Glow Discharge OpticalSpectroscopy and MassSpectrometry

Robert E. SteinerLos Alamos National Laboratory, Los Alamos, NM,USA

Christopher M. BarshickOak Ridge National Laboratory, Oak Ridge, TN,USA

Annemie BogaertsUniversity of Antwerp, Antwerp, Belgium

1 Introduction 22 The Glow Discharge 2

2.1 Fundamental Glow Discharge Processes 22.2 Radiofrequency-powered Glow

Discharge Operation 62.3 Pulsed Operation of the Glow

Discharge 72.4 Applications 8

3 Spectrochemical Methods of Analysis 103.1 Basic Requirements Necessary to

Obtain Optical Information 103.2 Atomic Emission Spectroscopy 123.3 Atomic Absorption Spectroscopy 143.4 Atomic Fluorescence Spectroscopy 153.5 Optogalvanic Spectroscopy 17

4 Mass Spectrometric Methods of Analysis 174.1 Basic Requirements Necessary to

Obtain Mass Abundance Information 174.2 Magnetic Sector Mass Analyzers 204.3 Quadrupole Mass Filters 204.4 Ion Trap and Fourier Transform Ion

Cyclotron Resonance Devices 214.5 Time-of-flight Mass Spectrometers 22

5 Conclusions 236 Disclaimer 23

Acknowledgments 23Abbreviations and Acronyms 23Related Articles 24References 24

Optical (atomic absorption spectroscopy, AAS; atomicemission spectroscopy, AES; atomic fluorescence spec-troscopy, AFS; and optogalvanic spectroscopy) and massspectrometric (magnetic sector, quadrupole mass analyzer,QMA; quadrupole ion trap, QIT; Fourier transform ioncyclotron resonance, FTICR; and time-of-flight, TOF)instrumentation are well suited for coupling to the glowdischarge (GD).

The GD is a relatively simple device. A potentialgradient (500–1500 V) is applied between an anode anda cathode. In most cases, the sample is also the cathode.A noble gas (e.g. Ar, Ne, and Xe) is introduced into thedischarge region before power initiation. When a potentialis applied, electrons are accelerated toward the anode.As these electrons accelerate, they collide with gas atoms.A fraction of these collisions are of sufficient energy toremove an electron from a support gas atom, formingan ion. These ions are, in turn, accelerated toward thecathode. These ions impinge on the surface of the cathode,sputtering sample atoms from the surface. Sputtered atomsthat do not redeposit on the surface diffuse into theexcitation/ionization regions of the plasma where theycan undergo excitation and/or ionization via a number ofcollisional processes.

GD sources offer a number of distinct advantages thatmake them well suited for specific types of analyses.These sources afford direct analysis of solid samples,thus minimizing the sample preparation required foranalysis. The nature of the plasma also provides mutuallyexclusive atomization and excitation processes that helpto minimize the matrix effects that plague so manyother elemental techniques. Unfortunately, the GD sourcefunctions optimally in a dry environment, making analysisof solutions more difficult. These sources also sufferfrom difficulties associated with analyzing nonconductingsamples.

In this article, first, the principles of operation of theGD plasma are reviewed, with an emphasis on howthose principles relate to optical spectroscopy and massspectrometry. Basic applications of the GD techniquesare considered next. These include bulk analysis, surfaceanalysis, and the analysis of solution samples. Therequirements necessary to obtain optical informationare addressed following the analytical applications.This section focuses on the instrumentation neededto make optical measurements using the GD as anatomization/excitation source. Finally, mass spectrometricinstrumentation and interfaces are addressed as theypertain to the use of a GD plasma as an ion source.

GD sources provide analytically useful gas-phase speciesfrom solid samples. These sources can be interfaced with avariety of spectroscopic and spectrometric instruments forboth quantitative and qualitative analysis.

Encyclopedia of Analytical Chemistry R.A. Meyers (Ed.)Copyright 2009 John Wiley & Sons Ltd

2 ATOMIC SPECTROSCOPY

1 INTRODUCTION

We have always questioned the physical makeup of ouruniverse, but only recently has science advanced to thepoint that some of the most fundamental questions canbe answered. Although all fields of science play a rolein information gathering, this is the specific charge ofthe analytical chemist. The field of analytical chemistry isbroad, and although no one area can claim to be morevaluable than another, the area of elemental analysis hasalways been critical to our understanding of the chemicaland physical properties of materials.

Many years ago, chemists characterized the world interms of the basic elements of fire, earth, air, and water.Today, we classify our world in a similar manner, butin terms of carbon, nitrogen, oxygen, hydrogen, and soon; all that has really changed are the techniques weuse – techniques that allow us to measure elemental andisotopic composition, elemental speciation, and spatialdistribution of atoms and molecules.

This article focuses on two techniques used in elementalanalysis: GD optical and mass spectrometries. We beginby describing the fundamental operation of a GD,including the sputtering and ionization processes. Next,we examine optical techniques that take advantage of theGD as an atomization source. Finally, we conclude bydescribing GD mass spectrometry, a powerful tool formultielement ultratrace analysis.

2 THE GLOW DISCHARGE

Although not always apparent, aspects of our everydaylives are permeated with technologies originally devel-oped to solve scientific problems. An excellent example isthe GD. Developed in the early twentieth century(1 – 3) asa spectrochemical ionization source used for fundamentalstudies of atomic structure, a variation of the same GDnow illuminates storefronts in the guise of the commonneon sign. Moreover, it is used for the deposition ofcoatings and the fabrication of integrated circuits in themicroelectronics industry, forms the basis of plasma TVs,and finds increasing applications in environmental andbiomedical fields.(4)

The GD is a simple device. Before delving intothe fundamental processes that characterize the glow,however, it is useful to define a few terms. A GD is anexample of a general class of excitation/ionization sourcesknown as ‘‘plasmas’’. This term refers to a partiallyionized gas with equal numbers of positive and negativespecies (ions or electrons) and a larger number of neutralspecies.(5) Often the terms GD and gas discharge areinterchangeable. A gas discharge is formed by passing an

electric current through a gaseous medium.(6) For currentto flow through this medium, a fraction of the gas mustbe ionized. In practice, one applies a potential gradientbetween a cathode and an anode; in doing so, electronsare accelerated toward the anode. As the electrons movethrough the gas, they collide with gas-phase atoms. Afraction of these collisions will be of sufficient energy toremove an electron from an atom, thus producing an ionand a secondary electron. The positively charged gas ionswill, in turn, be accelerated toward and impinge upon thenegatively biased sample cathode. Upon impact a varietyof species will be liberated from the surface. Gas-phasesample atoms are then free to diffuse into the plasmawhere they can undergo excitation and ionization. Thesedynamic processes are depicted in Figure 1.

A GD can be formed in virtually any vacuum cellthat can be equipped with an inlet for a support gas.A variety of geometries have been investigated in thepast. Table 1 lists some of the most common GD ionsources and their characteristics. A very popular sourceis based on a cross-shaped vacuum housing that providesports for spectroscopic viewing, sample introduction, gasintroduction, pressure monitoring, and ion extraction.The most widely used commercial source, mainly foroptical emission spectrometry but recently also formass spectrometry, is the so-called Grimm source(8) andderivations of it.

2.1 Fundamental Glow Discharge Processes

The fundamental processes occurring in the dischargedefine a number of discrete regions. In this article, onlythree regions – the cathode dark space, the negativeglow, and the anode region – are defined, because theother regions, though important, are often not visibledue to limited separation of the cathode and anode.

Sample cathode

Negative glow

Dark space

e−

e−

e−

e−e−

e−

Probe tip

Ar*

Ar*

Mo

Mo

Mo

AroAro

Aro

M+

M+

M+

Ar+

Ar+

Ar+

Sample

Figure 1 Illustration of various steady-state plasma processes.(Reproduced with permission from Ref. 7. John Wiley & Sons,Ltd, 1995.)

GLOW DISCHARGE OPTICAL SPECTROSCOPY AND MASS SPECTROMETRY 3

Table 1 GD ion sources and their characteristics

Voltage (V) Current (mA) Pressure (torr) Advantages Disadvantages

Coaxial cathode 800–1500 1–5 0.1–10 Can conform to various sampleshapes and sizes

Powders must be pressed

Penning ionization dominatedGrimm 500–1000 25–100 1–5 Depth profiling Flat samples only

Compacted powder samples High gas flow ratesHollow cathode 200–500 10–100 0.1–10 High sputter rate Complicated geometry

Intense ion beams for massspectometry

Charge exchange mechanismis important

Large localized atom populationsfor optical spectroscopy

Jet-enhanced 800–1000 25–30 1–5 High sputter rate Flat samples onlyCompacted powder samples High gas flow rates

For a complete treatise on this subject, the reader isdirected to several excellent articles.(9 – 11) The cathodedark space is characterized by low light intensity relativeto other regions of the plasma. This lack of luminosityarises from the absence of collisions and consequentlythe absence of excitation and the radiative relaxationevents that produce photons. The cathode dark spaceis located between the cathode surface and the negativeglow region. The presence of a large positive space chargein the cathode dark space causes the development of apotential gradient. The bulk discharge potential decreasesrapidly through this region, leading to its common name,the cathode fall. This large potential gradient affects theacceleration of electrons that can ionize the discharge gasspecies and liberate secondary electrons in the negativeglow,(5) which help to sustain the plasma. Radiativerelaxation of species excited in the negative glow regionyields the characteristic emission for which it is named.There is nearly charge-neutrality (equal positive andnegative charges) in this region, but the major chargecarriers in this region are electrons, because of their highermobility. As electrons collisionally cool, they slow down,decreasing their cross section for excitational collisionswith atoms. This smaller excitation cross section can resultin the Faraday dark space close to the anode, but whenthe distance between the cathode and anode is small,like in most GDs used for analytical applications, theFaraday dark space is not visible, and there is only asmall anode region, where the potential returns to zero.Figure 2 schematically shows these three regions, as wellas the potential and electric field distributions, positiveion and electron densities, and electron impact excitationrate (which corresponds roughly to the light intensity) inthese regions, as obtained from model calculations.(12)

2.1.1 The Sputtering Process

The GD is of particular utility when analyzing solidconducting samples. Most competing techniques used for

Negative glow

AnoderegionCathode

darkspace

−1000

−800

−600

−400

−200

0

V (

V)

−10

−8

−6

−4

−2

0

E (

kV c

m−1

)

0×1001×10112×10113×10114×10115×1011

n (c

m−3

)

0 0.2 0.4 0.6 0.8 1

z (cm)

0×100

1×1016

2×1016

3×1016

4×1016

Re-

exc

(cm

−3 s

−1)

Ar+e−

− +

(a)

(b)

(c)

(d)

Figure 2 Schematic diagram of the different spatial regions ina glow discharge (a), the corresponding potential and electricfield distributions (b), electron and Ar+ ion number densityprofiles (c), and electron impact excitation rate, which givesan idea about the luminosity (d), as obtained from numericalmodeling at 1000 V, 75 Pa, and 3 mA.

elemental applications (i.e. inductively coupled plasmamass spectrometry (ICPMS), flame spectroscopies, andgraphite furnace techniques) require a sample in solutionform to facilitate aspiration or introduction into the

4 ATOMIC SPECTROSCOPY

ionization source. The GD source possesses the inherentcharacteristic of producing gas-phase analyte atomsdirectly from the solid conducting sample material. Thisphenomenon, known simply as cathodic sputtering, can bemost easily described using a basic billiard ball analogy.Positively charged discharge gas ions are acceleratedtoward the negatively biased sample cathode. Beforeimpact, these high-energy ions recombine with Augerelectrons released from the cathode surface. The resultinghigh-energy neutral species impact the surface of thecathode, transforming their kinetic energy (KE) into thelattice of the sample, thus causing a cascade of collisionalevents, much like the breaking of a racked set of billiardballs. If the resulting energy transfer is sufficient toovercome a surface atom’s binding energy, the atom willbe released into the gas phase. The sputtering processliberates not only individual cathodic atoms but alsoelectrons, ions, and clusters of atoms and molecules. Thisprocess is illustrated in Figure 3.

Emitted electrons are accelerated across the cathodedark space into the negative glow where they cancontribute to excitation and ionization of gas-phaseatoms. The ions formed by the sputtering event do nottravel far from the cathode but are returned to the surfaceby the effects of the electric field. Once in the gas phase,the analyte atoms and neutral clusters are free to undergocollisions that may dissociate clusters and redeposit atoms

Deflected species

Incident ion

+ Sputtered atom

e−

Sample lattice

Figure 3 Illustration of cathodic sputtering process.

at the surface. A fraction of these neutrals diffuse into thenegative glow where they undergo excitation/ionization.

The effect of an ion’s impact on the sample latticecan be measured by the sputter yield, S, as shown inEquation (1)(13):

S = (9.6 · 104)(

W

Mi+t

)(1)

where W is the measured weight loss of the sample ingrams, M is the atomic weight of the sample, i+ is theion current in amperes, and t is the sputtering time inseconds. The ion current is related to the total current, i,by Equation (2):

i+ = i1 + γ (2)

where γ is the number of secondary electrons released,on average, by a single ion.

Much of the previous research involving sputtering hasused secondary ion mass spectrometry (SIMS). Typically,ion beams generated from these types of sources are moretightly focused and of much higher energy than in GDs.This is important to note since there are a number ofcharacteristics that affect the sputter yield for a particularsystem including the nature of the target, the nature ofthe incident species, the energy of the incident ions, andthe angle of the incident ion beam. An empirical formulathat is widely used for sputter yield calculations, andwhich describes the influence of incident ion energy andnature of incident species and target, is the Matsunamiformula.(14)

2.1.2 Excitation/Ionization Processes

Although the atomization or sputtering process createsspecies essential for atomic absorption and fluorescencespectroscopies, it does not supply the excited species andions needed for atomic emission and mass spectrometricanalyses. These excitation/ionization processes occur inthe collision-rich environment of the negative glow.Collisions that occur within this region not onlyprovide analytically useful species but are also integralin maintaining the stability of the plasma. Figure 4illustrates the three principal types of collisions thatoccur within the negative glow for the sputtered atoms,involving electrons, excited atoms, and ions. Excitationis dominated by electron impact (similar to Figure 4a),while ionization is governed by electron impact ionization,Penning ionization, and asymmetric charge transfer(Figure 4). In many cases, especially at low pressure andvoltage, Penning ionization accounts for the majority ofionization of the sputtered atoms.(7,15,16)

GLOW DISCHARGE OPTICAL SPECTROSCOPY AND MASS SPECTROMETRY 5

e− e−e−

e−

Mo

Mo

M0

Ar0

Aro

Arm

M+

M+

M+Ar+

(a)

(b)

(c)

Figure 4 Illustration of the three major ionization pathwaysfor sputtered atoms available in the GD plasma: (a) electronionization, (b) Penning ionization, and (c) asymmetric chargetransfer.

Electron ionization involves collision of an ener-getic electron with a gas-phase atom. Although it isof minor importance for the sputtered atoms,(15,16) it isthe dominant ionization mechanism in the plasma, morespecifically for the discharge gas atoms. If the electron isof sufficient energy, it can interact with an electron in thevalence shell of the atom, transfer enough energy to ejectthe electron from the atom’s electron cloud, and form anion and a secondary electron.(17,18) There is only a smallprobability that such a collision will result in ionization;this is zero below a threshold level, increasing at a rateof C1.127, where C is the excitational cross section, asthe electron energy increases.(19) This can be explainedusing classical collision theory. At the threshold, only thecomplete transfer of energy will result in ionization. Withincrease in energy of the electrons, collisions with onlypartial transfer of energy will result in ionization and thecross section will increase. At high electron energies, thetime and wavefunction overlap is too short for ionizationto occur, and the cross section begins to decrease.(6,19) Theaverage energy of electrons found within the GD is notadequat enough to ionize most elements or the dischargegas.(5,20,21) However, a Boltzmann distribution of energiespredicts the presence of a small percentage of electronswith sufficient energy to ionize all elements. Chapman(5)

has performed calculations of the Maxwell–Boltzmanndistribution of electrons to determine the percentage ofelectrons with energies above 15.76 eV, the first ionizationpotential of argon, the most common discharge gas. At anaverage energy of 2 eV, 0.13% of the electrons present areof sufficient energy to ionize argon. At an average elec-tron energy of 4 eV, the percentage increases to 5.1%. Amore thorough presentation of these calculations is givenby Chapman.(5)

Penning ionization, named after F.M. Penning whodiscovered the effect in 1925,(22,23) involves the transfer

of potential energy from a metastable discharge gas atomto another atom or molecule. If the first ionizationpotential of an atom or molecule is lower than theenergy of the metastable atom, ionization will occurwhen they collide. Ionization cross sections for mostelements are similar for the Penning process, resulting inslightly uniform ionization efficiencies. Metastable statesare reached through either the activation of a dischargegas atom to an excited state (typically by electronimpact) from which radiative decay is forbidden or bythe radiative recombination of discharge gas ions withthermal electrons. For argon, the most common dischargegas, the metastable levels are the 3P2 and

3P0 states withenergy levels of 11.55 and 11.72 eV, respectively. Table 2lists the metastable levels and energies of the dischargegases most commonly used to support GD plasmas.Metastable species are relatively long lived, existing formilliseconds under normal plasma conditions.(24) Theirlongevity within the plasma, along with a relatively largeionization cross section and energy sufficient to ionizemost elements, makes this process a major contributor toion production. Investigations of ionization processes insteady-state DC-powered discharges have indicated that40–80% of ionization occurring within these plasmas canbe attributed to the Penning ionization process.(15,16,25,26)

Penning ionization also affords the unique advantageof discriminating against the ionization of dischargecontaminants whose ionization potentials are greater thanthe metastable energy of the discharge gas atoms.

The third important ionization mechanism for thesputtered atoms is asymmetric (or nonresonant) chargetransfer with Ar+ ions. Note that resonant chargeexchange involves transfer of charge from one ion toan atom of the same species, whereas nonresonant chargeexchange involves the transfer of charge from an ionto an atom of a different species. The first process isimportant in the GD as energy loss mechanism for theAr+ ions on their way toward the cathode, and at thesame time as production process for fast Ar atoms, which

Table 2 Metastable spectroscopic notations, energies, andfirst ionization potentials for common discharge support gases

GD support gas Spectroscopicnotation

Metastableenergy (eV)

First ionizationpotential (eV)

Helium (He) 21S 20.6 24.523S 19.8

Neon (Ne) 1P0 16.7 21.616.6

Argon (Ar) 3P0 11.7 15.811.5

Krypton (Kr) 3P0 10.5 14.09.9

Xenon (Xe) 3P0 9.4 12.13P2 8.3

6 ATOMIC SPECTROSCOPY

can then also contribute to sputtering at the cathode.The second process, on the other hand, results in theformation of ions of the sputtered material. However,it is a very selective process and occurs only if there issufficient overlap between the (excited) energy levels ofthe initial ion and the created ion. Hence, it is a moreselective process than Penning ionization and electronimpact ionization. Therefore, it is strongly suggested thatvariations in relative sensitivity factors (RSFs) in GDMSare attributed to the occurrence or absence of asymmetriccharge transfer as ionization mechanism for the sputteredatoms.(27,28)

There are a number of other processes that play minorroles in ionization within the GD plasma. These processesinvolve fast Ar+ ion and fast Ar atom impact ionizationor associative ionization. Fast Ar+ ion and Ar atomimpact ionization are only important near the cathode,where the Ar+ ions and Ar atoms have gained enoughenergy from the electric field for giving rise to ionization.Integrated over the entire discharge region, they accountonly for a few percent of the total ionization of Ar gasatoms, but this minor contribution can still be importantfor sustaining the discharge.(16,29) Associative ionizationinvolves the combination of two metastable species, or thecombination of a higher excited species with a gas-phaseatom (i.e. the so-called Hornbeck–Molnar associativeionization), to form a molecular ion with the liberationof a secondary electron. This process contributes onlymarginally to ionization within the plasma, but it is theprincipal mechanism by which interfering metal argideions are formed.(5)

2.2 Radiofrequency-powered Glow DischargeOperation

Traditionally, most GD devices have used a steady-stateDC-powered source. However, about two decades ago,the utility of radiofrequency (RF)-powered discharges assources for atomic spectroscopies and mass spectrometryhas been investigated.(30,31) A major advantage offered bythe RF-powered plasma is the ability to directly analyzenonconducting samples such as ceramics and glasses.(32)

In the past, these types of samples were powdered, mixedwith a conductive matrix, and pressed into a pin or disk toallow analysis by the DC GD.(33 – 35) Utilization of an RFplasma allows the sputtering and subsequent analysis ofthese sample types without this added sample preparationstep.

Current cannot propagate through an insulatingmaterial; therefore, when subjected to an applied DCpotential, an insulator behaves in a manner similar tothat of a capacitor and begins to accumulate charge.With the application of a negative voltage, the surfacepotential of the insulating material decays to a more

t

t

DCoffset

+1

−1

0 t 2t 3t

(a)

(b)

Va

(kV

)V

b (k

V)

Figure 5 Electrode response to an applied square wavepotential: (a) Va, applied voltage and (b) Vb, response voltage.(Reproduced with permission from Ref. 5. John Wiley &Sons, Ltd, 1980.)

positive potential with time. This decay can be attributedto charge neutralization at the cathode surface. The DCdischarge will sustain itself until its threshold voltageis reached and the plasma is extinguished. Applicationof a high-frequency potential to a conductive materialadjacent to the insulator will allow positive charges thataccumulate at the nonconductor surface to be neutralizedby electrons during the positively biased portion ofthe cycle. Therefore, application of a high-frequencypotential will allow a negative potential to be maintainedon the nonconductor’s surface.

Figure 5 illustrates the phenomenon known as cathodeself-biasing that allows the discharge to be maintainedfor long periods of time.(5) Self-biasing is based uponthe mobility difference between ions and electrons. Asa negative potential is initially applied, the surface getscharged quickly, reaches a maximum, and begins to decaydue to the bombardment of positive ions (Figure 5b).When the potential is switched, electrons are acceleratedand bombard the surface much like the positive ionsduring the negatively biased portion of the cycle. Theelectrons, however, have a greater mobility than the moremassive positive ions; therefore, the surface potentialdecays more quickly. After a number of cycles, thewaveform will reach a steady negative DC offset. Intypical GDs used for analytical applications, where a largedifference in size exists between powered RF electrode

GLOW DISCHARGE OPTICAL SPECTROSCOPY AND MASS SPECTROMETRY 7

(i.e. sample) and grounded electrode (i.e. cell housing),this DC offset potential is approximately one half of theapplied peak-to-peak voltage and sustains the sputteringion current.(36,37) Indeed, because of this negative DCoffset, the sample will be continuously bombarded bypositive ions, but this charging will be compensated bythe bombardment of a high flux of electrons duringa small portion of the RF cycle.(36) Operation in thismode has proven useful for the analysis of materialssuch as nonconducting alloys, oxide powders, and glasssamples.(38)

2.3 Pulsed Operation of the Glow Discharge

The GD plasma can be operated in a modulated powermode as well as in the steady-state mode describedpreviously. Pulsed power operation offers some distinctanalytical advantages that make it attractive to theanalyst.(39 – 46) Steady-state discharges are power limitedbecause increased power application results in resistiveheating of the cathode, eventually jeopardizing sampleintegrity.(47) Modulation of the applied discharge powerpermits operation at higher instantaneous power whilekeeping the average power at an acceptable level. Thisoperation mode serves to increase the sputter yield byincreasing the average energy and/or the number ofincident ions, leading to higher signal intensities andhence better analytical sensitivities,(48) while allowing thesample to cool during the off portion of the dischargecycle. Modulated operation also provides temporalsegregation of discharge processes (see below).

Modulated GD operation relies on a microsecond tomillisecond square wave power pulse, with a duty cycleof 10–50%. These parameters allow sufficient time forcooling and for the removal of species from one pulsedevent before the next one is initiated. Figure 6 depictsa typical pulse sequence showing both discharge gas(Ar) and analyte ion signal profiles. It is apparent thatsignal behaviors for discharge gas and analyte speciesdiffer significantly. Upon power initiation, the dischargegas ion signal exhibits a sharp rise in intensity to amaximum(41,49) (Figure 6a). This ‘‘prepeak’’ results fromthe electrical breakdown of the discharge gas speciesupon power application. The short delay occurs becausethe acceleration of electrons and subsequent electronionization of the discharge gas are not instantaneousprocesses. In contrast, the analyte ion profile behavesdifferently (Figure 6c). A much longer delay occurs beforea more gradual signal increase is observed. This delayarises because sample atoms must first be sputtered fromthe cathode surface and diffuse into the negative glowregion before they can undergo ionization. The temporalcorrelation between the observation of discharge gas ionsand the appearance of analyte signal can be seen in the

figure. Both signals reach equilibrium conditions abouthalf way through the applied power pulse in the ‘‘plateau’’region. During this time regime, the plasma most closelyapproximates the behavior of a steady-state plasma.

Upon applied power termination, the two ion profilesagain show markedly different behavior. The dischargegas ion profile decreases quickly (≈500 µs) upon powertermination. Previous studies have explained this signaldecay as the rapid recombination of discharge gas ionswith thermal electrons to form metastable discharge gasatoms.(55,56) The analyte ion profile quickly increasesduring the ‘‘afterpeak’’ region, reaching a maximumshortly after pulse power termination. This signalthen decreases gradually to the baseline. Figure 6(b)represents the first derivative of the discharge gas ionsignal profile. The maximum and minima observed in thisportion of this figure represent the temporal locationof the respective maximum signal intensity increasesand decreases for the discharge gas signal. Interestingly,the maximum decrease in the discharge gas ion signalcorresponds to the afterpeak maximum for the analytespecies.(47) This correlation supports the theory thatat power termination electrons are collisionally cooled

5 ms

(a)

(b)

(c)

Int.

Int.

d(In

t.)A

/dt

Discharge power

Time

40Ar+

63Cu+

Figure 6 Temporal ion signal profiles: (a) discharge gasions, 40Ar+; (b) the first derivative of the 40Ar+ profile; and(c) sputtered analyte ions, 63Cu+. (Reproduced by permissionfrom King and Pan.(39) Copyright (1993) American ChemicalSociety.)

8 ATOMIC SPECTROSCOPY

and recombined with argon ions, forming metastablespecies. This increase in metastable argon atoms increasesthe probability of ionization via the Penning process,thus providing enhancements in the analyte ion signal.Nowadays, this explanation is generally accepted.(50 – 52)

However, it is not completely clear which recombinationprocesses are responsible for the formation of metastableatoms.(53,54) Indeed, on the basis of calculated electronand Ar+ ion densities and recombination rate coefficientsavailable from literature, recombination of Ar+ ionsand electrons seems not to be efficient enough. It issuggested(54) that dissociative recombination of Ar2

+ ionswith electrons might be the most efficient recombinationprocess in the afterglow, but this hypothesis needs furtherinvestigations.

Temporally gated separation and detection of speciesfound in these distinct plasma regions increase theutility of the GD devices.(47,57) The most analyticallyuseful region is the afterpeak. Data acquired within thistime regime offer two advantageous characteristics. Asillustrated in Figure 6(c), the signal intensity for theanalyte species increases substantially, thus potentiallyenhancing the sensitivity for analyte species in thisregion. The other advantage arises from the suppressionof electron-ionized interfering species.(40,41) The firstionization energies of the discharge gas and molecularcontaminants (e.g. H2O, N2, O2) are too high to allowionization through the Penning process. Upon powertermination, species that are ionized via electron impactare no longer excited. Thus, if the acquisition gate ismoved far enough into the afterpeak, contributions fromthese species will be minimized.

Finally, some other distinct advantages of pulseddischarges are the low overall sputter rates (in combi-nation with high transient sputter rates and hence highsignal intensities) opening possibilities in the field ofthin film analysis(58,59) (in fact, a wide dynamic rangeof layers, from several nanometers to tens of microm-eters thick, can be analyzed(60)), as well as the abilityto obtain structural, molecular, and elemental informa-tion of samples.(61,62) Indeed, depending on the extent ofinteraction with the plasma, the samples may undergo softchemical ionization, yielding molecular ions or they maybe completely atomized and ionized, yielding elementalinformation.

2.4 Applications

In the early 1970s, as GD instrumentation and tech-niques were being transitioned from research applicationsto routine sample analysis, spark source techniquesreigned as the analytical tool of choice for trace anal-ysis of solids. Spark source spectrometry was limited,however, by its expense, complexity, and unreliability.

These disadvantages associated with spark source spec-trometry facilitated the acceptance of the GD for theanalysis of solid samples.

2.4.1 Bulk Analysis

Historically, GD mass spectrometry and optical spec-troscopies have proven most valuable for the analysisof bulk conductive solid samples. As explained earlier,GD methods provide a representative gas-phase analytepopulation directly from samples in the solid form, thussignificantly simplifying sample preparation. GD tech-niques can also be used for the analysis of nonconductingsamples such as glasses, polymers, and ceramics.(63 – 65)

The first analyses of nonconducting materials involvedmixing a powdered sample with an easily sputtered,conducting matrix, such as high purity copper or silverpowder.(66 – 69) The mixture was pressed into a samplecathode using a die and hydraulic press system. Unfor-tunately, this complicated sample preparation, especiallyfor samples that were not in powdered form. Homo-geneity of the mixed sample also became a matter ofconcern for the analyst. Two other methods for theanalysis of nonconducting samples have also been used.These methods do not require sample mixing but ratheruse an RF-powered or a secondary cathode GD source todirectly sample solid nonconducting materials. Section 2.2describes plasmas powered by RF sources. Secondary, orsacrificial, cathode GD systems utilize a monoisotopicconductive mask (e.g. Ta, Pt) that is placed on top ofthe sample to be analyzed. A potential is applied to themask, which thus assumes the role of the cathode. Asmaterial is sputtered from the mask, a large portion ofit is redeposited on the surface of both the secondarycathode and the nonconducting sample. At this point,the layer of cathode material deposited onto the samplebecomes conductive and thus assumes the applied poten-tial and attracts impinging ions. As ions impinge uponthe cathodic layer, they ablate both the cathodic mate-rial and the underlying nonconductor, introducing bothspecies into the gas phase for subsequent analysis. It isclear that the mask material must not contain speciesthat are of interest because this will contaminate thesample and preclude accurate measurements. A numberof references are available describing in detail the use ofsecondary cathodes.(70 – 72)

2.4.2 Surface and Depth Profiling Analysis

A few decades ago, surface and depth profiling analysisby GD spectrometry has aroused great interest in theanalytical community. In a sense, GD is always a surfaceanalysis technique, acting as an atomic mill to erodethe sample surface via the sputtering process. Atoms

GLOW DISCHARGE OPTICAL SPECTROSCOPY AND MASS SPECTROMETRY 9

sputtered from the surface are subsequently measuredusing either optical or mass spectrometric techniques.GD sputtering consumes relatively large quantities ofthe sample in a relatively short time period (up tomilligrams per minute). This makes analysis of thicklayers (>10 µm) possible, but on the other hand, thinfilms of less than 10 nm thickness can be analyzedas well, by proper choice of plasma conditions.(73 – 75)

Figure 7 is a graphical illustration of a typical spectralintensity–time profile of a multilayer coating preparedby chemical vapor deposition of carbon steel. The outerlayer consists of vanadium carbide and the inner layerchromium carbide. It is obvious from Figure 7 thatsignal responses do not follow an ideal square wavepattern for appearance and disappearance. Clearly, themetal sample layer is not completely eroded at a well-defined time, but is scattered across a diffuse regionthat is gradually removed. This signal tailing can bea manifestation of nonflat crater profiles,(76,77) or itcan be due to the redeposition that occurs from theplasma. It has been estimated(76 – 78) that up to 67%of the sputtered atom population is returned to thesurface by collisions with argon atoms, to be resputteredbefore eventually escaping permanently from the surface.By calibrating erosion rates using standard layeredsamples, the thickness of layers can be determinedby analyzing the signal–time profiles, allowing thefull characterization of sample layers. Nowadays, GDs,mostly in the Grimm-type confirmation, are widely usedfor depth profiling and thin film analysis of variousmaterials.(73 – 75,79 – 82)

2.4.3 Analysis of Solution and Gaseous Samples

Analysis of solutions and gaseous samples has beenperformed using GD techniques, even though their major

Ligh

t int

ensi

ty (

arb.

uni

ts)

Sputtering time

Vanadium carbide Chromium carbide Steel

Fe

V C

VgCr

Figure 7 Qualitative analysis of multilayer coating. Vg,voltage. (Reproduced with permission from Ref. 79. PlenumPress, 1993.)

advantage lies in the ability to directly analyze solid-state samples. It is important to note that aspiration ofa solution directly into the GD will cause quenching ofthe plasma, making the removal of any solvent fromthe sample preferable. The simplest and most directmethod of doing this is to evaporate a solution ontoa conducting cathode, leaving a dried residue as thesample. A more elegant method for depositing solutionsamples onto a cathode involves electrodeposition. Thisapproach also permits analytes in large sample volumes tobe preconcentrated before analysis. Solutions containingsamples have also been mixed with a conductive powderand dried for subsequent analysis. This approach is similarto that used for the analysis of nonconductive powders.

Although limited by its difficulty, direct analysis ofsolution samples by GD techniques has been performedfor specialized applications. Strange and Marcus(83) haveused a particle injection system to introduce a solutionsample into the GD. Steiner et al.(84) have also utilizeda pulsed plasma TOF system to obtain concurrentelemental and molecular information for high vaporpressure liquid samples.

In recent years, there is an increasing interest in novelapplications for liquids and gaseous analysis, catalyzed bythe development of new types of GD plasma sources,often working at atmospheric pressure and/or withreduced dimensions. With respect to liquids analysis,the so-called electrolyte cathode atmospheric glowdischarge (ELCAD) was developed in 1992 by Czerfalviand Mezei.(85) An extensive review of this system,including its main operating parameters, mechanisms,and analytical performance, was recently published.(86) Inrecent years, Hieftje and coworkers have adopted (andmodified) this concept of the ELCAD, among others,for exploring the analysis of complex samples.(87,88)

Furthermore, a new design of a solution-cathode GD,with a smaller volume and a corresponding increase inpower density compared to the ELCAD-like designs,was also developed,(89) as well as another type of small-scale plasma, the so-called annual GD,(90) where aerosol isintroduced in the plasma through the cathode, and atomicemission is observed in the near-cathode region. Marcusand coworkers also developed several types of GDsources for liquids analysis, such as the liquid sampling-atmospheric pressure glow discharge (LS-APGD)(91) andthe particle beam (PB) sample introduction systemof liquid samples, to be used in combination withhollow cathode optical emission spectrometry (PB/HC-OES) and with GDMS (PB/GDMS).(92) For gaseousanalysis, a novel versatile DC atmospheric pressureGD (APGD) in He, interfaced with a TOFMS system,was developed by Hieftje and coworkers.(93) When thisdetection system is coupled with hydride generation, theanalytical performance of the hydride generation APGD

10 ATOMIC SPECTROSCOPY

is comparable to that of an ICP source. Detection downto 10 ppt could be realized.(93) Finally, some more GDsources were recently used for the analysis of gasesor vapors. Newman and Mason(94) introduced organicvapors into the flowing afterglow of a lower powerDC GD coupled to a quadrupole mass spectrometer,and Fliegel et al. evaluated the use of a pulsed GD-TOFMS system as a detector for gas chromatographyanalysis.(95)

3 SPECTROCHEMICAL METHODS OFANALYSIS

The goal of an analytical chemist is often to identify aparticular chemical species or to quantify the amount ofthat species in a sample. In a spectrochemical analysis,the chemist uses the intensity of radiation emitted,absorbed, or scattered by a particular species versus aquantity related to photon energy, such as wavelength orfrequency, to make such measurements.(96)

In this section, the basic requirements necessary toobtain a spectrochemical analysis are reviewed, alongwith three spectrochemical methods that are used toeffect the measurement: AAS, AES, and luminescencespectroscopy. Although there are many variations ofeach of these methods, we review only the classicalapproaches and the variations that have been used withGD devices. Table 3 summarizes the quantity measuredand gives examples associated with each measurementtechnique.(96)

Table 3 Classification of spectrochemical methods(96)

Class Quantity measured Examples

Emission Radiant power ofemission, 8E

Flame emission,DC arc emission,sparkemission,inductively coupledplasma and directcurrent plasmaemission,GDemission

Absorption Absorbance or ratioof radiant powertransmitted to thatincident,A = log(�/�0)

UV/visible molecularabsorption, IRabsorption, atomicabsorption

Luminescence Radiant power ofluminescence, �L

Molecular fluorescenceand phosphorescence,atomic fluorescence,chemiluminescence

UV, ultraviolet; IR, infrared.

3.1 Basic Requirements Necessary to Obtain OpticalInformation

For thousands of years, scientists have been performingqualitative analyses based on color, smell, taste, size,and shape. Although first-year college chemistry studentsare still taught to use their senses to help identifysubstances in qualitative laboratory, for the rest of usthese less-precise approaches have been replaced withchemical and instrumental methods that can measurenot only pure materials but also trace componentsin complex mixtures. Most materials do not give thisinformation spontaneously, however. Instead, to obtainchemical information about a sample, it is necessaryto perturb the sample through the application ofenergy in the form of heat, radiation, electrical energy,particles, or a chemical reaction.(96) The applicationof this energy often causes electrons in analytesto be excited from their lowest energy or groundstate to a higher energy or excited state. Severalspectroscopic phenomena depend on these transitionsbetween electronic energy states.(96) Information can beobtained by measuring the electromagnetic radiationemitted as the electron returns to its ground statefrom an excited state (emission), by measuring theamount of radiation absorbed in the excitation process(absorption), or by measuring the changes in the opticalproperties of the electromagnetic radiation that occurwhen it interacts with the analyte (e.g. ionizationor photochemical reactions).(55) Qualitative informationis extracted by observing a particular elementspecifictransition, whereas quantitative information is obtainedby quantifying the amount of radiation emitted orabsorbed in that transition. Figure 8 is an illustration usedby Ingle and Crouch(96) to depict the many processesinvolved in converting concentration information intoa number – the analytical chemist’s goal in making ameasurement. A sample introduction system presentsthe sample to the encoding system, which converts theconcentrations c1, c2, and c3 into optical signals O1, O2,and O3. The GD is slightly unusual (compared withother atomic sources such as the flame or the inductivelycoupled plasma) because it serves as both the sampleintroduction system (through sputter atomization) andthe spectrochemical encoder. The information selectionsystem (often a monochromator) selects the desiredoptical signal O1 for presentation to the radiationtransducer or photodetector. This device converts theoptical signal into an electrical signal that is processedand read out as a number. All spectrochemical techniquesthat operate in the UV/visible and IR regions of thespectrum employ similar instrumentation(96); the onlydifferences lie in the arrangement and type of sampleintroduction system, encoding systems, and informationselection system.

GLOW DISCHARGE OPTICAL SPECTROSCOPY AND MASS SPECTROMETRY 11

Spectrochemicalencoder

Sampleintroduction

system

c ′1, c ′2, c ′3

Numbera c1

i1, e1, f1

O1, O2, O3 O1

Controlsystem

Signalprocessing andreadout system

Opticalinformation

selector

Radiationtransducer

Analyticalsample

c1, c2, c3

Figure 8 Spectrochemical measurement process. (Repro-duced with permission from Ref. 96. Prentice Hall, 1988.)

3.1.1 Spectrometers

The optical information selector in Figure 8 sorts thedesired optical signal from the many signals produced inthe encoding process. Although it is possible to discrimi-nate against background signals on the basis of time andposition, most often discrimination is based on opticalfrequency (wavelength). The most widely used wave-length selection system is the monochromator, althoughthere are a variety of other systems, including polychro-mators and spectrographs, and nondispersive systemssuch as the Fabry–Perot, Michelson, Mach–Zender,and Sagnac interferometers. Our discussion is limited tomonochromators; for a thorough discussion of other typesof wavelength selection systems, the reader is directedelsewhere.(96)

Monochromators isolate one wavelength from thecountless number of wavelengths found in polychromaticsources. A monochromator consists of two principalcomponents: a dispersive element and an image transfersystem. Light is transferred from an entrance slit to anexit slit by a series of mirrors and lenses; along the wayit is dispersed into its various wavelengths by a grating(or sometimes a prism). To change wavelengths, onerotates the dispersive element, which results in differentwavelength bands being brought through the exit slit insuccession. One can easily calculate the angular dispersion(Da) (i.e. the angular separation (dβ) corresponding tothe wavelength separation (dλ)) of a grating by knowingthe angle of incidence, the angle of diffraction, the orderof diffraction, and the groove spacing of the grating.For practical purposes, however, it is more importantto calculate the linear dispersion, Dl = dx/dλ (a valuethat defines how far apart in distance two wavelengthsare separated in the focal plane), or the reciprocallinear dispersion, Rd (the number of wavelength intervalscontained in each interval distance along the focal plane).

A monochromator’s resolution is closely related toits dispersion in that dispersion determines how farapart two wavelengths are separated linearly while aninstrument’s resolution determines whether the twowavelengths can be distinguished. In many cases, theresolutions determined by the monochromator’s spectralbandpass, s, defined as the half-width of the wavelengthdistribution passed by the exit slit. If the slit width is largeenough to ignore aberrations and diffraction, a scan of twoclosely spaced monochromatic lines of peak wavelengthsλ1 and λ2 will be just separated (baseline resolution) ifλ2 − λ1 = 2s. Therefore, the slit-width-limited resolution�λs is given by Equation (3):

�λs = 2RdW (3)

where W is the slit width.(96) By adjusting the monochro-mator so that �λ0 = �λ1 and W = �λ/Rd, the image ofλ1 will be completely passed, while that of λ2 will be atone side of the exit slit.(96)

3.1.2 Detectors

Two detectors are most often used in atomic spectroscopy:photomultipliers (the most common) and multichanneldetectors. A photomultiplier is a more sophisticatedversion of a vacuum phototube. A cascade of electroncollisions with dynodes of increasing potential and thesubsequent ejection of electrons from each dynode’ssurface leads to the formation of an electrical currentproportional to the number of photons striking thedetector. The process begins with a photon striking acathode made of a photoemissive material (e.g. alkalimetal oxides, AgOCs, CS3Sb). If the energy of the photonis above some threshold value, an electron is ejected fromthe cathode. Only a certain fraction of the photons withenergy greater than threshold produce photoelectronswith sufficient KE to escape the photocathode. Thisfraction is called the quantum efficiency and is the ratioof the number of photoelectrons ejected to the numberof incident photons. After leaving the photocathode, aphotoelectron strikes the first dynode of the multiplier;this causes the subsequent ejection of two to fivesecondary electrons, which in turn are accelerated byan electric potential to a second dynode where theycause the release of two to five more electrons. Thismultiplication process continues until the electrons reachthe last dynode and impinge on the anode. A modernphotomultiplier tube might have 5–15 dynodes (made ofa secondary emission material such as MgO or GaP) in acascade. The result of this photomultiplication is that foreach photoelectron collected by the first dynode, a largenumber of charges is produced at the anode, within a fewnanoseconds. Photomultipliers can be operated either

12 ATOMIC SPECTROSCOPY

in analog mode, where the average current that resultsfrom the arrival of many anodic pulses is measured, orin photon counting mode, where the number of anodicpulses, and not photons, is counted per unit time.

A wide variety of photomultipliers are available withboth end-on and side-on viewing for adaptation to a widevariety of monochromators. Care must be given to thewavelength range over which one is working to ensure auniform response. Other concerns for the spectroscopist,all of which are beyond the scope of this article,include the quantum efficiency of the photomultiplier,the multiplication factor of each dynode, the operating(accelerating) voltages applied to the dynodes, and thedark current generated when a potential is appliedbetween the anode and cathode, with no photons hittingthe photocathode.

The second type of detector that is widely usedin atomic spectroscopy is the multichannel detector.These devices include early photographic detectors suchas photographic film or plates, as well as moderndetectors such as photodiode arrays and charge-coupledand charge-injection devices. The idea behind themultichannel detector is simultaneous detection ofdispersed radiation. Modern multichannel detectorsusually take the form of some sort of solid-state pn-junction diode device packaged in integrated-circuit formwith a large number (e.g. 256, 512, or 1024) of elementsarranged in a linear manner. These devices often havelinear dynamic ranges of 2–4 orders of magnitude.Limitations at the low end result from the noise associatedwith readout of a given dynode. Limitations at the upperend are the result of saturation; this is determined bythe number of electron–hole pairs that can be created.A typical saturation charge is 1–10 pC. The readeris referred to several excellent references for a morethorough description of multichannel detectors.(97 – 99)

3.2 Atomic Emission Spectroscopy

AES is the simplest spectroscopic method for determiningthe elemental composition of a sample, and is the logicalstarting point for a discussion on atomic spectrometry.Optical emission results from electron transitions occur-ring within the outer electron shells of atoms. Thesetransitions give rise to line spectra where the wavelengthof the lines relates to the energy difference of the levelsaccording to Equation (4):

�E = hcλ

(4)

Spectroscopists often categorize spectral transitionsaccording to term symbols. For a complete discussionof term symbols, the reader is referred elsewhere(100); abrief discussion follows.

Each electronic state of an atom has five quantumnumbers that define its electronic configuration. Theseinclude the principal quantum number, n; the orbitalangular momentum quantum number, l; the orbitalmagnetic quantum number, mi; the electron spin quan-tum number, s; and the spin magnetic quantum number,ms. According to Ingle and Crouch,(96) for many-electronatoms, the hydrogen quantum numbers can be thoughtof as describing the individual electrons, but they arenot ‘‘good’’ quantum numbers for the entire atom.Good quantum numbers are associated with operatorsthat commute with the total atom Hamiltonian. Theseinclude the resultant orbital angular momentum quantumnumber, L, produced by coupling the orbital angularmomenta of each electron, and the resultant spin quantumnumber, S. For atoms with weak spin–orbit interactions,L and S couple to produce a total angular momentumquantum number, J . A multiplet of closely spaced stateswith the same L and S values but different J values iscalled a spectroscopic term and is designated as n2S+1{L}J ,where n is the principal quantum number for the valenceelectrons, 2S + 1 defines the multiplicity, and J is thetotal angular momentum quantum number. Of all thepossible transitions between states, only a fraction ofthem are observed. From quantum mechanical principles,it is possible to derive selection rules that tell whichtransitions are allowed (i.e. those that occur with highprobability and give reasonably intense lines) and whichare forbidden (i.e. those that occur with low probabilityand give weak lines); this is beyond the scope of thisarticle. For this discussion, it is sufficient to note that termsymbols for almost all practical configurations have beentabulated(100) and tables of spectral line intensities havebeen assembled for nearly all the elements.(101)

In AES, the information relevant to an analysis can befound in the radiation emitted by excited analyte atomsdecaying from a nonradiational activation event. Theradiant power of this emission is a function of severalfactors, including the population density of the excitedatoms, the number of photons emitted per second by eachatom, the energy of each photon, and the volume of theemitting system. The reader is directed elsewhere for amore complete discussion of each of these factors.(96)

AES has the power to provide rapid, qualitative,and quantitative multielement analyses. Although aqualitative survey of the elements (i.e. a plot of theanalytical signal versus wavelength) in a sample maybe useful, more often the desired information is theconcentration of an analyte. Unfortunately, this is almostnever obtained directly as the result of an absolutemeasurement of an optical signal because the amplitudeand elemental concentration are seldom related in asimple way. Obtaining the desired concentration froman optical measurement usually involves calibration,

GLOW DISCHARGE OPTICAL SPECTROSCOPY AND MASS SPECTROMETRY 13

GD emission source

Planar cathodeCathode body

Hollow cathode

Hollow cathode

Cathode bodyAnode body Anode body

(Signal)

Preamp

109 Gain

Readout photometerand

high voltage supply

Anode

Photons Photons(–)

(–)

Photons

Gas

Gas

Insulators

Planar cathode discharge Hollow cathode discharge Hollow cathode discharge

Insulators

L1 L2 L3

(HV)

Monochromator

Iris

Figure 9 Schematic of an instrumental arrangement used at the Oak Ridge National Laboratory for AES. (Reproduced withpermission from Ref. 10. American Chemical Society, 1990.)

subtraction of blanks, comparison with standards, andother similar procedures.(96) Quantitative analysis of asingle element is most easily accomplished in AES bymonitoring the emission intensity as a function of theanalyte’s concentration under a given set of conditions(e.g. constant discharge gas pressure, voltage, and currentat a given wavelength). Standards, often provided by theNational Institute of Standards and Technology, providea range of concentrations over which a calibration curvecan be developed. The signal intensity of an unknownconcentration is then compared with the intensity of thestandards, thus providing the concentration of the analytein question.

Although simple in principle, quantification is compli-cated by a number of factors, including spectralbackground, incomplete wavelength separation, self-absorption, peak broadening, and so on, most of whichare beyond the scope of this article. When one considersthese complicated issues, it is clear why it is importantto control conditions precisely and use standards for themost accurate quantification.

Instrumentation used for emission spectroscopyincludes an excitation source, a sample container, a wave-length selector, and a radiant power monitor. Dependingupon the spectrochemical method, the excitation sourceand sample container may be separate components or

they may be combined, as is the case with the GD.Figure 9 illustrates one instrumental configuration usedin our laboratory at the Oak Ridge National Laboratoryfor AES. A 0.5-m monochromator serves as the wave-length selector and the combination of a photomultipliertube, preamplifier, and readout photometer comprisesthe radiant power monitor. Figure 9 also illustrates threecommon sources used with GD emission spectroscopy – aplanar cathode discharge and two versions of the hollowcathode discharge.

The planar cathode discharge is thus termed becausethe portion of the sample exposed to the discharge isflat. It is often contrasted with the coaxial or pin-typecathode used more commonly with mass spectrometryand described elsewhere in this article. With the devel-opment of the Grimm lamp in 1968(8) and its eventualcommercialization, the planar cathode discharge gainedwidespread use for emission spectroscopy. Althoughother planar cathode discharges have been developedin the past 30 years, the Grimm source still finds thegreatest application today. One interesting feature of theGrimm source is that it is an obstructed discharge (i.e.the discharge is confined to the sample by the extensionof the anode into the cathode dark space). Moreover,the vacuum in the anode–cathode inner space is lowerthan in the discharge region itself, necessitating a dual

14 ATOMIC SPECTROSCOPY

outlet pump with a larger throughput for the inner elec-trode space.(102) Another interesting feature of the Grimmsource is that the cathode is located outside of the sourceitself; this provides for easy sample interchange andmeans that the Grimm source is particularly amenableto the analysis of any flat conducting surface that can bebrought up to the source opening, such as metal sheets ordisks. Typical operating conditions for the Grimm sourceare 500–1000 V, 25–100 mA, and 1–5 torr. The relativelyhigh power produced by the source (12.5–100 W) meansthat the cathode is often water-cooled; this usually is not aproblem in emission spectroscopy, but makes interfacingthe Grimm source with a mass spectrometer (with its highvacuum requirements) more difficult.

Planar cathode discharges have been interfaced to avariety of commercial emission instruments.(10) Grimm-type sources find their greatest use in trace elementalanalysis of solids and in-depth profiling of layered metalsamples.(102) Detection limits by emission spectroscopyare of the order of 0.1 ppm.(10) Precision of the orderof 0.5–5% has been obtained for concentrations in the0.01–10 mg g−1 range.(102) Ablation rates range from 0.1 to3 mg min−1 depending upon the element, discharge area,current, and voltage.(102) At these rates, the Grimm sourceis ideal for thin layer analysis or in-depth profiling whereit may be necessary to profile from a few nanometers toseveral tens of micrometers in a relatively short time.(79)

Hocquaux(79) wrote an excellent chapter on thin filmanalysis by GD emission spectroscopy.

The other two sources shown in Figure 9 arehollow cathode discharges. Although the hollow cathodedischarge appears similar to the planar cathode phys-ically, the shape of the cathode cavity provides someproperties that make it appealing for atomic spectroscopy.This discharge derives its name from a cathode that hasbeen drilled out to form a cylindrical cavity closed atone end.(103) The so-called hollow cathode effect canbe visualized as a GD with two parallel cathode platesbeing brought sufficiently close to each other until thetwo cathode glow regions coalesce.(104) The result of thiscoalescence is an increase in current density that can beseveral orders of magnitude larger than a single planarcathode at the same cathode fall potential.(105) Couplingthis increase in current density with the longer residencetime that the analyte experiences in the negative glowregion (due to the cathode’s shape) results in a markedincrease in the intensity of radiation emitted compared toa planar cathode. In addition, background intensities arelow because electron number densities are low, resultingin a very high signal-to-background ratio.(106) To performan analysis using a hollow cathode discharge, it is neces-sary to machine the sample into the shape of a cylinder,or to place powder or metal chips into a hollow cathodemade of some inert material such as graphite. One can

also analyze solutions by drying a residue on the hollowcathode surface. Operating conditions vary widely, buttypically range from 200 to 500 V, 10 to 100 mA, and 0.1to 10.0 torr.(10) Detection limits have been reported in thepicogram range,(102) but more typical results are in thenanogram range. Although hollow cathode discharges arewidely used in atomic spectroscopy, the majority of thesedevices are light sources for AAS (see below). A typicalhollow cathode lamp (HCL) is depicted at the extremebottom right of Figure 9.

3.3 Atomic Absorption Spectroscopy

In absorption spectroscopy, spectrochemical informationcan be found in the magnitude of the radiant power froman external light source that is absorbed by an analyte.(96)

To obtain information relevant to measuring an element’sconcentration, however, it is necessary for the frequencyof the incident radiation to correspond to the energydifference between two electronic states of the analyteatoms being measured. Often, but not always, the atomsstart in their electronic ground state and are excited to ahigher lying electronic state by the incident radiation. Theadsorption of this radiation usually follows Equation (5):

A = − log T = − log MM0

= abc (5)

where A is the absorbance, T is the transmittance, a is theabsorptivity, b is the path length of absorption, and c isthe concentration of the absorbing species. This equationis commonly referred to as Beer’s law. To calculatethe concentration, one measures the incident radiation(M0) and the transmitted radiation (M), calculates theabsorbance, and relates the absorbance to concentrationusing a series of standards and calibration curves, similarto emission spectroscopy.

A typical instrumental configuration for an atomicabsorption spectrometer is shown in Figure 10. A hollowcathode, fabricated from the elements of interest, is oftenused as the source of incident radiation, although anelectrodeless discharge lamp may be used for someelements such as As, Se, or Te, where the emission from anHCL may be low. The HCL is focused to a point inside thedischarge and then refocused into the entrance apertureof the monochromator. To obtain the background signal,one can use a shutter, or alternatively modulate the HCLand measure the background during the off period. In thearrangement shown in Figure 10, a mechanical chopperis used to facilitate background subtraction by providinga reference signal to a lock-in amplifier. Transmission ismeasured first with the discharge off and then with thedischarge on, often for a range of currents and voltages.Using Beer’s law, absorbance is calculated for a series of

GLOW DISCHARGE OPTICAL SPECTROSCOPY AND MASS SPECTROMETRY 15

GD atom sourceChopper

HCL

L1 L2 L3 L4(HV)

Monochromator

(Signal)

Preamp Readout photometerand

high voltage supply

Probe

Sample holder

Quartz tube

Hollow cathode

Atomic absorption sputtering chamber

Gas

Gas 35 mm

Water coolingCathode Gas

Anode

Demountable hollow cathode

109 GainLock-inamplifier

Iris

Figure 10 Schematic of an instrumental arrangement used at the Oak Ridge National Laboratory for AAS. (Reproduced withpermission from Ref. 108. American Chemical Society, 1976.)

standards to produce a calibration curve; the absorbanceof an unknown is then correlated with its concentration.

Two different discharge configurations are shown inFigure 10 for atomic absorption. The one on the left-hand side is an atomic absorption sputtering chamberdeveloped by Gough.(108) A planar cathode is mountednear the top by pressing the sample against an O-ringthat provides the vacuum seal. The gas flow of the cellwas designed to provide transport of sputtered atoms intothe observation zone, 1–2 cm from the sample. In 1987,Bernhard(107) took the idea of gas-assisted transport ofatoms one step further, reporting on a design that usedgas jets aimed at the sputtering surface to significantlyincrease the sampling rate as well as the absorptionsignal in a sputtering chamber. A commercial atomicabsorption cell was designed based on this principle(Atomsource,(107) Analyte Corporation, Medford, OR),renewing the interest in AAS that began with Walsh(109)

more than five decades ago. The source on the right-hand side is a much simpler atom generator. It is basedon the direct insertion probe (DIP) design of King.(110)

The coaxial cathode in King’s original design has beenreplaced by a stainless steel ring that accommodatesa demountable hollow cathode (4.82 mm in diameter× 2.54 mm in length with a 3.18-mm hole at the center).The DIP facilitates alignment of the HCL emission, which

is focused through the orifice (i.e. the region of the highestatom density) and collected after it passes through awindow in a six-way vacuum cross. Typical operatingconditions for this source are 1.0–3.0 torr, 500–2000 V,and 2–15 mA. Detection limits for GD atomic absorptionare in the low-parts-per-million range. Although GDatomic absorption is not as widely used as flame orgraphite furnace atomic absorption, it has found its nichein applications where analysis by other atomic absorptionmethods (primarily solution-based) is difficult (e.g. theanalysis of materials that are difficult to dissolve).

3.4 Atomic Fluorescence Spectroscopy

Atomic fluorescence is similar to AAS in that both relyon an external light source to produce an analytical signalfrom an atomic vapor. In fluorescence spectroscopy, thesignal is contained in the emission of photons fromthe atom population after absorption of the incidentenergy. There are five basic types of fluorescence: reso-nance, direct-line, step-wise, sensitized, and multiphotonfluorescence.(96) For this discussion, it is not important todefine these five types, but only to say that the differenceslie in the excitation and relaxation pathways that eachfollows to produce fluorescence. Resonance fluorescence(where the same upper and lower levels are involved in the

16 ATOMIC SPECTROSCOPY

GD atom source

L1

Laser

Valve

Cathode Anode

(Signal)

109 Gain

Preamp Readout photometerand

high voltage supply

Fluorescenceobservation

window

Ceramic sleeve

Sample rodLaser path

Atomic fluorescence cell

L2 L3 L4

(HV)

Monochromator

Iris

Figure 11 Schematic of a conventional instrumental arrangement for a single-beam atomic fluorescence spectrometer.

excitation–deexcitation process so absorption and emis-sion wavelengths are the same) finds the most widespreaduse in analytical spectroscopy because the transition prob-abilities and the source radiances are the greatest forresonance fluorescence when conventional line sourcesare used.

Figure 11 shows a conventional instrumental arrange-ment for a single-beam atomic fluorescence spectrometer.Radiation from the source is focused into the GD. Fluo-rescence photons are imaged onto the entrance apertureof a monochromator that isolates the analyte fluorescencefrom background emission and fluorescence from otherspecies. Fluorescence is usually viewed at an angle of90° with respect to the excitation source to minimize thecollection of scattered source radiation.

One critical component of an atomic fluorescencespectrometer is the excitation source. HCLs, electrodelessdischarge lamps, and metal vapor discharge lamps have allbeen used successfully, although today most fluorescenceexperiments use a laser. Lasers are superior sourcesfor atomic fluorescence because they provide a fluenceseveral orders of magnitude greater than other sources,are tunable over a wide wavelength range, have spectralbandwidths much narrower than absorption line widths,and can be focused to very small spot sizes. Bothcontinuous wave (with chopping) and pulsed lasers havebeen used, with dye lasers finding the most use because

they can be tuned over a large number of wavelengths.An inductively coupled plasma has also been used asan excitation source; here, the excitation wavelength isgoverned by the analyte that is aspirated into the torch(usually at high concentrations).

Like AES and AAS, the ideal atomizer for fluores-cence would produce a stable population of atoms ofsufficient number density to make quantification of smallconcentrations practical. A fluorescence atomizer shouldalso produce minimum thermal excitation to limit analyteand interference emission, a potential source of back-ground. When this stipulation is met, Rayleigh scatteringfrom atoms or molecules determines the fundamentallimitation for background noise.

Most atomic fluorescence measurements have beenmade with flame atomizers,(96) but recently inductivelycoupled plasma has been used. Plasmas generally providebetter atomization efficiency and a larger population offree atoms than flames. When analyzing a solid sample,however, an atomizer, like the one shown in Figure 11, ismore practical.(111) This simple design consists of a Pyrexglass housing into which a 6.35-mm diameter samplerod is inserted through a ceramic sleeve and sealed tothe cell with O-rings. Quartz windows are glued intothe cell to allow a laser beam to pass. A fill port andan evacuation port are also provided, and the entirecell is pumped by a single rotary vacuum pump. A DC

GLOW DISCHARGE OPTICAL SPECTROSCOPY AND MASS SPECTROMETRY 17

power supply (Electronic Measurements, Eatontown, NJ)provides voltages up to 600 V and currents up to 200 mA.A typical operating voltage is 570 V at 25 mA. The cellis pressurized with argon to between 900 and 1000 Pa(7–8 torr).

More recent applications of atomic fluorescence by thisgroup(112) have produced GD cells constructed from highvacuum components such as ConFlat crosses and flangeswith sapphire windows, but the principles of operationremain the same – fluorescence is detected 90° to the laserbeam path by a photomultiplier tube. The fluorescencesignal is amplified by a wideband amplifier and processedby a gated integrator and boxcar averager; the result is afluorescence spectrum as a function of wavelength, whichis indicative of those elements in a GD cathode for whicha fluorescence transition is allowed.

3.5 Optogalvanic Spectroscopy

The final optical technique to be discussed is optogalvanicspectroscopy. The optogalvanic effect was first observedusing a weak, incoherent light source in 1928(113);light from one neon discharge affected the electricalcharacteristics of a nearby discharge. The process is quitesimple; in a discharge there is an equilibrium establishedbetween the neutral species and the corresponding ions.If some means of energy is added to the discharge, theequilibrium position can be displaced and the fractionof ions altered; this is the case when a photon isabsorbed by the gaseous atom or molecule. This permitsone to measure optical absorption by an all-electronicmeans, that is, without the use of photodetectors. Theelectrical circuit employed to monitor the optogalvaniceffect commonly includes a ballast resistor in serieswith the discharge resistance; the discharge impedancechange is usually monitored as a change in the dischargevoltage. The light source is modulated, and a lock-in amplifier is employed to measure the alternatingcurrent component of the discharge voltage inducedby absorption of light as the source wavelength isscanned. In theory, the atomization source could beany of the discharges discussed so far; in practice,however, we have found that the demountable hollowcathode operating in the same manner as it does foratomic absorption provides the greatest flexibility foroptogalvanic spectroscopy.(114)

4 MASS SPECTROMETRIC METHODS OFANALYSIS

Much like the spectrochemical techniques describedabove, mass spectrometry offers the analyst a methodfor determining the identity and quantity of a particular

species in a sample. This technique, however, providesanalytical information through the separation and subse-quent detection of charged species associated with thesample. Ions are generated in the source region andselected by mass-to-charge ratio (m/z) usually using elec-trostatic or magnetic fields.

In this section, the basic requirements necessary toobtain mass abundance information are described, alongwith five types of mass spectrometers: quadrupole massfilters, magnetic sector mass analyzers, QITs, FTICRdevices, and TOF mass spectrometers. This article focusesonly on variants that have been coupled to the GD source.Table 4 provides basic information and characteristicsthat are unique to each of the mass spectrometricsystems.

4.1 Basic Requirements Necessary to Obtain MassAbundance Information

Mass spectrometric systems in general require a numberof fundamental components to create, transport, separate,and detect ions, manipulate the resulting signal to accountfor system inconsistencies, and provide useful informationto the analyst (see Figure 12). A number of thesecomponents are considered in the following section.

4.1.1 Ion Source

There are a number of GD source geometries thatcan be implemented for a variety of analyses (seeTable 1). Some types were designed, and are well suited,for optical applications as outlined in the previoussection. Many of the advantages afforded by the opticallyapplicable sources stem from an enhancement of the atompopulation, often a result of increased sputtering rate ora confined viewing region that facilitates optical viewing.Mass spectrometry requires a different set of criteria,however. Sources designed for mass spectrometricapplications must provide a reasonable population ofanalyte ions (not atoms). These ions must then beextracted from the source region into the mass analyzer.For these reasons, one appropriate source for massspectrometry is the coaxial cathode. Ions generated by thecoaxial cathode are extracted through an ion exit orificein the anode. Ionization is dominated by the Penningprocess that leads to an ion population with a narrow KEspread relative to other GD source geometries.(115) Thisgeometry also facilitates sample introduction via a DIP,making the appropriate adjustment of the plasma relativeto the ion exit orifice trivial. In recent years, however,Grimm-type sources have also been combined with massspectrometric detection.

18 ATOMIC SPECTROSCOPY

Table 4 Mass analyzers and considerations for elemental analysis

Mass analyzer type Advantages Disadvantages

Magnetic sector mass spectrometer Commercially available Scan speedReasonable resolution Complex

CostQuadrupole mass filter Commercially available Limited resolution

RobustnessScan speedPeak hopping mode

QIT mass spectrometer Cost Complicated interface to GDsource

CID to remove interferences CostFTICR mass spectrometer High resolution Complex

CID capability Complicated interface to GDsource

Space charge limitedTOF mass spectrometer Simplicity Some ion extraction biases

Cost Poor isotope ratio measurementsSpeedSimultaneous data acquisitionGood resolution when operated

in reflectron modeSource space charge limited

Glow dischargeionization source

HV power supply(DC or RF)

Computer Dataacquisition

Ionoptics

Massanalyzer

Detector

0.1–10 torr 10−7–10−9 torr10−5 torr

Figure 12 Block diagram of typical GD mass spectrometersystem components.

4.1.2 Vacuum Systems

Mass spectrometric techniques impose stringent require-ments on a vacuum system. One advantage of interfacingany of the currently available mass analyzers to the GDas compared to an atmospheric pressure source arisesfrom the operating pressure of the ion source itself. Mostmass spectrometers operate at pressures from 10−5 to10−9 torr. The GD source also operates at a reducedpressure (0.1–10 torr with support gas), although notnearly as low as those required for mass analysis. Thepressure differences associated with GD mass spectro-metric implementation are overcome using differentialpumping schemes. These types of systems employ a seriesof pumping regions to reduce the effect of the requiredpressure drop. A typical scheme would involve pumpingthe ion source with a rotaryvane roughing pump. Thiswould evacuate the discharge cell to a pressure of approx-imately 10−3 torr in the absence of discharge gas. The nextdifferentially pumped region facilitates extraction of ions

from the ion source. This region is often evacuated using aturbomolecular pump or oil diffusion pump and maintainspressures of 10−4 –10−5 torr during discharge operation.The analyzer region, which includes the mass separationcomponents as well as the detector, is evacuated usinga turbo molecular, oil diffusion, cryo, or ion pump andmaintains pressures of 10−7 –10−9 torr.

4.1.3 Ion Optics

Ions generated in the GD plasma must be transportedefficiently and indiscriminately to the mass separationregion to facilitate accurate and precise analyticalmeasurements. The transfer of ions from the sourceto the mass analyzer is usually accomplished using anoptics system. One type of system used for ion transportis the Einzel lens.(18,116,117) This type of lens is basedon three conducting tubular lenses of similar dimensionmounted in series. Typically, similar potentials are placedon the first and third lenses while that of the middlelens is adjustable. Many times the middle lens is held atground potential. As an ion beam passes through this lenssystem, it is focused in a manner analogous to an opticalbeam, reaching a focal point on the opposite side. It isimportant to note that cylindrical lens systems such as theEinzel lenses produce a cylindrical ion beam rather thana planar one. This characteristic may limit their use forsome applications such as ion guides for magnetic sectorinstruments that require a ribbon-shaped ion beam.

A second type of ion optic is the Bessel box.(118,119)

This technology is used in conjunction with QMAs toremove photons and neutral species from the ion beam

GLOW DISCHARGE OPTICAL SPECTROSCOPY AND MASS SPECTROMETRY 19

while simultaneously limiting the KE spread of ionsentering the quadrupole lens region. If the energy spreadis very large, quadrupole performance deteriorates,resulting in degradation of mass resolution, peak splitting,and asymmetrical peak shapes.(120,121) A Bessel box isconstructed from a square entrance and exit electrodesurrounded by sets of electrodes on each of the other foursides. A center plate or cone is located within the boxparallel to the entrance and exit electrodes. Potentialsare applied to each electrode. These potentials can bevaried to permit the transmission of ions with a discreteKE. Photons and neutral species will not be affected bythese potentials and will proceed into the box linearlyand collide with the center stop, removing them fromthe beam. Ions entering the box with very little KE willbe repelled toward the entrance plate. If the energy ofthe ion is very large, it will not be steered around thecenter stop and will collide with the side electrode or exitplate. Only ions with the selected KE will travel aroundthe center stop, through the exit aperture, and into thequadrupole region. In most cases, one or two lenses arelocated behind the exit aperture to focus and transportthe selected ions.

Quadrupole lenses can also be used as ion guides whenoperated in an RF-only mode.(122,123) It is shown laterin this article that the application of both DC and RFpotentials provides a notch filter that can be adjusted andscanned to provide mass unit resolution. When only anRF potential is applied, the quadrupole acts as an ionguide, focusing all ions through quadrupole lenses.

4.1.4 Detection Systems

Three types of detection systems are routinely used forGD mass spectrometric measurements: Faraday cups,electron multipliers, and microchannel plates (MCPs).Detector selection is often independent of the massanalyzer in use.

The detector used most often for applications with highion abundance is the Faraday cup. Modern Faraday detec-tors are extremely quiet. When operated using high-graderesistors and amplification components, these detectionsystems offer state-of-the-art measurements with respectto signal-to-noise ratio. Although not currently availableon commercial GD mass spectrometric systems, recentdevelopments in multi-Faraday array detection systemsoffer increasingly precise measurements for scanninginstruments by negating the effects of source fluctua-tions on measurements. Each Faraday cup in the arrayis dedicated and positioned to measure a single isotopeat a given dispersion setting. This allows simultaneouscollection of the selected ions during the acquisitionsequence without scanning the mass dispersion device,virtually removing any dependence on fluctuations in

ion beam intensity arising in the source. Minimizationof source fluctuation effects afforded by multicollectorarrays is most important for applications involving themeasurement of isotope ratios.

For applications requiring optimum sensitivity, discretedynode electron multipliers operated in a pulse-countingmode are required. Operation in this mode registersa signal pulse for every ion impinging on the firstdynode. Each impinging ion generates a number ofsecondary electrons that are successively amplified byeach dynode. Overall gains of 106 –108 are common whenusing multipliers with 14–20 dynodes. After the pulse ofelectrons leaves the multiplier, it is amplified and proceedsto a discriminator that is set to remove pulses arising fromdark noise. The signal is then sent to a universal counterthat records each pulse, stores it for a given time, andpasses it to a computer-based data acquisition systemthat presents the data in a usable form. It is importantto keep the count rate low enough (

20 ATOMIC SPECTROSCOPY

the MCP, these electrons are collected by a positivelycharged electrode positioned parallel to the MCP. Theresulting signal is further amplified and manipulated usinga fast digitizing oscilloscope or any number of computer-based flash analog-to-digital converter (ADC) computerboards. Two or more MCPs can be arranged in a stackorientation to further amplify the signal, thus increasingthe system’s performance.

4.2 Magnetic Sector Mass Analyzers

Initially used by Aston(125) for his studies of gaseousdischarges, the magnetic sector mass analyzer, or massspectrograph as Aston called it, is based on the spatialdispersion of ions with different m/z that is effected whenthey traverse an electromagnetic field. The magnetic fieldacts as a prism dispersing monoenergetic ions of differingm/z values across a focal plane; see Figure 13(a).(126)

The radius of the curved flight path of an ion through amagnetic field is given by Equation (6):

rm =(

144B

)(mV

z

)1/2(6)

where B is the magnetic field strength in gauss, m isthe atomic mass of the ions (in atomic mass unit), V is theacceleration voltage of the ion before it enters into themagnetic field (in volts), z is the charge of the ion, and144 is a constant prescribed by the units.(126) Because asector instrument can be made to focus ions onto a plane,it can be designed as either a single or multicollectorinstrument. A single electronic detector is used whenoperating in the sequential acquisition mode. Differentm/z ions are brought into focus on the detector by varyingeither B or V . Simultaneous acquisition instrumentsutilize either a photographic plate or a detector arrayoriented in the focal plane. These systems allow thedetection of a suite of different m/z ions at a given Band V setting. Although more expensive, the array-basedsystems offer shorter analysis times and the potential formore precise measurements because source fluctuationeffects are minimized.

Mass resolution is another parameter that influencesthe ability of a system to solve an analytical problem.Mass resolution is a measure of the instrument’s abilityto separate ions having small mass differences. Massresolution in magnetic sector instruments is definedby ion beam focusing at the focal plane. Single-focusing instruments rely on direction focusing to increaseresolution. This is accomplished by narrowing theentrance and exit slit widths, thus reducing the width of theion beam.(126) Mass resolution becomes more importantwhen ions of the same nominal mass must be separated.

Magnet

GD plasma

GD plasma

0.1–10 torr

0.1–10 torr

Detector

10−8–10−9 torr

10−4–10−5 torr 10−7–10−8 torr

QuadrupoleDetector

Iontransferoptics

RF and DCvoltage

generator

(a)

(b)

Figure 13 Schematic of (a) a magnetic sector mass spectrom-eter and (b) a QMA.

Multisector instruments are quite complex and expen-sive to build and maintain. They provide adequate-to-excellent resolving power, especially when an electrostaticsector is coupled to a magnetic sector to provide double(momentum and energy) focusing. When operated inthe sequential detection mode, sector instruments arehindered by relatively slow scanning speeds that adverselyaffect analysis time and sample throughput. In the past, anumber of commercial instruments were available from avariety of vendors based on the sector design. Currently,however, their availability has become slightly limited.

4.3 Quadrupole Mass Filters

Since its development in the 1950s and early 1960s, thequadrupole mass spectrometer has become a powerfultool for the analysis of a variety of materials. Much ofits popularity stems from the time of its developmentwhen it was viewed as a more rugged, more compact,and more cost-effective alternative to magnetic sectormass spectrometry systems, albeit with compromisedperformance. The quadrupole mass filter is a variable