Zero-Fill

Zero-Filling is the process of adding data points to the end of the FID before Fourier Transform.

Zero-Filling can improve data quality by increasing the amount of points per ppm in the processed data to better define the resonance. The added data points have 0 amplitude so the only change of the processed data is more discreet data points.

The Zero-Fill menu is located under the Processing tab then Zero-Fill/LP. LP is Linear Prediction which is discussed here:
2D Linear Prediction

2D spectra are typically zero-filled in both dimensions to increase resolution.

In the direct detect dimension (F2 or t2), zero-filling to twice the data size or 4 times the data size will lead to slight resolution improvement. In the indirect dimension (F1 or t1), the data are often zero-filled to at least four times the amount of acquired points.


1) To accomplish 2D Zero-Filling, first choose the dimension to zero-fill. This is under Processing Menu, then Set Processing Dimension. First choose F2.

2) Next, open the Zero-Fill/LP tab.

2) On the top left, the window will show the original amount of data points of the FID. To the right under To Spectrum Size, choose a number larger than the Original. That number is the number that the Fourier Transform will operate on. If the Original is 1024, then To Spectrum Size is 4096, that would be a 4X Zero-Fill. Typically any number beyond 4X will not help, but the only disadvantage of a larger Zero-Fill is the data size/slower processing.

3) When using Zero-Fill, be sure the Spectrum Size is larger than the Original, else that is a negative Zero-Fill or truncation of the FID.

4) To best see the effects of the Zero-Fill, zoom in on 2 resonances that are close in chemical shift. As you change the To Spectrum Size, see how the shape of the resonances change and whether you can better resolve the 2 resonances. Note you should only see an effect in F2 dimension at this point.

5) Repeat the steps above for F1, by choosing F1 under Set Processing Dimension. On the To Spectrum Size option, typically you should choose at least 1024 (1K) no matter how many points were actually acquired. Often, it is suggested to make the data matrix square so if you chose 4K points in F2, then choose 4K in F1. That is not necessary, but it would give you the same digital resolution in both dimensions in homonuclear experiments (so COSY, ROESY, TOCSY).

6) As in step 4 above, zoom in on 2 resonances close in shift and observe the effect with the change of zero-fill. Also, note processing/redraw speed of the computer. If it is too slow, you might decrease the zero-fill. Once you add in linear prediction, then the speed will slow down more. As linear prediction is in the same menu as zero-fill you probably do them at the same time, but you can observe the effects from zero-fill separately before adding in the linear prediction.