Roadmap | The review process | Home

Discussion of dehydration melting


This discussion webpage was stimulated by the newly published paper Schmandt, B., S.D. Jacobsen, T.W. Becker, Zhenxian Liu & K.G. Dueker. Dehydration melting at the top of the lower mantle, Science, 344, 1265-1268, 2014.


June 20th, 2014, Don Anderson

Those are good points and that's a neat figure. Many of the arguments against melting also assume a passive sponge (Karato). Karato revised their original interpretation in light of new physical data and procedures..

Yes, the experiments are good, but the seismology comes across as an optional extra.

June 20th, 2014, Alexei Ivanov

Thank you, Don. Unfortunately I cannot assess the seismic data, but experimental part seems to me good.

As for the Bercovichi and Karato TZ water filter model, I do not think it is correct in the from it was originally suggested. The idea is stimulating, but it needs modification. Plumes are not required and moreover if a hot plume passes through the water filter, water does not fractionate into the hotter medium. This was a reviewer's comment on one of my papers. In the Bercovichi and Karato model, the TZ is a passive sponge. However, if TZ minerals are hydrated via fluxing from subducting slabs their volume increases so much that the formation of buoyant wet diapirs is inevitable.

Here is a Figure redrawn from Smyth J.R. Jacobsen S.D. Nominally anhydrous minerals and Earth's deep water cycle / Jacobsen S.D., Van der Lee S., eds. Earth's deep water cycle // Geophysical monograph, 2006, v. 168, p. 1-12. which shows preliminary results on the effect of water on the volume of ringwoodite, wadsleyite and olivine. Water has a more notable effect on volume than temperature.



June 20th, 2014, Don Anderson

Alexei - here is a short critique of the Schmandt et al paper, which is over-interpreted from a seismological point of view. The conclusions are the same as the assumptions. If the LVZ at 700 km is due to water leaking downward out of the transition zone then there are several obvious tests. Incidentally, this is the precise opposite of the water filter model of Bercovici and Karato, and it overlooks the fact the Karato "proved" that LVZs cannot be due to melting and that melts will drain away quickly..

A LVZ above or below the TZ does not prove that water came out of the TZ, even if TZ minerals can retain water, if somehow water can get there. A logical fallacy is that if the TZ can retain more water than other mantle minerals, they have to be water rich. This may or may not be true but there are many explanations for deep LVZs (CO2, dynamic effects [Ricard, Li, Weidner] that are associated with phase changes, small grain size etc.). The seismological tests include a deepening and sharpening of the 650, a LVZ above but not below flat slabs, more LVZs above inferred deep mantle upellings (plumes) than below flat slabs etc.These tests were not done or are negative.

The authors assumed whole-mantle convection from the start and invented a story to make this true. One needs neither water nor melting to make a LVZ below 650 but to prove it one way or another requires the above tests. Keshav and Presnall would argue for CO2, and Li, Weidner, Ricard and others would argue for dynamic melting. Note that a high seismic gradient below 650 km has long been a feature of global models and correlation of plate-reconstructed slabs and wavespeeds are best at 700-800 km depth. This is likely to be a density barrier, not a water table, quite apart from the seismology.

I am not impressed by models that are invented just to fit some new observations (like Just-So Stories, e.g. slabs entrain water out of the TZ upwards, and sinking slabs in the lower mantle entrain water downwards, because there are low velocity zones at the boundaries of the TZ). I like the idea that slabs displace material upwards that become broad ridge feeders but the Bercovici-Schmandt papers are very contrived. I also think that there is much water buried in the mantle but I would not quote this paper or its press releases as evidence.

June 20th, 2014, Alexei Ivanov

There is Russian saying that a drop can destroy the rock. Truly astounding in this paper is that Science editors discovered for themselves what has common knowledge, and common sense for others, for years.

In my view the paper is good. The authors conducted a new experiment and showed that under uppermost lower-mantle PT parameters, melting may occur under hydrous conditions. Similar experiments had beenconducted before with some controversial results. Using this experiment as evidence of melting at such a depth, they interpret seismic data in the framework ofhydrous melting model. Iit had already been done for the same region but for a shallower depth - above 410 km.

So the paper is an example of good scientific study and voila, Science editors finally discovered that Earth is round!

Any Science or Nature paper produces waves in the mass media that have typical shortcomings because journalists often (but not always!) do not understand the true meaning - they do not necessarily have the background for this. This paper was mainly highlighted in the newspapers as the discovery of an ocean of water in the mantle, that this ocean is primordial and not of cometary origin etc. Some Russian translations even claimed that the authors discovered ringwoodite. In other words, the true sense of the paper was misrepresented, but this is not the fault of the authors of the Science paper.

This why I am trying to write comments on essential publications (irrespective the journal where they were published) in our newspaper "Troitsky Variant – Nauka".

June 19th, 2014, Don Anderson

This was just published in the NY Times, based on the Schmandt paper. This illustrates why it is important to publish in Nature and Science.

June 19th, 2014, Don Anderson

They assume at the outset whole mantle convection and along with the water filter model conclude that water is entrained into the lower mantle, by deep sinking slabs, and is displaced out of the TZ by slabs from above. Plumes go right through the TZ without any effect. They assume narrow slab downwellings and diffuse upwellings, the precise opposite of the plume model. They ignore CO2 and dynamic melting (Li, Weidner etc.). In the water filter model it is assumed that broad upwelling currents dehydrate and melt as they pass through the 410-km discontinuity, leaving the water, the melt and the impurities behind. Diffuse downwellings dehydrate as they sink below 650 km. Unexplained is how water gets into the TZ or stays there.

The TZ water filter model & dehydration melting

Low-velocity anomalies (LVAs) are often simply attributed to excess temperature or water content, small grain size or decompression and dehydration melting, but the actual situation is much more complex and requires mechanisms for causing these phenomena. A horizontal LVZ can be due to the effects of CO2, ponding under a permeability barrier or a negative Clapyron slope boundary, shearing, metastable phases or the dynamic effects discussed in this paper. The presence of a LVA does not imply slab or water penetration into the lower mantle or whole mantle convection. LVAs can be formed at solid-solid phase boundaries and do not require the presence of melt.

The TZ water filter model (Bercovici, Karato) assumes that water is the main ‘impurity’ that lowers melting points and that the TZ is the major water reservoir in the mantle. In this model, the global mantle flow pattern is dominated by slab-related localized downwelling currents and diffuse upwelling flow. This is the precise opposite of the mantle plume model, which assumes narrow focused upwellings (plumes) and diffuse downflow (Bercovici and Karato). Melts and impurities accumulate above and below the TZ but not in the shallow LVZ. Alternatively, ancient ambient depleted mantle at the base of the TZ is forced up by the downward flux of subducting slabs and becomes the passive depleted upwellings that fuel mid-ocean ridges and near-ridge hotspots. Those that rise midplate interact with the surface boundary layer and pick up the impurities and chemical components that define midplate basalts (Anderson LLAMA model).

In addition to the low-velocity zones found near the top of the TZ, there are a few others that have been detected at 350-370 km and 600-700 km (Keshav; Bloom & Shen; Schmandt). In the water filter model it is assumed that broad upwelling currents dehydrate and melt as they pass through the 410-km discontinuity, leaving the water, the melt and the imputities behind. Diffuse downwellings dehydrate as they sink below 650 km. Whole mantle convection is assumed. TZ properties, however, are consistent with cold mantle accumulating above and depressing the 650-km discontinuity, displacing older warmer mantle upwards, elevating the 410; they are not consistent with whole mantle convection with through-going slabs and hot plumes. Alternative explanations of deep LVAs are CO2, segregated basalt, metastability, underplating and interaction of the seismic waves with phase changes.

June 17th, 2014, Michele Lustrino

I think it is an important contribution, because it recognises a relationship between wave speed anomalies and temperature. In several papers (e.g., Faccenna & Becker, 2010, Nature) it is assumed that red colours = hot mantle volumes and vice versa.

Now I read that the red can be associated with melt, i.e., something unrelated to the thermal state of the mantle. In other words, previous conjectures lose their feet and fall down.

Another story is if subducted slabs can pierce the TZ to enter the LM and again another story is the fate of these dense hydrous melts, where density could be even negative compared with the perovskitic matrix. (Among the others, Mg-Si perovskite now must be called bridgmanite, no longer "perovskite".) If this is true, these melts could downwell down to the CMB.

June 16th, 2014, Alexei Ivanov

It is generally considered that there are two "cross-over" regions in the mantle, where melt is denser than ambient mantle. One is more or less well defind and is located just above 410 km. This region is very thin (~ 10 km). Melt is trapped there unless it is H2O-rich (~2% or above that value). CO2-rich melt (5 wt% in experiments) is denser than ambient manlte and trapped there, but carbonatitic melt is probably not so dense and is probably not trapped.

Another cross-over region is expected at the base of the lower mantle, but due to poor knowledge about this region there are pro- and con-interpretations. Most researchers who deal with this region, think that the melt (either mafic or ultramafic) should be trapped at the base of the lower mantle. The thickness of this lower-mantle cross-over region is hundreds of kilometers. The uppermost lower mantle is denser than mafic melt, thus melt should not stagnate there.

last updated 21st June, 2014