Negative risk free rates were theoretically possible, but thought at the time to be stupid. See for example the development of the Cox Ingersoll Ross model, "The standard deviation factor, \sigma \sqrt{r_t}, avoids the possibility of negative interest rates for all positive values of a and b."

Many interest rate models had been in the past either criticised for theoeretically allowing negative rates, or specially developed to avoid them.

https://en.wikipedia.org/wiki/Cox%E2%80%93Ingersoll%E2%80%93...