The primary purpose of this paper is to review recent contributions from North American scientists to theory and observations of magnetic domain structures in magnetic minerals common to rocks.
Although rock magnetism deals with all the magnetic properties of rocks, a major goal of rock magnetists has been to understand the origins and properties of remanent magnetization (RM). A remanent magnetization is defined as the magnetization that exists in the absence of an external magnetic field. The analyses of RMs (of which there are more than 20 kinds) have led to the numerous, and well-known, successes of paleomagnetism.
The foundations of theoretical rock magnetism were laid down in an elegant paper by Neel [1949] dealing with the theory of thermal remanent magnetization (TRM) in an ensemble of identical noninteracting single domain (SD) (uniformly magnetized) particles. Neel handled the TRM problem by assuming that, on cooling a sample, the magnetization was in thermal equilibrium until it reached a blocking temperature. Below this blocking temperature the magnetic directions in individual grains essentially remained invariant. Providing one knew the energy states of the system, the partition function of the system could be obtained. Thus one could determine the magnetization of the ensemble at the blocking temperature through Boltzmann statistics. A theory for TRM emerged naturally by assuming the distribution remained invariant on cooling below the blocking temperature.
A major problem in applying Neel's SD theory is that the vast
majority of magnetic grains in most rocks are too large to be
uniformly magnetized. For example, Butler and Banerjee [1975]
calculated that equi-dimensional magnetite was SD only within a
very small size range, 0.06 to 0.08 
m, at room temperature.
(Grain size will be indicated by giving one dimension of an
equi-dimensional particle.) Below this range magnetite grains are
superparamagnetic (SP), i.e., relaxation occurs so quickly that the
net magnetization is always in thermal equilibrium. Immediately
above this range the grains were believed to contain two magnetic
domains, separated by a transition referred to as a domain wall.
On the average, as the dimensions of the grain increase, so do the
number of domains.
Although Neel [1955] also produced the first (of
[4]
many) multidomain (MD) theories for TRM, his MD theory and all
subsequent ones have proved to be inadequate. Nevertheless, with time
it became clear that, on the average, large MD grains (i.e., greater
than 50 
m in magnetite) do not contribute much to the stable
magnetization of rocks. The stable magnetization of interest to
paleomagnetists often appears to reside in ``pseudo-single domain''
grains, whose size and behavior is intermediate between that of
very small magnetically stable SDs and much larger, magnetically
``soft'' MDs. The term pseudo-single domain (PSD) was first
introduced by Stacey [1958] to describe the magnetic moment in a MD
grain associated with the sudden (Barkhausen) jump of a domain wall
as it becomes dislodged from a defect. Subsequently, a variety of
definitions for PSD grains and PSD behavior emerged. The subject
still remains murky today, but there is still a consensus that PSD
``behavior'' is intermediate between classical SD and MD ``behavior.''
Although there are different measures of ``behavior,'' the ones most
commonly appealed to are hysteresis loop parameters [e.g. Day et
al., 1977]. For example, a value of the saturation RM to
saturation magnetization 
0.5 is usually interpreted as
reflecting SD behavior, a 0.3 value as PSD behavior, and a 0.05
value as MD behavior. The size range for equi-dimensional PSD
magnetite, for example, is often taken to range from 0.05 
m to
approximately 15 
m; however, some grains in this size range
exhibit PSD behavior and some do not---for reasons that remain
unclear. Thus the importance of PSD grains to paleomagnetism
continues to motivate efforts to understand PSD grains.
To this end, Halgedahl and Fuller [1980, 1983] imaged domains using the Bitter pattern technique and found that grains with approximately the same shape, size, and composition sometimes exhibited different numbers of domains. Shortly thereafter, Moon and Merrill [1984, 1985] provided a theoretical explanation for these observations: several local energy minimum (LEM) states exist for most grains found in rocks, i.e., the number of domains present depends on the grain's history. Moreover, it was shown that domains were not regions in which the magnetization was perfectly uniform throughout. Because grains with identical size, shape and composition can have different numbers of domains, the partition function of an ensemble of identical grains will be different from that previously assumed and hence so will all the equilibrium properties of the ensemble. It follows that nonequilibrium properties also need to be recalculated, since the activation energies between LEM states must be taken into account. For example, the magnetic properties (e.g., RM) associated with the alteration of mineral ``A'' to mineral ``B'' sometimes will depend on the history of mineral ``A,'' because different histories can lead to different initial magnetic states. For reasons such as these, a considerable amount of attention has been expended by rock magnetists to understand domain structures and their relationship to remanent magnetizations and stability. In the remainder of this report, we briefly outline the progress made during the past few years.