278
ANALYTICAL CHEMISTRY, VOL. 51, NO. 2 , FEBRUARY 1979
a t a source temperature of 200 "C and a source pressure of 3 Torr. More fragmentation and less ionization (sensitivity) will be encountered in the PCI methods at a higher source temperature. T h e OH- NCI spectra of virtually all of the compounds studied here could be put to good analytical use. T h e (M 1)- ion is the most intense ion in the spectra of 22 of the steroids, and it is of sufficient intensity to allow facile identification of molecular weights for all save two of them. While some association ions are observed (referred to above as (M n)- ions), the amount is small in total. and absent for most of the compounds. By contrast, the (M + NHJ+ ion is found extensively in the NH, PCI spectra, and this ion will be markedly lower in abundance or perhaps absent a t higher source temperature. T h e OH- NCI results with OH containing steroids are particularly valuable, for such compounds generally show an (M - 1)-ion of useful intensity and ions produced by the loss of 18 additional mass unit for each OH in the molecule. In some cases the relative intensities of these ions give information about the positions of the OH groups relative to each other and to other groups in the molecule. As an example, such information allows one to distinguish between the diols (20) and (21), which are positional isomers. T h e OH- NCI spectra of cholesteryl esters of fatty acids such as cholesteryl stearate are particularly attractive, for they
contain peaks giving information about the molecular weight of the component ( ( M - 1)-),the identity of the acid present (RCOO-), and the presence of the cholesterol carbon skeleton ((XO-H,O)-). In CH4 and i-C4HIoPCI the fragmentation of these molecules is so extensive that no molecular weight information is available.
LITERATURE CITED A. L. C. Smit and F. H. Field, J . A m . Chem. Soc., 99, 6471 (1977). Y . Y. Lin and L. L. Smith, Biomed. Mass Spectrom., in press. A. L. C. Smit, M. A . J. Rosseto, and F. H. Field, Anal. Chem., 48, 2042 (1976). V . M. Bierbaum, C. H. DePuv, and R . H. ShaDiro, J . A m . Chem. Soc., 99, 5800 (1977) J H J Dawson and N M M Nibbering, J Am Chem SOC, 100, 1929 11978) - -, D. J. Cram, "Fundamentals of Carbanion Chemistry", Academic Press, New York, 1965. L. L. Smith, M. J. Kulig, and J. I. Teng, Steroids, 22, 627 (1973). R. Hiatt, "Hydroperoxides", in "Organic Peroxides", D. Swern, Ed.. Wiley Interscience, New York, 1971, p 46. 100, 1356 M. Meot-Ner (Mautner) and F. H. Field, J . Am. Chem. SOC., (1978).
+
\
RECEIVED for review August 4, 1978. Accepted November 6, 1978. The Rockefeller University portion of this work was supported in part by a grant from the Division of Research Resources, National Institutes of Health, and that of t h e University of Texas Medical Branch of National Institutes of Health Grant =HL-10160.
High Sensitivity, Continuous Flow Thermochemical Analyzer Richard S. Schifreen,' Carolyn Sue Miller,2 and Peter W. Carr" Department of Chemistry, University of Minnesota, Minneapolis, Minnesota 55455
A flow enthalpimeter for use with two fluid streams has been designed. It is based on a related device for use with one fluid stream and a bed of catalyst. Satisfactory results were obtained only when the two thermistors, which comprise part of a differential temperature measurement system, are subjected to the same net flow. I n addition, any volume located between the point of reagent mixing and the adiabatic column must be minimized. Assays for mineral acid, calcium ion, and nitrite based upon their reaction with tris(hydroxymethy1)aminomethane, ethylene bis(oxyethylenenitri1o)tetraaceticacid, and sulfamic acid, respectively, have been developed. Detection limits ranged from 10 pM to l mM, depending upon the heat of reaction and achievable base-line stability. Sample volumes of 120 pL and a flow rate of 2.5 mL/min allow for a throughput of nearly 60 sampleslh. I n general, sample volume, concentration, and flow rate affected the signal height and width in a fashion similar to that observed with an immobilized enzyme analyzer based on a related flow system. The chlef differences in behavior occur at low sample concentration.
In principle, thermochemical methods of analysis are applicable to any reaction which either generates or absorbs Present address, Clinical Chemistry Laboratory, H a r t f o r d Hospital, H a r t f o r d , Conn. 06115. 2Present address, Technical Services, Schilling Division, McCormick a n d Co., Inc., Salinas, Calif. 93901. 0003-2700/79/0351-0278$01 .OO/O
heat. These techniques, however, require a controlled environment and are often more tedious than other methods which provide the same information. Consequently, thermochemical analysis is often not considered for routine work when some alternative technique is available. Various attempts have been made to circumvent these limitations. In 1965, Priestly et al. introduced continuous flow enthalpimetry ( I ) . This approach was based upon mixing a reagent and sample stream in a plastic cylinder. Three thermistors monitored the temperatures of the sample, reagent, and product streams providing a differential measurement system. They recognized the need for a reaction chamber possessing high mixing efficiency and low thermal capacity and conductivity. This was accomplished by fitting the inlets to the small reaction cylinder with narrow inlets to give two opposed mixing jets. An isothermal instrument has been developed by Christensen et al. ( 2 ) . Their reaction vessel is immersed in a controlled temperature bath and maintained a t a constant temperature by a Peltier cooler and digitally controlled heater. Reaction heats are measured a t steady-state by t h e rate of heat addition required to maintain the pre-set temperature. This instrument is capable of measuring heats of reaction a t various pressures and quantitating large volume samples. Its primary advantage lies in t h a t it is independent of the heat capacity of the solution and of the characteristics of the calorimetric vessel as a result of the isothermal mode of operation. A third device has been developed by Peuschel and Hagedorn (3-5) and is being manufactured by Technicon C 1979 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 51, NO. 2, FEBRUARY 1979
Corp. (6). T h e sample and reagent streams are introduced into a 1-mL plastic and stainless steel cell agitated by a Teflon encased magnetic stirrer. A single thermistor continuously monitors the cell temperature. T h e entire apparatus is immersed in a controlled temperature bath which is necessary since only the cell temperature is monitored. T h e system is sensitive t o temperature changes of 5 m°C with a precision of less t h a n 1%. A wide variety of reactions, including precipitation, have been adapted to this instrument. Because of limitations in the cell design, however, only sample volumes of 1 m L or greater and high concentration can be analyzed a t a rate of 20-40 samples per hr. Censullo and Jordan described a method they term “peak enthalpimetry” (7).This system is similar to the one described here in t h a t a packed bed was used to provide mixing. Detection limits with this device were a t best a decade worse than those available with the present apparatus, and sample throughput a t comparable flow rates was lower. T h e flow enthalpimeter developed in our laboratory for use with immobilized enzymes meets the requirements of accuracy, precision, and throughput necessary for routine analysis without the need for an electronic temperature controller. In this study the instrument was modified t o mix a soluble reagent and sample stream while retaining the desirable operating characteristics of the immobilized enzyme reactor system. T h e reactions chosen for study in this work are all well characterized thermochemical systems. They are: the neutralization of T h a m (tris(hydroxymethy1)aminomethane) with hydrochloric acid ( 8 ) ,the complexation of calcium with EGTA ((ethylene bis (oxyethylenenitri1o)tetraaceticacid) (9,lO) and the oxidation of nitrite by sulfamic acid (11). At this time a precipitation reaction cannot be used in the flow system. T h e reactions involved in t h e various assays are: HCl-THAM H+ THAM + HTHAM’; AH,,oc = -11.3 kcal/mol
+
(1) C a - E G T A ( p H = 8.0)
Ca2+ + H2EGTA2-+ BTHAM CaEGTA2- + 2HTHAM+;
g
-30.6 kcal/mol (2)
N i t r i t e - S u l f a m i c Acid
NHzSOjH
+“ 0 2
e
Nz
+ HS04- + H20 + H+; = -95.6 kcal/mol (3)
T h e heat of the calcium-EGTA reaction is approximate since the EGTA exists in several forms at p H 8.0. T h e purpose of the present work was t o present details of the design of a new thermochemical flow reactor system and evaluate the analytical characteristics of the device in terms of its linear dynamic range, precision, and throughput. EXPERIMENTAL Reagents. The acid-base neutralization system employed a reagent buffer of 0.5 M tris(hydroxymethy1)aminomethane (Tham, buffer grade, Sigma Chemical Co., St. Louis, Mo. 63178), 100 mM reagent grade sodium chloride, and 1 mM ethylenediaminetetraacetic acid-disodium salt (EDTA, “Baker Analyzed”, J. T . Baker Chemical Co., Phillipsburg, N.J. 08865) adjusted to a pH of 8.5 with hydrochloric acid. The sample buffer was 100 mM sodium chloride. Samples were prepared by serial dilution of a hydrochloric acid solution calibrated against primary standard grade Tham (Sigma Chemical Co.). All samples contained 100 mM sodium chloride. The compleximetrjc reagent was 0.25 M ethylene bis(oxyethylenenitri1o)tetraacetic acid (EGTA, Sigma Chemical Co.) dissolved in a 0.5 M Tham (Sigma Chemical Co.) buffer adjusted to pH 8.0 with hydrochloric acid. Sample buffer was a 0.5 M Tham solution at pH 8.0. Samples were prepared by dissolving
279
[.-I
4
Figure 1. Schematic drawing of the flow enthalpimeter modified for mixing soluble reagents. (A) Sample buffer, (B) reagent buffer, (C) peristaitic pump, (D) pressure gage-pulse suppressor, (E) insubted water bath, (F) pre-equilibration coil, (G) injection valve, (H,, H2) matched equilibration coils, (I) mixing Tee connector, (J) reactor column, (K) sensing thermistor, (L) equilibration coil matched to H, and H, (M) reference thermistor, (0)ac phase-lock bridge
the appropriate amount of reagent grade calcium chloride dihydrate in the sample buffer. An interference study was performed by adding reagent grade magnesium chloride to the sample. The redox reaction reagent buffer contained 50 mM sulfamic acid (certified, Fisher Scientific Co., Fairlawn, N.J. 07410) dissolved in 0.1 M hydrochloric acid. The sample buffer was 0.1 M hydrochloric acid. Nitrite samples were prepared by weighing out the appropriate mass of reagent grade sodium nitrite into the sample buffer. No attempt was made to exclude oxygen since all reagents were discarded after 24 h. Deionized water was used for the preparation of all reagents. Instrumentation. The construction of a flow enthalpimeter using an immobilized enzyme reactor was described in detail earlier (12). The components used for the soluble reagent system were the same as described previously, although the system i s configured very differently. Figure 1 shows a schematic of the modified flow system. The sample (A) and reagent (B) buffer streams are pumped by a two-channel Masterflex peristaltic pump (C) (Model 7545 pump and controller, Model 7013 and 7013-20 heads, Model 6408041 tubing, Cole-Parmer Instrument Co., Chicago, Ill. 60648) through a pulse suppressor and pressure gage assembly (D) into the stirred, insulated water bath (E). The sample stream passes through the stainless steel pre-equilibration coil (F) and into the injection valve ( G ) . Both streams then pass through matched equilibration coils (HI) and (H2)constructed of stainless steel needle tubing (95 cm, 21 gage, 0.02-inch i.d., Precision Sampling Corp., Baton Rouge, La. 70815) and enter opposing channels of a Tee connector (I) (0.07-cm i.d. channels, Altex Scientific, Berkeley, Calif. 94710) where they are mixed. The flow then passes through a 1-cm Teflon fitting into the reactor column (4. I t is important that this fitting be drilled with a channel no wider than 0.07-cm id., giving a total pre-column mixing volume of 80 pL. A second fitting having a 0.128-cm i.d. channel yielding a pre-column mixing volume of 75 MLwas also evaluated. For all experiments, an “adiabatic” column 3.3 cm long with an internal diameter of 3 mm and an approximate volume of 0.42 mL was used and was packed with acid-washed 80-120 mesh solid glass beads (A. H. Thomas Co., Philadelphia, Pa. 19106). In some experiments the glass was treated with dimethyldichlorosilanp (DMDCS) by the method of Robinson (13). For this work, the glass was dried at 150 “C, rather than the recommended 500 “C, in order to remove bound water without disturbing the surface silanol groups. After passing through the reactor column, the solution passes the sensing thermistor (K) and through a third stainless steel equilibration coil (L). Finally, the buffer passes a reference thermistor mounted in a Tee fitting (M) and out to waste (N). The temperature difference measured by the thermistors is converted to a disbalance voltage by a differential ac phase-lock Wheatstone bridge ( 0 ) . Procedure. The instrument is operated in exactly the same manner as the immobilized enzyme reactor system. Five minutes before the first analysis, the stirring motor and pump are started to ensure full thermal equilibration as indicated by a flat base line. Flow rate is measured at the system outlet as the time
280
ANALYTICAL CHEMISTRY, VOL. 51, NO. 2, FEBRUARY 1979
Table I. Calibration Curve-Hydrochloric Acid- Tham Reactiona sample concenpeak tration, height ,b cv ,c mM moC % blank 0.1 0.5
Wl12,
mL
0.00 0.00
0.40 19 0.72 1.13 16 0.68 5 8.87 5.0 0.59 10 20.4 1.1 0.51 20 45.7 1.1 0.49 50 116 0.68 0.49 100 225 3.1 0.49 500 860 0.23 0.59 1000 1025 0.38 0.69 a Conditions: 120-pL samples; flow rate = 2.5 mL/min; Linear least squares pH = 8.5, 0.5 M Tham buffer. analysis (0.5-100 mM): slope = 2.28 i 0.02 (m"C/mM); Correlation coeffiy intercept = 1 . 1 3 i 1.06 (m"C). cient = 1.000; Student's t = 99.5. 1
required to fill a vessel of known volume. In this case, no sample dilution was necessary and the sample was merely drawn into the sampling valve and injected. Apparent temperature changes were calculated as the product of the measured resistive disbalance of the thermistors and their approximate thermal coefficient of -400 Q/'C. Peak width at half maximum height (q2) was converted to volume units via the measured flow rate.
iClrn?+
Figure 2. Representative recorder tracing for the determination of hydrochloric acid. Conditions: 120-bL samples, flow rate = 2.5 mL/min, pH = 8.5. Concentrations (mM): 1, 5, 10, 20, 20, 10, 5, 1 Table 11. Calibration Curve- -Calcium-EGTA Reaction' sample peak concentration, height,b CV,' Wl,2? mM moC % mL blank
0.00
0.1
0.37 1.05 2.36 13.6 28.2 58.2 150 301
0.5 1
5
RESULTS A N D D I S C U S S I O N P r e l i m i n a r y E x p e r i m e n t s . A number of preliminary
10
designs were tried. There are several important design factors which were discovered. First, both the signal and reference thermistors must be placed in series with respect to the total fluid flow. If this is not done, the base line wanders because of slight changes in flow rate since self-heated thermistors are quite flow sensitive (14). The alternative to this is to use very stable pumps or decrease the voltage applied to the thermistors (which was approximately 0.5 V peak to peak in this work) a t the expense of some loss in sensitivity and temperature resolution (15-1 7). Second, placement of the reference thermistor in an "X" connector at the point of mixing of the two fluid streams results in a negative pre-peak due to observation of the reaction which takes place around the thermistor. Third, attempts to design a reactor where one flow tube was joined to the column along its axis and the other tube joined to the column through a radial aperture gave very irreproducible peaks. This design was an attempt to avoid reaction outside an "adiabatic" column. We believe that this irreproducibility is a result of very inefficient radial dispersion and incomplete mixing. T h e system described above is a compromise (vide infra) design which avoids the above problems. It is interesting to note t h a t placement of both the signal and reference thermistors after the packed column introduces some thermal smoothing. Second, no double peaks were observed, indicating t h a t all heat produced in the column was completely discharged t o the heat sink. Linear Dynamic Range. All three chemical systems were evaluated with respect to their linear dynamic range. In general, the upper limit was determined by the concentration of the reagent in the reactant stream and the lower limit by the base-line noise and drift of the instrument. For all systems the linear range extended over a t least two orders of magnitude. T h e appropriate calibration curves and supporting data appear in Tables 1-111. Figure 2 shows a representative recorder tracing for the analysis of hydrochloric acid. Sample carry-over, as can be seen, is negligible.
100
20 50
14.2 16.7 4.0 3.5 2.2 1.2 2.4 1.9
0.61
0.61 0.62 0.61 0.59 0.59 0.59 0.59
10 mM Ca2+
28.2 2.4 0.59 32.1 1.3 0.59 a Conditions: 120-pL samples; flow rate = 2.5 mL/min; Linear least pH = 8.0, 0.25 M EGTA, 0.5 M Tham. squares analysis: slope = 3.02 i 0.01 (m"C/mM);y interCorrelation coefficient = cept = -1.04 ? 0.35 (m"C). 1.000; Student's t = 340. 1 mM Mg" 1 0 mM Hg2
+
Table 111. Calibration Curve-Nitrite-Sulfamic Acid Reaction' sample peak concentration, height,b cv,c WlI2, mM m3C YO mL blank
0.00
0.21 9.4 0.70 0.77 3.5 0.59 1.59 1.5 0.59 7.33 1.3 0.59 0.5 12.2 2.3 0.59 1.0 24.2 2.1 0.59 2 70.8 2.4 0.59 5d 10 125 6.1 0.59 a Conditions: 120-pL samples; flow rate = 2.5 mL/min; Linear 50 mM sulfamic acid, 0 . 1 M hydrochloric acid. least squares analysis: slope = 12.48 I0.07 (m"C/mM); Correlation coefficient y intercept = 0.13 i 0.27 (m"C). = 1.000; Student's t = 176. These data rejected as an outlier. 0.01 0.05 0.1
Sensitivity, as expected, was proportional to the heat of reaction. This accounts for both the high sensitivity and low limit of detection observed for the nitrite analysis. Precision is a function of the base-line stability. T h e high concentrations of T h a m used in several of the buffers increased flow noise and consequently increased both the noise and drift over the optimum levels of 50 b"C peak to peak and
ANALYTICAL CHEMISTRY, VOL. 51, NO. 2, FEBRUARY 1979
hC. COMCEMTSATION
(nY)
Figure 3. Calibration curve for hydrochloric acid at low concentration
showing the intercept effect. Solid line represents the least squares line for the entire calibration curve in Table I. Conditions: (0) Pre-column mixing volume = 30 pL, flow rate = 2.5 mL/min, pH = 8 5. (+) Pre-column mixing volume = 75 pL, flow rate = 2.5 mL/min, pH = 8.5. (x) Pre-column mixing volume = 75 pL, flow rate = 1.1 rnL/min, pH = 8.5
7 p"C/min, respectively, when water was present in both flow channels. As expected, the nitrite analysis, which did not utilize T h a m in either channel, had the best precision. This effect is probably due to the significantly greater viscosity of the 0.5 M Tham solutions over that of pure water. In addition, inexpensive, buffer grade Tham contains particles of an insoluble impurity having a diameter between 0.5 and 5 pm. These particles produce and deposit a yellow-brown color at the column inlet and gradually increase the system pressure drop. The column must be repacked when the pressure drop exceeds 20 psi in either channel at a flow rate of 2.5 mL/min. T h e problem may be eliminated by using primary standard grade T h a m or by passing the Tham solutions through a 0.45-pm Millipore filter. Examination of Table I1 shows that the determination of calcium with EGTA is much less susceptible to interference by magnesium than the corresponding EDTA based assay would be. Equimolar magnesium, in fact, only contributes a 14% interference in peak height over that obtained with a pure calcium sample despite the large excess of EGTA present. We have shown previously that calcium may be titrated thermometrically with this reagent without interference (18). Effect of Concentration and System Variables on Peak Width. Examination of Tables 1-111 shows that only the hydrochloric acid studies show significant variations in peak width as a function of concentration. Increased spreading at high concentrations due to incomplete conversion has been observed previously with the immobilized enzyme reactor system (12). This was, however, thought to be related to a slow zero-order kinetic process which could not occur in this system, even at the 500 mM or loo0 mM concentrations where complete reaction is stoichiometrically impossible. The similarity of the effect to that seen with the immobilized enzyme reactor, however, bears further study. It is clear that our understanding of microreactors is quite incomplete. T h e increase of peak width at low concentrations has not been observed with any other system. This increased peak spreading correlated with the presence of an intercept in the calibration curve which is a result of nonlinearity in the calibration curve a t low sample concentrations. Figure 3 compares this effect for several different configurations of the system. It appears t h a t reducing the pre-column mixing residence time, either by decreasing dead volume or increasing the flow rate, reduces the intercept. This may correspond to thermal interactions with the plastic fittings which have been observed previously (12) or to the amount of mixing and reaction which occurs before entering the adiabatic region. The direction of the intercept corresponds to a loss of sample or of heat. T h e area under a thermal peak should be independent of variations in peak width or asymmetry which result from
281
~ g ~ ; ~ ~y, ~