Mass Spectrometry for On-Line Monitoring ofPerfluoro Compounds Using Li+ Ion AttachmentTechniques
Toshihiro Fujii* and Sundaram Arulmozhiraja
National Institute for Environmental Studies, Tsukuba, Ibaraki 305-0053, Japan
Megumi Nakamura and Yoshiro Shiokawa
ANELVA Corporation, Yotsuya, Fuchu 183-8508, Japan
Ion attachment mass spectrometry is being developed forcontinuous measurement of perfluoro compounds (PFCs)found in the atmosphere as a result of semiconductormanufacturing processes. Studies were made on 5 green-house gases, CF4, CHF3, C2F6, SF6, and c-C4F8, to developimproved methods for PFC analysis, particularly at levelsfound in the atmosphere (the parts-per-billion concentra-tion range). The results demonstrate the feasibility ofperforming real-time measurements of the trace amountsof PFCs encountered in process facilities by generatingadduct ions from Li+ ion attachment. The identificationand detection of c-C4F8 is described as an example of theutility of this new method.
Reductions in the emissions of perfluoro compounds (PFCs),greenhouse gases produced during integrated circuit (IC) manu-facturing, are being actively pursued by semiconductor manufac-turers.1,2 The interest in accurate, rapid, and relatively low-costanalytical methods for the determination of PFC concentrationshas increased strongly in recent years mainly as a result of theregulations requiring abatement of PFCs in IC manufacturingfacilities.
PFCs, widely used as fluorine sources in semiconductormanufacturing processes such as CVD (chemical vapor deposi-tion) chamber cleaning and dry-wafer etching, can be emitted asgaseous byproducts, along with hazardous air pollutants (HAPs)and various other gases. There is a need for reliable methods3
that are capable of measuring PFCs in the exhaust gas in realtime.
Fourier transform infrared spectroscopy (FT-IR)4,5 and on-linequadrupole mass spectrometry6-8 are two analytical techniquescommonly used for developing PFC monitors. Air and exhaustmonitoring using FT-IR methods is a feasible, reliable, and cost-
effective way to support the environmental safety and health(ES&H) programs of a diversity of manufacturing companies.Zazzera4 and Gubner5 have shown that tool exhaust from semi-conductor processes, in addition to indoor air and other emissionsources, can be monitored using commercially available FT-IRequipment and published guidelines; however, FT-IR cannotmeasure concentrations of homonuclear diatomic species suchas F2. The ability to quantify F2 emissions without on-sitecalibration would make FT-IR more versatile and valuable.
A quadrupole mass spectrometer (QMS) would be very helpfulin determining the amount of specific PFCs present in air. In 1997,Ridgeway8 used a QMS with an electron-impact ion source tomeasure PFC concentrations. The main disadvantage of thismethod may be the interference of ion signals due to the highprobability of PFC dissociation. On-site calibration is essential foranalysis with a QMS.
The recently developed Li+ ion attachment mass spectrometry(Li+-MS) method provides mass spectra of quasi-molecular ionsformed by lithium ion attachment to the chemical species underhigh-pressure conditions.9-11 Results are obtained in the form ofa mass spectrum of Li+ adducts. As an example, the method wassuccessfully applied to the study of neutral species that emergefrom CH4/O2 microwave discharge plasmas,12,13 demonstratingthat Li+-MS combines sensitivity with the capability of quantifyingamounts of reactive products in detecting even free radical species.
To support the emission reduction objectives of the semicon-ductor industry, the development of standardized analyticalmethodologies is necessary. To meet this requirement, we appliedthe Li+ ion attachment mass spectrometric technique to thedevelopment of a direct, real-time monitoring method for PFCsemitted from IC manufacturing facilities that we expect willsignificantly improve these areas of PFC measurement technology.The compounds studied were sulfur hexafluoride (SF6), tetrafluo-romethane (CF4), trifluoromethane (CHF3), perfluoroethane (C2F6),and perfluorocyclobutane (c-C4F8).
(1) Langan, J.; Maroulis, P.; Ridgeway, R. Solid State Technol. 1996, 39, 115.(2) Maroulis, P. Semicond. Int. 1994, 17, 107.(3) Green, D. S. PFC Technical Update, SEMICON WEST 96;1996.(4) Zazzera, L. A.; Reagen, W.; Cheng, A. J. Electrochem. Soc. 1997, 144, 3597.(5) Gubner, A. E.; Kohler, U. J. Mol. Struct. 1995, 348, 209.(6) Stoffels, E.; Stoffels, W. W.; Tachibana, K. Rev. Sci. Instrum. 1998, 69, 116.(7) Stoffels, E.; Stoffels, W. W.; Tachibana, K. J. Vac. Sci. Technol. A 1998, 16,
87.(8) Ridgeway, R.; Maroulis, P.; Pearce, R. MICRO 1997, January, 45.
(9) Fujii, T. Mass Spectrom. Rev. 2000, 19, 111.(10) Fujii, T. Anal. Chem. 1990, 64, 775.(11) Fujii, T.; Ogura, M.; Jimba, H. Anal. Chem. 1989, 61, 1026.(12) Fujii, T.; Syouji, K. J. Phys. Chem. 1993, 97, 11380.(13) Fujii, T.; Syouji, K. Phys. Rev. E 1994, 49, 657.
Anal. Chem. 2001, 73, 2937-2940
10.1021/ac001200w CCC: $20.00 2001 American Chemical Society Analytical Chemistry, Vol. 73, No. 13, July 1, 2001 2937Published on Web 05/26/2001
EXPERIMENTAL SECTIONApparatus. All of the experiments were performed using a
modification of the apparatus (ion attachment mass spectrometer,IAMS) produced by the ANELVA Corp. (Fuchu, Japan) with aquadrupole mass spectrometer and a lithium ion emitter.
The mass spectrometric setup used for the experiments, inwhich sampling was performed at atmospheric pressure, is shownschematically in Figure 1. In this setup, a stream of gas fromambient air is directed into a Li+ ion attachment reaction chamber(RC); the RC is fixed to the main vacuum envelope, which housesa quadrupole mass spectrometer.
Reaction Chamber. The RC constructed for this work is amodification of the previously reported lithium ion source10 usedfor lithium ion attachment mass spectrometry. It consists of alithium ion emitter, a reaction region, and a repeller. The repellerelectrode is made of a 20-mm-diameter stainless steel disk placed2 mm behind the bead and at a distance of 10 mm from theaperture. The reaction region is a cylindrical tube (i.d., 160 mm)with the lithium ion emitter centered within the chamber. Theadduct ions of sample gas from the RC pass through the apertureand are transported to the skimmer having a 1-mm-diameterorifice.
Lithium Ion Intensity. An emission current of Li+ ions ofabout 6 10-7 A was measured at an emitter current of 4.4 A, aN2 gas flow of 8 mL/min passing through the RC at 100 Pa, andwith an ion multiplier gain of 3.8 103. At a fixed pressure, theLi+ current increases approximately linearly with the emittercurrent in the pressure range of 10-100 Pa.
Mass Analysis. A lens (Figure 1) was used for focusing theion beam into the QMS. The rods were biased below ground byconnecting the dc rod-driven circuit to separate dc power supplies.A channeltron electron multiplier detector was used for iondetection.
Vacuum System. The vacuum chambers were evacuateddifferentially by three turbomolecular pumps (TMPs). The flowrate of sample gas introduced through the gas flow line into theRC was adjusted to maintain a pressure of 100 Pa in the RC. Thelens region ahead of the QMS was maintained at < 3 10-3 Pawith a 360 L/s TMP; the QMS region was evacuated to 7 10-5Pa with a 100 L/s TMP.
Procedure. c-C4H8 standard gas and N2, admitted to the RChousing through needle valves (NV), were used to check thesystem. Sensitivity optimization was performed by varying theoperating conditions, and the linearity of the QMS response inthe range of 3.0 10-8 g/s, which is based on the dilution factorof the gas flows, was confirmed by systematically changing theflow rate of the c-C4H8 gas entering the RC.
Data Analysis. In PFC analysis, the Li+ adduct mass spectraalso depend on differences in the Li+ affinity of the analytemolecules. This effect showed in experiments with cylinderscontaining 100 ppm SF6, 100 ppm CF4, 100 ppm C2F6, and 100ppm CHF3. In addition, the QMS also leads to a mass discrimina-tion effect, because at constant ion energy, the transmissiondecreases with increasing mass.
RESULTS AND DISCUSSIONPerformance and Response Characteristics. The c-C4F8
example was selected to show the performance and responsecharacteristics possible with this analytical instrument in theoptimal case. Figure 2A shows a typical mass spectrum in themass range up to m/z 210 and Figure 2B, a background spectrumwith no sample in the N2 carrier gas. Only peaks due to Li+ adductmolecular ion, with no fragmentation, are observed for the testcompound, c-C4F8 (Figure 2A); this spectrum is completelydifferent from the 70 eV electron ionization (EI) mass spectrum(not shown), which shows many fragment ions, such as C2F4+,C3F5+, and CF+ ions but no molecular ions.
The c-C4F8 sample, diluted by 12 mL/min of N2, was introducedat a rate of 1.8 10-6 g/s (0.012 mL/min of c-C4F8) from theneedle valve at a temperature of 25 C. The measurement of the
Figure 1. Schematic representation of the experimental setup: astream of sample gases from gas flow lines with needle valves (NV)was introduced into the reaction chamber (RC); TMP, turbomolecularpump; RP, rotary pump; QMS, quadrupole mass spectrometer.
Figure 2. (A) Spectrum obtained from 1000 ppm c-C4F8 in the N2gas, showing c-C4F8Li+ (m/z 207). For comparison, the 70 eV electronionization mass spectrum was measured; the abundant ions observedand their relative intensities are C2F4 (100), C3F5 (87), CF (54), CF3(25), CF2 (13), and C3F3 (6). The background mass spectrum (B) istaken with no c-C4F8 in the N2 carrier gas.
2938 Analytical Chemistry, Vol. 73, No. 13, July 1, 2001
adduct ion current yielded a value of 4.4 10-9 A, which is theoutput of the ion multiplier with a gain of 7.8 102. Thus, thesensitivity coefficient is 2.4 10-3 A/(g/s).
Linear Response Range. Signal response was linear over thechosen range of 7 10-10 to 1.6 10-6 g/s for c-C4F8. A plot offour introduction rates versus peak heights produced an es-sentially straight line up to 1.6 10-6 g/s, where more than 5%of the total Li+ reactant ions are used for the attachment. For alinear response, the sample size must ensure a large excess ofreactant ions in the cationizing chamber.
Sensitivity. The minimum detection amount (mda) was deter-mined by assuming that the actual noise level of the systemultimately limits the sensitivity. With a signal-to-noise ratio of 3,an ion multiplier gain of 7.8 102, and a Li+ emission current of6 10-7 A, the mda was calculated as 1.4 10-12 g/s, assumingthat the capability of the ion detection system in the electrometeris around 1 10-14 A. This value leads to a minimum detectionconcentration limit of 7 ppb (applying this concentration toc-C4F8, a peak 3 times higher than the noise level is produced);therefore, the detectable concentration was calculated as 7 ppb(v/v). Minimum detection amounts for other fluorinated com-pounds are the following: sulfur hexafluoride (SF6, 22 ppb),trifluoromethane (CHF3, 28 ppb), perfluoroethane (C2F6, 98 ppb),and perfluoromethane (CF4, 4.5 ppm).
Background Mass Spectra. When the nitrogen gas from thecylinder passed to the mass spectrometer, the background massspectrum (Figure 2B; scan speed, 5 s; scanning mass range, 1-210amu) generally showed a moderate-intensity peak at m/z 25corresponding to H2OLi+ as well low-intensity peaks at m/z 39,51, 53, and 63 corresponding to other ions. These were alwayspresent in the full-scan mass spectrum as minor, but possiblyinterfering, ion signals. The H2OLi+ peak is probably caused bythe ever-present water in the nitrogen carrier gas; m/z 39 (O2-Li+), and 51 (CO2Li+) may also be assigned as adduct ions of N2gas components. The m/z 53 peak is caused by Li+ attachmentto the frequently used solvent ethyl alcohol, which was presentin the flow line.
The appearance of alkali metal ion contaminations at m/z 23,39, 41, 85, 87, and 133 is dependent on the heating condition; thedistribution of these peaks primarily depends on conditioningtemperature and time. Presumably, the limit of detectability isdetermined mainly by the amount of impurities (Na+, K+, Rb+,Cs+) in the emitter bead; the content of Na+, K+, and Rb+ ions inthe bead is reduced to a negligible amount after weeks of heating.
Relationship between the Li+ Ion Affinity and the Intensityof the Adduct Ions. The ionization efficiency in the (M + Li)+
ion formation depends strongly on the Li+ affinity of the molecule,which ranges up to 50 kcal/mol. Although Li+ affinities of manymolecules are not currently available in the literature, they canbe calculated.14 This is useful in predicting which molecules canbe detected at a low-level concentration.
The mass spectrometric observation of molecular ions formedby the attachment reaction of Li+ ions with molecules mayestablish a relative order for the sensitivity and an empiricalrelationship between the Li+ ion affinities and the intensity of theadduct ions. The relative sensitivity can then be estimated for theorganic compound of interest when the Li+ ion affinity is known.
To determine experimentally the relationship between the(M + Li)+ intensity and the Li+ ion affinities of chemicalcompounds and to investigate to which extent the empiricalrelationship can be used for the determination of the relativesensitivity for the chemical species of interest, the relative ioncurrents of the (M+Li)+ ions formed by the attachment reactionwere examined for several fluorinated compounds with Li+
affinities ranging from 10 to 30 kcal/mol. These species werechosen partly because of their heavy involvement in the global-warming process.
The measured Li+ adduct intensities of formulated gas mix-tures of several fluorinated compounds and SF6 are listed in Table1: The sample flow rate was 12 mL/min. The data showedpronounced trends due to differences in Li+ affinity. Compoundswith high Li+ affinity were readily detected at low levels, butunsatisfactory results were obtained for nonpolar molecules ormolecules with low polarizability, such as CF4. The Li+ ionaffinities, which are used for the derivation of the empiricalrelationship, were obtained from theoretical calculations.
We performed theoretical calculations of Li+ ion affinities ofthe 5 greenhouse gases, CF4, CHF3, C2F6, SF6, and c-C4F8. Toconfirm the accuracy of the theoretical results, we calculated theLi+ ion affinities of two similar fluorinated molecules, CH3F andCH2F2, for which experimental values are available in theliterature.15 We used the basis sets 6-311+G(2d) and 6-311+G(3df)with the B3LYP functional to calculate the Li+ ion affinities; bothzero-point and thermal energy corrections were made and wereincluded in the affinity values. The results for all of the compoundsstudied are listed in Table 1, together with the experimentalvalues. The calculated affinities for the CH3F and CH2F2 moleculescoincided well with experimental values. All of the theoreticalcalculations were performed using Gaussian 98.16
(14) Smith, B. J.; Radon, L. Chem. Phys. Lett. 1995, 245, 123.
(15) Staley, R. H.; Beauchamp, J. L. J. Am. Chem. Soc. 1975, 97, 5920.(16) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.;
Cheeseman, J. R.; Zakrzewski, V. G.; Montgomery, J. A. Jr.; Stratmann, R.E.; Burant, J. C.; Dapprich, S.; Millam, J. M.; Daniels, A. D.; Kudin, K. N.;Strain, M. C.; Farkas, O.; Tomasi, J.; Barone, V.; Cossi, M.; Cammi, R.;Mennucci, B.; Pomelli, C.; Adamo, C.; Clifford, S.; Ochterski, J.; Petersson,G. A.; Ayala, P. Y.; Cui, Q.; Morokuma, K.; Malick, D. K.; Rabuck, A. D.;Raghavachari, K.; Foresman, J. B.; Cioslowski, J.; Ortiz, J. V.; Baboul, A.G.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.;Gomperts, R.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.; Peng, C.Y.; Nanayakkara, A.; Gonzalez, C.; Challacombe, M.; Gill, P. M. W.; Johnson,B.; Chen, W.; Wong, M. W.; Andres, J. L.; Gonzalez, C.; Head-Gordon, M.;Replogle, E. S.; Pople, J. A. Gaussian 98, Rev. A.7; Gaussian, Inc.: Pittsburgh,PA, 1998.
Table 1. Intensities of the Adduct Ions (M + Li)+ andLi+ Ion Affinities of M (Fluorinated Compounds)a
Li+ ion affinity, kcal/mol
compd I(M + Li)+ 6-311+G(2d) 6-311+G(3df) expt.
CF4 13 12.1 12.3CHF3 2100 19.4 19.6CH2F2 26.6 26.8 26.5CH3F 30.0 30.1 31.0C2F6 590 16.9 17.2c-C4F8 8300 20.6 21.1SF6 2600 16.3 16.9
a Affinities calculated at B3LYP functional using two different basissets.
Analytical Chemistry, Vol. 73, No. 13, July 1, 2001 2939
A plot of MLi+ ion intensities versus Li+ ion affinities is shownin Figure 3 for several fluorinated compounds. I(MLi+) is the Li+
adduct ion current obtained experimentally (in arbitrary units);Li+ ion affinities (in kcal/mol) were calculated at the B3LYP/6-311+G(3df) level. Disregarding from SF6, it is apparent thatI(MLi+) increases with increasing affinity values.
From the exponential least-squares fit of Figure 3, the followingequation was derived with A, the Li+ ion affinity in kcal/mol
The correlation coefficient for the plot, excluding SF6 data, is 0.998,showing that, for a set of similar molecules, I(MLi+) generallycorrelates well with Li+ ion affinities over the energy regionexamined. SF6, however, exhibits an inversion; the Li+ ion affinityfor SF6 is smaller than that of C2F6, but its I(MLi+) is larger. Thisinversion cannot yet be explained. It may be partially due to thesteric factor: some collisions will be more effective than others,depending on certain directional orientations of the molecule. Atthis time, the theory does not explain the considerable variationsin the steric factor.
CONCLUDING REMARKSThis study demonstrates the feasibility of measuring PFC
concentrations in air by generating ions by Li+ ion attachmentand mass analysis. The analytical method exhibited the followingfeatures: (1) high sensitivity, with the ability to detect 7 ppbc-C4F8 molecules in air; (2) the ability to accept high-capacitysample introduction from atmospheric pressure and to allow easycoupling of various sample introduction sources to the massspectrometric analysis; (3) the opportunity for real-time detectionof any PFC species, including radical intermediates; (4) the abilityto identify compounds by the generation of ions that do notfragment, which is especially useful for the determination ofmolecular weight; and (5) analysis of mixtures where no frag-mentation is desired.
Compared with FT-IR, which has quickly become establishedas a prime analytical method for tracking emission profiles fromexhaust systems within the semiconductor industry, the presentmethodology has both merits and limitations. The MS methodmay be more sensitive in most cases; however, the most importantconsideration concerns the relative sensitivity for each speciesand byproduct gas, which is required to accurately characterizeand quantify exhaust stream content. The relative sensitivity forsome species, such as nonpolar molecules, is too low to bedetected with this method. On the other hand, in FT-IR, inter-ference from various gases could suppress the ability to analyzecertain gases. FT-IR has a more limited dynamic range comparedto the MS method; moreover, a few pollutants cannot be detectedby FT-IR.
Hence, it can be concluded from these considerations that MSand FT-IR are complementary analytical techniques for real-timePFC exhaust monitoring. These two methods are valuable to thesemiconductor industry for a better understanding of the processand for supporting the PFC emission reduction programs;however, further investigation is needed to demonstrate thefeasibility of lithium ion attachment to PFC concentrations in areal air sample, which may have a large amount of moistureand interfering molecules. The possible interference should beconsidered carefully.
Received for review October 11, 2000. Accepted April 11,2001.
Figure 3. Plot of I(MLi+) vs Li+ ion affinity for several compoundshaving different Li+ ion affinities under the normalized condition thatthe partial pressure of the compounds is 100 ppm.
I(MLi+) ) 0.002 100.313A (1)
2940 Analytical Chemistry, Vol. 73, No. 13, July 1, 2001