Anal. Chem. 1998, 70, 833-837
Articles
Hardware Modifications to an Isotope Ratio Mass Spectrometer Continuous-Flow Interface Yielding Improved Signal, Resolution, and Maintenance Keith J. Goodman*
Department of Food Science and Human Nutrition, Iowa State University Ames, Iowa 50011
A home-built combustion interface was constructed to improve signal, resolution, and maintenance of a continuous-flow gas isotope ratio system. Chromatographic peak shapes were preserved by minimizing changes in tubing diameter and dead volumes. A single piece of fused silica capillary was used to connect the gas chromatograph (GC) to the isotope ratio mass spectrometer (IRMS), thus eliminating extraneous combustion furnace and water trap fittings. Analysis of a standard mixture of hydrocarbons yielded a 2-fold increase in signal over a slightly modified conventional system. Column efficiency, expressed as trennzahl (TZ), improved significantly (Student’s t-test 95% CI) by an average factor of 1.4 for replicates analyzed under similar conditions. The design is robust, requires less maintenance, and reduces leaks because the number of connections is minimized. Benefits of this system are transferable to virtually all commercially available continuous-flow systems. Continuous-flow isotope ratio mass spectrometry (CF-IRMS) remains unparalleled for rapid, high-precision isotope analysis of pure compounds and complex mixtures.1 CF-IRMS systems offer improvements in analysis time and sample size over conventional gas isotope ratio mass spectrometry (GIRMS). Recent developments in reaction interfaces have greatly expanded the breadth of elements and compounds analyzed by this technique. The earliest demonstrations of continuous-flow technology for high-precision work were achieved through interfacing an elemental analyzer to a GIRMS for the analysis of nitrogen,2 followed by carbon3 a couple of years later. With this technique, however, high-precision isotope ratio determinations were limited to micrograms quantities of materials. The first implementation of a gas chromatograph (GC) combustion interface4 enabled routine isotope ratio determinations at high precision for nanogram quantities of carbon-containing compounds and complex mixtures. * Current address: Metabolic Solutions Inc., 7 Henry Clay Dr., Merrimack, NH 03054. (1) Brenna, J. T. Acc. Chem. Res. 1994, 27, 340-346. (2) Preston, T.; Owens, N. J. P. Analyst 1983, 108, 971-977. (3) Preston, T.; Owens, N. J. P. Biomed. Mass Spectrom. 1985, 12, 510-513. (4) Barrie, A.; Bricout, J.; Koziet, J. Biomed. Mass Spectrom. 1984, 11, 583588. S0003-2700(97)00887-1 CCC: $15.00 Published on Web 01/27/1998
© 1998 American Chemical Society
Within the next decade, development of a liquid chromatograph (LC) interface expanded applications to nonvolatile compounds for high-precision analysis of carbon.5 During the same period, further modifications to the GC reaction interface ushered in rapid on-line analysis of oxygen.6-8 Addition of a reduction step, subsequent to the combustion furnace, produced an instrument capable of determining nitrogen isotope ratios in mixtures of nitrogen-containing compounds.9,10 By 1995, continuous-flow hydrogen analysis was achieved for systems implementing equilibration11 and on-line reduction.12 Recent introduction of an online pyrolysis interface has made possible position-specific carbon isotope analysis of complex organic mixtures.13 Several requirements govern all reaction interface designs: (1) All compounds must be pure prior to conversion in the reaction interface. Contaminants are indistinguishable from the original compound once molecules are converted to the desired small molecule, such as CO2 for the isotopic analysis of carbon. (2) Any potentially interfering products must be removed before effluent enters the tight ion source of the isotope ratio mass spectrometer. This point is most evident when analyzing 13CO2 generated from combusted organic compounds. If not trapped, water generated during the combustion process participates in an ion-molecule reaction to produce the contaminant +H12C16O2 at m/z 45. (3) Solvents should be diverted to prevent overloading or premature degradation of the limited reaction capacity of the interface. (4) An open split or other suitable means for restricting carrier gas flow into the IRMS is required. Despite tremendous success expanding continuous-flow applications through advances in interface chemistry, interface design has overlooked the requirements for optimizing chroma(5) Caimi, R. J.; Brenna, J. T. Anal. Chem. 1993, 65, 3497-3500. (6) Prosser, S. J.; Brookes, S. T.; Linton, A.; Preston, T. Biol. Mass Spectrom. 1991, 20, 724-730. (7) Brand, W. A.; Tegtmeyer, A. R.; Hilkert, A. Org. Geochem. 1994, 21, 585594. (8) Giesemann, A.; Jager, H.-J.; Norman, A. L.; Krouse, H. R.; Brand, W. A. Anal. Chem. 1994, 66, 2816-2819. (9) Preston, T.; Slater, C. Proc. Nutr. Soc. 1994, 53, 363-372. (10) Merritt, D. A.; Hayes, J. M. J. Am. Soc. Mass Spectrom. 1994, 5, 387-397. (11) Prosser, S. J.; Scrimgeour, C. M. Anal. Chem. 1995, 67, 1992-1997. (12) Tobias, H. J.; Goodman, K. J.; Blacken, C. E.; Brenna, J. T. Anal. Chem. 1995, 67, 2486-2492. (13) Corso, T. N.; Brenna, J. T. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 10491053.
Analytical Chemistry, Vol. 70, No. 5, March 1, 1998 833
tography.14 Commercial reaction interfaces cause peak broadening by introducing flow disturbances into the gas stream through changes in tubing diameter.15,16 Loss of resolution can result in undetected coeluting peaks or loss of accuracy for overlapping peaks when conventional detection and integration algorithms are employed.17 Reported here is a combustion interface design optimized to preserve chromatographic integrity while maintaining the required functions of solvent diversion, combustion (for analysis of carbon), and water removal.14 The entire system was built from inexpensive off-the-shelf components, and features of this design can be adapted to other commercial systems and on-line chemistries. EXPERIMENTAL SECTION Instrumentation. A VG Optima (Fisons, Wysenthawe, UK) isotope ratio mass spectrometer was used exclusively for this work. Instrument control and data analysis were provided by vendor-supplied software (Optima V2.01). The source pressure was 1.1 × 10-6 mbar when in continuous-flow mode. The ion source was held at 3.5-kV acceleration voltage and tuned for optimum linearity, resulting in an absolute sensitivity of 1000 molecules/ion. The CO2 calibrated gas standard (vs Pee Dee Belemnite) was introduced from the bellows during a GC run. GC Conditions. The GC used was a Hewlett-Packard 5890A (Hewlett-Packard, Palo Alto, CA) equipped with a split/splitless injector held at 250 °C and operated in split mode. Separation of the hydrocarbon mixture was achieved on a Chrompack CP-Sil 5CB 25 m × 0.25 mm × 0.25 µm (Chrompack, Rariton, NJ) fused silica capillary column. High-purity grade He (99.999%, Air Products, Des Moines, IA) was the carrier gas with a 20 cm3/ min split flow and a column linear velocity of 33 cm/s. The GC oven was programmed from 90 °C (hold 0 min), ramp 8 °C/min to 105 °C (hold 0 min), ramp 23 °C/min to 185 °C (hold 2.3 min). Standard Mixtures. Separate stock solutions of nonane, decane, undecane, dodecane, and methyl decanoate, >99% purity (Sigma, St. Louis, MO), were prepared gravimetrically and diluted to 0.43 mmol in Optima hexane (Sigma, St. Louis, MO). Individual solutions were stored in glass screw-capped tubes (10 mL) fitted with Teflon liners. A working standard mixture was prepared by diluting appropriate quantities of each of the individual solutions with hexane to yield an isomolar mixture between 3.5 and 8.6 µmol. δ Notation. The standard notation for expressing isotope ratios at natural abundance is the δ notation,18
δPDB (‰) )
(
)
RSPL - 1 × 1000 RPDB
where RSPL and RPDB refer to the ratio 45/44 for the sample and standard, respectively. For carbon isotopic analyses, isotope ratios are calculated against the international carbonate standard Pee (14) Goodman, K. J. Abstracts of Papers, 210th National Meeting of the American Chemical Society, Chicago, IL, August 20-25, 1995; ACS: Washington, DC, 1995; Paper 0.39. (15) Sternberg, J. C. In Advances in Chromatography; Giddings, and Keller, Eds.; Marcel Dekker: New York, 1966; Chapter 6. (16) Goodman, K. J.; Brenna, J. T. Anal. Chem. 1992, 64, 1088-1095. (17) Goodman, K. J.; Brenna, J. T. Anal. Chem. 1994, 66, 1294-1301. (18) Craig, H. Geochim. Cosmochim. Acta 1957, 12, 133-149.
834 Analytical Chemistry, Vol. 70, No. 5, March 1, 1998
Figure 1. Schematic of a conventional combustion interface (CCI). See text for details.
Dee Belemnite (PDB), where RPDB ) 0.011 2372. Isotope ratio determinations were achieved using a CO2 reference gas calibrated against PDB. Conventional Combustion Interface (CCI). A complete description of the conventional VG Isochrom II is available elsewhere.19 Issues relevant to this work are discussed here and presented in Figure 1. One distinguishing characteristic among interface designs is the way in which the solvent is diverted away from the reaction furnace. This design coordinates delicately balanced opposing flows of He (at C1 and C2) with an air-actuated needle valve (a) (SGE, Austin, TX) in order to divert solvent from the combustion furnace. During backflush mode, the needle valve opens, and GC effluent is diverted away from the combustion furnace, through the three-way union (b) (SGE, Austin, TX), and toward the FID. Once the solvent has eluted, the needle valve closes, and column flow is restored through the combustion furnace (d). The combustion furnace is comprised of a quartz tube, 61 cm × 0.5 mm × 0.8 mm, filled with oxidized copper pellets and held at 800-850 °C. The outlet at the exit of the combustion furnace (g) functions as an open split by venting the furnace to atmosphere. A portion of the effluent, comprised of purified bands of CO2 and H2O in He carrier, is drawn through the transfer capillary and into the water trap (e) on its way to the IRMS. The water trap is a 100-cm length of stainless steel tubing immersed in a -80 °C 2-propanol bath cooled with a Cryoprobe (f) (Neslab Cryocool Immersion Cooler CC-100II, Neslab Instruments Inc., Newington, NH). Assembly of the Isochrom II was performed as recommended by the manufacturer with the following exceptions: (1) The standby valve assembly/standard introduction system was completely abandoned and the 100-µm transfer capillary was connected directly to the CRT7 valve (the continuous-flow inlet valve to the isotope ratio mass spectrometer). (2) The original water trap was replaced with the assembly described above. The original water trap consisted of 70-100 cm of stainless steel tubing wound around a copper probe immersed in liquid nitrogen (LN2). The temperature of the water trap was maintained at -100 °C by balancing heat input through a resistive heater with the cooling of the LN2 reservoir. (3) Unlike the tubing in the LN2-based water trap (h), tubing for the modified water trap was carefully wound around a template of appropriate circumference to prevent crimping. This modification served to orient the coiled stainless steel tubing and butt connectors on the same plane rather than at right angles, as indicated by the arrows (h). All aforementioned modifications were meant to improve results for the conventional (19) Eakin, P. A.; Fallick, A. E.; Gerc, J. Chem. Geol. (Isotope Geosci. Sect.) 1992, 101, 71-79.
Figure 2. Schematic of the single-capillary interface design (SCID). See text for details. Figure 3. Replicate analysis data for a hydrocarbon mixture using the single-capillary interface design (SCID).
setup without altering changes in tubing diameter or the number of connections. (4) The carbon dioxide gas standard was introduced from the bellows, not through the conventional standby valve assembly. Introduction of the standard into the continuousflow stream was found to consume excessive amounts of the highpurity helium carrier gas and carbon dioxide standard without a noticeable improvement in precision or accuracy. Single-Capillary Interface Design (SCID). A diagram of the single-capillary interface design (SCID) is given in Figure 2. The backflush valve from the previous configuration was replaced by a four-port, two-position air-actuated rotary valve configured with 1/ -in. fittings, a 6-in. extension, and a Valcon T rotor (a) (Valco 16 Instruments Co. Inc., Houston, TX). Valve connections enabled column flow to be switched between the flame ionization detector (FID) and combustion furnace. An auxiliary flow of He (i) was set to approximately match the column flow. The auxiliary He minimized flow disturbances during valve actuation and purged the tubing when operating in column flow divert mode. This port was also used to inlet oxygen for recharging the combustion furnace. The furnace capillary (b) consisted of a 2-m length of 0.32-mm-i.d. deactivated fused silica (J+W Scientific, Folsom, CA) extending continuously from the rotary valve (a) to the open split (g). Preparation of the furnace capillary required two or three pieces of 30-cm × 0.1-mm-o.d. copper wire (Aldrich Chemical Co., St. Louis, MO) to be carefully inserted 0.25 m from the capillary end connected to the rotary valve. The furnace was charged by oxidizing the Cu to CuO at 500 °C after assembly of the interface. The furnace capillary was threaded through the existing quartz tube (61 cm × 0.5 mm × 0.8 mm) to hold it in place during operation. Union tees 1/4 in. × 1/16 in. × 1/16 in. (SGE, Austin, TX) (c), identical with those used for the conventional setup, served to anchor the furnace capillary in place. A 75-cm length of Teflon tubing (1.2-mm i.d., 2-mm o.d.) was centered on the remaining length of capillary extending from the combustion furnace (c), to protect the water trap region of the interface. The section of capillary with the Teflon guard was coiled to form two 9-cm-diameter loops and placed into a 2-propanol bath (e) maintained at -100 °C with a cryoprobe (f). Assembly of the open split (g) was achieved by threading the 0.1-mm-i.d. transfer capillary 3-5 mm into the end of the 0.32-mm-i.d. furnace capillary. An additional sheath flow of He gas at the open split was not necessary as long as the flow through the system was sufficient to accommodate flow requirements into the mass spectrometer. A piece of fused silica capillary (h) of dimensions identical to those of the SCID capillary (b), but without the copper wire, was used to connect the rotary valve to the FID. Replicate Analysis. Replicate analysis of the hydrocarbon mixture was conducted with the SCID. Isotope ratios were
calculated using default parameters for the vendor-supplied software. Replicate injections were made over a 4-day period. Data are reported as the mean ( standard deviation (1 SD). Signal. A comparison of signals for the two interface configurations was determined for replicate injections of the hydrocarbon mixture under closely matched conditions. To standardize flow rates into the mass spectrometer, the transfer lines for each configuration were calibrated by monitoring the m/z 44 signal for a sampled gas mixture of CO2 in helium. Peak areas were determined by curve-fitting the raw ASCII data for m/z 44 using a commercially available software package (Peakfit, V3.10, Jandel Scientific). Resulting peak areas for the components of the hydrocarbon mixture were used to quantify the difference in signals for the two configurations. Results are expressed as the mean ( 1 SD. Column Efficiency. Column efficiency was quantified using trennzahl20 (TZ) or “separation number”. TZ is a measure of column efficiency that determines the number of resolved peaks able to fit between two peaks of a homologue pair. The calculation for TZ is given below
TZ )
(
∆tR
w0.5(A) + w0.5(B)
)
-1
where w0.5(A) and w0.5(B) are the peak widths at half-maximum and ∆tR is the difference in retention times for two consecutive peaks of a homologue pair. TZ determinations were performed for each of three configurations under closely matched GC conditions for the hydrocarbon mixture. The first configuration served as the ideal case, where the effluent was directed only through the rotary valve to the FID. The second and third configurations employed the SCID and CCI, respectively. Each injection yielded three TZ measurements of the four n-alkanes in the hydrocarbon mixture. Results were compared using a Student’s t-test at the 95% confidence level. RESULTS AND DISCUSSION Replicate Analysis. Repeat injections of the hydrocarbon standard mixture were performed to monitor the stability and lifetime of the SCID. Figure 3 presents results for replicate analysis of the standard mixture acquired over a 4-day period. Deviations in δ 13C ‰ are plotted on the ordinate. Injection number is plotted on the abscissa. No trends are apparent in the (20) Jennings, W. Gas Chromatography with Glass Capillary Columns; Academic Press: New York, 1980.
Analytical Chemistry, Vol. 70, No. 5, March 1, 1998
835
Table 1. Replicate Analysis Summary Data Obtained with the Single-Capillary Interface Design (SCID)a n ) 38 (1.9-2.6 nmol on column)
nonane decane undecane dodecane methyl decanoate a
n ) 80 (0.8-0.98 nmol on column)
mean
1 SD
mean
1 SD
-32.96 -33.35 -29.76 -32.88 -31.94
0.12 0.12 0.16 0.07 0.11
-33.07 -33.29 -29.53 -32.52 -31.74
0.18 0.13 0.09 0.07 0.09
Means and standard deviations of individual components are given.
Figure 4. Trenzahl results for nonane-decane (non-dec), decaneundecane (dec-und), and undecane-dodecane (und-dod) analyzed using three independent configurations. Solid bar, analysis using GC and flame ionization detector. Shaded bar, analysis using GC and single-capillary interface design and the IRMS. Open bar, analysis using GC and the conventional combustion interface. Error bars represent the 95% confidence interval.
data. A summary of the replicate analysis data is given in Table 1. The data are grouped into two categories based on quantity of material injected. System precision was δ < 0.2‰ (1 SD), where typical manufacturer specifications for 3- or 10-kV instruments are δ e 0.3 and 0.2‰, respectively. Manufacturer specifications, however, are usually determined for far less than the 590 independent determinations shown here. Signal. A comparison of signal yielded a 1.96 ( 0.11 (mean ( 1 SD)-fold increase in peak area for the SCID over the CCI. This is likely due to the elimination of the stream diluting and splitting effect caused by the CCI open split and makeup gas assemblies (see Figure 1, c2 and g). Column Efficiency. Figure 4 shows TZ results for the three configurations mentioned above. The TZ is given on the ordinate, and the n-alkane combinations are labeled on the abscissa. Error bars indicate the 95% confidence interval. The TZ for the FID configuration represents the maximum preservation of peak shape for the system. Column efficiency for the SCID and CCI is significantly less than that for the FID. However, the SCID better preserved column efficiency, as indicated by the significant improvement in TZ over the CCI. Comparison of Peak Shapes. The limit of the chromatographic efficiency is determined by peak distortions contributed by the injector, column, detector, dead volume, and additional connections.21 The injector and column are standard in this CFIRMS system and, therefore, should not distort peak shapes beyond what is typically achieved by GC. Figure 5 compares signal decay for normalized peaks generated using the three configurations and a CO2 pulse. The STD trace was generated (21) Guiochon, G.; Guillemin, C. L. Rev. Sci. Instrum. 1990, 61, 3317-3339.
836 Analytical Chemistry, Vol. 70, No. 5, March 1, 1998
Figure 5. Comparison of signal decay for a CO2 pulse (9), an analyte peak detected with the FID (0), an analyte peak detected with the IRMS after the analyte passed through the SCID (b), and an analyte peak detected with the IRMS after the analyte passed through the CCI (O).
by pulsing CO2 gas through the changeover valve into the ion source of the IRMS. This reveals the lower limit for signal decay achievable with the IRMS system in continuous-flow mode. The FID trace was generated by an analyte peak (nonane) detected by the FID after the analyte passed through the column and the rotary valve. This peak illustrates the minimum contribution to peak broadening caused by the chromatographic process in this system. The SCID trace was generated by the analyte peak detected by the IRMS after the analyte passed through the column, rotary valve, and the SCID combustion interface. This plot illustrates the degradation in peak shape caused by the SCID combustion interface when compared with the FID. This extra broadening is likely due to flow restrictions in the furnace region rather than the cryogenic-based water trap, because peak broadening is reproducible when the trap is operated at ambient temperature. The CCI trace was generated by the analyte peak detected by the IRMS after the analyte passed through the column, rotary valve, and CCI combustion interface. This trace reveals the contribution of dead volumes and changes in tubing diameter to peak broadening in the CCI. Maintenance. The SCID offers an additional advantage over the CCI in that it is easier to assemble and maintain. Assembly of the CCI requires leak-free connections at critical junctions, such as the three-way union (Figure 1, b) and the combustion furnace. Assembly of the three-way union is cumbersome for the unskilled operator and is a frequent source of leaks. In contrast, capillary connections to the rotary valve in the SCID system are simpler, because they are functionally similar to connections made to a GC injector or detector. The capillary connection to the combustion furnace in the CCI system is at a region where the temperature reaches >450 °C (Figure 1, c2 and g), which is beyond the upper limit of most ferrules. The SCID avoids this problem because the one-piece combustion furnace extends continuously through this high-temperature region. Solvent diversion in the CCI requires a delicate balance between opposing flows that are typically below the working range of most needle valves. Inadequate backflush flow allows solvent to enter the combustion furnace, resulting in a decreased furnace lifetime. Excessive backflush flow compromises analyte peak shape and signal. With the CCI system, any changes in the GC that alter column flow (i.e., changes in the column dimensions or temperature program) will require the backflush flow to be reoptimized. The SCID can easily accommodate changes that
alter carrier gas flow, because solvent is diverted through the timed actuation of a rotary valve. The original LN2-based water trap for the CCI required refilling every 4-6 h. It is possible to adopt an automatic refill system connected to a larger LN2 dewar, but the cryoprobe system offers the advantage of a low-maintenance system at a competitive price. ACKNOWLEDGMENT This work was supported by the Center for Designing Foods to Improve Nutrition, which receives funding from the Iowa
Agriculture and Home Economics Experiment Station, with partial funding provided from the USDA special grants program and the Hatch Act. The author acknowledges helpful discussions with Drs. T. Corso and H. Tobias.
Received for review August 14, 1997. Accepted December 8, 1997. AC970887Q
Analytical Chemistry, Vol. 70, No. 5, March 1, 1998
837
