Next: Soap film experiment Up: Data analysis Previous: Data analysis

I will show plots of , , and , for seven different driving amplitudes. I will compare the results from this analysis with the results presented in [1] (appendix C). The present analysis based on Kolmogorov's choice of which data should be counted, whereas our earlier analysis was heavily influenced by Richardsons view.

The implementation of the particle tracking algorithm we used for the data aquisition gives an over-representation of particles which are close by each other, but this should not give rise to any concern. Whether we take ten or thousand samples should the probability distribution of the velocities be identical.

Lengths are measured in mm and times are measured in ms, in the graphs presented in this and the following section.

Table 5.1: Driving amplitudes and the corresponding reduced control parameter.

Figure 5.5: Left: as a function of R. Right: The gradients of the graphs in the plot on the left as a function of .

Figure 5.2 shows as a function of R (on the left), and locally defined powers of R that fit the function (on the right). None of the curves have even something close to a constant gradient over a whole decade. The maximum gradient occurs at a particle separation equal to one wavelength ().

Figure 5.3: as a function of R. This analysis is limited to look at relative particle tracks from the moment the two particles were closest to each other (and closer than ) and onwards.

The data plotted in figure 5.3 scales roughly as R to powers in the range 0.8 to 1. This is equivalent to the powers slightly below 2 you see for allmost a decade in figure 5.2. So even though we have different rules for which data should be analysed, we get similar results (just as Kolmogorov's turbulence theory fits Richardson's data quite well).

Figure: Left: as a function of R. Right: The gradients of the graphs in the plot on the left as a function of .

I do not have an interpretation of , but it shows a much nicer behaviour than , so I would like to know if it is possible to relate it to existing turbulence theory.

Figure 5.4 shows as a function of R (on the left), and locally defined powers of R that fit the function (on the right).

is roughly proportional to for .

Figure: Left: as a function of R. Right: The gradients of the graphs in the plot on the left as a function of .

Figure 5.6: divided by as a function of R. This should be a horizontal line for wave turbulence.

Figure 5.5 shows as a function of R (on the left), and locally defined powers of R that fit the function (on the right). There is no distinct power law in these data.

Weak turbulence theory predicts that is proportional to (see definition of b in section 1.2.4), and thus that is proportional to . Figure 5.6 shows the data divided by the predicted function. From and up to the maximum value for R does the weak turbulence prediction fit with the data.

The factor needed to fit the data to the function depends on the energy dissipation in the system (please see [14] for the details).

Figure 5.7: The upper graph shows the number of data points per mm in the old analysis, and the lower graph shows the same data for the analyses made for this thesis. Both of the graphs are for a driving amplitude of .

Figure 5.7 shows the densities of data points used for some of the graphs in the preceding figures. It is not a big surprise that there are more data available for the analyses made for this thesis, if you consider that I have not ruled any of the available data out beforehand.

Next: Soap film experiment Up: Data analysis Previous: Data analysis

sparre@cats.nbi.dk