electronic convergence

[spintron]

I am trying to relax a metal-insulator heterostructure. I am having problems
conveging the total energy at every electronic iterations. The total energy seems to increase. In the end the loop is broken because of the threshold 1e-5 . Any insights on this?
confree energy TOTEN = -167.52894716 eV
free energy TOTEN = -168.36355883 eV
free energy TOTEN = -168.23184756 eV
free energy TOTEN = -167.37089993 eV
free energy TOTEN = -167.41203008 eV
free energy TOTEN = -167.32645766 eV
free energy TOTEN = -167.31684814 eV
free energy TOTEN = -167.31138188 eV
free energy TOTEN = -167.31040229 eV
free energy TOTEN = -167.30997861 eV
free energy TOTEN = -167.30998027 eV
free energy TOTEN = -167.309980 eV
free energy TOTEN = -168.38991456 eV
free energy TOTEN = -208.59069684 eV
free energy TOTEN = -182.56234786 eV
free energy TOTEN = -170.66164975 eV
free energy TOTEN = -169.04249581 eV
free energy TOTEN = -169.09522469 eV
free energy TOTEN = -168.87834461 eV
free energy TOTEN = -168.81930097 eV
free energy TOTEN = -168.81395638 eV
free energy TOTEN = -168.79267845 eV
free energy TOTEN = -168.78923394 eV
free energy TOTEN = -168.77743142 eV
free energy TOTEN = -168.77571713 eV
free energy TOTEN = -168.77480724 eV
free energy TOTEN = -168.77476356 eV
free energy TOTEN = -168.77475368 eV
free energy TOTEN = -168.774754 eV