broader than that from the stock solution

sample. The reason for such a delay and broadening is evident. In the case of the aqueous

sample, after a drying step, the analyte is uniformly distributed as a very thin layer over the

sampling hollow surface and can be evaporated almost instantly as the temperature of the

surface reaches the required value. When atomizing a solid sample, the temperature of the

bulk of the matrix raises at slower rate compared to a thin layer, and the release of the analyte

from the entire volume of the melted sample droplet is a prolonged process governed by the

diffusion of silver and silver compounds from within the droplet bulk to the droplet surface.

It is worth mentioning in this connection, that the matrix frequently influences the rate of

atomization and causes both the peak height and the peak shape to change, though the peak

area is not affected and depends only on total number of atoms in the analytical zone during

the atomization period [70].



0.5 -

0.0 -

0 20 40 60

Time, s

80 100

Figure 24. Temporal signal profiles: peak 1 aqueous stock solution; peak 2 solid
reference material SRM 2709.

I t

Table 6. Comparison of results (in pg g-') for the determination of silver in soil reference
materials (data are 95 % confidence limits; n=3)

Reference material Certified value Determined value RSD, %

SRM 2709 0.41 0.03 0.35 0.10 9.4

SRM 2710 35.3 1.5 31.0 3.7 5.7

SRM 2711 4.63 : 0.39 4.75 0.25 1.1


The results of the determination of silver in SRM samples are presented in Table 6.

Based on Student's t-test at the 95% confidence level, there were no significant differences

between the measured values and the certified values of the silver concentrations in the

SRM's. The RSD's for the analysis of all SRM's were within 10%. The results indicated that

both the precision and the accuracy of the method were satisfactory compared with the

certified values.


The LEAFS-diffusive tube ETA method has been developed for the direct

determination of femtogram levels of silver in sea water. Insignificant matrix interference,

causing only a 2-fold depression in the analytical signal obtained from the sea water sample

relative to the one from the aqueous standard of the same silver concentration, were corrected

by using a two-plot calibration procedure. No signal depression was observed when

analyzing solid standard reference materials (soils). A good agreement of the results of the

analysis with the certified SRM's values demonstrated the accuracy of the proposed method.

The sensitivity of the method can further be improved by the use of an isothermal

furnace with a graphite filter instead of the non-isothermal one used in the present work.

Also, the use of different types of graphite with different structure and porosity or other

porous refractory materials for on-line partitioning of vapors of complex compositions can

further minimize matrix interference and improve the accuracy of the analysis when no

adequate standards are available.



Plasmas, optically induced in the gas phase, were first described by Maker, Terhune,

and Savage in 1963 [83]. Since then, the majority of studies have been spectroscopic in

nature. Only in the past several years has laser induced breakdown spectroscopy (LIBS) (or

laser ablation -LA) been used extensively as a practical method for the trace analysis of gas,

liquid, and solid samples. Advancements in laser technology over the past three decades has

made powerful lasers, needed for LA, more accessible, reliable, and easier to maintain. The

recent popularity of LA as an analytical technique has generated a considerable number of

publications [84-88]. However, the details of the breakdown process are still unclear and

are still being investigated [89-91].

Breakdown occurs when a powerful laser is focused onto a solid surface. Power

densities of the order of 109 W/cm2 or greater are usually sufficient to cause breakdown in

which high energy plasmas with high electron temperatures (104-105 K) and high electron

densities (10"5-10"1 cm-3) are formed [92]. A phenomenological description of the plasma

and post-plasma evolution is given in [87]; the sequence of processes involved is shown in

Figure 25. When a short laser pulse strikes a surface, the surface temperature instantly

increases past the vaporization temperature for any material. This occurs due to one- or

Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EJ6IMSOCC_FH0U5L INGEST_TIME 2013-01-23T15:56:31Z PACKAGE AA00012973_00001