Igneous microstructures from kinetic models of crystallization
Introduction
The extreme variability of natural igneous microstructures in terms of crystal size, shape, number, and position is due to the interplay of competing kinetic processes operating during crystallization (e.g. nucleation, growth, aggregation, and grain boundary annealing). It has proven challenging to decipher the contribution of each kinetic process to the final microstructure, including the role of the intensive parameters (e.g. temperature, time) governing crystallization. Present knowledge of the crystallization kinetics of solidifying magmas is mainly limited to relatively short duration melting experiments that are most applicable to rapidly cooled magmas (i.e., lavas) and not coarse-grained, slowly cooled igneous rocks that record a fuller and more dynamic history of magmatic activity. To augment empirical results, some of which have been available from the late 1700s (e.g. Hall, 1798), a theory of crystallization kinetics was adapted from classical kinetic theory (e.g. Kirkpatrick, 1981), but this has proven too broad based to be of immediate practical use in predicting the styles of igneous microstructures. That is, this atomic scale theory is inherently incomplete for silicate melts and, in the light of sparse kinetic information from actual magmatic systems, is exceedingly awkward for direct application as a forward model of microstructural development.
An alternative approach is to view igneous microstructural development as a grain-scale process with fundamental rules describing the transient appearance and morphological evolution of crystals, which is consistent with but does not resolve atomic scale processes of diffusion and bonding during crystallization. This rule-based kinetic model of crystallization relies on long established general observations from crystallization experiments, from post-mortem studies of solidified magmas, and the results of both crystal size distribution (CSD) theory and CSD measurements on rocks. The thrust of the present work is to demonstrate that a stochastic numerical algorithm for forward modeling of igneous microstructural development can be used to delineate the relative contributions of individual kinetic processes in forming realistic igneous microstructures using various kinetic models of crystallization.
In brief, a 3D space of melt containing randomly orientated and randomly positioned crystals is represented by an ensemble of discrete volumetric elements; changes in time, and increasing crystallinity, are achieved by discrete temporal steps. In assuming that cooling and crystallization can be decoupled, as discussed below, the design of the algorithm is greatly simplified and crystallinity can be expressed solely as a function of time. Functions describing the change of crystallinity with time are derived using the Avrami method by employing any number of exponential variations (including constant) in functions describing the rates of nucleation and crystal growth (Marsh, 1998). An investigation of the sensitivity of the algorithm to variations in the input parameters precedes the evaluation of the microstructures generated by the various kinetic models. Three basic kinetic models are investigated: constant rates of nucleation and growth, exponential nucleation rate with constant growth rate, and exponential nucleation rate with crystal-dependent dispersive, but individually constant, growth rate. An extension of the algorithm is also demonstrated by modeling multiply saturated mineral growth, which simulates basalt crystallization involving simultaneous plagioclase and clinopyroxene nucleation and growth. Aside from understanding the kinetics of crystallization, the computed microstructures are invaluable as 3D samples of test materials necessary for input into models developed to obtain macro-scale physical properties under conditions unobtainable by experiment. Failure behavior, elastic moduli, permeability, and grain-scale melt flow patterns including flow channelization are critical for continuum scale models of magmatic processes and can, thus, be determined throughout the melting interval over a wide range of pressure.
Section snippets
Background
Previous studies of igneous crystallization predicting crystal size and number were hampered by having to solve simultaneously for the thermal regime in conjunction with either scaling laws (e.g. Brandeis and Jaupart, 1987) or classical kinetic theories of nucleation and crystal growth (Spohn et al., 1988, Tomaru, 2001). These models are useful to compare with laboratory measurements, but because they are unable to predict the spatial representation of the evolving microstructure they cannot
Decoupling cooling and crystallization
Most igneous rocks are holocrystalline (not glassy) regardless of the cooling regime in which they have formed. This suggests that for any given time scale imposed by cooling, nucleation and growth automatically adjust locally to attain full crystallinity (Marsh, 1998). Although this is not always strictly true, it is a very good approximation, especially for basaltic compositions. This directly implies that crystallization can be decoupled from cooling, which can be demonstrated through
Kinetic model
During crystallization, a variety of kinetic processes can influence the final igneous microstructure. The present approach is to consider two of the most basic processes, heterogeneous nucleation and crystal growth, and to test simple models of these processes in reproducing reasonable microstructures. The essential feature of this approach is the imposition of a function describing the variation of bulk crystallinity with time. The utility of employing a crystallinity function, in addition to
The stochastic algorithm
The crystallizing domain is a cubic volume of length L on a side that is discretized into a number of smaller cubic volume elements (voxels) each of width W, containing either solid or liquid (see Fig. 2). A crystal is composed of multiple adjacent solid voxels and the number of voxels per side of the domain is m. The crystallization time is discretized into a number of time steps nt of equal time duration Δt. At the beginning of a crystallization simulation no nuclei or crystals are present in
Sensitivity analysis
The simulated microstructure is, clearly, sensitive to the degree of spatial and temporal discretization used in the stochastic algorithm. Judicious choice of the discretization parameters is required for the simulation to approximate the analytical solutions for numbers and lengths of crystals (validation) and also to minimize error when the microstructure is input into other physical models. The number of crystals in the domain and the spatial resolution of crystals are controlled by two
Monominerallic crystallization
As a simple first example of the stochastic algorithm, we simulate monominerallic crystallization of rectangular prisms of aspect ratio 1 : 1.5 : 3.5 in three realizations (r = 3) per simulation. The domain and crystallization times are discretized with the following parameters: m = 96, nt = 15, α = 0.46875, and β = 1.5. Three kinetic models are tested: constant nucleation and growth rate (a = 0), exponential nucleation and constant growth rate (a = 4 and a = 8), and exponential nucleation and dispersive growth
Polyminerallic crystallization and percolation
The stochastic algorithm can be easily extended to include crystallization of more than one mineral phase. It is common in most magmas for several distinct solid phases to grow simultaneously. Basalt crystallization is simulated with a simple model of simultaneous plagioclase and clinopyroxene (or olivine) crystallization in which both phases equally share the total crystallinity with time. Plagioclase is represented by a lath shape of aspect ratio, 1 : 2 : 5 and clinopyroxene by a stubby prism
Conclusions
A stochastic algorithm is developed to predict the 3D spatial representation of the igneous microstructure during progressive crystallization involving simultaneous nucleation and crystal growth. Of the kinetic models of crystallization tested, exponential nucleation rate with an exponential factor, a, of eight and constant crystal growth rate yields a microstructure in which the CSDs remain approximately log-linear (as often found for rocks) throughout the crystallization interval. For smaller
Acknowledgements
Reviews by R. Amenta and A. Mock are greatly appreciated. This work is supported by NSF Grant OPP-0440718 to Johns Hopkins University.
References (30)
Three-dimensional computer modeling of fabric evolution in igneous rocks
Comput. Geosci.
(2001)- et al.
Permeability and melt flow in simulated and natural partially molten basaltic magmas
Earth Planet. Sci. Lett.
(2005) - et al.
Acquisition, optimization and intrepretation of X-ray computed tomographic imagery: applications to the geosciences
Comput. Geosci.
(2001) A method for three-dimensional reconstruction of macroscopic features in geological materials
Comput. Geosci.
(1998)- et al.
Numerical models on the onset of yield strength in crystal-melt suspensions
Earth Planet. Sci. Lett.
(2001) - Amenta, R.V., 2004. Computer modeling of igneous hypidiomorphic textures using crystal prisms and plates with...
Kinetics of phase change: I
J. Chem. Phys.
(1940)- et al.
Rejuvenation of the Fish Canyon magma body: a window into the evolution of large-volume silicic magma systems
Geology
(2003) - Boudreau, A., 2003. Vapor separation and migration in a solidifying and compacting crystal pile. Eos Trans. AGU,...
- et al.
The kinetics of nucleation and crystal growth and scaling laws for magmatic crystallization
Contrib. Mineral. Petrol.
(1987)