Basically a three way catalyst has two operating modes during which two distinct types of reaction occurs within it: -
Reduction - With an abscence of O2, achieved with a feed gas of lambda <1 (deals with NOx)
Oxidization - With an excess of O2, achieved with a feed gas of lambda >1 (deals with HC & CO)
To enable each reaction to occur the fuelling has to be modulated rich/lean in a sort of sine wave. This is called lambda dither and, the average over a given time is lambda=1. This is achieved by the front sensor, which ECU uses to control the amplitude of the rich/lean excursions.
During lean operation the catalyst behaves like an O2 store and, upon the introduction of rich feedgas there is a delay before this oxygen us used up and reduction reactions can begin. This is not ideal for the efficient reduction of all 3 pollutants. To effectively reduce all 3 pollutants the exit gas of the catalyst needs to be lambda=1.
A rear sensor is thus employed to give fuelling control for the exit gas, however, the type of control is slightly different. Instead of affecting the amplitude the ECU uses the rear sensor to effect a bias by holding the time that the fuelling remains rich or lean. This is called bias.
So basically the front sensor controls to lambda = 1 and the rear sensor controls the switchover/hold time for the rich/lean excursions.
MS