The medical ultrasound imaging systems have been improved by taking into account the nonlinear wave propagation. Since the backscattered nonlinearities are a function of the transmitted signal, enhancing these nonlinearities means to find the best excitation. In order to reach the solution, one way consists in modeling the ultrasound system by parallel subsystems (one channel per harmonic component). Therefore finding the optimal wave for harmonic generation means to create a matched filter for each channel. Time reversal process is well-known to optimize the transmitted wave thanks a physical matched filter. However, it is not well-adjusted, since the time reversed signal will be composed of the second harmonic components and the nonlinearities will double the frequency. Therefore, the harmonic backscattering will thus be due to a linear process and so will not optimize the nonlinearity generation. Our objectives is thus to find the optimal wave to locally optimize the harmonic generation by taking into account this frequency doubling and without increasing the transmit power. As described by the time reversal, the first wave at 2 MHz propagates to the medium. The backscattered signals came mainly from high impedance inclusion behind another inclusion. The backscattered second harmonic components at 4 MHz are extracted from a Hammerstein model. The second harmonic signal is time reversed and then frequency shifted at the fundamental frequency (2 MHz) by two steps. First, the phase is evaluated from the second harmonic signal. Second, the second harmonic signal is demodulated with the phase estimated previously. Finally, the time reversed signal is retro-propagated. In this study, it is shown that the nonlinear components can be increased by 15 dB in comparison to the time reversal process and conventional imaging. Moreover, the SNR can be increased by 7dB. By maximizing the harmonic generation, the nonlinear time reversed waves focus better on the inclusion and wave coherence is preserved. From this point of view, the optimization process can be viewed as an extension of the time reversal principle to nonlinearites. This finding is very interesting even if the wave phase is a critical point.