Fig. 3 shows cryo-SEM images of frozen droplets and
conventional freeze-dried coffee. The droplets feature a
highly homogeneous structure, where the darker regions
correspond to pores where ice crystals existed before
being sublimed off in the SEM preparation stage. The
homogeneity in Fig. 3(b) is clearly superior to that in
conventional product shown in Fig. 3(a). The regularityin crystal formation indicates that the temperature field
is uniform, as expected for the low Biot number regime
(discussed below). The high magnification image of the
droplet surface in Fig. 3(c) shows the presence of a thin
outer skin of high solids content, generated by evaporation
of water from the surface and resulting in a barrier
to further mass transfer. This skin was observed on all
droplets imaged and will result in a lower water activity
at the surface than in the bulk solution: measurement
of aw at the surface of the droplets is therefore not
straightforward and approximating aw to zero is not
unreasonable.
Key parameters in microstructure development are
the solution composition, temperature and cooling rate.
These cannot be controlled independently and it was not
possible to carry out an extensive study of these relationships
in this study. However, the impact of composition
(10 and 50 wt.% coffee) and cooling rate (via um) over the
temperature range ambient to 15 C is illustrated in
Fig. 4 and summarized in Table 1. Under these conditions
the droplets nucleated spontaneously. The droplet
in Fig. 4(a), generated at a lower cooling rate, features
crystal sizes with a larger average hydraulic radius, rh a
wider range of crystal sizes and a smaller aspect ratio,
compared to the droplet in Fig. 4(b), frozen at a faster
rate, which indicates the existence of fewer nucleation
sites. Numerous nucleation studies have shown that the
nucleation site density increases with increased melt cooling
rate (Hindmarsh, 2003). These data confirm that
spray freezing (and subsequent drying) yields a smaller
average pore size and narrower distribution than freezing
coffee solutions on a plate, for which Pardo et al. (2002)
have reported average pore sizes for 40 wt.% coffee solutions
of 42 lm. The effect of composition is very evident
for the 10 wt.% droplet in Fig. 4(c), where the lower coffee
concentration results in less freeze concentration and
less inhibition of crystal growth, hence larger crystal sizes
(average rh = 20.4 lm) and notably thinner walls.