Capillary condensation, invasion percolation, hysteresis, and discrete memory R. A. Guyer Department of Physics and Astronomy, University of Massachusetts, Amherst, Massachusetts 01003 K. R. McCall Department of Physics, University of Nevada, Reno, Nevada 89557 Received 18 December 1995 A model of the capillary condensation process, i.e., of adsorption-desorption isotherms, having only pore- pore interactions is constructed. The model yields 1hysteretic isotherms, 2invasion percolation on desorp- tion, and 3hysteresis with discrete memory for interior chemical potential loops. All of these features are seen in experiment. The model is compared to a model with no pore-pore interactions the Preisach modeland to a related model of interacting pore systems the random field Ising model. The capillary condensation model differs from both. S0163-18299606725-2 I. INTRODUCTION Hysteresis is often seen in a measured quantity’s response to a probe. For example, the displacement u ( t ) response of an oscillator to a periodic driving force F ( t ) is hysteretic. In this case the hysteresis in u versus F is due to a damping induced phase difference between u ( t ) and F ( t ). Hysteresis with discrete memory H/DMis less familiar although it is seen in a wide variety of physical and model systems. 1–5 Discrete memory or end point memory refers to the memory and memory erasure that occur at the extrema of the probe protocol. In magnetism the probe is the external field; in capillary condensation the probe is the chemical potential. A very restricted set of hysteresis models also have discrete memory. An elegantly simple explanation of H/DM is given by the Preisach model, 6–8 a model for the behavior of a large number of noninteracting, individually hysteretic units. In many instances where the Preisach model is used the basic condition for its application, noninteraction or indepen- dence, is absent. For example, the Preisach model has been extensively used to describe the capillary condensation pro- cess, i.e., adsorption-desorption isotherms. 9 However, the in- vasion percolation feature seen in many isotherms 4 provides compelling evidence for pore-pore interactions and suggests that a Preisach model has circumscribed validity. 10,11 More sophisticated models have been developed for adsorption-desorption isotherms 12,13 and for systems that possess H/DM. 14 The purpose of this note is to describe a capillary condensation model CCMthat, in marked con- trast to the Preisach model, owes all of its behavior to the interaction of individually nonhysteretic units, and to show that this model possesses H/DM. The CCM possesses the essential features necessary to give the desired qualitative behavior of adsorption-desorption isotherms although it has been stripped of much of the detail that would make it a practical model for inverting data to learn the pore space geometry. In Sec. II the capillary condensation model is de- scribed. In Sec. III results of numerical simulation of the CCM are presented. The essential content of the CCM and its relationship to other models, e.g., the random field Ising model, are discussed in Sec. IV. II. MODEL The capillary condensation process is characterized by the geometry of the pore space, the rules for the evolution of fluid configurations in the pore space, and by the chemical potential protocol to which the fluid is subjected. For defi- niteness and simplicity we adopt the following model. aThe pores, cylinders of length b , are the bonds of a square lattice L L . There are N =2 ( L / b ) 2 pores. bThe pores are in the presence of a fluid at chemical potential , i.e., the pore space is connected to a chemical potential reservoir. cA pore is either empty, in state =0, or filled with fluid, in state =1. A pore in state 0 has a thin liquid film on its surface and vapor in its interior. A pore in state 1 is filled with liquid and may have a meniscus at its ends depending on the state of its near neighbors the six pores with which it shares a node. dTo the i th pore we assign a radius r i from the prob- ability density ( r ) and a critical value of the chemical po- tential i from the probability density ( ). There may be a correspondence between i and r i so that r dr = d . 1 To avoid details that depend on ( r ) and ( ) but are ines- sential to our discussion we map the spectrum of critical pore chemical potentials onto the interval 0,1by ordering 1 ,..., N from smallest to largest and using p i = ordinal number of i )/ N . We map the reservoir chemical potential to the appropriate point in the sequence p 1 ,..., p N and con- tinue to denote it as . eThe fluid in the system is carried through a chemical potential protocol ( t ). Here t simply indicates that the chemical potential evolves in time. This time evolution oc- curs so slowly that the fluid configurations in the pore space are taken to be a sequence of stable or metastable fluid con- figurations C ( t ) =C ( t m ) =C m consistent with the chemical potential protocol, ( t ) =( t m ) = m . fOn chemical potential increase m+1 m . If m p i m+1 , then i =0 1. PHYSICAL REVIEW B 1 JULY 1996-I VOLUME 54, NUMBER 1 54 0163-1829/96/541/184/$10.00 18 © 1996 The American Physical Society