Sample convection in liquid-state NMR: Why it is always with us, and what we can do about it

[Display omitted] •Most liquid NMR samples in most experiments convect.•Horizontal temperature gradients cause convection with no critical lower limit.•Attempts at sample temperature control tend to increase, not decrease, convection.•Convection rates can be reduced by reducing sample tube inner diameter.•Sapphire tubes greatly reduce convection under normal conditions. Many NMR experiments on liquids suffer if the sample convects. This is particularly true for applications, such as the measurement of diffusion, that rely on spatial labelling of spins. It is widely assumed that, in most well-conducted experiments with stable temperature regulation, samples do not convect. Unfortunately this is not the case. It is shown here that typical NMR samples show measurable convective flow for all but a very narrow range of temperatures; convection is seen both above and below this range, which can be as small as a degree or so for a mobile solvent such as chloroform. This convection is driven by both vertical and horizontal temperature gradients. Measurements of convection velocity are presented for a range of samples, sample tubes, probes, and temperatures. Both decreasing sample tube inner diameter and changing sample tube material from glass to sapphire can slow convection markedly, with sapphire tubes being particularly effective. Such tubes are likely to be particularly helpful for accurate measurement of diffusion by NMR.