T1 noise

T₁ noise is streaks found in the indirect dimension(s) of multidimensional spectra. It is normally found around the most intense peaks in a spectrum and can obscure crosspeaks. How t₁ noise arises, and what can be done to reduce it, are discussed below.

In the 2D COSY spectrum below the vertical streaks are what is referred to as t₁ noise. Its called t₁ noise because it appears in the t₁ dimension, the indirectly detected dimension that is sampled with the incremented delay. Notably, the t₁ noise appears mainly around the sharpest most intense peaks.

t₁ noise is caused by instrumental instabilities. In early NMR spectrometers there was a long list of possible causes¹, but technological improvements have reduced most of these instabilities so greatly that in modern spectrometers most are no longer detectable. The causes that do remain are temperature variation, building vibrations, and lineshape degradation. To reduce their impact modern NMRs are equipped with temperature control systems that limit temperature changes to around 0.1^oC, pneumatic legs that physically isolate the magnet from its surroundings, and ever more complex and sophisticated shim systems. It should also be noted that sample spinning causes t₁ noise and so samples should not be spun when recording multidimensional spectra.

The way in which instrumental instabilities cause t₁ noise is illustrated in the figure below. The top left panel shows a series of rows from a 2D experiment after the first fourier transformation in the directly detected dimension. The intensity of the peak decreases in each successive row because it is modulated by the frequency of the peak in the second dimension, and so follows a sinusoidal decay. To perform the second fourier transformation, the same point in each of the rows is extracted to obtain an FID-like signal. The bottom left panel shows the points extracted from the center of the peak in the top left panel. In this perfect example the points all fall on the dotted black line showing the expected decay. This example would not produce any t₁ noise.

In the panel at the top right of the figure, the peaks in successive rows do not line up exactly due to instrumental instabilities. When points are extracted from each row some do not fall on the expected decay curve, as shown in the bottom right panel. Fourier transformation of this data leads to spurious signals that we see as t₁ noise.

From the example on the right of the figure it can be easily seen how the sharpest and largest peaks are more prone to t₁ noise. A small movement of a sharp peak will produce a larger drop in intensity than the same movement of a broad peak. Additionally, the larger the peak the larger the deviation from the expected decay. This is why t₁ noise tends to come from large, sharp peaks.

The figure above shows the impact of movement of the resonance over the course of a multi-dimensional experiment. This is most likely due to temperature variation. Changes in shimming during the experiment will affect the peak shape rather than its position, making the peak broader or altering its phase. These changes will also result in the extracted points not appearing in their correct places, thereby producing t₁ noise.

There are a few things users can do to prevent t₁ noise. Including dummy scans at the start of the experiment will enable the system to come to thermal equilibrium once pulsing starts, and so minimise peak movement. Avoiding moving around the magnet while it is recording will minimize vibrations and changes in shimming. There are some processing tools to remove t₁ noise. For homonuclear spectra like COSY and NOESY, symmetrisation² can remove the streaks, while for heteronuclear spectra, like HSQC and HMBC, baseline subtraction³ can be useful.

In some cases what looks like t₁ noise is actually due to poor processing and can be corrected after data acquisition. Poor phasing in the indirect dimension will lead to streaks around the large peaks. This can be identified by positive noise streaks on one side of the peak and negative streaks on the other. Another cause may be a poor choice of apodisation. Too harsh a window function may lead to "sinc wiggles", sinusoidal oscillations around large sharp peaks. Reprocessing with a less resolution enhancing window should remove the streaks.

References

1. A.F Mehlkopf, D. Korbee, T.A Tiggelman, R. Freeman

Sources of t₁ noise in two-dimensional NMR

J Magn Reson. 1984;58(2):315-323

2. G. Facey

NMR - Removing t₁ noise from Homonuclear 2D NMR Data

YouTube, uploaded 2013 April 5

3. G. Facey

NMR - Removing t₁ noise from Heteronuclear 2D NMR Data

YouTube, uploaded 2013 March 26