There is much interest on the numerical and practical optimization of the doping profiles in semiconductor devices. This interest is mostly driven by the power semiconductor industry, which is in continuous search for higher breakdown voltage semiconductor switches that have lower on-state resistance, and high current handling capacity and switching speeds. Among the existing power semiconductor devices, insulated-gate bipolar-transistors (IGBTs) are becoming more and more attractive because of their high blocking voltage capabilities and on-state current. Although these devices suffer from a relatively lower switching speed compared to current state-of-the-art metal-oxide-semiconductor field-effect-transistors, IGBTs have already become the preferred switching devices in many power applications, such as induction heating applications, traction motors, etc. [1]. In this work we present numerical finite element optimization results for a realistic 70 mm-wide punch-through IGBT structure. The optimization focusses on the two-dimensional doping distribution of the gate, emitter and collector junctions, and of the drift and buffer layers in order to increase the breakdown voltage of the IGBT. The optimization method is based on an adjoint space technique that our group has recently developed for the design of nanoscale semiconductor devices, and which is adjusted here to compute the doping sensitivity functions of the breakdown voltage and on-state resistance in IGBTs [2-4]. These doping sensitivity functions show how sensitive the breakdown voltage and on-state resistance are when we add one additional acceptor or donor impurity at a particular location inside the semiconductor. Since the doping sensitivity functions provide information about the gradient of the breakdown voltage and on-state resistance with respect to the doping concentration they are instrumental for the numerical and practical optimization of IGBTs. As we will show at the conference, we are able to increase the breakdown voltage from approximately 300 V to over 700 V without deteriorating the on-state resistance of the transistor. In the full presentation we will describe how the adjoint method can be used to calculate the doping sensitivity functions of the breakdown voltage, on-state current and voltage, and on-state resistance. We will also present details about the numerical implementation of the optimization method, which is based on a modified gradient optimization appropriate for semiconductor devices. Note that most traditional optimization approaches such as the steepest descends method, the Broyden–Fletcher–Goldfarb–Shanno algorithm, or similar algorithms fail because of the doping concentration varies over many orders of magnitude, which vary can from region to region inside the same device.