Based on in situ optical measurements and off-line analyses for four c
oals, the basic features of single-particle pulverized coal char combu
stion have been elucidated as a function of carbon conversion. Two reg
imes can be clearly defined: one at low carbon conversion, where the r
eacting particle populations have properties that are nearly time inva
riant and a second regime at higher carbon conversion where the distri
bution properties change dramatically. At low carbon conversion, there
is a broad distribution of single-particle combustion rates, reflecti
ng the heterogeneity in the parent fuel. Particle-to-particle reactivi
ty differences are shown to be the primary cause of the broad temperat
ure distribution for Pocahontas coal char. At high carbon conversion,
carbon-rich particles can be distinguished statistically from inorgani
c-rich particles by in situ measurement of their spectral emissive fac
tors at 800 nm. In each case where char carbon conversion proceeds pas
t 50-60%, many particles are observed to undergo large temperature dec
reases resulting from a loss of reactivity, referred to as near-extinc
tion events. Near-extinction is generally observed to occur before lar
ge changes are observed in the particle optical properties, suggesting
that deactivation occurs when the particles are still carbon-rich. Pl
ots of particle temperature vs emissive factor conveniently illustrate
and summarize the process of char particle combustion to high convers
ion. These plots reveal two distinct stages in the combustion lifetime
of a char particle: (1) a rapid combustion stage at low carbon conver
sion, followed by (2) a deactivation and near-extinction at roughly co
nstant optical properties, initiating a final burnout stage that occur
s slowly and at low temperatures. The two-stage nature of the char com
bustion process significantly lengthens the time required to achieve h
igh carbon conversion, and the existence of two stages cannot be predi
cted by conversion-independent kinetic models. More realistic char oxi
dation models are needed that account for fuel heterogeneity and conve
rsion-dependence effects.