Filling copper into TSVs (Through Silicon Via) is achieved by copper electroplating in an acid copper sulfate bath with combination of several additives. The additives are generally classified into accelerator, suppressor or leveler. It is assumed that suppressor and leveler cover a plating surface and both additives suppress electrodeposition, while accelerator obstructs suppressor adsorption and prevents the suppression of electrodeposition. However, the filling mechanism is not well understood. For the TSV filling, we consider that leveler has an essential role because extreme bottom-up deposition is available with only leveler addition though significant non-uniformity is observed. We assume that the extreme bottom-up is caused by diffusion-adsorption based mechanism. To achieve more detailed understanding, mathematical modeling and numerical simulation are discussed, but there are many unknown parameters to perform the simulation. Additive size is one of important parameters we want to know, because coverage of suppressing additives adsorbing on the plating surface dominates the electrodeposition rate. Therefore, in this study, sizes of additives were attempted to measure in an aqueous condition using an AFM (Atomic force microscope).

High-speed AFM (Research institute of biomolecule metrology Co., Ltd. , Japan ) is a good tool for measuring soft material in an aqueous solution, and we employed it for the observation of the plating additives. Figure 1 shows an example image. PEG (Polyethylene glycol) of Mw. 3000 was observed on a mica substrate. The high-speed AFM can take one frame image very quickly (maximum 20 frames per sec.), but each image is not that reliable as shown in figure 2 which shows an identical particle observed by the high-speed AFM. Averaging was needed to determine the particle size. Probe tip of the AFM was also problematic. Less than 10nm of tip radius was guaranteed by the manufacturer, but precise shape was unknown. Then, we decided to measure DNAs (deoxyribonucleic acid) as a reference, and the tip radius of the probe was estimated to be typically 5.5nm. With image averaging and probe tip compensation, additive radiuses were calculated assuming hemisphere shape. Radiuses of PEGs of Mw. 3000 and 20,000 were estimated to be 1.8nm and 3.3nm, respectively. Interestingly, radius of PEI (Polyethylene imine) of Mw. 10,000 was estimated to be 1.3nm which was smaller than PEG of Mw. 3000.