Abstract: In this paper, using panel data from 2006 to 2015 of the manufacturing industry, and from the perspective of industry heterogeneity, the generalized least squares method and the variable coefficient model were used to test the difference effect of industrial agglomeration on the green total factor productivity in manufacturing industry, heterogeneous hysteresis effect, heterogeneous interaction effect and heterogeneous short-term fluctuation effect. The research showed that in most industries industrial agglomeration significantly promoted the green total factor productivity, but the effects of some industries were not significant or even negatively correlated. This showed that industrial agglomeration did not show the scale effect in all industries, and the congestion effect in some industries exceeded the scale effect. From the perspective of lagged effect, each industry showed a heterogeneous variation with lag period:positive(negative)direction effect gradually increased, gradually weakened, first increased and then decreased, first decreased and then increased, always remained insignificant and so on. From the perspective of interactive effect, the smaller the average size of the business sector, the more likely to get positive effects from industrial agglomeration; in most industries, the state-owned property rights and asset specificity all played a restraining role to a certain extent in the green productivity effect of industrial agglomeration. However, on the contrary, in a few industries such as transportation equipment manufacturing and pharmaceutical manufacturing, state-owned property and asset specificity played a positive role; in the short term, industrial agglomeration in most industries did not effectively promote the growth of green TFP. Only the short-term effects of industrial agglomeration in the manufacturing of textile, clothing, shoes and hats, non-metallic mineral products and special equipment manufacturing industry were significant. Meanwhile, the long-term and short-term effects of most industries showed a differentiated feature. Therefore, when formulating a policy of agglomeration, the government should not neglect the differences among different industries and adopt a "one-size-fits-all" measure. Instead, the government should formulate targeted measures and dynamic policies according to the specific impact laws and the agglomeration stages in different industries.