In Chinese Glazes, we learn from Nigel Woods that the cobalt used for underglaze blue & white underglazes and blue glazes came in a range of chemical compositions and grades of purity. Thus, there are many shades of blue due to the quality of cobalt-containing stone as well the overlying glaze.
In the same book, Nigel presents a lovely Chinese blue stoneware glaze which, in addition to cobalt, contains iron and manganese “impurities”.
In fact I’m personally not fond at all of glazes and underglazes containing only cobalt as a coloring oxide. Pure cobalt often comes out as a garishly blue color. In the triaxial blend below, I take a nice clear glaze (Sue’s Clear) with added 1% Cobalt Carbonate. Then I blend with 1.5% Red Iron Oxide (bottom left) and 1.5% Manganese Dioxide (bottom right). The resulting colors on the bottom row are much more pleasing to my eye.
The full image can be viewed here: http://www.derekau.net/wp-content/uploads/2014/12/BLUE_TRIAX_ALL.jpg
Having not fired cone 6 since college, I started by first testing a number of clear cone 6 glazes on https://glazy.org
Out of these tests, there were two glazes that I preferred. The first, Sue McLeod’s Clear, is a soft clear with minimal clouding and has B2O3 at 0.18 which, according to Matt’s information, is ideal for cone 6 glazes.
The second glaze is a shop glaze available at the Wellsville Creative Arts Center called WCAC Celadon Clear. With B2O3 at 0.45, it is really high in boron and possibly less durable than the lower-boron clears I tested. However, WCAC Celadon Clear is by far the clearest glaze I’ve tested, almost like a layer of pure glass or honey. Even on dark stoneware it’s really clear with almost no clouding.
Being new to cone 6, I was curious as to the effect of boron levels on clear glazes. So, I created two biaxials, both with R2O fixed at 0.2. In the first Sue’s Clear inspired biaxial, B2O3 is set at 0.18. In the second biaxial inspired by Celadon Clear, B2O3 is doubled to 0.36.
Each biaxial resulted in a nice clear, with the higher Boron clear being almost completely transparent and glossy, while the Boron 0.18 clear is translucent and soft.
Standard Cone 6 Porcelain Body #551
Same chart but with words describing each test glaze:
Best Clear at B2O3 0.18
The best clear resulting from the B2O3 0.18 biaxial is here: C6 R2O 0.2 B2O3 0.18 Best Clear
B2O3 0.36 Biaxial
In order to test the effect of higher B2O3 levels, I doubled the amount of Boron in the initial biaxial from 0.18 to 0.36 while maintaining the same R2O:RO ratio. I also made the boundaries of the tests a little higher (see map comparison). I was surprised to see that the only clear glazes in the 0.36 Boron test appear much farther down (lower in Si & Al) in the chart. But the “clear” region is still in the same Si:Al Stull region.
After testing WCAC Celadon Clear and seeing the results of my B2O3 0.36 biaxial, it seems there is definitely a region of very glossy, very clear glazes at higher boron levels.
Coincidentally, I tested an old glaze recipe posted to the Clayart mailing list by Laura Speirs in 1996: https://glazy.org/recipes/21102 As with the WCAC Celadon Clear, the Speirs recipe is also very high in Boron (0.51), and it also fires very clear and glossy:
VC Easy Glossy
One afternoon I began discussing the WCAC Celadon Clear with a WCAC member, Nancy Alt. I was very surprised to discover the interesting history of this glaze.
In 2009 Nancy Alt had visited Val Cushing’s home and purchased a vase with a lovely blue-green celadon glaze. Nancy asked Val if he could share the glaze recipe, and he not only shared it but converted it from cone 9 to cone 6 (the temperature Nancy was firing). Val’s email is copied below. It shows the extremely generous nature of this amazing potter and teacher:
From: Val Cushing
Subject: Re: celedon glaze
Date: May 12, 2009 at 12:46:57 PM EDT
To: Nancy Alt
This glaze is one I made for C/9 oxidation electric firing, so that it would appear to be a blue green celadon. I have revised it for you to be the same color and texture only for C/6 ox. electric . I will give you two to try , first VC Pale Emerald, C/6 , glossy , blue/green , celadon looking. as follows………. Kona F/4 feldspar 24, Ferro Frit 3134 24, Dolomite 4, whiting 14, barium carbonate 2, zinc oxide 2, flint 24, and EPK 6. ADD TO THAT , 1/2 % COPPER CARBONATE for blue green. VC/easy glossy, C/6 ox. , electric , celadon looking , green. Cornwall Stone 46, Gerstley Borate 20, Ferro Frit 3124 26, Ball Clay 8. — add 2% copper carb. and 1/2 % red iron oxide for celadon looking green color. Test these two Nancy and if the color is not exactly what you expected let me know and we can make a revision. We may have different “tastes” about color , but we can get what you want…My pale emerald should be quite a bit like the glaze on the jar of mine you now have. and THANK YOU . Val
So it turns out that the glaze I liked so much, WCAC Celadon Clear, was actually a Val Cushing recipe called “Easy Glossy”. I checked Cushing’s Handbook for the recipe and didn’t find it. Nor could I find similar recipes in the Glazy database. So it’s quite possible this is a newly discovered Val Cushing glaze recipe.
However, the WCAC Celadon Clear had been modified from the original “Easy Glossy”, most notably subbing Gerstley Borate for Gillespie Borate. I wanted to see not only the original recipe but also the color variations that Cushing was working with. So I created a triaxial blend.
Below is the triaxial blend using Copper Carbonate and Red Iron Oxide.
Some of Robert Tichane’s glaze tests and reproductions of Chinese Glazes donated to the Freer and Sackler Galleries: https://
archive.asia.si.edu/ collections/edan/ default.cfm?searchTerm=tich ane&btnG.x=0&btnG.y=0&btnG =Search
It’s important to wear a NIOSH certified mask whenever using dry glaze materials.
I guess mixing up glazes isn’t that big of a deal, but I’m sharing my technique just in case there are some absolute beginners out there.
I find it easier to use a digital scale, see my article here.
Glazes “don’t travel well”, in other words materials, application, and firings vary from studio to studio. Even for well-known glazes, it’s important to first make a small tests. For these tests, I use 50g or 100g of material and apply the test glaze to a number of different clay bodies.
I use the Glazy Batch Calculator on my phone which will show you the subtotals for arbitrary amounts of total glaze materials.
Once I’m happy with a test, I mix up a larger batch of 1-2Kg. 1Kg is enough material to glaze small cups, 2Kg is a good amount for small bowls. These larger tests should reveal any problems with glaze suspension (is bentonite required?), application (cracking, peeling, etc.), and fired glaze defects. Once you have some nice results with 1Kg, you can finally move on to a big bucket of 5-10Kg.
Mixing up a test
Flat test tiles require the least amount of glaze for application. Here’s my article about how I make test tiles.
Sieve Mesh Size
For “natural” glazes containing large-grained materials or ashes, or in cases where homogeneity is not a concern, it’s fine to use a larger screen of 60-80 mesh. But in all other cases I use 120 mesh or smaller. Small mesh size is very important for glazes that contain small amounts of very important materials such as coloring oxides (e.g. cobalt and iron). But it’s also important to ensure that materials are adequately broken up and mixed (such as clays).
Below you can see two tests of the same batch of glaze fired in the same kiln. The glaze on the left was applied after passing the materials three times through an 80 mesh screen. The glaze on the right is the result of passing that same glaze once more through a 120 mesh screen.
Poorly dispersed colorants like iron are easy to see in fired glazes. But keep in mind that other “invisible” glaze ingredients like clays, feldspar, etc. also need to be well-dispersed and mixed in order to ensure the glaze melts properly. If you use a 60-mesh screen for tests and then a 120-mesh screen for large glaze batches, there will be differences between the fired results.
After years of using simple balance scales to measure out glazes, I finally decided to invest in a better setup. I couldn’t find any triple-beam scales for sale in Jingdezhen, so instead I purchased a cheap 200-gram digital scale from a local shop. I was delighted at how much simpler and faster it was to mix up tests with the digital scale. It was only a few months later when I compared the digital scale to my old balance scales and discovered that the digital scale was consistently inaccurate, even just after calibration.
After having wasted 600RMB, I decided to just buy the best reasonably priced scales I could find. The only imported brand in my price range and available in China was the Ohaus Scout Pro line. I purchased two- one for tests and measuring colorants (model SP202, up to 200 grams with 0.01 gram readability) and one for mixing up bigger batches of glaze (model SP4001, up to 4000 grams with 0.1 gram readability).
The SP202 is very accurate, great for when you are making very small test batches. The scale can also be used to measure colorants for big batches of glaze.
I use the SP4001 to directly measure out 1-3kg batches of glaze, or for measuring out each ingredient in larger glaze batches.
After a couple years, the Ohaus scales are still performing very well, especially considering that they are stored on the glazing patio and subjected to the weather. The scales cost me much more than I wanted to spend, but they are well worth the money.
- If you’re looking to purchase scales for small glaze batches but don’t have a lot of money to spend, go for a triple-beam scale. A good triple-beam will be much more trustworthy than a cheap digital scale.
- If you only have enough money to buy one digital scale, get a 200-gram scale, preferably with .01 readability. This will allow you to make accurate test glazes, as well as accurate colorant additions to larger batches of glazes.
- If you do buy a digital scale, don’t forget you will need to calibrate it from time to time. (I do so each glaze-making session.) You will need accurate calibration weights in order to so, adding to the final cost.
A lot of potters in China still seem to mix glazes the old-school way- one cup of this, two cups of that. And strangely enough this technique seems to work pretty well for complex traditional materials. Being a foreigner I tend to make things overly complicated. I am also terrified of mixing up a great glaze but not remembering the exact composition or, even worse, not knowing how to adjust it if it doesn’t come out right. So I rely on a lot of line blend and triaxial glaze testing.
Line blends are a useful tool for comparing two different recipes. Commonly, the two recipes are different glazes. For instance, you can blend two different celadon glazes in different proportions to create a new celadon recipe. Line blends are also commonly used for testing additions of coloring oxides, for instance the effect of incrementally adding iron oxide to a clear glaze.
The range of the line blend is arbitrary- you can start each variable from 0% and go up to 100% or you could choose any range in-between.
For example, when adding red iron oxide to a reduction-fired clear glaze like Leach 4321, a line blend from 0% to 10% in 1% increments is sufficient to see the gradual transition from clear to blue celadon (1%), light green celadon (2%), dark green celadon (3-4%), brown, and tenmoku (7-10%).
To illustrate a line blend, here is a sample using two RGB colors blending by opacity. If you already know the result of the outlying 100% blends, you could remove them from the test. However, each firing is different and I usually leave in the 100% blends regardless.
You could mix each test in a line blend individually, but a much less time-consuming method is to only make the left and right-most 100% solutions. Using a syringe, you can easily create each mixed blend.
Ian Currie popularized volumetric blending. (See his article here.) First, mix the same weight of each glaze you will be blending. Second, add some water to the glazes and sieve thoroughly. Third, add water to each glaze so that their volumes are equal. Now the two glazes are ready for blending.
Below is an illustration of the blends in a 20ml syringe. For my test tiles, 20ml is the minimum volume I need in order to completely cover a test tile (including a second dip). See here for more information on how I make test tiles.
As you can see, for 10% increments using a 20ml syringe we need at least 20+18+16+14+12+10+8+6+4+2 = 110ml of each 100% recipe. Preparing 100 grams of glaze material for each glaze should be sufficient.
Each time before you take glaze into the syringe, be sure to re-mix the glaze and confirm it has not settled.
Although I use 20ml test batches, it’s easier to use a 30ml syringe so there is extra room for drawing in air, making it easier to mix the glazes.
Here is a line blend of Dolomite added in 2% increments to a Chinese porcelain stone. Normally, I would test from 4-14% Dolomite, as I know from experience that this is the most useful range. Below are the tests from 4-12% Dolomite in which the porcelain stone turns from a satin-matte to light celadon glaze.
However, it’s often good to test beyond the limits. We might expect that even more Dolomite added to this glaze stone will result in a runnier, more transparent celadon/clear glaze. But ceramics is much more complicated than simply mixing two colors together.
Here is the same test carried further, from 16% to 20% Dolomite. We’ve gone from a creamy satin-matte to celadon and now to Dolomite matte with crystals.
This test also shows the importance of making small increments when doing line blends.
Ceramics recipes are a complex interaction of multiple variables. Triaxial blends involve three variables, making them more useful for exploring ceramic glazes and bodies.
Below is a four-row opacity triaxial blend of three colors (red, green, and blue) in RGB with a range of 0%-100% for each color. The outer edges of the blend are simply line blends of two colors, while the middle area of the blend contains mixes of all three variables.
The size, variables, and ranges of a triaxial are completely up to you. A four-row triaxial with ranges starting at 0% is of limited use in testing mixes of all three variables. You could adjust the four-row triaxial for ranges of 20%-80% so that each test includes at least some percentage of each variable. Or, you could move up to a five-row triaxial for even more results.
Below is a five-row opacity triaxial blend of three colors (red, green, and blue) in RGB with a range of 20%-100% for each color. Because the range starts at 20%, each test includes all three variables. If you instead wanted the outer blends to represent line blends, the ranges would start at 0%.
Moving up to six and eleven-row triaxials you will notice that the number of tests we are creating is growing very quickly. Because the range is from 0%-100%, the outer edges of the triaxial represent simple two-variable line blends.
Based on these simple RGB color-blend triaxials, it might seem as if larger numbers of rows are unnecessary since we can easily infer the range of colors from a smaller four or five-row triaxial. However, as mentioned before, ceramics recipes are a complicated mix of ingredients and sometimes results do not transition gradually from one test to the next. Using too small of a triaxial with too large of an increment might miss important changes in glazes. For instance, if Whiting were one of our variables in an eleven-row triaxial, the difference between each Whiting increment would be drastic.
The number of tests required for a given triaxial size can be determined by the triangular number sequence. For a triaxial with only one row we of course need only one test. But the number goes up very quickly as we add rows: 2 rows has 3 tests, 5 rows has 15 tests, 11 rows has 66 tests, and so on.
If we want to create a triaxial where each variable is in the range of 0%-100% and tests are in 10% increments, we will need an 11-row triaxial with 66 tests. Unfortunately, it takes a lot of time to create so many test tiles! It would be nice if we could create meaningful tests while at the same time reducing the size of the triaxial.
Reducing Triaxial Size
Often we already have idea of the ranges we want for each variable in a triaxial.
For example, say we’d like to find a nice Chinese blue & white qinghua underglaze. Nigel Wood’s Chinese Glazes shows that qinghua is not simply cobalt blue but rather a complicated mix of oxides. We decide upon three basic variables for our triaxial: cobalt (top), iron (left), and manganese (right). In order to do a comprehensive test, we decide to use steps of 10% for each variable. Usually we would need a full 11-row triaxial for this type of test, necessitating 66 tests. However, we decide that the left side (a line blend of cobalt and manganese) and the bottom side (a line blend of iron and manganese) are not interesting to us, so we can leave those out. Now we are left with only 45 tests, in other words a 9-row triaxial.
Volumetric Blending for Triaxials
In our blue & white (qinghua) underglaze triaxial we still have 45 tests to make. But instead of mixing each test individually, we can use Ian Currie’s volumetric blending to create intermediate tests.
Starting at the top of the triaxial pyramid, remove each adjacent glaze. Now we are left with only 15 glazes to mix. To produce the intermediate glazes, simply mix the two glazes in a 1-to-1 relationship by volume.
Note that the more times you need to mix a glaze, the more total glaze material you will require. For the type of triaxial below I usually mix 200 gram batches in order to ensure that I have enough glaze. My test tiles only require 20ml of glaze each.
We can further reduce the number of glazes to prepare, with the trade-off of more numerous and more complicated volumetric blends. In the example below, we are only left with 6 glazes to make.
The actual fired result of this triaxial is much less balanced than the computer-generated RGB diagrams. It’s apparent that cobalt oxide is a much stronger colorant than both iron oxide and manganese. It seems the most interesting results are in the bottom two rows of the triaxial. It might be interesting to “zoom in” on the bottom portion of the triaxial to refine the color even further.
In designing the tests, we also take into consideration that the glaze that covers an underglaze will affect the resulting color. So for each test tile, the top half is covered with a basic transparent glaze (Limestone), while the bottom half is covered in a traditional chinese glaze (灰釉). The narrow unglazed band in the middle gives us further information.
The changes in coloring oxide triaxials are usually straightforward- colors gradually shift. Results are not so certain, though, when performing triaxial tests on other glaze components such as glass formers, melters, stabilizers and opacifiers. Sometimes huge changes in a glaze can occur within only one or two-percent changes of the recipe. So ideally, for each glaze test we would create a 51-row (range of 0%-100% in 2% increments) or 101-row (range of 0%-100% in 1% increments) triaxial. But a 51-row triaxial needs 1,326 tests, while a 101-row triaxial needs 5,151 tests!
So in designing triaxials with small percentage increments it is often necessary to eliminate vast swaths of the triaxial.
For example, in the partial triaxial below I am searching for a nice teadust glaze. The triaxial is based on 2% increments of Chinese glaze stone, whiting, and silica. From experience and prior testing, I have already determined fixed percentages for ingredients not included in the test (red iron oxide and talc), and I have set fixed ranges for the three variables. For instance, from past tests I know that I do not want too much or too little whiting- a range of 10%-16% is enough. In this manner I have pared down a 1,326 test triaxial to only 13 tests.
As you can see in the results below, there is indeed a great range of glazes even within 2% increments. If I had created a smaller triaxial with larger 10% increments I might have entirely missed the teadust crystal effect.
Further reducing size
By plotting out all of the tests (or even just the extremities) in the large triaxial above on a Silica/Alumina chart, we can further reduce the number of glaze tests. From experience, it’s obvious that a cone 10 glaze with 60% Whiting and no Feldspar or Silica won’t work out very well. But by looking at the chart, we can see entire areas of the test triaxial that are “out of bounds” for a good glaze and thus probably don’t need to be included.
Having said that, there are some interesting glazes (like Shinos) out of the ranges of commonly accepted glaze limits. And in my example above, the tests that look most to me like teadust are quite high in silica and fall just inside the blue “underfired” zone.
For more information, see R.T. Stull’s original article in Transactions of the American Ceramic Society, Volume 14, pages 62-70. Also see Matt Katz’s Introduction to Glaze Formulation Online
Tests build upon each other. Coarse triaxials can be later refined, and specific glazes targeted within smaller ranges. The more experience you have, the more you know where to look. If you’re just starting out, I recommend a large, 11-row triaxial of Potash Felsdpar, Silica, and Whiting which will reveal a range of glaze types, from celadons to mattes. Once you have a good feldspar/silica/whiting glaze you could try adding coloring oxides, stabilizers like ball clay or kaolin, opacifiers, etc.
Triaxial Worksheets Download (PDF)
I’ve tried all manner of methods for making test tiles- thrown, extruded, and slab. Each type has advantages, but I’ve finally come to the safest, most economical and useful method for my needs.
Assuming you have a fast way to make slabs, using this technique you can easily make a couple hundred test tiles in an afternoon.
Hand-rolled slabs work fine. The irregular edges can be saved and used to make the supporting triangles. To attach the supports I use dry clay trimmings and pure vinegar. The triangles are dipped into the vinegar slip for about two seconds, then wiggled into place on the test tile until the two slabs are firmly joined. (I have tried this with all of my clays, from porcelains to stonewares, without problems. However, your clay may differ.)
Just after joining, the test tiles should be covered and dried slowly for the first day or so. You may notice some cracking along the join line, but the join should be strong enough to last through bisque and glaze firings.
If you have casting slip, the fastest way that I have found to make test tiles is to make a slipcast slab using a large, flat plaster table. Supports are joined to tiles using the same casting slip.
These test tiles will not fall over (especially important as I often put tests in public kilns). They can be stacked very closely in the kiln, and the height can be varied by choosing either the long or short end of the triangular support. Because they are flat, they require less glaze for coverage (handy when using very small batches of material) but also show more area of glaze than an extruded column. The hole in the base allows the tile to be hung on a glaze bucket or on a display. I stamp the back of the tile with a code for the type of clay, and there is ample space to write information about the test using an underglaze pencil.
The tiles should be quite thick in order to better absorb and support thicker layers of glaze.
After dipping the face of the tile into a glaze, I dip one side (not the top) again in order to see different thicknesses. (Double-dipping the top of the tile often results in the thicker glaze running down into the thinner area, sometimes resulting in a completely even coat after firing.)
Using flat, square tiles I can dip twice while only using 20ml of glaze. This is especially useful when doing triaxials or biaxials using the syringe method.
The back of the tile has a lot of space for writing glaze information. Before firing use an underglaze pencil (I use Amaco Black Pencils), after firing a permanent marker can be used to record the date and temperature.
Another advantage of these square tiles is that they photograph very well.
For my purposes, the best size for a test tile is 6cm or 7cm square. This size is large enough to get a feel for what the glaze looks like, while small enough to be covered by only 20ml of mixed glaze (including a second dip).
Conveniently, the supporting triangles are half the width of the tile, which allows for first cutting the slab into 6cm or 7cm columns.
I know it seems really anal, but this cutting pattern saves clay and ensures the tiles all rest at the same angle, allowing for more accurate comparisons and tighter packing in the kiln.
Jingdezhen Porcelain Stone
There are a number of types of porcelain stone mined throughout Jingdezhen and the surrounding countryside. Some are more suitable for making porcelain clay, while others are traditionally used for glazes. It is difficult to know how similar modern-day porcelain stone is to traditional materials. During the few years I have lived in Jingdezhen, some mines have been closed off to private mining, while others have simply run out of material. Those still operating often mix poor-quality material with good material in order to increase production. And plaster is added in ever increasing amounts in order to make the porcelain stone bricks less likely to break in transit.
Below are three types of porcelain stone fired to 1310 Celsius in a reduction atmosphere. From left to right: San Bao porcelain stone, Yaoli glaze stone #1, Yaoli glaze stone #2.
Most porcelain stone is made from a combination of rocks. The stones below are used to make Yaoli glaze stone. On the left is the more common, dirtier stone, in the middle is the higher quality stone, while the right is the mixed, washed and purified stone before adding plaster. All examples were fired to 1310 Celsius in reduction atmosphere.
San Bao Porcelain Stone (三宝瓷石)
Below are some simple tests of porcelain stone from San Bao Village (三宝瓷石). This stone is often used for making traditional porcelain clay, but it can also be used for glazes. (However, usually “glaze stone”, or 釉果， is used for glazes.)
The only chemical analysis I have found for San Bao Porcelain Stone comes from 陶瓷艺术釉工艺学 published by 江西高校出版社:
SiO2: 7.013, Al2O3: 17.64, Fe2O3: 0.69, CaO: 0.54, MgO: 0.09, K20: 4.02, Na2O: 4.68. LOI 2.01
(The SiO2 amount is most likely a typo, they probably intended to write 70.13.)
However, this analysis can only be used as a basic guideline. There are noticeable differences between the porcelain stone produced by various families in San Bao, and quality seems to change every year.
Traditional glaze recipes usually call for whiting, but unless fired very carefully, porcelain stone has a tendency to carbon-trap when whiting is the flux. Replacing whiting with wollastonite eliminates this problem.
Below are a few glazes containing San Bao porcelain stone, silica, wollastonite and kaolin. It is quite easy to make a very nice celadon with porcelain stone.
Yaoli Glaze Stone (瑶里釉果)
Yaoli Glaze Stone (瑶里釉果) is traditionally used for creating glazes. As with San Bao porcelain stone, there are a number of families mining, cleaning, and creating the glaze stone bricks sold in Jingdezhen.
陶瓷艺术釉工艺学 has the following analysis:
SiO2: 73.99, Al2O3: 15.55, Fe2O3: 0.37, CaO: 1.76, MgO: 0.33, K20: 2.88, Na2O: 2.63. LOI 2.88
But again, the material varies from seller to seller and from year to year.
At the left is a simple melt test of 75% glaze stone with 25% Wollastonite and 80% glaze stone with 20% Wollastonite. The 20% Wollastonite version is already a perfectly useable celadon.
Below are a few simple celadons made with glaze stone, silica, wollastonite, and kaolin.
A test of 80% porcelain stone and 20% kaolin (New Zealand halloysite). This porcelain is particularly white and fairly translucent. It is quite nice to throw but fragile when bone dry. Mixtures of various types of porcelain stone with between 10-40% kaolin produce porcelain bodies suitable for a range of temperatures.