Signal Enhancement in Nuclear Magnetic Resonance Spectrometry

(13) West, R., Hunt, R. H., Whipple,. R. O., Ibid., 76,310(1954). J. A. Magnuson. E. W. Knaub. Silicone Products Department. General Electric Co. Wate...
2 downloads 0 Views 320KB Size
(4) Dostal, P., Cermak, J. Novotna, B., Collection Czech. Chem. Commun. 30, 34 (1965). (5) Fritz, J. S., Schenk, G. H., ANAL. CHEM.31, 1808 (1959). (6) Jay, R. R., Ibid., 36, 667 (1964). ( 7 ) Klimova, V. A., Zabrodina, K. S., Shitikova, N. L., Izv. Akad. A-auk SSR, Ser. Khim. 1 , 178 (1965); CA 62, 11152d (1965).

(8) Magnuson, J. A., AXAL. CHEM.35, 1487 (1963). (9) Pike, R. M., Fournier, A. F., Rec. Trav. Chim. 81, 475 (1962). (10) Rochow, E. G., Hurd, D. I., Lewis, R. N., “The Chemistry of Organometallic Compounds,” p. 169, Wiley, New York, 1957. (11) Stetzler, R. S., Smullin, C. F., ANAL. CHEM.34, 194 (1962).

(12) Tabern, D. L., Orndorff, W. R., Dennis, M. L., J . Am. Chem. SOC.47, 2043 (1925). (13) West, R., Hunt, ‘R.H., Whipple, R. O., Ibid., 76,310 (1954). J. A. MAGNUSON E. W. KNAUB

Silicone Products Department General Electric Co. Waterford, N. Y.

Signal Enhancement in Nuclear Magnetic Resonance Spectrometry SIR: one of the serious limitations of experimental nuclear magnetic resonance (NMR) spectrometry is its relatively low sensitivity as compared with, for example, ultraviolet spectrometry. This factor creates problems in the case of scarce or sparingly soluble materials. Solubility or availability of material can thus rule out numerous worthwhile studies. Recently, attempts have been made t o improve the signalto-noise ratio of data provided by KMR spectrometers and thus to increase the effective sensitivity of the instrument. The technique of using a multichannel storage device to integrate the signal over a number of sweeps is one such method (for example, the computer of average transients, (CAT) Alnemotron Corp., 45 South Main St., Pearl River, L’.Y.). Also a simple and inexpensive automatic procedure which makes use of the spectrometer integrator has been developed in this laboratory (2). Both of these techniques take advantage of the fact that over a long period of time random noise will average to zero. I n this work we have employed the integrator method t o record a portion

Figure 1.

of the proton spectrum of a material which was present a t low concentration. The results of a further effort to suppress noise by the use of numerical smoothing procedures are presented. The methods used were all variations of the weighted moving average technique which has been described in the literature with a suggestion of its applicability to the treatment of XMR data (3). In practice, the procedure involves averaging a set of data containing a fixed number of observations which occur at equally spaced intervals. Successive new data sets are obtained by advancing in unit intervals, and the averaging is repeated. The mathematical expression for this operation is

EXPERIMENTAL

All spectra were recorded on a Varian model A-60 N M R spectrometer using the 50 C.P.S. sweep width setting.

k=n

Ij(smoothed)

= CSIj+k k = -n

(obs.)

+

where 2 n 1 is the number of points over which the average is taken and the Ckfs are weighting factors which include normalization. By proper choice of the Ckfs a variety of filtering functions can be obtained.

Vinyl proton region of methyl vinyl ketone at 60 mc./sec.

Sweep width, 50 cps.

Upper spectrum, -50% in CDCh and lower spectrum, -.06% Brackets indicate the region studied

in CDCIS

Figure 2. studied VOL. 37,

NO.

Integrator spectrum of region

12, NOVEMBER 1 9 6 5

1609

. .

Figure 4.

Figure 3. Results of smoothing a single sweep with equal weighting of points From left to right, numbers of points used were 3, 5, 7, and 9

Integration of the signal was performed for 20 seconds every 0.1 C.P.S. and the results were charted by the instrument recorder. Sideband calibration was performed with a Hewlett-Packard 2005 audio oscillator and He\Vlett-Packard 5 2 2 ~frequency counter. A total of four forward sweeps was recorded as described. The integrated signal intensities were measured by hand and the values transferred to paper tape for processing. The studied consisted Of ketone in terated chloroform The ketone concentration was approximately 0 . ~ 6 % by weight, At this concentration, the vinyl protons were undetectable when the spectrometer was operating a t maximum gain. Figure 1 shows the proton spectrum of methyl vinyl ketone and a hjgh-gain Scan of the sample studied. That portion of the spectrum which was for this study was the indicated high field doublet which exhibits a splitting of 1,1 c,p,s. A typical integrator sweep of the two peaks is shown in Figure 2. CALCULATIONS AND RESULTS

Smoothing of the signal was accomplished through the use of an IBM 1620 I1 digital computer for which simple Fortran programs were written. The two types of simple smoothing procedures applied t o the data were the unweighted moving average and the

Triangular smoothing of single sweep

Numbers of points used were 5, 7, 9, and 1 1

symmetrical triangular weighted moving average. In addition, the cubic-quartic set of least squares convolute integers was applied to the data. The use of these integers corresponds to fitting a cubic or quartic equation to the 2 n 4- 1 Points and evaluating the function a t the mid-point (1, 3, 4). The results of these smoothing efforts are illustrated in Figures 3, 4, and 5 . That the unweighted moving average function is the least effective in removing noise from the signal is apparent. The least squares and the triangular convolute functions appear to produce somewhat similar results. AS is expected, increasing the number of points included in the weighted moving average resulted in a smoother curve, but the two peaks were correspon,-Jingly less well resolved, The effect of averaging four scans and smoothing the result is shown in Figure 6. I n this case, the triangular function included 7 points and the 11point set of integers was used for the least squares smoothing. Comparison of these curves with the previous results demonstrates that averaging several sweeps significantly improves the signalto-noise ratio. Increasing the number of points in the weighted moving average and averaging several scans before smoothing have similar results. However, averaging several sweeps has the advantage of eliminating low fre-

quency noise which is not too well suppressed by smoothing a single sweep. An interesting feature of these two peaks was uncovered by the smoothing process. The base line difference between the two sides of the peaks suggests that the detector phase was not adjusted properly* This condition was not easily noticed in the integrator spectrum because Of the low signal-to-noise ratio. CONCLUSIONS

The procedure described above is particularly well suited for use with the automated integrator spectra, because they already consist of a sequence of discrete intensity measurements a t uniform frequency intervals. For efficient use, however, the spectrometer output should be automatically transferred directly to tape. Our results rather clearly demonstrate that a much more dilute sample could easily have been studied using the described technique. If necessary, the signal-to-noise ratio could be further increased by any one of, or all of, three means : increasing the density of observations would permit the use of more points in the moving average while not increasing distortion of the peaks;

Figure 6. Results of averaging four sweeps and smoothing the resulting data Figure 5.

Least squares smoothing of single sweep

Numbers of points used were 7, 9, 1 1, and 1 3

1610

ANALYTICAL CHEMISTRY

Triangular 7-point smoothing was used on the left curve and 1 1-point least squares smoothing was used on the right curve

increasing the time allowed for signal integration would improve the signal-tonoise ratio a t an early stage; using a greater number of scans would also produce much the same effect. In summary, any method leading to increased time averaging results in an increased signal-to-noise ratio. The signal enhancement achieved here is not necessarily optimal or even typical. In particular a greater density of points would be desirable from the standpoint of resolution. The density of observations obtained in the present study was limited by the necessity of recording distinct lines.

With automated data collection devices, this restriction would be removed and any desired density of points could easily be obtained. Nevertheless, these results do suggest that the use of the integrator technique coupled with curve smoothing has the potential of dramatically improving the sensitivity of NMR spectrometers, and our belief is that application of this method merits further attention. ACKNOWLEDGMENT

The authors thank Roger W. Crecely for his assistance in several phases of this work.

LITERATURE CITED

(1) Guest, P. G., “Numerical Methods of Curve Fitting,” p. 349 ff., University Press, Cambridge, England, 1961. (2) Mayo, R. E., Goldstein, J. H., Rev. Sci. Znstr. 35, 1231 (1964). (3) Savitzky, A., Golay, M. J. E., ANAL. CHEM.33,1627 (1964). (4) Whittaker, E., Robinso? G., “The Calculus of Observations, p. 291 ff.,

Blackie and Son, Ltd., London and Glasgow, 1948. J. M. READ,JR. J. H. GOLDSTEIN Department of Chemistry Emory University Atlanta, Ga. WORKsupported by National Institute of Health Grant No. GM108 48-07.

A Device for Automatic Gradient or Stepwise Chromatography Application to the Separation of Peptide Mixtures by Gradient Chromatography on Dowex-1 SIR: The separation of peptide mixtures on Dowex-1 columns with a p H gradient made from volatile buffers has been previously described (for review see ref. 6). Schroeder and Robberson (7) have now improved their system by eliminating sharp drops in pH during the gradient from the starting basic medium of pH 9.4 to the acid medium of pH 2. The slightly sigmoid pH curve obtained is reproducible and has been shown to give excellent separation of peptides. Although complex mixtures can be separated using this resin ( 6 ) , its special value probably lies in its use to further fractionate simple mixtures of peptides obtained from initial separations on Dowex-50 columns. We have found it advantageous to rechromatograph many zones from Dowex-50 chro-

matograms on Dowex-1 and our experience has been that such a treatment results in peptides having virtually no contamination. To facilitate routine work of this type we have developed an automatic system for production of the new gradient described by Schroeder and Robberson (7) which includes automatic regeneration of the column after completion of the chromatogram. EXPERIMENTAL

The resin, Dowex-1 (200 to 400 mesh) was prepared as previously described (6). The buffers used were those described by Schroeder and Robberson ( 7 ) . Columns (0.6 X 60 cm.) were poured after equilibration of the resin with the starting buffer

of p H 9.4 and were thermostated a t 36” C. Table I shows the conditions used for chromatography. For the automatic production of the gradient a system of solenoid valves connected to a multiple sequence timer was used. Figure 1 depicts the system employed. The solenoid valves numbered 1, 2, 3, 4, and 5 in Figure 1 (Allied Control Co., Inc., New York, N. Y., Catalog No. 21384, 6-watt, 24volt 60 cycles, orifice 3/32, p.s.i. 80, stainless steel, normally closed) were connected a t the “In’, end to the buffer vessels A , B, C, D,and E. The rrOutJ’ ends were connected to the inlets of a mixing vessel. This mixer was a cylindrical borosilicate glass vessel (diameter 35 mm., height 40 mm.) with a male joint ST 40/35 having a borosilicate glass cover with a female joint

Table I. Conditions for Chromatography Column 0 . 6 X 60 cm. Temperature 36’ C. Flow rate 40 ml./hour Volume of mixer 40 ml. Fraction size 2 . 0 ml. Equilibrating buffer pH 9.4 Time on

Development A. pH 9.4 buffera B. pH 8.4 buffer C. pH 6.5 buffer D. 0.5N acetic acid E. 2.0N acetic acid

ml.

sequence timer, minutes

10 30 40 60 100

15 45 60 90 150

Volume,

Regeneration F. pH 9.4 buffer 80 120 a The composition, preparation, and storage of the buffers are as described by Schroeder and Robberson ( 7 ) .

U

I

I

I I

I 1

I I

Fraction collector

1 Stirrer

L _ . - _ _ _ _ _ _ _ I_ - _ _ _ - _ - _ _ _ Figure 1 .

I

I

I

L

I I

--_-_____

I -I

Diagram of the gradient producing system VOL. 37, NO. 12, NOVEMBER 1965

1611