Hi, according to Varian grad shimming manual (maybe Bruker has a manual like that too?) chloroform in deuteroacetone is used for shimming for the best lineshape. The recommended recycling delay is 6-12 seconds and the steps are:

• shim on z1, z2, x1, y1 manually
• set up the grad shimming pulse sequence
• record the shim map
• gradient shim on Z shim set
• shim x1, y1, xy and zy (do not touch Z shims)
• repeat gradient shimming

Perhaps a couple of first times you'll need to shim higher order x and y's. But you need to know which one's to adjust on your magnet (for example on our magnet I've been advised not to change X3 and Y3). I occasionally adjust xy, x2-y2 and third order x and y's except x3 and y3.

Take a look at this shimming manual by Acorn NMR, it explains typical interactions of shims.

If this is for small molecule work, you could try spinning your sample at about 4 Hz. This might get your line widths down.

There are two problems with using the standard lineshape sample for mapping: long 2H T1, and convection. The lineshape sample is in acetone, which has very low viscosity and boiling point, so convection is typically a big problem. In the gradient shimming experiment, two gradient echo images are acquired, with short and long echo times, and convection can significantly reduce the signal in the long echo experiment. If the spins move in the echo delay, they experience different effective gradient strengths in the encoding and decoding steps and the signal is thus not refocused. For that reason, DMSO or D2O samples are normally preferred for mapping - convection is minimised, but you still have a strong 2H signal.

As Kirk indicated, your shim maps seem noisy. This could come from poor parameter settings - what changes have you made from the standard gradshim1d2h parameter set? The most basic things you need to optimise are the 2H pulse length, and the number of scans. Note that the 2H 90 degree pulse needs to be in the prosol table, as since xwinnmr 3.5 gradshim has used getprosol for the 2H 90 degree pulse, and the pulse in the parameter set will be ignored. The pulse program then uses p1*0.15 for the excitation pulse. You probably want a minimum of 32 scans for reasonable S:N, more if you are at low field (300 MHz or less). Note that the D1 used is very short (consistent with the small flip angle excitation pulse), so 64 scans might take e.g. 15 seconds per echo measurement. You can find more details about parameter settings in the guide that was published in the Bruker Spin Report - first Google hit if you search for: gradshim manual spinreport.

For iteration files, larger window sizes should improve the hump, but if the window is too large the effect of the noisy points at the ends of the shim map will distort things. Does a window of 15 work with the above maps? Normally you want a larger window, as this will optimise the hump, but not so large that the resulting half width becomes broad, but running with a smaller window may indicate if the problem is with the whole maps, or just that the edges have too low S:N.

Finally, if the starting shims are not good, your maps will not be either. If you are doing the mapping without spinning, the off axis shims need to be reasonably good as well as the Z-shims, so adjust x, y, xz, yz by hand if you are going to run nonspinning. The usual advice is that you should run the mapping as you will do the shimming, i.e. if you run all your samples spinning, generate the maps whilst spinning too.