The study by Desiderio et al. presents a systematic exploration of the effect of viscosity and density differences on the preservation of primordial mantle material as BEAMS and the generation of LLSVP-like accumulations of recycled oceanic crust (ROC). Overall, I think that this is a nice contribution and should be suitable for publication after minor revisions. In my view the most significant issue is the omission of internal heating, which at least needs to be physically justified.

Line 28: I do not think that it’s reasonable any longer to claim that the nature of the LLSVPs is strongly debated. Masters et al. (2000) described the anti-correlation of bulk and shear wavespeeds in the lowermost mantle and there is no way to reconcile this with a purely thermal origin. More recent work by Lau et al. (2017) using tidal constraints and by Moulik and Ekstrom (2016) using normal mode and body wave tomography make a compelling case for high intrinsic density within (at least part of) the LLSVPs as well. It seems that this debate was vigorous in the 2000s but has been long since settled.

Line 48: Seismic-tomography images is awkward. Suggest “tomographic images” instead.

Line 120: The choice to run models without internal heating needs to be justified physically as it seems inconsistent with what we know about the energetics of mantle convection, and the paper aims to understand the preservation of heterogeneity under Earth-like mantle evolutionary scenarios. Under ancient, hotter mantle conditions, mantle heat production was larger than the present day value and the omission of mantle heat production seems even more unreasonable. In general, heat production will destabilize convectively isolated ‘blobs’ in the mantle because they will heat up and become more buoyant and less viscous (e.g. Becker et al., 1999). It’s not clear to me, given all of the complexities of composition-dependent buoyancy and  viscosity considered here, how the choice to run models that are not earth-like in terms of the energetics of convection can be justified. I understand that HPEs are incompatible in Bridgmanite, so they may be concentrated in the later products of magma ocean crystallization. But they are also even more incompatible in ferropericlase and moderately incompatible in CaPv, and they have to go somewhere to conserve mass during magma ocean crystallization! The recycled oceanic crust should also be enriched in HPEs and the harzburgitic residue depleted. The authors could comment on whether this could affect the dynamics of ROC accumulation in the lowermost mantle.

Line 130: The treatment of basalt/eclogite buoyancy in the deep mantle is simplified and the dynamics may be quite different if the authors used thermodynamic lookup tables for pyrolytic and basaltic compositions from Stixrude and Lithgow Bertelloni (2024)

Section 2.4: The temperature dependence of effective Clapeyron slopes is not considered. I wonder if the authors might at least comment on this? The equations of state developed by Stixrude and Lithgow-Bertelloni (2011, 2024) predict that the effective Clapeyron slope will change over time in a cooling mantle, which will affect the tendency towards layering and certainly the viability of BEAMS preservation.

Line 170-175: These dynamics might be quite different if mantle heat production was included.

Line 197: The mathematical symbol used here is showing up as both a greater than and less than sign, which I find confusing. Why not just define a threshold of f=0.3? Presumably the results are not very sensitive to the choice of this value?

Figure 3: I found this figure extremely confusing. There’s too much going on for me to understand quickly what’s being shown and it would almost be easier to look at a data table – the opposite of what a figure is supposed to accomplish. I would find it clearer to have four separate panels with contour plots, and perhaps use the same color scale for all four panels to highlight trends. It’s too hard for me to see the trends in the four different quantities on the same plot with four different symbols and four different color maps. Also, why abbreviate everything? The figure would be much more interpretable if the authors show the regime boundaries and give each region a nice simple label like ‘Mixed Primordial’ or ‘deep ROC layer’. You could also annotate the axes – buoyancy ratio and viscosity ratio.

212: I think that radially-averaged depth profiles is redundant and maybe incorrect. I would call this the azimuthal average or horizontal average.

213: regime -> regimes

Section 4.3: It would be nice for the authors to also address the preservation of primordial noble gas signatures in the OIB source. Are the authors arguing that the BEAMS are a likely source of primordial noble gas (He, Ne, Xe) isotopes? If so, is this at odds with the idea that BEAMS represent the early crystallization products since the noble gases are highly incompatible during solidification?

443: There is also a new global attenuation tomography paper (published while the present manuscript was in review) that shows an interesting feature around 1000 km depth: Sun et al. (2025) A high attenuation layer around 1000 km depth. Earth and Planetary Science Letters  669, 119577.

See detailed review in attached file