Testing Tamahagane – Traditional Japanese Steel

Patreon

Thank you Patreon supporters! Through Patreon support I was able to pay for metallography and for the consumables for making coupons and running the CATRA experiment. If you want to support more knife steel research please join us at Patreon.com/KnifeSteelNerds

Video

Here is the video version of the following information:

What is Tamahagane?

Tamahagane is the name used in Japan for steel made with their centuries-old process. There are several videos available online to see the process such as this one. It starts with “iron sand” which is a naturally occurring form of iron mostly composed of magnetite, Fe₃O₄, though can also contain some hematite (Fe₂O₃) and other constituents like quartz and titanium dioxide [1]. One nice thing about magnetite (as the name suggests) is that it is magnetic, so it can be relatively easily separated from other minerals with magnets. The iron sand is then “smelted,” which is the process of reducing the oxide ore to the base metal, in this case, iron. The smelting is performed in a traditional furnace called a tatara. A clay tub is used which is layered with charcoal and the iron sand [2]. The charcoal acts as a reducing agent to turn the magnetite into iron, and also is a carbon source to turn the iron into steel (steel is iron plus carbon). The charcoal reduces the steel because carbon monoxide and carbon dioxide forms when it burns in air. The carbon monoxide (CO) reduces the magnetite to iron [3]. At the end the clay vessel is broken and the tamahagane is removed, which is in the form of rough and porous sponge-like pieces that also contains slag (large undesirable inclusions). The swordsmith separates pieces by “grade” which corresponds to the approximate content [4]. The grade is determined by color and also the fracture appearance [4]. The swordsmith then forges the pieces gently, and then folds and forge-welds them multiple times to eliminate porosity, remove some slag, even out the carbon content, and break up inclusions and impurities.

The “sponge-like” pieces of tamahagane after smelting. Picture from Wikipedia.

Soo all of that describes the manufacturing process. Ultimately tamahagane is a simple carbon steel. It is even simpler than a typical carbon steel like 1095 or White #1, because it is lacking intentional additions to those steels of manganese and silicon. So the steel has carbon of some content (described next) and then only trace amounts of other elements.

While there are a few individual swordsmiths that make their own tamahagane, most of the tamahagane produced in Japan comes from Nittoho tatara in Shiname Prefecture, the last of its kind. All tamahagane production had ceased by World War II. But this facility was restarted in 1977 and continues to produce the traditional steel for Japanese swordsmiths.

Published Tests of Tamahagane

Carbon Content and Composition

The cutting edge of Japanese swords has typically been found to be around 0.5-0.8% carbon [2, 5-9]. However, there are different types of construction used in Japanese swords, such as the “kobuse” method that uses a low carbon core and a higher carbon sides and cutting edge. Thus, the low-carbon pieces of tamahagane can be used in these low-carbon portions, giving the blades greater ductility and reducing the chance of catastrophic breakage. Measurements of the core carbon content have returned values between 0.01% [7-8] and 0.2% [2]. Sometimes steel with a carbon content in between these two extremes is used on the sides [2,7-8].

Schematic of “kobuse” construction from [3]

However, the carbon content is not always perfectly distributed throughout. Analysis of a modern kogatana knife (utility knife) produced by Sadanao Mikami had a range of carbon from 0.6% on one end to 2.0% on the other [5-6]. The uneven carbon content can come from distributions within the original pieces, mixing of low carbon and high carbon pieces (often done intentionally), and performing an insufficient number of forge-folding operations to even out the carbon. They reported that the average carbon content was 1.03% for the entire knife [5]. There were also small amounts of other elements:

Composition of modern kogatana by Sadanao Mikami [5]

Microstructure and Hardness

Japanese swords are produced to have a “hamon,” a visible line that distinguishes between a soft spine and the hard cutting edge. Thick clay is applied to the blades at the spine to slow the quench rate so that it does not harden. The hard phase at the edge is martensite, and the soft phase is pearlite. At the hamon transition, there is a combination of martensite and pearlite. The low carbon core contains large amounts of soft ferrite. The pearlite content of the core is determined by the carbon content, more carbon means more pearlite.

Images of microstructure from a “modern” sword (1945) and an “old” sword (~600 years old) at different locations [9]

Micrographs of a ~600 year old sword [7]

Studies on the hardness of Japanese swords correspond to these differences in microstructure and carbon content. Typically, the hardness at the edge is between 600-900 HV [2,3, 5,7, 9] and the spine is between 100-300 HV [2, 3, 7, 9]. This corresponds to about 55-67 Rc at the edge and 0-30 Rc at the spine (Note: Rockwell C hardness only reliably goes down to 20).

Microhardness values of ~600 year old sword [7]

Microhardness of a “Koto” style Japanese blade (before 1596 AD) [3]

Microhardness from the 1945 and 600 year old swords [9].

Impurities, Inclusions, Forge Welding, and Folding

A study was performed along with swordsmith Sadanao Mikami on the effect of forging and folding on impurities and inclusions in tamahagane [10]. The process is shown in the figure below:

The forge welding and folding process used in Ref. [10]

The swordsmith started by hammering chunks of tamahagane into sheets, then cutting those sheets into chips. The chips are sorted by carbon content by hitting them with a hammer. High carbon chips break and low carbon chips don’t. The low carbon chips were stacked together and forged together into a bar, then cut and restacked six times (A1-A6); this was repeated with the high carbon chips (B1-B6). Then the low and high carbon bars were stacked together and forge welded together, followed by cutting, restacking, and forge welding six more times (C1-C6). The microstructure became more evenly distributed through the forge welding process as the carbon content became more equalized, as shown in the following images:

Microstructure evolution of tamahagane after forge-folding operations [10]

One common claim of folding and forge welding is the elimination of inclusions, oxides and slag that are present in the original tamahagane pieces. The researchers found that while bands of FeO were present in A1 and B1 from oxides present in the original pieces, they were not found in A6 or B6. In other words, the FeO bands were eliminated by further forging and forge welding. They were originally present in bands from forge welding flat pieces together. They also found silica (SiO₂) and Fayalite (Fe₂SiO₄) inclusions in the steel. The silica particles were less than 1 micron and the Fayalite inclusions were 16-20 microns in length in the A1 and B1 samples. The thickness of the Fayalite bands was significantly smaller than the FeO, about two microns.

Iron oxide bands forge welded into tamahagane during initial forge welding

The Fayalite was reduced from 16-20 microns long in the A1/B1 specimens to 14-15 microns in the A6/B6 specimens. These were further reduced to 13.1 microns in the C1 specimens and down to only 8.2 microns in the C6 specimens. So while forge welding and folding did not significantly reduce the volume fraction of Fayalite inclusions, they were reduced in size, more isotropic (round), and more evenly distributed after that process.

Inclusions in tamahagane steel in the C1 condition (one forge weld of low and high carbon steel), left, and after six folding and forge welding operations, right [10]

In a study on a kogatana knife produced by Sadanao Mikami had relatively homogenous oxide inclusions in the size range of 5-10 microns [5]. They reported a volume of 4-7% inclusions.

Oxide inclusions in modern tamahagane knife [5]

Analysis of a ~600 year old sword [7] found an inclusion content of 0.8-1.9%, which they found to be much higher than the content in a modern production steel (labeled “ordinary steel”).

Volume percent of inclusions in ancient sword vs modern steel [7]

Micrographs of inclusions in ancient sword [7]

Analysis of another ancient sword found an inclusion volume of 1.1% [3]. The authors claimed that this inclusion content is low enough to avoid a “significant reduction in its toughness.”

Inclusions in pre-1596 AD sword (cross-hair scratch added intentionally) [3]

Strength and Toughness

Tensile tests of a sword by swordsmith T. Takaichi were performed and compared with a conventional carbon steel with similar carbon content of 0.55% C [2]. They found that in testing of the steel before heat treating that the conventional steel (S55C) had greater utility than the tamahagane sword steel (S1). When observing the fracture surfaces, they found oxides in the tamahagane and they speculated that it was these oxides that led to the lower ductility.

Stress-strain curves and fracture appearance of tamahagane sword steel (S1) and conventional steel (S55C) [2]

In tests of a ~600 year old sword and a 1945 sword, they measured the strength and ductility of the swords with a 4-point bending test [8-9].

4-point bend test of tamahagane swords [9]

They found the ancient sword to form a crack at around 0.88 ton-force, but the blade did not fully break and continued bending until final fracture. The 1945 sword broke at 2.62 ton-force and did not have the load drop from cracking that the ancient sword had. They calculated the bending strength of the swords as 2552 MPa for the ancient sword and 4645 MPa for the 1945 sword. They claimed that these results are “in a range of performance tool steel” [8]. The hardness of the cutting edges were similar in the two swords. Instead, they said the difference in measured strength was because the 1945 sword had a 5 mm hardened area while the ancient sword had only 1 mm of hardened area (see earlier microhardness charts).

I didn’t find much other toughness testing of tamahagane so comparisons between this traditional steel and modern carbon steels are in short supply. I found one study of miniature charpy specimens taken from a Japanese sword [11], but they did not compare with similar tool steels. Furthermore, the values at the edge of the knife where we would probably be more interested were essentially zero:

Image taken from [11]

Effect of Carbon Content on Final Properties

While there seems to be a consensus that the cutting edges of tamahagane Japanese swords are in the range of 0.5-0.8% carbon or so, I am not sure if the carbon content was and is that consistent. There was the one study on the tamahagane knife where they found carbon values as high as 2% within the same knife that read 0.6% in other locations [6]. I also found a study on different iron sand use at the Nittoho tatara (the last major facility remaining), and found that they had moved to a different source of sand that resulted in higher carbon and lower impurities [4]. So I think there is reason to believe that carbon content of tamahagane has gone up in recent decades.

The carbon content is very important to steel properties. One of the best known is the effect of carbon on the resulting as-quenched hardness. The higher the carbon content, the higher the hardness, though it levels out around 0.6% carbon:

Image from [12]

Carbon has an equally strong effect on impact toughness, even when tested at the same hardness for different low alloy tool steels:

Higher carbon also gives steel more iron carbide (cementite) which gives it greater wear resistance. This can give knives greater edge retention because edge wear is slowed. However, for a sword this wear resistance is not really beneficial, and toughness is a much more important property. So the 0.5-0.7% range of carbon would be about optimal for a sword, given the high combination of hardness/strength and toughness that you get from this microstructure. But for smaller knives, kitchen knives, etc made with tamahagane higher carbon could be beneficial for wear resistance.

1.25%C steel with significant carbide content (white particles)

0.6%C steel with almost no visible carbide

My Tamahagane for Testing

I recently went to Japan and took a class from Yuya Nakanishi, also called Masahiro. You can watch a video of our trip to Japan and making the “kogatana” (a utility knife). We took home our two knives that we made with him. However, I also wanted to test the steel for toughness which would be difficult with just the knife, as the standard toughness test uses rectangles. So I asked him if I could have a piece for toughness testing and he let me have one. This piece had already been through the forge welding and folding process. I also asked him what temperatures to use for heat treating and he told me 780°C for austenitizing and 180°C for tempering. I normalized and annealed the piece first, and followed his recommendations by austenitizing for 10 minutes, quenching in water, and tempering twice for one hour at each time.

Image from https://morinokyoto.jp/masahiro_en/

Carbon Content

I measured the broken toughness specimen with LECO combustion for carbon and sulfur content and got 1.06% carbon and 0.001% sulfur. This carbon content puts it into a range similar to White #2 (1.05-1.15%C). This is a good range for a utility knife, though as explained above would be on the high side for a sword. I am not sure if this is because Masahiro intentionally used higher carbon tamahagane for these smaller knives, or if it is due to carbon fluctuations, or higher carbon tamahagane being made at the Nittoho tatara. If I hear back from Masahiro I will update the article. This is very close to the carbon content in the kogatana by Mikami of 1.03% [5], which perhaps provides evidence that the carbon content of tamahagane has increased from older blades. I wasn’t able to measure other elements because of the size of the steel that I had.

Micrographs

The carbide structure looks consistent with the measured carbon content of 1.05%. It has a significant amount of small and evenly distributed carbides, but a smaller volume than the higher carbon 26C3. It looks closer to O1 and 52100, though the carbides are even finer in the tamahagane.

Micrograph of the heat treated tamahagane

O1 steel

52100 steel

Looking at a lower magnification micrograph, the carbide structure appears relatively consistent throughout. There are no visible layers from the forge-folding process. There is a “bright spot” in the upper left portion of the image. Higher magnification reveals that this is not carbide clusters but either an etching artifact or retained austenite.

Lower magnification of tamahagane

Impurities and Inclusions

The inclusion content of the tamahagane provided to me is very low, lower than reported inclusion contents in the journal articles that I cited. Using threshold measurements I estimate the inclusion content to be around 0.1%. I am not sure why this steel is so much “cleaner” than those analyzed in the other studies.

Polished (unetched) surface of tamahagane. Black particles are inclusions.

There is a small amount of orange, square/rectangular particles in the steel. These are likely titanium nitride (TiN) as these particles are typically yellow/orange and angular [13]. Iron sand typically has TiO₂ in it [1], and the Mikami blade had 0.03% Ti in it [5], so the presence of TiN is not entirely unexpected. Titanium and nitrogen have a strong affinity for each other, and these particles are often found in steels with small titanium additions.

Titanium nitride particles (TiN)

Hardness

The hardness of my toughness coupon was 64.6 Rc. This is relatively high in hardness, maybe even a bit high for a utility knife. However, it is within a reasonable range, and would provide very good cutting and edge retention. This lines up closely with tempering of other simple high carbon steels, such as 1.25% carbon steel 26C3, shown below. The 26C3 is somewhat lower in hardness, though this steel was tempered in another oven. My furnace is very good at maintaining a consistent temperature during tempering, so “overshooting” is less common.

I am not able to measure the hardness of the edge of the knife that I made at Masahiro Tantojo, as it is only hardened at the edge. It would require cutting it apart and using microhardness measurements to see. Since the knife was tempered by eye by the swordsmith it could be significantly different in hardness. The tempering was performed for a very short period, until it reached a temperature where water would sizzle when he touched the blade. My guess is that it is higher in hardness, somewhere between 65-67 Rc. However, if the carbon content is significantly lower in the blade it is possible that it would also be lower in hardness.

Toughness

Given the relatively high hardness of 64.6 Rc, the toughness was not chart topping. I measured 5.8 ft-lbs. However, this is a relatively high value for that level of hardness, as you can see when compared against other low alloy steels here:

As you can see the toughness falls right on the same line as 26C3 (green line in the upper image). So the tamahagane appears to be behaving similarly to conventionally produced low alloy high carbon steels. Overall I think this is a good result considering that some studies I cited before showed a potential reduction in toughness from oxides.

Edge Retention

Comparing the edge retention is a bit tricky since the kogatana design is different than the standard test knife that I use for CATRA testing. This knife is thinner behind the edge, about 0.010″ (0.25 mm) after sharpening, while our test knives are around 0.015″. This gives the thinner knife an advantage for a given edge angle, as shown in this earlier study on 154CM steel:

There is also the complication that I do not know the hardness of the kogatana knife, my guess is it is in the range of 65-67 Rc. When I tested it with CATRA I got 370 mm, which would also lend credence to the hypothesis that it is on the upper end of hardness.

There aren’t many low-alloy steels I have tested that would be in that hardness range. The light grey dotted lines show the approximate relationship between hardness and edge retention. So you will notice that the tamahagane is on the same line as most of the other low alloy steels, especially after accounting for the knife being thinner behind the edge. This is not too surprising given that most of the low alloy and carbon steels have tested in a similar range after accounting for hardness.

Tim Zowada Information

Tim is a bladesmith who makes his own bloomery steel using iron sand from Lake Superior. He was an author on one of the cited papers above along with Dr. John Verhoeven. He had the following comments after reading my report:

Here are a couple things that might be helpful. In general:

1. The final carbon content is determined by the average initial carbon content and the number of folds. More folds decrease the carbon content. The outside of the billet will decarburize during forging. That decarb layer is folded in with each weld, decreasing the average carbon content. Plus, the oxidation of the billet (scale) will decrease the overall size of the billet during all the folding and welding. For high quality steel, the typical number of of folds will be between 10 and 15. It ends up being a balancing act based on experience. Yoshindo Yoshihara prefers about 0.7% carbon for his swords. His brother prefers 0.6%. Maybe I have that backwards. The Art of the Japanese Sword (Yoshihara) is an excellent book that you should own. I’m not sure why the carbon content varied so much in your blade. My best guess is a low number of folds. A quick nital or ferric chloride etch would tell. You would be able to see lighter and darker bands of varying carbon content. Typically, the carbon content is pretty much homogeneous through the bar by about eight folds.

2. High hardness and low abrasion resistance is an advantage for my straight razor work. At 65 HRC, the blades hold up very well against whiskers. But, with the low abrasion resistance, a clean leather strop will remove enough metal to keep them sharp. They don’t need to go to the stones very often.

3. Like many things, the final carbon content and quality of the steel will vary from smith to smith, and batch to batch. I think what you got was a good representation of high carbon (1.0%) tamahagane.

Summary and Conclusions

Reviewing the prior studies on tamahagane, researchers of this material have been pretty realistic about its properties. At its best, tamahagane performs like other simple carbon steels. The potential for higher oxide inclusion content potentially limits its toughness, though in my own experiments it was roughly in line with the toughness you would expect for a steel made with modern methods. It is very interesting to learn about traditional methods for making steel and how they compare to the more advanced methods that we have developed since.

---

[1] Tanii, Hiroshi, Tadahiro Inazumi, and Keiichi Terashima. “Mineralogical study of iron sand with different metallurgical characteristic to smelting with use of Japanese classic iron-making furnace “Tatara”.” ISIJ international 54, no. 5 (2014): 1044-1050.

[2] Okayasu, Mitsuhiro, H. Sakai, and T. Tanaka. “Mechanical Properties of Samurai Swords (Carbon Steel) Made Using a Traditional Steelmaking Technology (tatara).” Journal of Material Sciences & Engineering 4, no. 2 (2015): 1-6.

[3] Verhoeven, J. D., and Tim Zowada. “Comparison of Two Swords of Antiquity: The Japanese Sword and the Muslim Crucible Damascus Sword.” Metallography, Microstructure, and Analysis 12, no. 6 (2023): 934-943.

[4] SUZUKI, Takuo, and Kazuhiro NAGATA. “Effect of the charge of” Komori” iron sand on the properties of” Tamahagane” steel produced by” Tatara” operation.” Tetsu-to-hagané 85, no. 12 (1999): 911-916.

[5] Das, Ananda Kumar, Takuya Ohba, Shigekazu Morito, and Muneo Yaso. “Evidence of Lath Martensite in High-C Japanese Sword Produced from Tamahagane Steel by Tatara Process.” In Materials Science Forum, vol. 654, pp. 138-141. Trans Tech Publications Ltd, 2010.

[6] Matsumoto, Chihiro, Ananda Kumar Das, Takuya Ohba, Shigekazu Morito, Taisuke Hayashi, and Go Takami. “Characteristics of Japanese sword produced from tatara steel.” Journal of Alloys and Compounds 577 (2013): S673-S677.

[7] Yaso, Muneo, Toshifumi Takaiwa, Yoshihiro Minagi, Kunichika Kubota, Shigekazu Morito, Takuya Ohba, and A. K. Das. “Study of microstructures on cross section of JAPANESE SWORD.” In European Symposium on Martensitic Transformations, p. 07018. EDP Sciences, 2009.

[8] Yaso, Muneo, Yoshihiro Minagi, Toshifumi Takaiwa, Kunichika Kubota, Tsuyoshi Kanaizumi, Takuya Ohba, Shigekazu Morito, and Taisuke Hayashi. “Study of Metallurgy and Mechanical Property on Japanese Sword.” In Materials Science Forum, vol. 738, pp. 222-227. Trans Tech Publications Ltd, 2013.

[9] Yaso, Muneo, Toshifumi Takaiwa, Yoshihiro Minagi, Tsuyoshi Kanaizumi, Kunichika Kubota, Taisuke Hayashi, Shigekazu Morito, and Takuya Ohba. “Study of Japanese sword from a viewpoint of steel strength.” Journal of Alloys and Compounds 577 (2013): S690-S694.

[10] Takami, Go, Takuya Ohba, Shigekazu Morito, and Ananda Kumar Das. “Microstructural observation on materials of the japanese sword under fold-forging process.” In Materials Science Forum, vol. 654, pp. 134-137. Trans Tech Publications Ltd, 2010.

[11] Misawa, Toshihei, and Shin-Ichi Komazaki. “Ductile-brittel transition evaluation of Japanese sword and weld metals using miniaturized impact specimens.” In European Structural Integrity Society, vol. 30, pp. 119-125. Elsevier, 2002.

[12] Krauss, George. “Martensitic transformation, structure and properties in hardenable steels.” Metallurgical Society AIME,(1978): 229-248.

[13] Descotes, Vincent, Thibault Quatravaux, Jean-Pierre Bellot, Sylvain Witzke, and Alain Jardy. “Titanium nitride (TiN) germination and growth during vacuum arc remelting of a maraging steel.” Metals 10, no. 4 (2020): 541.