Page 218 - Machine Learning for Subsurface Characterization

188 Machine learning for subsurface characterization

energy is emitted at the same frequency. When this spin-flip occurs, the nuclei
are said to be in “resonance” with the field; hence, the name for the technique to
excite the nuclei to higher energy states is referred as the nuclear magnetic
resonance (NMR).
When the EM radiation is switched off, the excited nuclei release their
excess energy (NMR relaxation/decay signal) and return to low energy levels
by two relaxation processes. Relaxation is a measure of deterioration of
NMR signal with time during the conversion of the excited nonequilibrium
population to a normal population at thermal equilibrium. For material
characterization the NMR relaxation/decay measurement is analyzed in
terms of two processes, namely, T1 (spin-lattice) and T2 (spin-spin)
relaxations. T1 relaxation is the loss of signal intensity due to the interaction
with environment resulting in the relaxation of the components of the
nuclear spin magnetization vector parallel to the external magnetic field,
whereas T2 is the broadening of the signal due to the exchange of energy
with neighboring nuclei at lower energy levels resulting in the relaxation of
the components of the nuclear spin magnetization vector perpendicular to the
external magnetic field. T1 relaxation indicates the inter- and intramolecular
dynamics. T2 relaxation involves energy transfer between interacting spins
via dipole and exchange interactions.
For fluids contained within pores, both T1 and T2 relaxations depend on
surface relaxation in addition to primary controls of viscosity, composition,
temperature, and pressure. Surface relaxation occurs at the fluid-solid
interface between a wetting-pore fluid and the rock-pore walls, which is very
different from the individual bulk relaxations of pure fluid and solid. Surface
relaxation dramatically reduces T1 and T2, such that surface relaxation is the
dominant contributor to T2 relaxation. Surface relaxation depends on the
surface-to-volume ratio, which is determined by the mineralogy and is a
measure of permeability. T2 relaxation of a pore-filling fluid is considered a
combination of bulk, surface, and diffusion relaxation, wherein the surface
relaxation dominates. Surface relaxation occurs at the fluid-solid interface
and is affected by pore shape, pore network characteristics, and mineralogy.
Bulk relaxation is affected by the fluid type, the hydrogen content, and its
mobility. Diffusion relaxation occurs due to the nonzero gradient of the
magnetic field exciting the sample/reservoir under investigation.
The rate of NMR T2 relaxation of fluids in pores most significantly depends
on the frequency of collision of hydrogen nuclei with the grain surface, and this
is controlled by the surface-to-volume ratio of the pore in which the hydrogen
nuclei are located. Collisions are less frequent in larger pores, resulting in a
slower decay of the NMR signal amplitude and allowing the characterization
of fluid-filled pore size distribution. When all pores are assumed to have a
similar geometric shape, the largest pores have the lowest surface to volume
and, thus, the longest T2. However, subsurface rocks have pores of varying
sizes containing more than one fluid type, so the T2 relaxation consists of