A practical consequence of the behavior observed at higher current density is that the instabilities grow from random perturbations caused by the initial kinetic roughening regime. This implies that growth stability can not be improved by, for example, starting from a very smooth surface. To suppress instabilities, other strategies must therefore be used. We have examined two strategies, pulse plating and additives, and describe below nanoscale observations of each.
Pulse plating is a well-known electrodeposition technique for improving surface roughness [120]. Figure 65 and Appendix Video B.12 show a set of pulsed ex- periments where the current is on for one second followed by a five second waiting time with no applied current. The off time allows diffusion to replenish the solution near the electrode, which in turn allows for uniform growth during subsequent pulses
of times less than the transition time. The current density used implies a transition time of over 3 s from the previously observed experiments. Figure 65a shows that growth instabilities do not develop, even with asperities present on the underlying substrate. By tracking the max, min, and average height in Figure 65b, we see there is initially little growth in distance between the max and min when pulsing. Further confirmation comes when considering the point-wise normal speed distribu- tion, which is shown in Figure 65c. The RMS roughness shown inFigure 65d has an average growth exponent, which includes the off time, of 0.48 during the four total cycles. This growth exponent is approximately akin to random uniform deposition and means the surface is governed by physics that should not develop large asperities like its continuous growth counterpart. The local current density increases from 142
A/m2 to 539 A/m2 for the last four cycles shown based on the rate of average height growth for the total period including off time. We include the off time in the velocity calculation as the diffusive behavior is governed by the total time. This increase in growth rate is due to a bubble that moves into the imaging region after the first 4 cycles (i.e., at 24s) (it ultimately can be seen at the end ofAppendix Video B.12), causing the liquid layer to thicken as the silicon nitride membranes bow out into the vacuum. The data in Figure 65 show that pulse plating can be used to grow in a stable mode. By varying the applied current, the period, and on/off time ratio, one can vary independently the diffusive limit and the reaction kinetics [120].
Additives can be used to control growth morphology during electrodeposition. Here, we show the effects of lead on copper plating. Figure 66a and Appendix Video B.13 shows a set of experiments where the electrolyte includes saturated PbSO4; the surface appears smooth, compared to Figure 62. The growth rate is constant during deposition as seen in Figure 66b and stagnant when the current
Figure 65: Galvanostatic Pulsed deposition with average current density of 288A/m2 and local average growth rate of 13nm/sincluding off time (equivalent to 340A/m2 ) in 0.1M CuSO4 and 0.18M H2SO4. The local current density when only considering the pulse on time is 2040A/m2. (a) Images recorded in bright-field conditions extracted from the video sequence found in theAppendix Video B.12. Times in seconds since current flow began. (b) Average, maximum, and minimum growth height as a function of time. (c) “Heat Map” of normal growth speed with white being fastest and blue/purple being slowest (need to redo with a color scale) (d) Log-log plot of RMS roughness of interface as a function of time. Best fitting straight lines are shown with exponents indicated.
is off, with a constant difference between the max and min heights. Figure 66c
reveals a very consistent point-wise growth velocity along the entire edge. The RMS roughness calculation shown in Figure 66dhas an average growth exponent of 0.02, and even shows an initial smoothing of the substrate. The growth clearly does not have a large amplification. Presumably the lead is acting as a surfactant that slows the reaction rate, keeps the system reaction limited, and allowing surface tension to continue to smooth the surface. This hypothesis is supported by the fact that the local current density as measured via the growth rate is lower when lead is present compared to when plain copper. See the supplement for details.
In conclusion, we have quantified the transition to morphological instability during the early stages of electrodeposition of copper, using liquid cell electron microscopy to provide the spatial and temporal resolution necessary to probe the critical early stages of growth. We have further quantified the exploitation of the initial roughening regime via pulse plating and the use of additives as a means to postpone the onset of these diffusive instabilities. Quantitative analysis of the videos shows we can pinpoint the onset of diffusion limited growth explicitly, even when macroscale approximations fail to account for local variations in current density. The quantitative measures also reveal the growth mode prior to the transition time where the surface roughens in a reaction limited growth regime. This suggests that an initially flat surface can not help to suppress instabilities and other means must be used. Detailed understanding of early stage growth will lead to the development of control schemes and geometries to be exploited in nanofabrication.
Figure 63b shows δi for the low current density deposition given in Figure 64. The key result seen in this figure is that δi is essentially constant throughout the entire deposition and never shows a rise. Although the absolute value of δi is larger
Figure 66: Galvanostatic deposition with average current density of 216 A/m2 for 10 s followed by 10sof no current, repeated. Average local growth rate is 14nm/s(equivalent to 381A/m2 ) in 0.1M CuSO
4and 0.18M H2SO4saturated with PbSO4(a) Images recorded in bright-field conditions extracted from the video sequence found in the Supplemental Appendix Video B.13 with detected edge as overlay. Times in seconds since current flow began. (b) Average, maximum, and minimum growth height as a function of time. (c) “Heat Map” of normal growth speed with white being fastest and blue/purple being slowest (need to redo with a color scale) (d) Log-log plot of RMS roughness of interface as
compared to the high current density case, the relative value and time derivative are what provide insight into an oncoming diffusive instability. This also shows that the initial phase of the growth shown in Figure 62is the transient, reaction-limited regime theorized to exist before the onset of diffusion limited growth where we have roughening, but not the formation of ramified features.
7.3.1. Copper Electrodeposition Partial Conclusions
We have investigated and quantified the transition to morphological instabilities dur- ing the electrodeposition of copper. Liquid cell electron microscopy with devices like the nanoaquarium provide access to the spatial and temporal resolutions necessary to probe the critical early stages of growth. Quantitative analysis of the obtained videos shows we can pinpoint the onset of diffusion limited growth explicitly, even when macroscale approximations fail to account for local variations in current density. By considering the variations in point-wise current density (as measured by the speed of the normal velocity along the interface), we can see the rise of the diffusive instability well before the transition is visible in the interface. The quantitative measures also reveal the phase of growth prior to the transition time where the surface roughens in a reaction limited growth regime. Detailed understanding of early stage growth will lead to the development of control schemes and geometries to be exploited in nanofabrication.