Assignment 3
NMR Guided Reading
February 4, 2000
Today we are going to learn about processing parameters. Basically this involves zero-filling and choosing the right window function (and window function parameters). The appropriate reading in Derome (which will be quite necessary today) is pp. 22-26.
You need to first obtain the allylglycine sample in D2O (courtesy of Dr. Erickson) from my office and lock and shim on it. Obtain a 1D-spectrum of the sample using the standard D2O parameters. Fourier transform (no exponential multiplication) and phase as well as possible. Take a look at expansions of each of the peaks, switching between points and lines to observe how well the individual peaks are resolved. There is one set of peaks upfield (not terribly far upfield) which are really messy. Plot out a nice expansion of this multiplet for comparison later on.
Now, change the acquisition parameters so that only 1/4 of the data points are collected. Acquire a new spectrum and look at all of the multiplets again. They should be considerably lumpier. By looking at the points in the spectrum, you will see why. They are not resolved well enough in the time domain. There will also be significant truncation in the FID signal leading to nasty baseline wiggles on large peaks (these wiggles are sometimes referred to as feet).
Upload the data again and reprocess it, but this time we will change how we process it. Go to the Commands menu, hold down the option key and click on 1DFT. A new window will pop up with several parameters. (In fact, any command with the zero-minus symbol next to it, will give a parameter window when opened with option-click.) Click on FFT for the type of transform and click on zero fill (x1). The size of the data set will now be doubled prior to Fourier transform. When you click on ok, you will now have all of these actions done. Look now at the multiplets with points as well as interpolated data. You will see that the digital resolution is better (there are more points) but the peaks probably do not look considerably better. The wiggles will still be there as well. That is because the FID is chopped off still. Look at Figure 2.17 in Derome on p. 23.
We will now apply an apodisation function (cutting off the feet). Upload the data, bring up the 1DFT parameter window and apply all of the parameters from before plus the window function exponential multiplication. When you click on EM another parameter, line broadening, will show up. Make it 0.4, and click on ok. Now the new spectrum should have the wiggles decreased and in addition should have better signal to noise (compare the signal to noise to the original spectrum without any window function applied). However, a drawback to the EM function is that it broadens lines. Look carefully at the busy multiplet that we plotted. You will probably see some line broadening in the new spectrum. Repeat the processing with a higher LB of about 2.0. Now compare the multiplets. There should be a huge change. The optimum line broadening is the natural line-width of the resonances in question. Transform the spectrum again without in window functions, measure the line width of a reasonably well-resolved line and then apply this line broadening. You should see the optimum increase in S/N ratio with minimal line broadening effects.
There is a graphical feature on MacNMR to look at the exponential multiplication function. Upload the data and then click on the exponential command while holding down the option key. The screen changes to show the EM window. Now you can change the line-broadening factor and see the function that will be applied (the red line in over the FID). Unfortunately, with MacNMR you cannot see the effect of the the function on the data until actually applying it by clicking on ok. Do that with a large value for LB. You will now see that the FID goes to zero faster. Do a transform and you will see the effect.
Now, supposes your signal to noise ratio is just fine, but you want to enhance the resolution of the individual peaks. We would like to do that with this sample to see the crazy multiplet better. What we need is the Lorentz-Gaussian function described in Derome. This will increase the importance of the data at the end of the FID resulting in a more detailed spectrum. First, however, we need to collect as much good data (points that have signal in them) as possible. This will require that you increase the number of data points collected. You should collect enough so that roughly the last 1/4 of the data set looks more or less like noise.
Once you have done that we will apply the Lorentz-Gauss function described in Derome. With many software packages this is done with a single function in which there are two different parameters to change. However, with the MacNMR software, this procedure is done by sequential application of the Lorentzian transformation (EM) followed by the Gaussian transformation (GM). Therefore, do the following. Obtain the necessary FID. Find the optimal LB (since you are acquiring more data this may be different than that found before. Now, upload the data again and do an EM. However, instead of using the optimized LB, use the negative of it (-0.2 instead of 0.2).
Whoa, that looks horrible doesn't it. The end of the spectrum warps down or up because there is some DC-offset in our electronics that we must correct. Therefore, upload the data, do a baseline correction (command menu) and repeat the EM with the negative LB. Still looks funny, but now the FID is at least symmetrical about the axis. We will now use the second function (the Gaussian function) to make it look more normal again.
Go to the command menu and option click on Gaussian multiplication. Set the Gaussian coefficient to approximately 0.2 and observe where the function cuts off now. Click ok and look at the results on the peaks in the multiplet -- it should be much better resolution. You can play around with both of these parameters (LB and GB) to get very dramatic results. This comes at the cost of signal to noise ratio.
As you do this, look at various peaks in the spectrum. What is often true is values which are good for some peaks, distort others too much. You will see some really funny lineshapes when you play around this way.
All right, that's it for now. We will spend a little bit of time with this spectrum again. We are going to do some NMR simulation using the program GNMR. If we have enough time, we might start this today.