Dynamic Nuclear Polarization
The first suggestion on how to increase nuclear polarization beyond Boltzmann equilibrium by "dynamic" means was made by Overhauser in 1953. He predicted an increase in the steady-state polarization of one type of spins I, if the resonance of another type of spins S would be "saturated" by ESR/NMR-suitable irradiation. In detail, his arguments were based on the hyperfine coupling between spins of conduction electrons and nuclear spins in a metal; but it was soon realized that formally the same mechanism should enhance the NMR signal from a (liquid) solvent or solute by adding a radical, and saturating its ESR transition.
However, the Overhauser effect is based on coupled relaxation equations: its efficiency is field-dependent and decreases sharply when going to the field values common in modern NMR spectrometers.
An early review, that could be read as an introduction: K.H. Hausser and D. Stehlik, Dynamic Nuclear Polarization in Liquids, Adv. Magn. Reson. 3, 79 (1968). In a stricter sense, Dynamic Nuclear Polarization refers to transfer of polarization in solids from electron spins on spatially fixed paramagnetic centers to the nuclear spins. These methods were developed from 1957 onwards, mainly in Saclay and Berkeley, and had as motivation the development of nuclear-polarized targets for particle physics. The two principal processes for polarization transfer are the "solid effect" and "thermal mixing". Both can be efficient at liquid helium temperatures. The solid effect is essentially a microwave-stimulated "forbidden transition" in a dipolar-coupled two-spin system (the picture on the Home page is an example), whereas thermal mixing is described in terms of "spin thermodynamics": it makes thermodynamic assumptions about the behaviour of large numbers of weakly coupled spins. In particular, there is supposed to be dipolar coupling between the electron spins: their concentration must not be too low. In recent polarized targets, the sample material is a glassy frozen solution of stable radicals at concentrations of 10-20 mM: thermal mixing is expected to dominate.
The extension to liquid samples was developed in the early 2000's by a group working at Amersham Health (now GE Healthcare) in Malmö, who discovered that it is actually possible to transform a frozen-beads sample into a polarized liquid solution at room temperature by dissolving it rapidly in superheated water. Their method is now usually called dissolution DNP.