A barely remembered anecdote, the recent talk of detecting nuclear explosions, and a Web of Knowledge search combined to bring this paper up on my screen:
Seismic tomographic inversion of Russian PNE data along profile Kraton 
The acronym ‘PNE’ is made more explicit in the next result:
Origin of upper-mantle seismic scattering – evidence from Russian peaceful nuclear explosion data 
That’s right – PNE stands for ‘peaceful nuclear explosion’, a phrase from a simpler time when people seriously advocated using nuclear explosives to dam rivers, create new anchorages, and release gas from underground reservoirs (which, shockingly, turned out to be radioactive). Both the USA and USSR engaged in this interesting pastime, but the Soviets detonated 239 PNEs against the Americans’ 28, a disparity that is probably due more to the relative power of each country’s citizenry (I suspect even in those days. if given the choice very few people would fancy a close-up view of mushroom clouds) than a difference in official enthusiasm.
Anyway, it seems some Russian geologists had the bright idea of using PNEs to undertake controlled source, wide-angle seismology on a scale not achieved before or since. They arranged to have a series of nuclear explosions set off along a number of lines (with names like 'Quartz' and 'Kraton') stretching several thousand kilometres across the USSR, along which they also deployed seismometers. Unlike reflection seismology, where only the reflected seismic energy coming directly upwards from buried discontinuities such as boundaries and faults is collected, wide-angle seismology deploys seismometers a long way from the source in order to pick up refracted seismic energy; because the speed of sound generally increases with greater depth (and pressure) in the Earth, due to Snell’s Law of Refraction downward-travelling waves eventually curve back upwards towards the surface. If you get far enough from the source, not only will a seismometer pick up these deep-diving waves when they reach the surface, but because of their higher average speed these waves will actually reach that seismometer before ones that have travelled at shallower depths. The distance from the source at which the ‘first arrivals’ from different depths first appear provides a great deal of information about the velocity structure, and hence the physical structure, of the subsurface rocks along the survey line (simply put, earlier appearance of deep waves=faster velocities – although obviously it’s a bit more complicated than that, especially when you factor in the presences of sharp discontinuities in wave-speed across lithological boundaries).
Methodology wise, there is quite a large overlap between these wide-angle techniques and earthquake seismology. However, because the source in this case is man-made, you minimize some of the major uncertainties which limit resolution when your source is a natural earthquake: the exact time and location that the source wave was generated, and the nature of the original signal (as I’ve discussed in the context of the North Korean test, man-made seismic signals are generally much shorter and simpler than an earthquake signal). The frequency of the signal is generally higher, too, which also improves resolution. There is however, one major drawback: the airguns (or less commonly nowadays, conventional explosives) used in seismic experiments produce orders of magnitude less energy than an earthquake does, meaning that pretty much all of the waves penetrating below the crust-mantle boundary (Moho) –about 35-40 km below the surface in normal continental crust - dissipate before they can get back up to the surface. The figure below shows the inferred paths of seismic waves picked up along a wide-angle survey line in the Aleutian Islands (the source is located in the centre, the scales are in km) .
Whatever your feelings about the rationality of using nuclear bombs in this way, you can’t deny that they’re going to generate a hell of a lot more energy, allowing signals from deep in the mantle to be received back at the surface. Here’s a similar figure for the ‘Kraton’ PNE seismic line (from . Note the change in horizontal as well as vertical scale: whereas seismometers 2500 km away from the source are fairly pointless for normal wide-angle surveys, for a PNE shot this is about the point that first arrivals from the 660 km discontinuity start arriving. Which I can’t help being a little bit impressed by.
The second of the two papers I’ve cited concentrates on the uppermost mantle, 100-200 km beneath the Earth’s surface. The waves travelling through this region appear to be travelling slower than in the overlying rock, and are also being scattered a lot, resulting in noisy first-arrivals data (the relationship between the arrival time and the distance from the explosion is not linear). The authors propose that this is because the mantle at this depth is partially molten. It seems that the unique data generated by these 30-40 year-old experiments is still reaping dividends, now that scientists worldwide can access them and apply the latest processing tricks.
 L Neilson et al., Geophysical Research Letters 26(22), p 3413-16, 1999.
 L. Neilson et al., Geophysical Journal International154, p 196-204, 2003.
 D Shillington et al., Geochemistry Geophysics Geosystems 5, Q10006, 2004 [doi]