Onderzoek naar interacties tussen plasma en ultradunne materialen voor atomair etsen met hoge doorvoer in toekomstige knooppunten

With the exponential growth of data generation, the storage and processing capabilities must scale at a similar pace within constrained space, cost, and power budget. The reduction of the dimensions of the memory and processing elements has been maintained until now by a combination of improvements in lithography, etching and the introduction of new materials. The next generation of conventional lithography, called High Numerical Aperture Extreme Ultraviolet Lithography (“High NA EUV”), will force the use of extremely thin layers, typically a dozen atom thick, whose main purpose is to be used as a protection layer of the material underneath. Two major issues arise from it. The most obvious one is the extreme etch resistance these few atom-thick films will need to have to the plasma used to transfer the patterns defining the device to the underlying materials. Also, at such extreme thickness, materials can become transparent to plasma, thereby failing in their purpose as a protection from the areas which should not be etched. For many combinations of materials, this is the final physical limit of plasma etching as it used to be performed for the last five decades. Thankfully, some alternative ways of etching at such dimensions have already been worked on, among which atomic layer etching (ALE). This approach is based on the difference in the amount of energy needed to induce desorption of volatile molecules from a surface, and is usually introduced as a two-step cyclic process: saturate a surface with some defined molecule and then trigger desorption by ion bombardment with adequate energy. The choice of the correct molecule to desorb from the surface of the material to be etched and not from the protection material, along with a fine control of the energy of the incoming plasma species, is defining the good performance of an etch process. The main drawback of such an approach is its very poor throughput: each step of every cycle is several second long and each cycle etches few Angstroms, making atomic layer etching a slow and costly process. Increasing the efficiency of each step quickly becomes bottlenecked by the time needed to replace one gas with another in the entire volume of the etch chamber. Improving the throughput of atomic layer etching will therefore require a deep understanding of the plasma/ material interactions both during the steady and the transient states of gas switching as the later will represent most of the process time.