We study the origin of the bi-stability jump in the terminal velocity of the winds of supergiants near spectral type B1 .
Observations show that here the ratio v _ { \infty } /v _ { esc } drops steeply from about 2.6 at types earlier than B1 to a value of v _ { \infty } /v _ { esc } =1.3 at types later than B2 .
To this purpose , we have calculated wind models and mass-loss rates for early-type supergiants in a T _ { eff } grid covering the range between \mbox { $T _ { eff } $ } = 12 ~ { } 500 and 40 ~ { } 000 ~ { } K .
These models show the existence of a jump in mass loss around \mbox { $T _ { eff } $ } = 25 ~ { } 000 ~ { } K for normal supergiants , with \dot { M } increasing by about a factor five from \mbox { $T _ { eff } $ } \simeq 27 ~ { } 500 to 22 500 K for constant luminosity .
The wind efficiency number \eta = \mbox { $ \dot { M } $ } \mbox { $v _ { \infty } $ } / ( L _ { * } / c ) also increases drastically by a factor of 2 - 3 near that temperature .

We argue that the jump in mass loss is accompanied by a decrease of the ratio v _ { \infty } /v _ { esc } , which is the observed bi-stability jump in terminal velocity .
Using self-consistent models for two values of T _ { eff } , we have derived v _ { \infty } /v _ { esc } = 2.4 for T _ { eff } = 30 000 K and v _ { \infty } /v _ { esc } = 1.2 for T _ { eff } = 17 500 K. This is within 10 percent of the observed values around the jump .

Up to now , a theoretical explanation of the observed bi-stability jump was not yet provided by radiation driven wind theory .
To understand the origin of the bi-stability jump , we have investigated the line acceleration for models around the jump in detail .
These models demonstrate that \dot { M } increases around the bi-stability jump due to an increase in the line acceleration of Fe iii below the sonic point .
This shows that the mass-loss rate of B-type supergiants is very sensitive to the abundance and the ionization balance of iron .

Furthermore , we show that the elements C , N and O are important line drivers in the supersonic part of the wind .
The subsonic part of the wind is dominated by the line acceleration due to Fe .
Therefore , CNO-processing is expected not to have a large impact on \dot { M } but it might have impact on the terminal velocities .

Finally , we discuss the possible role of the bi-stability jump on the mass loss during typical variations of Luminous Blue Variable stars .