Composition and phase changes observed by magnetic resonance imaging during non-solvent induced coagulation of cellulose

Abstract

The coagulation of cellulose from solution in N-methylmorpholine-N-oxide by water was studied by stray field magnetic resonance imaging. Changes in composition and diffusion were readily observable at different depths within the cellulose gel. It was found that the composition changed relatively quickly, due to diffusive exchange of water and solvent, such that cellulose precipitation occurred deep within the bi-phasic region of the phase diagram. The expected spinodal decomposition mechanism was supported by electron microscopic observations of the cellulose morphology. Large increases in diffusion coefficient and relaxation times were attributable to the phase separation and the progressive development of pores within the gel. Moreover, it appeared that the composition changes and evolution of morphology occurred over separate time-scales.

Introduction

It is widely recognised that the mechanism by which a polymer solidifies from solution plays a major role in defining the resulting morphology. This is particularly important in the manufacture of polymer membranes, since morphology strongly influences the transport properties [1], [2]. In particular, whether the cross-section appears dense or porous on the scale of nanometres to micrometres affects permeability, selectivity and, consequently, the potential applications of the membrane. Dense membranes are generally less permeable but provide more stringent selectivity of smaller species, compared with more porous counterparts.

The various processes that occur during polymer coagulation from solution have, therefore, received much attention. Comprehensive reviews have been given by Van de Witte et al. [3] and Wienk et al. [4]. In general, coagulation can be brought about by different methods involving changes in temperature (thermally induced phase separation), evaporation of a volatile solvent (dry-casting) or diffusive exchange with a non-solvent (immersion-coagulation or wet-casting). However, while these methods appear outwardly different, they depend on common underlying molecular processes, which can be recognised in terms of the locations on the phase diagram where the phase separation takes place and the corresponding morphologies. These processes are demonstrated in Fig. 1, for a hypothetical polymer/solvent/non-solvent system.

Route A represents vitrification. As the polymer concentration increases, the solution becomes progressively more viscous; chain motion and diffusion slow down and the polymer plus any residual solvent effectively reverts to a glass. This results in a dense morphology with little or no porosity. Vitrification is frequently encountered during ‘dry-casting’, where a volatile solvent is removed by evaporation. However, it may also occur in non-solvent induced coagulation if the ‘outward’ diffusion of solvent is significantly faster than the ‘inward’ diffusion of non-solvent.

In example B, phase separation occurs in the metastable region between the binodal and spinodal lines, at higher polymer concentration than the critical point. Nucleation and growth of non-solvent droplets occurs, which results in the formation of liquid-filled pores in a continuous polymer matrix. This route is common for the immersion-coagulation of many polymer/solvent/non-solvent systems, but can also occur during dry-casting, where a small amount of less volatile non-solvent is included in the polymer solution, along with the more volatile solvent.

In example C, phase separation occurs in the unstable region of the phase diagram, bounded by the spinodal line. Here, polymer-rich and polymer-depleted phases separate initially by the progressive growth of ‘concentration waves’ of constant wavelength but increasing amplitude, in a process known as ‘spinodal decomposition’. This results in a mutually continuous, interwoven network of polymer domains and pores.

Route D enters the metastable region at a polymer concentration below the critical point. In this case, the nucleation and growth of polymer particles produces a granular morphology.

While the above examples have been presented in terms of isothermal composition changes, which may occur during dry- or wet-casting, similar processes can also be described for thermally induced phase separation. Parallels between vitrification as a result of composition change and melt processing are obvious. However, phase separations in other regions of the phase diagram can also be initiated by temperature changes, if a composition is miscible at one temperature but immiscible at another. Examples of this include the upper- or lower critical solution behaviour of some polymer solutions.

In spite of the considerable research effort into non-solvent induced polymer coagulation, a significant problem has been the absence of a suitable method for observing the composition changes taking place within the polymer. Hence, much of the present understanding is based on mathematical models of the diffusive exchange of solvent and non-solvent that are expected during coagulation, along with post-mortem observations of the coagulated polymer morphology. Whilst sophisticated mathematical models have been developed [5], [6], [7], [8], [9], [10] doubts can arise as to whether these models adequately reflect the physical processes that occur during coagulation. Moreover, these models depend on accurate ‘in-put’ data, concerning diffusion coefficients and how they change with composition, which may not be available for many systems.

We recently demonstrated the capabilities of magnetic resonance imaging (MRI) for following composition and phase changes during polymer coagulation [11]. MRI is a powerful tool for studying both the distribution of liquids in porous or absorbent solids and movement due to flow or diffusion. The basic principles have been described by Callaghan [12], Blümler et al. [13] and Blümich [14]. By applying a magnetic field gradient across the sample, the resonant frequency of the nuclei under observation (typically protons) encodes for position, while echo attenuation can be used, for example, to determine molecular motion due to diffusion.

The technique of stray field imaging is a variant of MRI, which uses a high permanent magnetic gradient to observe samples in one-dimension [15], [16], [17]. This method overcomes many of the problems associated with imaging solid-like systems, which undergo restricted molecular motion, giving relatively fast transverse nuclear relaxations. It can produce relatively high resolution profiles and is particularly useful for observing planar homogeneous structures, which change in one-dimension only. Previous studies include water diffusion in zeolite powder beds [18], [19], sandstone rock plugs [20], [21] and fibrous cement tiles [22], solvent diffusion in polymers [23], [24], the setting of paint films [25] and water diffusion within cellulose membranes [26].

The findings from our MRI investigation of the widely studied cellulose diacetate (CDA)/acetone/water system agreed with previous suggestions [8], [9], [10], [27] that the coagulation occurred by nucleation and growth of non-solvent filled pores (equivalent to path B in Fig. 1). This work also demonstrated the rapid formation of a relatively impermeable ‘skin’ of CDA in contact with the coagulant, which restricted the subsequent diffusive exchange of water and acetone. This led to a much slower composition change deeper within the polymer and the formation of larger pores, away from the CDA-bath interface.

In the present work, we have used similar experimental methods to study the different system, based on the coagulation of cellulose from solution in N-methylmorpholine-N-oxide monohydrate (NMMO/H₂O). This system is the basis of the lyocell process for making regenerated cellulose fibres and films [28], [29] and has been the subject of much previous academic and industrial research. The onset of cellulose solubility appears when the water content of NMMO is less than about 16% by weight [30]; although, for most purposes, a composition close to or slightly dryer than NMMO monohydrate (i.e. 86.7% NMMO and 13.3% water by weight) is used [31], [32], [33], [34]. The cellulose can be regenerated from solution by immersion-coagulation in water, which allows the non-solvent to diffuse into the cellulose and NMMO to diffuse out.

The NMMO/cellulose system also well exemplifies the current position regarding investigations into the evolution of polymer morphology. Much effort has been made to analyse the prevailing sub-microscopic structure of the regenerated cellulose by electron microscopy [35], [36], [37], [38], [39], [40], [41], [42], [43] and scattering methods [37], [38], [39], [41], [42], [43], [44], [45], [46], [47]. However, while there has been some study of the effects of solution composition and spinning conditions on the resulting regenerated cellulose [31], [32], [33], [34], [48], [49], [50], [51], there has been little or no investigation into the composition changes which accompany the regeneration process.