The Snowflake Warrior Vase: Experiments to understand snowflakes and their suspension

Snowflake Warrior Vase, possibly Beijing, China, about 1825-1875. Gift of Benjamin D. Bernstein. 57.6.10.

I was intrigued and somewhat baffled when introduced to the Snowflake Warrior Vase and other snowflake glass objects in the Museum’s collection by Dr. Shelly Xue, our 2019 Carpenter Foundation Fellow for Asian Glass, and Astrid van Giffen, associate conservator at The Corning Museum of Glass. I was aware of the vase in our collection but can honestly say I had never given the background glass (the interior layer of glass) much attention. They were researching the Snowflake Warrior Vase and similar snowflake glass objects (read more about the Warrior Vase carvings and snowflake glass here) and asked me, a glassmaker, to suggest how such an unusual glass might have made.

Harry Seaman discussing samples with Dr. Shelly Xue (middle) and Assistant Conservator Lianne Uesato, in the Conservation Lab.

Looking closely, the inner layer of the Snowflake Warrior Vase is a glass unlike any other. This colorless glass not only contains a great density of fine bubbles (seeds) but also a significant amount of undissolved inclusion material. This “snowflake” material is in suspension evenly throughout the seedy glass. The seeds and inclusion material show up in varying densities in different objects, but the Snowflake Warrior Vase is a stunning example with a significant density of both bubbles and snowflake material.

Detail of the Snowflake Warrior Vase showing ruby and snowflake glass.

What are the snowflake inclusions? How were they introduced to the colorless glass? While little is known about how snowflake glass is made, some suggest that “stone powder” or local quartz sand is thrown into the glass. The same silica source that the glassmakers used in making the base glass would be a readily available and predictable product to use for the snowflakes.

Harry Seaman viewing a microscope image of snowflake glass.

Our theory is that the inclusions were a sand-like material melted into suspension during a fritted glass re-melt. Specifically, a pot of colorless glass was melted and then ladled into water where it cooled quickly and fractured into small granular particles (frit). It was then dried and sorted for size. A quantity of this frit was then mixed with the sand-like inclusion material and re-melted in the furnace. This process of using frit as our base and mixing in the inclusion material cold in a controlled and measured way allows the inclusions to be controlled for size and dispersion, hopefully avoiding clumping and creating a more homogenous mixture. The newly re-melted frit-glass should be both a stable and predictable base glass that will trap tiny bubbles amongst the granular particles. Together these processes should create a seedy, inclusion carrying glass that has predictable physical properties, such as working qualities and a predictable co-efficient of expansion, important to ensure compatibility with the dense colored glass to be layered over it.

Melting a glass like snowflake glass produced some practical questions: would the sand and frit melt together to create a snowflake-like effect? Would different sands affect how the glass and seeds would look? In an attempt to reverse-engineer the glass, we designed a few simple glass melting tests. For these experiments, we used Bullseye Glass Co.’s standard clear in coarse particle size (Bullseye code 1101-03) as our base frit.  We melted three batches:

Frit and sand mixture before the melt.
  1. A control melt of the frit by itself with nothing else added.
  2. A test of the frit with sand from Boshan, China, where snowflake glass was made in the 1960-80s.
  3. A test of the frit with common domestic silica. 

We hoped melts 2 and 3 would yield a glass with a combination of bubbles and un-melted silica suspended throughout a colorless glass. We weren’t sure what effect the different types and sizes of sand might have. 

The crucible in our electric kiln.

Each melt after our control consisted of just over 1000 grams of glass, of which about 8 percent by weight was sand. The glass was mixed with the sand with a misting of water and soap to help the silica evenly homogenize throughout the glass. The mixture was loaded into a ceramic crucible about 225mm high and 120mm wide (about the size of a large travel mug) and heated up to 2200 degrees Fahrenheit over 24 hours, then samples were gathered from the melted glass. 


Our control melt was fully melted with some seeds noticeable throughout. Our second and third melts both yielded stratified melt results with the following characteristics:

  • At the top of the melt a heavy 20mm layer of foamed glass full of the un-melted silica.
  • Below the foam, a glass with both many more noticeable seeds and inclusion material in suspension throughout. 
  • In the lower half of the crucible, the bubbles and inclusion material began to diminish and eventually become nearly absent. This, along with the heavy quantity of bubble and un-melted silica on the top surface (foam) leads me to believe that the silica was buoyant enough or carried upward by the rising of the bubbles during the melting process. 
Close up of a glass sample from the heavy foam top layer of melt 2.

When comparing melts 2 and 3 to our control there are many more bubbles. The sand seems to contribute to the number of bubbles in the glass melt.

Though not conclusive, the initial round of experiments produced promising results. We were able to melt material in suspension within a seedy glass. The samples do have some differences; most visually evident was that the in-house sand seemed to contain more clumps or clusters of particles, while the Boshan sand melted into a more homogenous suspension. Both seemed very similar to the snowflake glass in appearance when looked at in small samples, but in larger samples, the difference become more noticeable suggesting a few other avenues to investigate with further test melts. What effect on bubble size, quantity, and inclusion dispersion would changing the frit size have? What other inclusion materials and particle sizes might look more snowflake-like in a large sample? How might different melting cycles improve glass appearance and homogeneity? Finally, how might base glasses of different compositions affect the final appearance?

With so many new questions it’s hard to say where our next steps will lead. This unique and amazing 19th-century glass embraces some qualities most glassmakers would consider unacceptable and instead utilizes them in an ingenious and creative way. It helps tells a story around the Snowflake Warrior Vase and becomes a unique background behind the cameo carving of so many other objects.

On your next visit to the Museum, I challenge you to find the Snowflake Warrior Vase in the historical galleries and to look closely for yourself. You can also see more images of the vase and other snowflake glass objects on our website.

Snowflake warrior vase, possibly Beijing, China, about 1825-1875. Gift of Benjamin D. Bernstein. 57.6.10.
Click the image to view in 360°.



I would like to thank everyone involved with this research project, in particular Dr. Shelly Xue, Astrid van Giffen, and the CMoG collections staff.

1 comment » Write a comment

Leave a Reply

%d bloggers like this: