The emergence of labor division in a multi-agent system is analyzed by the method of statistical physics. We consider a system consisting of N homogeneous agents. Their behaviors are determined by the returns from their production. Using the Metropolis method in statistical physics, which in this model can be regarded as a kind of uncertainty in decision making, we constructed a Master equation to describe the evolution of the agent's distribution. We introduce an earning function including learning by doing to describe the e ect of technical progress and a formula for competitive cooperation. And we also introduce two order parameters to characterize the division behavior. The model gives us the following interesting results:
(1) As a result of long-term evolution, the system can reach a steady state. (2) When the parameters exceed a critical point, the division of labor emerges as the result of phase transition.
(3) Although the technical progress decides whether or not phase transition occurs, the critical point is strongly a ected by the competitive cooperation. From the above physical model and the corresponding results, we can get a more deeply understanding about the labor division.