General issues: The abstract is 53 words too long for ApJ, and needs to be shortened at the revision. %MJK Done. decimated run with A=16 missing. Could be done meanwhile waiting for the report. MJK: made k.dat file today and queued the run. MJK: Has now run nearly 1000 time units, kinematic range nicely covered already. %MJK Inserting now numbers to table 1 and 3. E_0 issue: Matthias almost call a halt to everybody's work today as he though the whole leakage issue is because of simultaneous usage of kinetic and magnetic forcing producing an E_0, as the forcings could be correlated. We had a discussion over zoom, and looked at my purely magnetic runs with exactly similar parameters but kinetic forcing disabled, and saw no difference whatsoever in the results. Two possibilities: both types of runs have E_0, or then this is not the cause of the mean fields. As we are now avoiding any kind of conclusive statement of the reason for the mean fields anyway (quite specifically even declare that they are not due to the leakage), I do not see any reason to delay the submission because of this. If Matthias now finds E_0 from some of the runs, then we can add that explanation later. AP: Matthias to commit his removal of E_0 and MJK to run spot checks of the runs. And, now that I think back, this issue was even discussed in Nordita at some point. I remember checking this specifically, but perhaps not completely thoroughly, and then I did not find u and b being correlated, hence no E_0. I think there was even some alteration to the forcing functions made by Matthias himself because of this issue, due to which I had to run a whole bunch of simulations again. a1 magnetic forcing case now run. Observations: test problems get unstable, even daainit=50 is too long. Plateaus hard to find for any measurement. Strong mean fields are, again, generated instantaneously. The fields are larger than in the decimated case, but smaller than in the runs included in the paper with the standard forcing. Trying to find plateaus and compare the transport coeffs., however, indicates that the measurements do not change qualitatively. Hence, removal of the emf_0 does not provide a remedy to the mean fields, and does not change the measurements considerably. %MJK Should we now mention this in the revision? Items related to the beginning of Intro %MR: depends on the whole tensor %would %AB: do you really think you can stabilize it with anything? %MR: I think we cannot conclude from a negative \eta_xx that the whole tensor is not positive definite. %AB: we just talk about one component (see work with Long Chen) %MJK Text is weak enough, case closed. Model: non-converged discussion: %MR: We may drop \meanUU here. %AB: why don't you do this then? %MJK This has been re-instated now. Units: %MR: S needs a unit %MJK Unit is [t], for which we have not defined any unit either, but have used the turnover time in the plots. %MR: a general remark on the expected shape of the mean fields - homogeneity of turbulence. %MR: I meant to explain here that ideally we have homogeneity and hence constant coeffs -> \meanBB ~ exp(kz+\lambda t), % but there are limitations if \meanBB is dynamically relevant. -- We can do this later. Endless discussion chain: %MR: then the referee would ask: Why haven't you done it this way? %MJK because our intention was not study this phenomenon in the first place, but the slower dynamo instability seen before elsewhere, and claimed to change in nature by adding magnetic forcing. So we have a clear answer to the referee. %MR: and to the reader? %MJK Obsolete. Another non-convergent discussion: %MR: I mean are they diffusive or inductive? %MJK negative eta_{yx} is inductive. In FMHD it is positive, here negative, so how to even try to classify them? %MR: That's what I mean: the reader would not know. %MJK Declare this obsolete. Another dispute: %MR: unclear: first we say that it is negligible ("also holds ... here") and then we say, it is not. %MJK In the dispersion relation it is not "summed" with eta_{yx} as you say, but when you "fit" coefficients %MJK from the time series, e.g. using the method of moments, all coeffs should be fitted for, especially %MJK if they are non-zero. %MR: I think it is confusing for the reader to intersperse the discussion of the dynamo excitation conditions % with remarks on a rather unrelated fitting procedure to measure the coefficients. % At best, this can be in a footnote. %MJK Check out if this is clear enough. Argument about the dynamo wavenumbers %MR: How is this justified for the FK runs? %MJK All the modes decay with the same rate, I do not know which one should I choose there. Even if you would come up with a better idea NOW I would not start computing them again. %MJK Clearly enough explained to close this discussion. Items related to 2.5 and A.1. %all those wavevectors for which $k_{x,y,z} \leq k_{\rm min}=2$. %NS: abs || as both = and - are removed simultaneously; also /k_1 %AB: could there be a (slightly) different cutoff for positive and negative values that explains the asymmetry in the appendix? %NS: these are integer numbers in k.dat and identical in +ve and -ve %NS: checked again k.dat file for this %NS: would numbers of elements of +kx and -kx be relevant? %AB: Yes, the numbers of +kx and -kx should be the same. I'm talking about Fig.9 %NS: Yes, the numbers of +kx and -kx are the same, but the difference is %NS: there are 12 elements each for kx = +3 and -3, and only 4 elements %NS: each for kx = +4 and -4 (thus total 32 vectors); similarly for ky, kz %AB: Does this explain the m=1 asymmetry in the first panel of Fig.9? %NS: Not sure I understand; red curves in both panels appear m=2 to me. %NS: maybe I will perform a better run and update during revision %AB: There is a dominant m=2, yes, but the first figure also has an m=1; %AB: the red bump at 135 is stronger than at 315, for example. %NS: it seems the following line plot was moved to removed, %NS: but i thought this is what Axel wanted to see. You may choose %NS: between this and contour plots, or keep both if needed %NS: I see that the appendix has one coloumn format which led to %NS: this extreme fig font/size. Adjusted but not sure what is optimal. %NS: the line plot with phi at theta=pi/4 %AB: what causes the m=1 asymmetry? %AB: Here too, use xticks=8 %NS: done; isn't it closer to m=2, possibly due to gaps in the ring? %AB: I'm not quite sure I understand the difference between left and right. %NS: will check this all again, maybe also with better run %MJK Nishant to run a model to see whether m=2 vanishes. %MJK Not urgent. Items related to 3.2.3 (Aspect ratio) %shows anomalously high values of the turbulent %MR: not true for the values in Table 1 %MJK Used to be much worse as daainit was 100 instead of 50, but I removed the last 50 time units. Now definitely less anomalous, but still not agreeing with the trend. Even then, when I look at the time series, all the fluctuations are still on the high side, and pdfs look funny - hence the idea to repeat this with smaller time step. %MJK Now re-run, but the halved time step does not have a big influence. %MJK Runs remain anomalous. %NS: agreed with most; modified last sentence and removed comments. %NS: a curious question (for later): qktfm gave +ve eta_12 when we %NS: wrote numerical paper SJ15; agreement with theory was perfect, %NS: except for eta_12; now all components seem to agree with theory; Why? %AB: will you move it to todo? %MJK Still we do not understand this.