> ----------------------------------------------------------------------
> Report of Referee A -- LE17401/Bhat
> ----------------------------------------------------------------------
> The manuscript reports Direct Numerical Simulations of the
> compressible magneto-hydrodynamic equations. According to the authors,
> the main finding is that the small-scale dynamo phase is followed by a
> large-scale dynamo phase. The authors call the large-scale growth
> "quasi-kinematic", although there is little evidence of that (and no
> proper definition of this term). The paper is poorly written and lacks
> precise and clear evidence of the results that are advertised. Here
> are more detailed comments:

> - Why use compressible equations to describe effects that seem
> completely independent of compressibility?

In the incompressible case, the pressure depends nonlocally on all points
in the domain. This requires global communication between two time steps,
which is inefficient and avoided by our approach. There is no significant
difference in the results between the two cases.

> - In the bottom panel of figures 1 and 2, it is incorrect to say that
> three stages can be clearly identified. In figure 2, the second stage
> does not correspond to a plateau, and starting from t=50 the curve
> looks like a decaying exponential.

%AB: an exponentially changing growth rate does not make much sense either.
%AB: maybe we should compare linear and log plots, as I did in the appendix.

> - The authors associate their "quasi-kinematic" dynamo to an
> exponential growth of the large-scale field, together with a linear
> increase of the magnetic intensity at the forcing scale (M4). However,
> this contradicts the theory of large-scale dynamos: in standard alpha
> effect theory, the small-scale field is proportional to the
> large-scale one and both grow exponentially.

%AB: the standard theory is not right.

> - When there is shear, the magnitude of the alpha effect coefficient
> typically depends on the shear intensity.

AB: yes.

> - There is some inconsistency between the theoretical estimates and
> the numerical values of the growth rate, but instead of discussing
> them the inconsistencies are hidden in some "efficiency factor".

AB: one expects the conventional theory to be imprecise.
AB: OK, we should discuss this.

> To conclude, the results advertised in the title and abstract are not
> present in the numerical data (no well defined plateaus) and the
> interpretation is both qualitatively and quantitatively inconsistent
> with the well-established large-scale dynamo theory. For these reasons
> I cannot recommend this paper for publication in PRL.

> ----------------------------------------------------------------------
> Report of Referee B -- LE17401/Bhat
> ----------------------------------------------------------------------
> The authors report the exponential growth of large-scale magnetic
> fields after the saturation of small-scale magnetic fields, which was
> referred to as "a novel quasi-kinematic large-scale dynamo". However,
> as the authors also noted in their introduction, the saturation of
> magnetic fields depends on scales (occurring first on small scales),
> so I do not see how their finding is particularly novel.

> Also, the manuscript presents different simulations depending on the
> presence of a prescribed shear, helical/non-helical forcing, etc.
> However, I do not find a systematic analysis on the dependence of
> their results on shear parameter, magnetic Reynolds number, etc. For
> instance, Figures 1 and 2 compare the cases with shear and without it,
> but apart from presenting growth rates for different modes against
> time, I am unsure what to learn from these two figures in terms of the
> role of shear in dynamo. Another example is Figure 5 which shows three
> points for three different shear values.

The goal is not to learn something about shear, but to show that the
novel feature in the late saturation behavior is independent of shear.

> The discussion is based on the quasi-linear analysis, which obviously
> requires the values of turbulent transport coefficients (\eta_T,
> \alpha). I am unsure about the accuracy of their estimated values for
> these coefficients to validate the agreement of the quasi-linear
> analysis and computational results.

AB: Maybe we should shorten the presentation of mean-field theory.
AB: People think we verify anyhing here.

> In Figures 3 and 4, the data should be presented for the same range.

One is a semi-log plot, so one cannot compare Figs.3 and 4.`

> In summary, I do not find the results not sufficiently significant,
> and thus regret that I cannot recommend this manuscript for the
> publication in Physical Review Letters.

AB: we need to rework to clarify the signficance.