Parameter sensitivity analysis for mean field games of production (Q2673509)

A couple of years ago Chan and Sircar have proposed an interesting mean field game system to model the production of an exhaustible resource (see [\textit{P. Chan} and \textit{R. Sircar}, Appl. Math. Optim. 71, No. 3, 533--569 (2015; Zbl 1318.49072); \textit{P. Chan} and \textit{R. Sircar}, SIAM Rev. 59, No. 3, 588--615 (2017; Zbl 1369.91146)]). Using the words of the authors (of the current paper), ``in this game, players control the rate at which they will sell from their current stock, and they must leave the game once that stock goes to zero. The optimal strategies are determined by the market price. In equilibrium, the market price is determined by the average of all the players' strategies. When the demand depends linearly on the production rate (or on the price offered), one can determine the equilibrium by solving [the mean field game system]''. In this model a particular small parameter \(\varepsilon>0\) appears, that measures the rate at which the firms compete. When \(\varepsilon=0\), every firm is a monopolist (and in such a scenario the PDE system decouples, hence it is much easier to solve), if \(\varepsilon\) is very large, all firms are nearly interchangeable in the eyes of consumers. In their original paper, Chan and Sircar have proposed a numerical scheme to solve the PDE system associated to the mean field game. This method consisted in a formal asymptotic expansion of the solutions with respect to the parameter \(\varepsilon\). This expansion was entirely formal as to justify it, one would need to show that the solution is actually smooth with respect to this parameter \(\varepsilon\). In the paper under review, the authors fill this gap, and they provide the rigorous regularity results for the solutions with respect to the parameter. This is not a trivial task, and the main analysis is carried out at the level of various linearized systems of forward-backward type. The philosophy behind this is similar to the sensitivity analysis that one can carry out to show the well-posedness of the master equation arising in mean field games (see for instance [\textit{P. Cardaliaguet} et al., The master equation and the convergence problem in mean field games. Princeton, NJ: Princeton University Press (2019; Zbl 1430.91002)]).