ACKNOWLEDGMENT The authors thank Wolfgang Kladnig for numerous helpful discussions. The authors also thank Rudolf Schoenholzer for technical assistance during the course of this work.
LITERATURE CITED (1) C. R. Adams, A. F. Sartor, and J. G. Welch, "Standarization of Catalyst Test Methods", S.W. Weller, Ed., ACHE Symposium Series, American institute of Chemical Engineers, 70, 49 (1974). (2)"Methods of Analysis for Fluid Cracking Catalysts", W. R. Grace & Co., Baltimore, Md., 1972,p 32. (3)Carl L. Rollinson in "Comprehensive Inorganic Chemistry", Vol. 3,J. C. Baiiar, H. J. Emeleus, R. Nyholm, and A. F Trotman-Dickenson, Ed., Pergamon Press, New York, 1973,pp 623-769. (4)J. J. Phillipson in "Catalyst Handbook", Springer-Verlag. New York, Inc., 1970,pp 46-63.
(5) F. B. Sherman, V. A. Klimova, Yu S.Khodakov, V. S.Nakhshunov, I. P. Yakovlev, and Kh. Minachev, lzv. Akad. Nauk. SSSR., Ser. Khim., 764 (1972);Anal. Abstr., 24, 1445,1973. (6)A. P. Bolton, J. Cafal., 22, 9 (1971). (7)H. Hattori and T. Shiba, J. Cafal., 12, 1 1 1 (1968). (8)D. J. C. Yates, Catal. Rev., 2, 113 (1968). (9)R. M. Pearson, J. Catal., 23, 338 (1971). (10)R. M. Pearson, Abstracts, First Chemical Congress of the North American Continent, Mexico City, Nov. 1975,Petr 2. (11)N. C. Rasmussen and Y. Hukai, Trans. Am. Nucl. SOC., 12,29(1967). (12)P. K. Aditya and A. K. Batra, lndian J. Pure Appl. Phys., 7 , 323 (1969). (13)M. Heurtebise, H. Buenafama, and J. A. Lubkowitz, Anal. Chem., 48, 1969
(1976). (14)M. Heurtebise and J. A. Lubkowitz, J. Radioanal. Chem., 31 503 (1976). (15)D. S.Maciver, H. H. Tobin and R. T.Barth, J. Catal., 2, 485 (1963).
RECEIVEDfor review May 11,1976. Accepted August 30,1976. This work was supported by Grant No. 31.26-S1-0394 from the Consejo Nacional de Investigaciones Cientificas y Tecnpl6gicas (CONICIT).
Analysis of Hydrocarbon Fractions by Carbon and Proton Fourier Transform Nuclear Magnetic Resonance Spectrometry Harry'C. Dorn' and David L. Wooton Department of Chemistry, Virginia Polytechnic Institute and State University, Blacksburg, Va. 2406 1
A combined 'H and 13CFourier transform nuclear magnetic resonance (NMR) approach is employed in a model study of phenanthrene. The NMR method utilized a reference compound contalning carbon and hydrogen whlch Is added to the sample of interest. The H / C ratio, weight percent carbon, and hydrogen data obtained for phenanthrene are presented, and posslble sources of error In this approach are indicated. Factors influencing quantitatlve 13Cmeasurements In the Fourler transform mode are also discussed.
It has been recognized for many years that both l H and 13C nuclear magnetic resonance spectrometry are useful tools for quantitative characterization of the various types of carbons and hydrogens in a given hydrocarbon skeletal framework, (e.g., aromatic carbons, alkyl hydrogens, etc.). A number of procedures have been devised for detailed characterization of coal and/or petroleum fractions in terms of a hypothetical average molecule using either IH NMR alone or in combination with 13C NMR (1-5). The NMR spectral data are normally used in conjunction with other analytical data, such as elemental combustion analysis and molecular weight measurements, to obtain a number of important molecular parameters. For example, the method described by Knight (2) requires lH and 13C NMR, elemental combustion, and molecular weight analytical data for complete characterization of petroleum fractions. A potential problem exists in approaches of this type if direct comparisons are made between the measurements usually carried out on solutions (NMR and molecular weight) and the elemental combustion data. That is, the bulk elemental combustion C,H data may not represent the actual material dissolved in the solution NMR measurements. This is an important consideration in the case of complex mixtures of the type encountered in coal and petroleum research. Thus, an alternative method allowing determination of solute weight percent carbon and hydrogen in solution could have certain advantages and complement the normal elemental combustion data. 2146
*
Another major consideration in quantitative l3C NMR studies, concerns the experimental mode(s) employed in obtaining the 13C spectral data. The advent of Fourier transform (FT) and proton pseudo random-noise decoupling techniques are very useful for increasing the signal-to-noise ratio per unit time for 1% signals; however, a number of factors usually alter the simple relationship between intensity and the number of nuclei a t resonance using these experimental modes. The major factors influencing quantitative measurements in the FT mode have been discussed by several authors (6-8) and are listed here: 1)Nuclear Overhauser effects. 2) Spin-lattice relaxation recovery times (Tl's) and corresponding dependence on pulse repetition rate. 3) Adequate digitization in the frequency domain FT spectrum. 4) The effects of limited pulse strength ( H I )on the 13Cspectral region (AFHz) to be covered (6).
Of these, the nuclear Overhauser effect, 1, and TI dependence on pulse repetition rate, 2, are the two of primary concern in quantitative 13CFT measurements. Fortunately, it has been found that paramagnetic relaxation reagents which are added to the sample provide a convenient means of suppressing but not equalizing the Overhauser effects, 1, and a t the same time decrease T1 values (9-12). The resulting decrease in 71' values allows more rapid pulse repetition rates, 2. A second method of avoiding the effects of 1and 2 involves the use of a gated lH decoupling sequence (13)where the l H decoupler is only on during the time the l3C signal is monitored. In this sequence, a lH decoupled 13C spectrum is obtained with suppression of the nuclear Overhauser effects if rather low duty cycles for the l H decoupler are employed as well as long pulse repetition intervals. The pulse repetition intervals are normally four to five times longer than the carbon Tl's of the sample. The extension of these quantitative 13C PFT techniques to studies involving petroleum and/or coal fractions has apparently not been undertaken. Ih view of the considerations above, a model study was initiated aimed at 1)developing a complementary I3C and lH NMR approach for determining a number of molecular pa-
ANALYTICAL CHEMISTRY, VOL. 48, NO. 14, DECEMBER 1976
rameters, such as, H/C ratio, weight percent C and H, and 2) an examination of methods for obtaining quantitative 13C NMR data in the FT mode. In order to test the accuracy and precision of the combined l H and 13CFT approach, a model study was initiated. The model compound chosen was phenanthrene, an aromatic hydrocarbon commonly encountered in coal liquefaction processes (14-16). 'H and 13CPT Method. The combined lH and 13Cmethod employed here relies on the use of a reference material containing carbon and hydrogen which is added to the sample of interest (phenanthrene). Although p-dioxane was found to be a satisfactory reference for this study, other compounds could be employed. The 13C FT spectra are recorded under conditions which assume that the area of each 13Csignal within the spectrum (sample and reference) will be the same for equal numbers of nuclei at resonance; that is, the techniques previously noted are used to suppress the "extra" factors normally influencing 13Cquantitative measurements in the FT mode. Integration of the 13C spectrum provides the area ratio between the sample and the reference. In a similar manner, the l H FT NMR spectrum is then recorded and integrated for the same sample used in the I3C measurements. In this case, relatively long repetition intervals were also employed to allow complete recovery of the lH magnetization. The l H and I3C spectra obtained under these conditions provide the H/C ratio for the sample directly.
The parameters and symbols used in Equation 1are defined as: (Htot) = total number of hydrogen atoms per sample molecule. (Ctot) = total number of carbon atoms per sample molecule. (H), = total number of hydrogens per reference molecule (e.g., 8 for dioxane). (C), = total number of carbons per reference molecule (e.g., 4 for dioxane). (IHtOt)/(IH),= the integrated l H intensity ratio for the sample l H signals vs the reference. (IC),/(ICtot)= the integrated 13C intensity ratio for the reference vs. the sample 13C signals. Although the use of Equation 1does not require a known weight of the reference or the sample, the weight percent carbon and hydrogen can be obtained from the IH or 13C NMR data if weighed amounts of reference and sample are used. The equations applicable under these conditions are given below.
wt%H=-
Htot
(mol wt)
x 100
(3)
(4) w t % C = - mol wt x 1200
(5)
The parameter (mol), is the number of moles of reference (dioxane) in the NMR sample. All other terms in Equations 2-5 have been previously defined. The total weight percent hydrogen and carbon calculated by means of Equations 2-5 can be directly compared with values obtained from elemental combustion data. However the advantage of the combined l H and 13C NMR approach is more keenly appreciated when these equations are recast in terms of individual carbon and hydrogen types (e&, Carom, Harem Calkyl, etc.). Equations 6 and 7 illustrate this point for the aromatic carbons in the sample of interest.
\"I
barom wt % carom =x 1200 mol wt
(7)
The Car,, and ICaromrepresent the number of aromatic carbon atoms per sample molecule and integrated intensity for the sample 13C aromatic region, respectively. Similar equations can be written for any carbon or hydrogen types provided that the corresponding l3C and/or l H chemical shifts can be assigned and separately integrated. Further characterization of the sample requires molecular weight data normally obtained via vapor pressure osmometry measurements. With the molecular weight data and use of equations, such as 2,4, and 6, the number of carbons or hydrogens per sample molecule of a given type (e.g., Carom)can then be easily calculated. The number of carbons and hydrogens of a given type are useful in obtaining a number of important additional molecular parameters. As an example, the number of aromatic rings per sample molecule (R,) can then be determined using equations presented by Knight ( Z ) , if the total number of aromatic and substitutable aromatic ring carbon atoms per molecule are known.
EXPERIMENTAL A Jeolco PS-100 nuclear magnetic resonance spectrometer was used to obtain the IH and 13Cspectra a t 100 and 25.1 MHz, respectively. The spectrometer was used in the Fourier transform mode for 13Cand 'H with a Jeolco EC-100 data system. For the 'H spectra, data were also obtained in the continuous wave (CW) mode. The NMR spectrometer was used with an internal deuterium lock system operation a t 15.35 MHz. The NMR samples and reference were contained in 8-mm tubes. In addition, a concentric 3-mm capillary tube containing deuterium oxide was placed inside the 8-mm tubes for lock stabilization. The IH gated 13C F T NMR experiments were performed with hardware and software provided by Jeolco, Inc. In this mode, the proton decoupler is on only during the time interval ( t l that the magnetization is being monitored and is off for a much longer time interval (7'). The time intervals for t and T were 1.31 and 5.0 s, respectively, with N = 8K. A total of 200 free induction decays were collected and time-averaged in each I3C NMR experiment. In the IH F T NMR measurements, the data acquisition time ( t ) and number of data points ( N )were the same as those employed in the corresponding I3C NMR measurements. In this case 1-20 free induction decays were collected for Fourier transformation. In the 'H and I3C F T NMR experiments, the flip angles used were 90" rf pulses, 50 and 20 I S , respectively. The integrations were carried out with the standard software provided by Jeolco, Inc. The choice of the reference is dictated by a number of factors, such as possible IH and/or 13C spectral overlap with sample of interest, solubility in typical NMR solvents, volatility, lack of chemical reactivity, etc. Although p-dioxane was used in the present study, a number of other compounds are being examined as possible reference materials. Tris(acetylacetonato)chromium(III) (Cr(acac)a) was used as the relaxation reagent; however, potentially "better" relaxation reagents ( I 7) have been suggested which minimize interaction between the metal chelate and substrate molecules (reference and sample). The NMR samples were prepared by weighing the reference ( p dioxane), sample (phenanthrene), and Cr(acac)a directly into the NMR sample tube. The NMR sample used in obtaining the data reported in Tables I and I1 consisted of 0.3865 g of p-dioxane (spectrograde, MCB), 0.6451 g of phenanthrene (Eastman Kodak), 0.0261 g of tris(acetylacetonato)chromium(III),and 1.6 g of methylene chloride (Fischer) added as the solvent. The lH and 13C F T NMR measurements were carried out within a period of 1-3 days. The capped NMR sample was stored a t -10 "C in a cold storage room during the interval between the lH and 13C NMR measurements.
RESULTS AND DISCUSSION A number of preliminary lH and 13C FT NMR experiments were carried out in order to establish the appropriate experimental conditions for this study. The results of these pre-
ANALYTICAL CHEMISTRY, VOL. 48, NO. 14, DECEMBER 1976
2147
Table I. Relative I3CFT NMR Signal Areas for Phenanthrene 61 CZ,C3 c4 c9 c11 ClZ Relative areas 1.0a 2.08 f 0.04 1.00 & 0.03 1.05 f 0.03 1.00 f 0.04 1.00 f 0.04 kPPm 128.9 126.9 123.0 127.2 132.4 130.7 a Values relative to C1; standard deviations are based on 4 repetitive 13C FT NMR measurements. 1% assignments based on data presented in Ref. 19; I3C chemical shifts referenced to TMS using the relationship, GTMS = ddjoxane 67.4 ppm.
+
lr
Table TI. ‘H and 13C FT NMR Data for Phenanthrene Phenanthrene H/Cratios w t % C w t % H C,,,, H,,,, 5.EiC 14.2d 9.fje NMR values 0.69 f 0.04a 96.0b 5.66 14 10 Actual values 0.71 94.34 a Value determined from Equation 1; standard deviation is based on 3 and 4 repetitive lH and 13C FT NMR measurements, respectively. Value determined from data only; Equation 5. C Value determined from lH data only; Equation 3. Value based on assumed molecular weight and Equation 4. e Value based on assumed molecular weight and Equation 2.
Y
lill I
150
liminary experiments were informative in providing insight regarding potential sources of experimental error. Although definite conclusions were not reached with respect to all the experimental parameters that were varied, a number of guidelines were established and are outlined below. 1) Variation of the relative concentration of sample and reference suggest that the IH and I3C integrated intensity ratios between the sample and reference, (e.g., (Ic,/Ic))should be in the range of 0.1 to 10 for accurate data. 2) The use of just the gated lH decoupling sequence without addition of the paramagnetic relaxation reagent, (Cr(acac),) was found to be less desirable than employing both techniques simultaneously. The advantage of using both techniques was not only in terms of the precision and accuracy of the calculated H/C ratio, but the practical problem of the long pulse repetition times ( T )necessary for this sample without the relaxation reagent. However, it should be mentioned that the addition of the paramagnetic relaxation reagents may not be necessary in certain coal or petroleum samples if paramagnetic materials are already present in relatively high concentration (18).
3) The precision of the results improved somewhat when the numbers of data points ( N )were increased in the FT data collection while leaving the same sampling frequency constant (e.g., increase in data acquisition time). This was particularly evident in the cases where Cr(acac)3 was not added to the sample. Apparently line broadening in the case of added Cr(acac)3helps provide better signal representation for a fixed value of ( N ) . Preliminary 13C FT NMR experiments utilizing the gated 1H decoupling sequence with variable delay times (5“ = 0-12 s) indicated that the integrated areas did not change for values of 2’ 2 3-5 s with Cr(acac)3. The relative areas for the individual 13C signals are presented in Table I ( T = 5 s) and indicate the quantitative “line to line” self-consistency of this method. Based on the insight provided by these initial experiments, the experimental conditions presented in the Experimental section were employed in obtaining the data in Table 11.The actual values reported for the Car,, and Ha,,, are based on knowledge of the actual molecular weight. In an unknown sample this would have to be obtained from molecular weight measurements (e.g., vapor phase osmometry). A typical 13C spectrum is also illustrated in Figure 1. The uncertainty indicated for the values in Tables I and I1 2148
130
Y I U
110
80
70
50
PPM
Flgure 1. Fourier transform I3C NMR spectrum far phenanthrene, methylene chlaride, and pdioxane
are not unreasonable when compared to similar NMR determinations (3).The Car,, and H,,,, allow an estimate of the number of aromatic rings per molecule using the equations presented by Knight, a value of 3.2 was obtained. This value is based only on the 1H and l3C NMR data and assumed molecular weight. A typical 13C spectrum i s also illustrated in Figure 1. For comparative purposes continuous wave (CW) IH NMR spectra were also recorded for the same sample used in obtaining the data in Table 11. For this sample and other samples as well, the H/C ratios from the CW 1H and FT 13C spectra were generally higher than those obtained by the dual FT approach.
CONCLUSIONS The combined ‘H and 13C FT NMR method described in this model study of phenanthrene provides a convenient approach for determination of the H/C ratio, weight percent carbon, and weight percent hydrogen. The results of this study suggest that a reasonable level of accuracy and precision can be achieved if careful attention is given to the experimental variables. Although this study was limited to phenanthrene, this approach has successfully been applied to a solvent refined coal (SRC)sample (18).A possible extension of this approach could be envisioned to include 33S,170,and I4N NMR data with the appropriate reference material. This could in principle provide weight percent data and functional group analysis of these elements. However, sensitivity limitations of these “NMR active” nuclides will probably limit this extension.
LITERATURE CITED (1) R. A. Frledei and H.L. Retcofsky, Cbem. Ind. (London),1866, 455. (2) S. A. Knight, Chem. Ind. (London), 1987, 1920. ( 3 ) D. R. Clutter, L. Petrakls, R. L. Stenger, and R. K. Jensen, Anal. Chern., 44, 1395 (1972). (4) J. K. Brown and W. R. Ladner, Fue/(London), 39, 87 (1967). (5) R. B.Williams, “Symposlum on Composltlon of Petroleum Oils, Determination and Evaluation”, ASTM Spec. Tech. Pub/., 224, 168-84 (1858). (6) R. Freeman and H. D.W. Hill, J. Magn. Reson., 4, 366 (1871). (7) J. 8 . Stothers, in “Topics in Carbon-13 NMR Spectroscopy”, G. C. Levy, Ed., Wiley-interscience, New York, 1974, pp 232-233. (8) Q.A. Gray, Anal. Chem., 47, 546A (1975). (9) G,N. LaMar, J. Am. Chem. Soc., 83, 1040 (1971).
ANALYTICAL CHEMISTRY, VOL. 48, NO. 14, DECEMBER 1978