UHV-TEM

Ultra high vacuum transmission electron microscopy

Electrochemical deposition of Copper Electrochemically deposited copper is used for metallization in integrated circuits, and a detailed understanding of nucleation, growth and coalescence during electrodeposition is essential in optimizing the final microstructure. A detailed understanding of kinetics is best achieved by real time observations, but liquid phase processes like this are difficult to study in real time, whether by TEM or by scanning probe techniques. One solution for TEM is to confine the liquid between electron transparent membranes in a cell. Here we describe how such a cell can be constructed and used to make dynamic observations of the nucleation and growth of copper clusters during electrodeposition. This experimental technique, which requires only a conventional microscope, should be applicable to a broad range of dynamic phenomena at the solid-liquid interface. Design of a liquid cell The reaction takes place in a cell in which liquid is confined in an electron-transparent layer. The cell is made up of two Si wafers glued face to face. Each wafer is coated with 100nm Si3N4 and etched from the back to leave a 100x100 µm Si3N4 viewing window. An SiO2 layer maintains a distance of 0.5-1 µm between the wafers. A polycrystalline Au working electrode is deposited across the viewing window and over a via, and connected through the wafer (resistivity 0.005Ωcm) to an external contact. The upper wafer includes two reservoirs (large holes), which are capped with thick (1 mm) glass spacers. Liquid is introduced with a syringe and flows between the viewing windows by capillary action. The cell is then sealed by gluing sapphire lids over the spacers. A gold wire is placed as counter electrode in one reservoir, and a Cu reference electrode is placed in the other reservoir. A heat-curing epoxy is used to glue the wafers and spacers, and a UV-cured epoxy is used for the sapphire lids. The purpose of the spacers is to separate the UV-cured epoxy from the electrolyte so that the epoxy can set properly. To improve the image quality, an imaging energy filter (Gatan, Inc.) reduces the background of electrons scattered inelastically from the windows and liquid. The spatial resolution is ~5nm, limited by the thickness of liquid (and windows) which must be imaged, and images can be recorded at video rate (30 frames per second) with high signal to noise ratio. Figure 1: A schematic diagram of the components of the cell (with the viewing window enlarged for clarity) and a photograph of the cell with optical micrograph of the viewing window. Electrochemistry in small volumes Does electrodeposition in the restricted volume available in the cell faithfully mimic the standard process? Here we compare cyclic voltammograms recorded in the TEM cell with data from a larger electrochemical cell. Both show typical peaks for diffusion-limited deposition and stripping of Cu. The similar current density suggests that deposition occurs only on the electrode and not elsewhere in the TEM cell, and ex situ examination confirms that 3D clusters grow uniformly over the electrode. Figure 2:Cyclic voltammograms for electrodeposition of Cu on polycrystalline Au, recorded in the liquid cell using a Cu reference electrode (black) and in a large volume of the same electrolyte (red). The scan rate was 25mV s-1 and the electrolyte contained 0.2M CuSO4 + 0.05M H2SO4. The current density was calculated using the cell’s measured electrode area of 210µm x 10µm. The variation in peak potential and the slightly different peak current density are due to the ohmic drop over the solution in the cell’s narrow channel. Also shown are copper clusters of density 6.5x108 cm-2 formed after galvanostatic deposition in a cell at 5 mAcm-2 for 4 seconds. The image was recorded with a focused ion beam using secondary electron contrast. Nucleation dynamics Video sequences recorded during deposition are shown here and compared with a simple growth model which applies under the constant-current conditions used in these experiments. In the model, the growth of each cluster is limited by its share of the total current, and (as long as there is no significant depletion of copper ions in the solution) the current is shared between clusters according to their surface areas. Even though this simple model does not include convection and diffusion, it provides a good match for early growth, up to ~2 seconds. This is significant in allowing calculations of cluster size distribution, important in understanding film uniformity at coalescence. At later times, we expect diffusion-limited kinetics due to depletion of ions in the solution, and we do in fact see a reduction in growth rate after 2 seconds. Video 1 (slow_deposition_rate.mp4) and Video 2 (fast_deposition_rate.mp4)    Figure 3: Still from Video 1 (slow_deposition_rate.mp4)    Figure 4: Still from Video 2 (fast_deposition_rate.mp4) Videos showing nucleation and growth of copper clusters on a polycrystalline gold electrode during electrochemical deposition and stripping, at current densities of 5 mAcm-2 and 50 mAcm-2. The field of view (horizontal distance) is 6.3 µm. The electron beam passes through the electrolyte, the SiN membranes and the Au electrode, which all contribute a fairly uniform background. The grain size of the electrode is small (15nm) and individual grains are not visible. As the current is applied a change of contrast occurs, caused by charging or by motion of the liquid, but this rapidly fades and nucleation events are then visible. Individual Cu clusters show as dark regions due to their higher electron scattering. Progressive nucleation occurring at 5 mAcm-2 is clearly visible in the first sequence. Nucleation is complete after 2-3 seconds but no further growth occurs after about 3 seconds due to depletion of cupric ions in the electrolyte. After about 8 seconds the current source is turned off then a reverse (stripping) current is applied, and the nuclei dissolve back into the solution. At 50 mAcm-2, the same processes are visible but with a much shorter time scale: in this case nucleation requires less than 0.4 seconds. The nucleation density is higher and the nuclei are smaller. Detailed analysis of the growth kinetics show that depletion of the electrolyte becomes important at a much earlier time at this high growth rate. Figure 5: The radii of several individual clusters measured from the first video, fitted by a growth simulation using the experimentally measured parameters. The fit is good at small times while for large times growth is limited by depletion in the liquid, an effect not included in the simulation. "Homogeneous nucleation" We can investigate preferred nucleation sites for individual Cu clusters by comparing the same area of the electrode after four separate deposition experiments, with the Cu stripped between experiments. Each time, nuclei appear in different positions with no evident correlation. Nucleation of Cu on Au occurs preferentially at steps. In our experiments the grain size in the electrode (15nm) is much smaller than the separation between nuclei (400nm) so the density of surface defects potentially available (steps, grain boundaries and triple points) is much higher than the actual nucleus density. Thus the uncorrelated arrangement shows that nucleation is not dominated by a small population of preferred sites. If many sites have similar nucleation probability, we may be seeing the blocking of potential sites by the diffusion field of a previously formed cluster, a situation similar to homogeneous nucleation. Figure 6:Images of the same area of electrode recorded after four deposition experiments, and a composite image with the positions of clusters superimposed (different colours). Galvanostatic deposition at 50 mAcm-2 was carried out in 0.3M CuSO4. Summary It is possible to measure and analyse the nucleation and growth of individual copper clusters during electrochemical deposition using a technique which allows real time observation of liquid phase growth. On polycrystalline Au, nucleation shares characteristics of a homogeneous process with many equivalent sites available, and cluster growth is by equipartition of the available material flux over the cluster surface areas. Electrochemical measurements and analysis suggest that we can reproduce conventional electrodeposition at short time scales with appropriate growth rates. The technique developed here, with its combination of good time resolution, large area of analysis and TEM analytical capabilities, is complementary to scanning probe techniques and is especially useful in analysing sub-second, 3D growth processes.