Take Home Lesson - III

“The moral of this section is that there are numerous contributions to the error in a quantitative measurement made by FTNMR, and while each of them may be reduced to 1% or so in apractical fashion, the combined error is still likely to be significant. I am always skeptical of measurements purporting tobe accurate to better than a few percent overall, unless they come with evidence that careful attention has been paid to the above details.

Taken from Derome (p.172)

Reblog this post [with Zemanta]

V. Integration - II

Continuing from Integration - I

4. The spectrum should have a S/N of atleast 250:1 for the smallest peak to be integrated. The S/N measured for the quartet in Figure V-3, is 213, which is close to fulfilling this criterion. We can improve this number by choosing a line broadening that is approximately equal to the true (natural) linewidth of the peaks.

5. The baseline should be flat. Distortion due to phase problems should be corrected. Baseline distortion due to non-optimum parameter selection causing a baseline roll are not discussed here.

6. The peaks need to be sufficiently digitized, as discussed in Section II. If the linewidth at half-height is 1 Hz, you need a digital resolution of less than 0.5 Hz.

7. The same area should be included or excluded for all the peaks. For example, all peak integrals should be measured +/- 5 Hz around each peak, not +/- 20 Hz around one peak, +/- 10 Hz around a second peak, etc. Spinning sidebands are included in this category, and should consistently be either included or excluded.

The reason the integrals for the phenyl region is so inaccurate for the spectrum in Figures V-1 and V-2 is because the recycle time or pulse repetition time was too short. Obviously, if you did not know the phenyl region represented 5 instead of 4 protons, you could draw an erroneous conclusion about the structure of this compound from the integrals in Figures V-1 and V-2.

image025

Reblog this post [with Zemanta]

V. Integration- I

This section will show you how to obtain good integrals. Examples of spectra of 0.1% ethylbenzene in CDCl3 from the Unity+300 and VXR-S 400 NMR spectrometers are given in Figures V-1 and V-2 respectively. Both spectra were taken using the default parameters for acquiring 1H spectra. If we assign an integral of 3.00 to the CH3 triplet, then the phenyl region integrates to 4.52 protons, while the CH2 qurtet integrates to 2.08 protons. Thus the integral for the phenyl protons is 9.6% too small, while the integral for the CH2 quartet is off by 4.0%. The 9.6% error for the phenyl protons is not due to spectrometer error; it is because we have chosen parameters for acquiring the spectrum which guarentee we will get inaccurate integrals.

The accuracy of the integrals obtained for most routine spectra is usually about 10-20%. This accuracy is sometimes sufficient especially if you already know what the compound is. However, this accuracy is usually not adequate to determine the exact number of protons contributing to a given peak, nor is it sufficient for quantitative applications such as kinetic experiments or assays of product mixtures where one demands an accuracy of 1-2%. For example, 20% accuracy is not sufficient to decide whether two peaks have a relative ratio of 1:3 or 1:4. Obtaining 1-2% accuracy can be achieved but you need to be aware of the factors that affect integrations. These factors are discussed below.

1. There should be no nuclear Overhauser effect contributions or any other effects that selectively enhance certain peaks. This is a problem only with X nuclei such as 13C.

2. No peaks should be close to the ends of the spectrum. The spectral width should be large enough such that no peak is within 10% of the ends of the spectrum. This is because the spectrometer uses filters to filter out frequencies that are outside the spectral width. Unfortunately, the filters also tend to decrease the intensities of peaks near the ends of the spectrum. For example, at 299.96 MHz, if two peaks are separated by 7 ppm, a spectral width of atleast 2100 Hz is sufficient to get both peaks in the spectrum and prevent foldovers. However, to avoid distortion of the integral intensities because of filter effects, the spectral width should be set 10% larger on each side, 210 Hz, giving a total spectral width of about 2520 Hz (8.4 ppm). Although the standard setup parameters on the Unity+300 and VXR-S 400 should easily satisfy this criterion, you should be prepared to make the spectral width larger if necessary.

image023

3. The pulse repetition (recycle) time should be atleast five T1s. Data should be collected under conditions which ensure that all the nuclei can fully relax before the next FID is taken, i.e., if 90 degree pulse width are used, relaxation delays of 5xT1 of the longest T1 of interest are necessary. In the case of ethylbenzene, the longest T1 of interest is 9.8 sec for the phenyl protons, so the relaxation delay when using 900 pulse width should be at least 49 sec (5x9.8). In Figure V-3, is shown the spectrum of 0.1% ethylbenzene in CDCl3, using a 90 degree pulse width, realaxation delay of 60 sec, taking 32 transients, and using a line broadening of 0.1 Hz ( for comparison, the default parameters on the Unity+300 use a 45 degree flip angle and a recycle delay of 5.1 sec). If we assign an integral of 3.00 to the CH3 triplet, then the phenyl region integrates to 5.17 protons, while the CH2 quartet integrates to 1.98 protons. Thus, the integral for the CH2 quartet is off by only 1.0%, while the integral for the phenyl prtons is now 3.2% too high. The errors for both the phenyl protons and the CH2 protons are now comparable; they reflect a good choice for the recycle delay for this sample. However, the 3.2% error for the phenyl protons is still higher than the 1-2% error range one needs for accurate quantitaive measurements. In this particular example, the signal from the residual proton (CHCl3, 7.29 ppm) contributes to the integral of the phenyl region making it higher than the true value. This problem can be solved by measuring the spectrum in a solvent such as CD2Cl2 whose residual proton signal will not overlap with any of the ethylbenzene protons.

Reblog this post [with Zemanta]

Take Home Lesson - II

At some point, you may take a spectrum and wonder why the signals are so weak. Well over 75% of the time, the problem is not with the spectrometer, but with your sample. You can test this quickly by taking a spectrum of 1H or 13C standards such as ETB (or menthol). In this way, you can save yourself needless frustration on the instruments by identifying problems that are due to a bad sample.

 

 

Reblog this post [with Zemanta]

IV. Signal-to-Noise Measurement - II

Choice of a noise region must be consistently applied for standard samples, and for 0.1% ethylbenzene (ETB), we have chosen 5 to 3.5 ppm. It is also critical that the noise be amplified if necessary in order to make an accurate measurement using a ruler. There is not much point in

trying to compare the ratio of noise to a few millimeters high to a peak 20 cm high; plot the noise such that it is several centimeters high. If the instrument performs the signal-to-noise measurement and calculation, the relative intensities are not as critical. Both the Unity+300 and VXR-S 400 NMR spectrometers perform the S/N calculation by software commands.

The last point concerning sensitivity is demonstrated during the 13C{1H} part of the checkout. The checkout requires that you acquire three spectra, using 1, 4, and 16 transients. Signal-to-noise increases as the square root of the number of transients; to double the signal-to-noise you must take four times as many transients.

When using a concentrated sample such as 30% menthol for 13C, or when running routine 1H spectra, the number of transients or scans is quite small, so the point discussed above may not seem important. However, suppose you are in the following situation: you have only a few mg of research sample, and after collecting a 13C{1H} spectrum on the Unity+300 for 2 hours, you get peaks with S/N of only 5:1. Since the peaks are barely visible above the noise ( and you may have missed quarternary carbons), you want to re-collect the spectrum to get a S/N of 50:1, a value most typical for carbon NMR. Unfortunately, this will take 10x10x2 = 200 hours! In such cases, you should consider using a higher field strength spectrometer, which will be able to give you a spectrum with good S/N in much less time.

 

Reblog this post [with Zemanta]

IV. Signal-to-Noise Measurement - I

               The first spectrum described in the checkout involves determining a signal-to-noise measurement, or S/N. S/N is an important criterion for accurate integrations (see section V), and is also one of the best ways to determine how sensitive an NMR spectrometer is. In general, a higher S/N specification means that the instrument is more sensitive.

A typical result for the Unity+300 checkout is S/N = 87 and is shown in Figure IV-1, while a typical result for he VXR-S 400 checkout is S/N = 109 and is shown in Figure IV-2. If you measure S/N for the ethylbenzene standard that is significantly lower than these values, write your observations in the spectrometer logbook and inform S. Chandrasekaran .

S/N measurements for proton spectra are always determined using a sample of 0.1% ethylbenzene in CDCl3 (ETB). It is important that the spectrum be acquired under standard conditions:

1. 900 pulse width

2. Line Broadening = 1.0 Hz

3. Spectral Width = 15 to 5 ppm

4. A sufficient relaxation delay (at least 5xT1)

5. A sufficient digital resolution (less than 0.5 Hz/point)

6. One transient acquisition

Optimum signal-to-noise ratio for any sample is achieved using a line broadening equal to the linewidth at one-half height. When this line broadening is applied, the linewidth at half-height doubles, i.e., it is the sum of the natural linewidth at one-half height plus the line broadening applied. The equation used for calculating S/N is:

S/N = (2.5 x A) / Npp

A = height of the chosen peak Npp = peak-to-peak noise

Peak-to-peak noise means exactly that - a measurement from the most positive to the most negative positions for the noise.

As shown below, the wildest differences occur with low probability, so there seems to be a large gap between them and what appears to be the main part of the noise; will-power must be exercised in order to truly measure this gap and not to dismiss it in our minds as “spikes” rather than the incoherent noise it is. True instrumental defects resulting in spikes can be dismissed, but only after several measurements have proven a spike is always present.

image019

image021

 

Reblog this post [with Zemanta]

Take Home Lesson - I

Knowledge of correct lineshapes allows you to decide quickly whether your sample is correctly shimmed. You have to decide whether the return (a better lineshape) is worth the time spent achieving that lineshape.

 

image015

image017