Ceramic surfaces: reduction-cool reds (and browns?)

For a video of a recent lecture including some of the material below: Vimeo YouTube

** note that most images below can be clicked to link to full-resolution versions **

I’m using this post to collect and expand upon some tidbits I’ve posted on IG/FB, with additional and higher-resolution images. The main subject is reduction-cool reds, a specific type of surface color/texture that can form on the surfaces of iron-bearing clays when we fire them in reduction and delay re-oxidation by down-firing (stoking small amounts of fuel during the initial stages of cooling). As an appendix I also present some images of a shiny brown type of surface that can arise when reduction cooling fails to produce red-on-black results, and speculate as to what went “wrong” in the case of this example.

All of the images presented in this post pertain to “test tiles” (actually, square-ish cups that we make by extruding square tubes and adding a slab bottom, usually around two inches per side) that we fire in the train kiln at Utah State University. This is joint research with Dan Murphy, John Neely, and the MFA students at USU. The tiles shown/discussed in this post are all made from Dan’s “50-50” clay body, an even mix (by weight) of Lizella and Holmes-Fredericksburg fireclay, which is roughly 3% iron oxide (computed as Fe2O3) by weight. We find that we tend to obtain reduction-cool reds on these tiles most reliably when they are fired in the back of the kiln, on surface areas that are not covered by particulate fly ash (often on the back/bottom sides, and sometimes on the inside surfaces). We typically end up at cone 7-10 in this part of the kiln.

Before proceeding let me note that there seem to be many distinct kinds of red surfaces one can obtain with iron-bearing clays, especially in atmospheric kilns. The comments here are really focused on what we call reduction-cool reds, which are vibrant and even a bit “fuzzy” in appearance (definitely matte, perhaps eggshell but not glossy or even waxy) yet do not rub off under normal cleaning procedures, often appearing as patches on an otherwise black or dark-gray background. This is all on unglazed clay with 2ish-plus percent iron oxide by weight. In terms of process, we tend to think of the key to obtaining reduction-cool reds as keeping the kiln in reduction as it cools after shutdown, down to a temperature of something like 1650-1400F (opinions and practices vary widely) before we allow the atmosphere to reoxidize. Some air will always be leaking into the kiln, but as long as we keep stoking fuel at high temperatures the resulting combustion burns up much of the oxygen; hence the “reduction cool” strategy should effectively delay reoxidation of the kiln atmosphere until we stop stoking. Nearly identical reduction-cool black+red surfaces can apparently be obtained with a wide range of clay bodies in many different kilns, although people who strive for reduction-cool reds tend to work with train kilns and it’s not clear that you can get precisely these sorts of reds in a kiln with significant residual salt/soda. It does not seem that any particular kind of wood is required to obtain these reduction-cool reds but I don’t know of any systematic study of that. I also don’t know of any systematic study of water reduction in this context.

And as one last aside, let me note that a lot is known about red “flashing” on porcelain and porcelaneous stonewares and it appears to involve completely different crystal structures and formation processes than reduction-cool reds.

The images below (Figures 1 and 2) show typical examples of reduction-cool red surfaces on our 50-50 test tiles.

Figure 1
Figure 2

Figure 1 is a regular photograph of several tiles sitting on a table after unloading, showing patches of reduction-cool red on black backgrounds, mainly on the bottom surfaces of the tiles. Figure 2 (which is a detail of Figure 11 below) shows a  150x close-up of a reduction-cool red surface, taken using an optical microscope. The splotchiness of the red (with areas of more- and less-intense red) as well as the black “grains” are typical of this type of surface.

The next image (Figure 3) shows an optical microscope image taken at high (1000x) magnification, of a cross-section of a reduction-cool red surface.

Figure 3

Note that the redness is confined to a very thin outer skin, which sits atop a thin glassy black layer, which in turn sits on a “grey” zone of reduced-looking mature ceramic interior to the wall of the tile (the grey zone is actually a jumble of domains of varying composition, themselves ranging from clear to black in color). The small black grains that we noted in the surface view (Figure 2) appear to be regions of the glassy black layer that poke through (are not covered by) the outer red skin — we’ll return to this business a bit later, but first I want to show and explain something more about the cross-sectional view.

The next images show another cross-sectional view of a reduction-cool red surface (different part of the same tile as the 1000x image in Figure 3), first as a 150x optical micrograph with a yellow dotted box (Figure 4) and then as a scanning electron microscope (SEM) image (Figure 5) of roughly the area in the yellow dotted box. The SEM image (taken with a CBS detector in Z-contrast mode, for those in the know) has a 30 micron scale bar in the lower right corner; it shows a region containing the glassy-black layer and thin red skin at the outer surface of the tile. It is already visible in Figure 5, if you zoom in, that the red skin has a somewhat feathery/snowflakey structure — again, more on that later when we look at some surface views.

Figure 4

F124 outer crust 2kx CBS-Z
Figure 5

But staying with the cross-section for now, the following screencap (Figure 6) collects together a set of data that was obtained using an energy dispersive spectroscopy (EDS) tool on the SEM. On the left side there is an SEM-type image of the area being studied, with a green arrow indicating a particular “line cut” through that area. On the lower right (between the large graph and the row of little graphs) is a black-and-white stripe of image showing a closeup of the region around the line cut. The large graph above the image stripe contains a number of curves (color coded as indicated in the legend) that show the levels of various atomic species detected by the EDS tool along the line cut shown by the green arrow. The different curves are broken out (and individually scaled) in the row of little graphs at the bottom of the screencap. The main thing to notice here is that the iron (red) curve stays low along most of the line cut but jumps high when it hits the very outer layer of surface. It’s hard to be very precise, but this presumably corresponds to the thin red crust and perhaps part of the glassy black layer.

outer crust 10kx-line1-screencap.jpg
Figure 6

This sort of data indicates that there is an elevated concentration of iron at the very surface of the fired clay. The following screencap (Figure 7) shows data for another line cut in the same region.

outer crust 10kx-line2-screencap
Figure 7

Let us move on now to some surface views. If we think of the cross-sectional images in Figures 3 and 4 as “side views” then these next images are views “from the top.” They are just like the views shown in Figure 2 but we’ll go to higher magnifications. Figure 8 shows an SEM image (secondary electron with a TLD detector in immersion mode, for those in the know) of an area just around the edge of one of the small black glassy grains visible in Figure 2 (zoom in on Figure 2 if you’re not sure what I’m talking about). There is a 5 micron scale bar in the lower right corner.

Figure 8

The relatively smooth region towards the lower-right in Figure 8 corresponds to one of the glassy black grains, while the “rubbly” region in the rest of the image corresponds to reduction-cool red. To use a visual analogy, the glassy black grain is like a frozen lake while the reduction-cool red is like a rocky shore (just speaking of “texture” in the SEM image). Using the EDS tool on the electron microscope we can estimate that the composition of the glassy black grain is mainly silica with perhaps 4 wt% iron, while the reduction-cool red region is something like 50 wt% iron. Recall that the clay we started with before firing is about 3 wt% iron oxide, and that the cross-sectional studies above show that after firing the extreme outer skin of the ceramic has much higher iron content than the interior. It seems reasonable to hypothesize that something about the reduction-cool process causes a migration of iron from the interior of the tile wall out to its skin — iron gets pulled out from the depths of the clay to its surface. As an alternative, one might guess that extra iron gets deposited on the surface from gas-phase iron from the kiln atmosphere, but up to now I have found much more support in the materials science literature (see below) for the pulling-out mechanism. Also, consider the SEM image below (Figure 9), which zooms in with even higher magnification on a “frozen lake” region of the kind we discussed above in Figure 8:

Figure 9

If you look closely at the little clusters of “rocks” in Figure 9, it seems reasonable to interpret that these are aggregates of crystals that appear to be busting out through the otherwise glassy surface — one gets the impression that there are more crystals lying below the surface and that these are the tip of an iceberg, so to speak. If we assume that these are the same types of crystals that cover the rocky shores (the reduction-cool red regions), then these would be iron oxide. This is not a proof by any means but I think that images like this (which are easy to find) are good evidence to favor a pulling-out mechanism as opposed to deposition from gas-phase iron.

The iron oxide crystals seen in Figures 8 and 9 are extremely small — looking at the scale bars, most of the iron oxide crystals look to have dimensions on the order of tens of nanometers. They are also piled high, corresponding to the feathery/snowflakey appearance we noted in the cross-sectional view (Figure 5). It is generally known that red iron oxide (hematite) powders appear brighter and more vibrant when the grain size is very small. The very fine structure of the iron oxide cyrstal regions we observe by electron microsopy thus relates plausibly to the vibrant reduction-cool red color we see by eye.

To finish up the main part of this post, I would like to provide references to some scientific papers that support the idea of iron being pulled out to the surface of the clay by the reduction-cool process. The following paper by Smith and Cooper is part of a series of works by Reid Cooper’s research group that have carefully studied the mechanisms by which reduced iron (Fe2+) dissolved in an aluminosilicate melt can be reoxidized when the melt is placed in an oxidizing atmosphere:

Reference 1: “Dynamic oxidation of a Fe2+-bearing calcium-magnesium-aluminosilicate glass: the effect of molecular structure on chemical diffusion and reaction morphology,” Donald R. Smith and Reid F. Cooper, Journal of Non-Crystalline Solids 278, 145-163 (2000).

It is key to this work that the iron in the melt is initially reduced, and then the atmosphere is switched to an oxiding one at some fixed temperature. Speaking a bit loosely, the results of these (and related) studies suggest that if reoxidation occurs at a sufficiently high temperature that the melt is still fairly liquid, oxygen from the atmosphere diffuses into the melt and oxidizes iron atoms wherever they are buried. If on the other hand the reoxidation occurs at a lower temperature (around the “glass transition” temperature, where the melt has solidified enough to become tacky), oxygen cannot effectively penetrate into the melt and instead iron is pulled up to the surface to be oxidized directly by atmospheric oxygen. I should note that the chemical composition of the silicate melts used in these and related studies differ somewhat from those of the clays we use in ceramics, however, similar sorts of “pulling out” mechanisms have been observed by various authors using a pretty wide range of materials. In particular I would point to the following reference by Burkhard and Müller-Sigmund, regarding surfaces that I believe are a lot like reduction-cool reds, which form on basaltic lava under certain cooling conditions:

Reference 2: “Surface alteration of basalt due to cation-migration,” Dorothee J. M. Burkhard and Hiltrud Müller-Sigmund, Bulletin of Volcanology 69, 319-328 (2007).

If we assume that the findings of this research can be directly applied to reduction-cool ceramics, we may draw the following conclusions. First, the iron in our clay (at least, the iron near the surface of our clay) needs to be reduced at high temperatures. Second, after we close up the kiln and start cooling, we don’t want to allow that iron to re-oxide while the clay surfaces are still too hot, while they are still molten and susceptible to inward diffusion of oxygen — hence the need to stoke fuel during the early phases of cooling while the kiln temperature is still high. Third, at some point while the clay surfaces are passing through a “tacky” temperature zone, we need to revert to an oxidizing atmosphere in order to pull iron out and oxidize it to make reduction-cool reds. Now, exactly what these temperature ranges are will depend on the details of the clay composition and perhaps the length of firing (because there are fluxes in the kiln atmosphere during firing, coming from the wood we burn, and these presumably build up gradually in the surfaces of our pots and lower their glass transition temperature). But experience tells us that it isn’t super critical to be very exact about when we switch from reduction cooling to reoxidation — that switch just needs to happen at not too high and not too low a temperature.

It seems reasonable to infer from the scientific work that we might produce the most vibrant reduction-cool reds if we reduce the iron in our clays very strongly at high temperatures, and then re-oxidize them with a very oxidizing atmosphere towards the high end of the range in which the pulling-out mechanism can occur. We have not yet tested this systematically, but the idea might be to try down-firing only to something like 1700F (just a guess) and then opening up the kiln to see if we can fire in oxidation to try to hang at that temperature for some time (a few hours?). Kind of like a reverse strike firing.

There is a potential secondary reason that we need to keep the kiln in reduction during the early (highest temperature) phase of cooling, which has to do with iron sequestration in crystals. Some aluminosilicate crystals can nucleate from the melt at temperatures above our target zone for reoxidizing reduction-cool reds; the exact types and amounts of crystal formed will depend upon details of the clay composition, but some types (e.g., mullite and plagioclase) will have a tendency to incorporate iron more strongly if it is oxidized than if it is reduced (roughly speaking, because oxidized Fe3+ can more easily substitute for aluminum Al3+ in the crystal structures, while reduced Fe2+ is too big because it carries an extra electron). This could potentially “sequester” iron out of the melt early in the cooling (at high temperatures), locking it up in crystals so that it isn’t available later on (at lower temperature) to undergo the pulling-out reoxidation mechanism. I’m not sure if this is really relevant to reduction cooling with dark clays, but scientific studies do suggest that iron capture in mullite crystals is responsible for the fact that many porcelains are whiter when fired in oxidation than when fired in reduction.

Appendix: Shiny browns in “failed” reduction cool?

In this last part of the post I include some images of one particular test tile from a reduction-cooling firing, which failed to develop the characteristic red-on-black surfaces and instead has large areas of shiny brown. Before really investing too much in the interpretations offered below we should really do more studies, but I think we do learn something interesting already from this one particular example.

The following image shows a pair of (already cut-up) test tiles, one red-on-black and another brown. They are both the same 50-50 clay body and were both subjected to reduction cooling.

F118 F124 together.jpg
Figure 10

We note right away that the interior of the clay is yellow/ochre for the tile with shiny brown surface, versus gray for the tile with reduction-cool red. This suggests that the shiny brown tile experienced a much more oxidized firing than the red-on-black one. What I find quite interesting now is to look at the surfaces at higher magnification:

F124 composite.jpg
Figure 11
F118 composite.jpg
Figure 12

As we noted above, reduction-cool red surfaces generally have the structure of a red skin on a thin glassy black layer, on top of the bulk clay. If you look at the closeup of the shiny brown surface (Figure 12), it appears to actually have some sparse splotchy red on top of a thin and patchy glassy black layer, with the underlying yellow/ochre showing through here-and-there. One could intepret this in the following way. Suppose the tile was fired (for whatever reason) under oxidizing conditions, yielding a yellow/ochre/buff color for the bulk clay. Then the reduction it experienced during the closing-up of the kiln and the early (higher temperature) part of the reduction cooling downfire acted to develop a thin, partial skin of moderately reduced silicate melt. When the kiln was allowed to reoxidize, a small amount of iron was pulled out to the surface of the reduced patches and turned red there. In the magnified view provided by an optical microscope the very sparse red-on-black takes on a somewhat purple hue, but when we zoom out and look as if with our bare eyes this combines with the yellow/ochre of the underlying clay to look brown. The following images are full-resolution views of the shiny brown surface:

F118 cut tile exterior side 101x EDF from stitched 3D fine glare.jpg
Figure 13
Figure 14

Of course we shouldn’t over-generalize from the details of this one example, but I think it is nice to see one concrete way in which already-known behaviors of iron can produce shiny brown surfaces in a reduction cool-type wood firing.

Part of this work was performed at the Stanford Nano Shared Facilities (SNSF), supported by the National Science Foundation under award ECCS-1542152.

6 thoughts on “Ceramic surfaces: reduction-cool reds (and browns?)

  1. Dr. Mabuchi,
    Is the mechanism at work the the formation of corundum (α-Al2O3), and a lattice of iron (in various states; most likely hematite, α-Fe2O3) forming around it?

    As you report, the red colour structures are “drier” and more matt-surface with rough crystals. If these areas have more alumina than silica by ratio, is corundum (crystalline Al2O3) or “iron-red” spinel what is developing in the cooling with oxygen? I’ve seen examples of hematite inclusions in quartz crystals and they are very similar in colour.

    Because this colour can develop regardless of cooling speed, or atmosphere, I thought about a series of articles written in 2004 by Yoshihiro Kusano, Jun Takada, and others. They were investigating the development of the iron red blush on pots fired with the hidasuki technique. Obviously, potassium is a factor in hidasuki, but the findings present a helpful starting point.
    If I understand correctly what I have read, the papers I have been able to access on analysis of hidasuki color all note the presence of corundum, as a precipitate. Further, they were able to describe hematite formation around the corundum crystals. The surfaces of these fired results vary in stability because the toughness of corundum is sensitive to crystallographic orientation (which may or may not be random). If the crystals develop in a certain way, they can be broken down in contact with other hard surfaces.

    I would speculate that, in heavy reduction at the firing’s peak to maximize the amount of FeO in the clay body, reduction cooling with spotty or late minimal re-oxiding of the available surface FeO preserves the “dark background,” of the clay and a different matrix against which the red color in the corundum structures forms: so we have a “negative hidasuki,” flash, like the Sabattier effect in photography.

    The paper on epitaxial growth of epsilon-phase iron (ε-Fe2O3; thought to be responsible for the visual characteristics of Jian dynasty iron-bearing glazes), discusses a situation where “partial melting” occurs on the surface of pots fired with the “Hidasuki mechanism.” This creates a condition where mullite is replaced by corundum and hematite. As I understood it, the paper describes a surface where the corundum acts like a base-layer of glue, and the hematite covers it. The result is a singular crystalline structure that is a “sandwich” of hematite and corundum: α-Fe2O3/α-Al2O3/α-Fe2O3.

    I think, given the findings of the papers, certain high-iron content clay bodies favour the production of surface corundum crystals with hematite formations because of their high alumina/low silica ratio. In other words, if airborne fluxes are not always a factor here, there may be so much FeO (flux) present with the right amount of Al2O3 on the surface of the sintering ware, that corundum with hematite and/or ε-Fe2O3 make various “sandwich crystal” formations on the cooling pots. The occasional rubbing off would then be a result clay bodies that have an abundance of iron on the surface of the cooling pot, or the crystallographic orientation of the corundum structures is such that it results in a weaker surface.

    The samples of Bizen clay used in the tests are described as having “2-3% wt of Fe2O3.” If corundum is the culprit, could we begin piecing the theory together by conducting an experiment to test whether the silica, or alumina in high-iron content clay bodies develops this particular red-coloured surface?

    For example: select a “simple,” off-white clay body formula whose chemical analysis is known, with a relatively close ratio of alumina and silica. Make one batch of test tiles of the body itself without additions (the control). Make one family of tests where RedArt (to simulate the addition of iron and silica) is added to the body in 1% increments up to 8%. Make another family of tests where Laterite (to simulate the addition of alumina and iron) is added to the body in 1% increments, up to 8%. Granted, there are other factors in the materials, but this should give us a rough start. If the iron content alone isn’t the factor, then similar tests would need to be conducted with potassium spars as another variable factor -since that is the flux responsible for corundum formation according to the research papers. Needless to say, the kiln, fuel, firing rate, timing and duration of body reduction, cooling atmosphere and cycle, etc., would have to be the same.

    Thank you for your time and consideration.
    John C. Boyd

    1. Hi John,

      Thanks very much for your comment. I’m quite familiar with the work by Kusano and colleagues – I have discussed it with them (they visited Stanford in 2017 and I visited them in Japan in 2018) and have imaged the type of corundum-hematite structures they discovered, in flashed porcelain-type clays using optical, electron and ion microscopies. These reduction-cool reds are entirely different! There really isn’t any evidence of surface corundum on the reduction-cool reds, it’s basically all iron oxides with some admixture of magnesium. I think the mechanism described by Reid Cooper and colleagues really gives a quite plausible explanation for the effect, with lots of anecdotal support from the SEM imaging and EDS studies I’ve done. With this mechanism as the explanation of the effect, the most important practical factor for achieving strong reduction-cool reds on pots should be delaying re-oxidation of the kiln atmosphere during cooling until the temperature is below the glass transition temperature of the ceramic surfaces. It seems like people I know who fire this way manage to get pretty reproducible results when firing in known kilns with known materials. At this point I feel like the reduction-cool reds are pretty well understood, so I’m more focused on trying to understand what exactly the browns are that occur when your reduction-cool “goes bad” because you either re-oxidize too hot or fail to fully reduce your pots during firing.


Leave a Reply