Wootz Damascus steel blades

ELSEVIER

wootz

Damascus Steel Blades

J. D. Verhoeven,” A. H. Pendray,.+ and E. D. Gibson* *Department of Materials Science and Engineering and Ames Laboratory, Iowa State University, Ames, IA 50011 USA; and +ABS Master Bladesmith,

Williston, FL 32696 USA

Wootz Damascus steel blades contain surface patterns produced by bands of cementite particles which are generated in situ as the blades are forged from small ingots. A process for making these blades has recently been developed

which involves making ingots in a gas-

fired furnace followed by forging to blade shapes. This study presents a series of additional experiments which provide strong evidence that the mechanism responsible for the formation of the aligned cementite bands is similar to the mechanism that produces banded hypoeutectoid steels. That mechanism attributes the selective formation of ferrite bands to microsegregated alloying will form in ultraclean small amounts of the
elements. The results of this study show that the cementite bands hypereutectoid steels (I’ and S levels
lective coarsening of cementite particles during the thermal cycling of the forging process. The particle coarsening is induced to occur preferentially in the interdendritic regions of the alloys by the very small additions of the carbide-forming elements. 0 Elsevier Science Inc., 1996

that match both the attractive surface patterns and the internal microstructure of wootz Damascus blades. The method consists of melting an iron charge in a small sealed clay graphite crucible inside of a gas-fired furnace and forming an ingot by furnace cooling. These ingots will be called GF (gas-fired) ingots. The composition of the metal charge is controlled to match that of the 17- to B-century wootz Damascus -O.lw/,P, and small reblades (=1.6”/,C, siduals of S, Mn, and Si; see the top line of Table 1. It is covered with slag material consisting of soda-lime glass, oyster shells, and green leaves, with some charcoal included to control the C level. The relatively high I’ level of these hypereutectoid steels makes them severely hot short and nearly impossible to forge. A rim heat treatment was developed in which the ingots were held at =12OO”C in iron oxide for several hours, thereby producing a protective rim of nearly pure iron around the ingot which allowed the hot short interior to be forged.

INTRODUCTION In a recent article [l] we discussed the nature of genuine Damascus blades. The term genuine was used to distinguish the blades from pattern-welded Damascus blades. These genuine Damascus blades were made in Damascus, Syria, from small crucible steel ingots produced in southern India. This steel is generally termed wootz steel. Rather than call blades made from this material genuine Damascus blades, it seems more appropriate to call them wootz Damascus blades because the term wootz clearly identifies the process of manufacture and avoids judgment of which type of blade is truly genuine [2]. Most of the high-quality wootz Damascus blades in museums and private collections were made in the 1718th centuries and the macrostructure, microstructure, and chemistry of these blades have been characterized [l]. A method has been presented [3] which can produce blades from small steel ingots 9 MATERIALS CHARACTERIZATION 37:9-22 1996 0 Elsevier Science Inc., 1996 655 Avenue of the Americas, New York, NY 10010

1044~5803/96/$15.00 PI1 SlO4&5803(96)00019-8

1. D. Verhoeven et al.

10 Table 1

Measured Compositions of Various Ingots in Units of Wt.%: Combustion Analysis Used for C, 0, and N and Emission Spec. Analysis Used for Remaining Elements

Steel

C

Mn

Si

-S

P

Cu

Avg. liter” [8] GF (3191) [3]

1.60 0.06 1.66 0.27

0.04 0.02 0.14 0.025

0.11 0.058

0.05 0.042

GF (7593) VIM (3-40-l) VIM (5-103)

1.48 0.13 1.57 0.10 1.51 0.11

0.14 0.09 0.10

0.025 0.074 0.024

0.01 0.01 0.05

0.012 0.034 0.046

0

N

0.007

0.0090

0.0110

co.005 0.018 0.018

0.0094 0.0020 0.0055

0.0122 0.0162 0.0135

Cr

Ni

MO

Al

0.06

0.01 0.04

0.01

0.02 co.01 co.01

0.01 co.01 0.05

co.01 co.01 co.01

“Average of 8-11 wootz Damascus blades available from the literature [S]. GF, gas fired; and VIM, vacuum induction melting.

When these ingots were forged to blades at temperatures below the A,, temperature it was found that cementite particles had formed throughout the interior and were clustered into aligned bands parallel to the forging plane. The aligned bands of cementite particles were shown to match those which occur in wootz Damascus blades. (Note: Evidence has been presented [4] which indicates that ancient bladesmiths used some type of a rim heat treatment to enable forging of the high-P hotshort ingots.) Initially it was found that blades made from the GF ingots would produce aligned bands of cementite and good surface patterns only about 40% of the time, but this lack of reproducibility has now been overcome essentially. The problems with unsuccessful blades can be partitioned into three categories: (1) Blades made from domed ingots. Occasionally the surface of an ingot will bulge upward as a result of excessive cavity porosity. Blades made from such ingots do not produce aligned bands of cementite particles. Experiments have shown that domed ingots are produced by very slow cooling (rates slower than furnace cooling), but they are also occasionally found in furnace cooled ingots. (2) Blades showing excessive graphite stringers. In these blades a large portion of the cementite bands are replaced by graphite stringers located where the bands would be and having the same spacing and alignment as the bands. (3) Blades in which the cementite does not align, but forms nearly exclusively on the austenite grain boundaries present in the final forging cycle.

Problem 1 has been eliminated either by using forced air cooling of the ingots or by simply discarding the occasional domed ingot formed with furnace cooling. Problem 2 has been eliminated as explained in the second part of the experimental section. Experiments have shown that problem 3 can be eliminated by proper control of the crucible charge and careful process control of the melting and freezing processes. Proper control of the charge is discussed in the sixth part of the experimental section. Process control was achieved by monitoring a type-B thermocouple located at the outside bottom of the clay graphite crucible. An optimum time/temperature melting cycle has been established by repeated experimentation. It consists of a rapid heat up of the charge, holding for a period of 2040 minutes between 1440°C and 1480°C followed by cooling at furnace cooling rates or faster. Experiments employing a second thermocouple (protected with an alumina tube) in the melt showed that during the 20- to 40minute period the melt temperature lagged the control temperature by up to 20°C on heating and then fell below the control temperature on cooling by up to 20°C until after the thermal arrest. Research on successfully reconstructed wootz Damascus blades [5] established that the aligned bands of cementite particles formed in situ during the forging of the blades and that the spacing of the bands corresponded roughly to the spacing of the primary dendrites in the solidified ingots which would be expected after the deformation. It was therefore indicated that a type of carbide banding, similar to the fer-

Wootz Damascus Steel Blades

rite banding that occurs in rolled hypoeutectoid steels, was producing the aligned microstructure. A thermal cycling (TC) technique was later used [6] on reconstructed Damascus blades and two museum-quality wootz Damascus blades which provided support for the banding idea. First, the blades were austenitized to the point where all of the cementite particles in the bands producing the damask pattern were dissolved and the blades were quenched. In both types of blades it was found that subsequent TC caused the reprecipitation of the cementite particles in a band morphology similar to the original blades. If, however, the blades were austenitized very hot, 12OO”C, for long times, 18h, the banding would not return on cycling, thus indicating that homogenization of slowly diffusing impurity elements such as Mn, Si, S and P plays a key role in the banding (Note: The fact that the rim heat treatment, also done at 12Oo”C, does not eliminate cementite banding is thought to be due to combination of the fact that the time temperature is shorter and that the interdentritic spacing of an ingot is considerably larger than in the final blade, and diffusion times increase as the square of diffusion distance.) It was concluded [6] that the TC experiments support the theory that the aligned bands of cementite in wootz Damascus blades are produced by a type of carbide banding, but that the specific elements involved in the banding mechanism were not known. A long series of additional experiments has been carried out which support the banding mechanism and identify specific elements that produce the banding. These experiments are described sequentially in the Experiments section. They involve additional studies with GF ingots and a series of controlled experiments in a laboratory environment using vacuum induction melting (VIM) so that atmosphere conditions and starting compositions can be carefully controlled. Preliminary work was done on the sealed crucibles in the GF furnace with an internal thermocouple as just described. The internal time/temperature

11 record upon cooling, which corresponded to the optimum external time/temperature conditions established for successful ingots in the GF furnace, was matched on the laboratory VIM ingots. Surprisingly, it was found that blades forged from the controlled laboratory VIM ingots could not be made to produce surface patterns or aligned bands of cementite particles. This led to a series of additional experiments in GF ingots which eventually demonstrated that the key to band formation is inclusion of small additions of carbide-forming elements. The fairly extensive set of VIM experiments have been documented elsewhere [7] and they are discussed only very briefly here because they were all done without addition of the carbide-forming elements.

EXPERIMENTS ELIMINATION

OF RIM HEAT

TREATMENT

The rim heat treatment was required because the relatively high I’ level used to match the compositions of wootz Damascus blades forms the low-melting steadite structure (melting point -95O’C) and causes hot shortness in forging. Several GF ingots were made with the P lowered to the 0.020.03 range and it was found that the ingots were no longer hot short. Blades formed from these ingots without the rim heat treatment continued to produce good surface patterns and aligned bands of cementite particles. Therefore the ingots made for this study generally used P levels in the 0.02-0.03w/, range and no rim heat treatment. (Note: Although the average I’ level of 11 ancient Damascus steel blades is reported as 0.11% [8], one of the blades in the average had a level of 0.03% and a recently analyzed blade [9] also had a level of 0.03%. PROBLEMS

WITH

GRAPHITE

FORMATION

It is fairly common to find stringers of graphite in blades after forging. Such stringers are also occasionally found in ancient blades, as has been illustrated and

J. D. Verhoeven et al. Table

2

Outline of Procedures

for Preparation of the Gas-Fired (GF) and the Vacuum-Induction-

Melting (VIM) ingots Component

VIM ingots (no slag)

GF illgofs

VIM ingots (slag)

Crucible

Clay graphite

Alumina.

Clay graphite

Charge*

Armco + Sore1 + clear

Armco + Mn + Si +

Armco + Mn + Si +

green glass + oyster shells + leaves Melt process

2040

minutes between

1440 and 148O”C, cooling rate at arrest temp. = 5 -10°C minute

FeS + FeP

FeS + FeP + tempered glass

Vacuum outgas 158o”C/5 minutes, backfill with

Vacuum to 8OO”C,backfill

N2 gas; cooling rate at

with N2 gas, 158o”C/5 minutes; cooling rate

arrest temp. = 5-lO”C/ minute

at arrest temp. = 5-lO”C/ minute

*Annco: A high purity electrolytic iron previously discussed [S]. Sorel: A high purity iron + carbon alloy produced by QIT America.

discussed elsewhere [4]. The stringers are aligned along the forging plane with roughly the same spacing as the cementite bands. They are colinear with the MnS stringers. This result presents fairly conclusive evidence that they are forming in the interdendritic regions (IRS). The as-solidified ingots are loaded with fine microporosity positioned in the IRS. Recent work [lo] using hot isostatic pressing has shown that the graphite is nucleating on the interdendritic micropores in the ingots. Experiments have found that this graphite problem can be eliminated by addition of a forging cycle employing temperatures above the A,, temperature (=102O”C) with adequate deformation to close up the microporosity. VIM EXPERIMENTS Initial VIM experiments used high-purity alumina crucibles charged with Armco iron and additions of Mn, Si, I’, and S to achieve the desired compositions. Ingots weighed 145Og and were 77mm in diameter. The initial VIM ingots displayed an annular ring of cavity porosity around an island at the center of their top surface, whereas the GF ingots had a flat upper surface with radially directed surface dendrites and no cavity porosity. The GF ingot appearance was matched in the VIM ingots by vacuum outgassing the melt for around 5 minutes at =158O”C. The melt procedure involved heating to =lOOO”C under vacuum, back-

filling with nitrogen gas, heating to =158O”C and outgassing, backfilling again with nitrogen gas, and then controlled cooling to match the cooling curves of the GF ingots. Table 2 clearly illustrates the differences between the GF and VIM ingot processing. A total of 10 VIM ingots were prepared and forged to blades. None of the ingots displayed a good surface pattern. Both a VIM ingot and a GF ingot were sectioned through their centers by electrospark cutting and the macrostructures and microstructures were essentially identical. The secondary dendrite arm spacing was measured on several GF and VIM ingots and found to be essentially the same, at around 300km. Because of the similarity of macrograin structure and the secondary arm spacing of the two types of ingots it is concluded that the microsegregation accompanying dendrite solidification would be the same in both types of ingots given the same alloy composition. The chemical compositions of 10 GF ingots and four VIM ingots have been measured and found to be quite consistent. Table 1 presents the composition of two GF and two VIM ingots. In several cases GF and VIM ingots were forged in pairs to insure the same forging conditions and the GF ingots consistently produced patterns while the VIM ingots did not. Longitudinal sections of blades from GF ingot 7593 and VIM ingot 5-103 are compared in Fig. 1. The cementite particles in the VIM ingots failed to line up in bands and generally

13

Wootz Damascus Steel Blades

gave the appearance that they had formed on the set of austenite grains present in the final forging treatment. As seen in Table 1, the VIM ingot 5-103 had the level of Cu and Ni raised to 0.05”/,. This was the only VIM ingot that had Cu and Ni added and this was done to match the higher level of these elements sometimes reported in wootz Damascus blades. The experiments show that in spite of a close match in the dendrite structure, the macrostructures and microstructures, and the chemical composition of the elements in Table 1, the VIM ingots do not produce aligned bands of cementite particles when forged to blades with the same forging schedule which does produce the aligned cementite bands in the GF ingots.

(a)

TC EXPERIMENTS

WITH

BLADES

just discussed it has previously been shown [6] that wootz Damascus blades and the reconstructed blades made from GF ingots can be fully austenitized to remove the cementite bands, quenched, and then thermally cycled and the cementite band structure will reappear if several thermal cycles are employed. To evaluate the performance of the VIM ingots, pieces of blades forged from three of these ingots were thermally cycled along with the blade from a GF ingot with the time/temperature schedule

As

(b) FIG. 1. Longitudinal sections of blades forged from (a) GF ingot 7593 and (b) VIM ingot 5-103. Etch: picric plus boiling picric acid. X100.

0

200

400

600

800

Time FIG. 2. Recorded trace of thermocouple

1000 1200

1400 1600

(sec.)

attached to the four samples given the TC (thermal cycle) treatment.

14

f. D. Verkoeven et al.

(a)

3(b). This latter result has been reproduced several times in blades from GF ingots. Two conclusions are evident: (1) The bands of cementite particles are reforming in the IRS in blades made from GF ingots and (2) the bands cannot be made to form in the blades of the VIM ingots with this cyclic heat treatment. (Note: Studies of as-forged blades have also shown that MnS stringers occur within the carbide bands, which, consistent with conclusion 1, means that the carbide bands of the forged blades also occur in the IR.) TC EXPERIMENTS

(b) FIG. 3. The banded cementite structure that reformed in the blade from ingot 7593 following the TC treatment. (a) Picric etch followed by boiling picric etch showing the well-formed cementite bands. (b) Higher-magnification micrograph illustrating the MnS stringers consistently found within the cementite bands. Picric etch only. x 100.

shown in Fig. 2. The samples were 1.5mmthick sections cut transverse to the blade axis. (This treatment of austenitizing for 3 or 10 minutes at llOO”C, water quenching, and thermal cycling (TC) six times as shown in Fig. 4 will henceforth be called the TC treatment. Metallographic examination found a fully martensitic plus retained austenite structure after both the 3- and lominute hold times.) No evidence of band formation was found in any of the three VIM ingots. The GF ingot produced a distinct set of bands as is illustrated in Fig. 3(a). Examination of the darker etching regions seen within some of the bands showed that the darker etching was caused by stringers of MnS, as is illustrated in Fig.

WITH

INGOTS

Small (1.5mm-thick) pieces of both GF and VIM ingots were given the same TC treatment as applied to the blades. Metallographic examination confirmed that the samples contained only plate martensite and retained austenite after the 1100°C austenitization and quench. After the six-cycle heat treatment similar to Fig. 2, it was found that in the GF ingots clustered arrays of cementite particles were formed in the IRS as seen in Fig. 4. Figure 4(a) is a lowmagnification picture illustrating a dark etching constituent which appears to lie in the IRS of the dendrites that had formed in the ingot. This conclusion was confirmed by etching the sample with Oberhoffer’s etch, which even more clearly revealed the location of the IRS. Figure 4(b), a highermagnification micrograph of the highlighted square region on Fig. 4(a), shows that the dark regions are clusters of cementite particles. This relatively simple test has been applied to a large number of GF ingots which have produced good blades and in all cases results similar to Fig. 4 have been obtained. It is also interesting to note that whenever microporosity is found in these ingots [dark patch in Fig. 4(b)] the surfaces of the pores are surrounded by clusters of cementite particles. The VIM ingots consistently failed to produce clusters of cementite particles with this cyclic heat treatment. Cementite particles form throughout these ingots but are present randomly and one cannot detect the location of the IRS from them.

Wootz Damascus Steel Blades

15

which good surface patterns were obtained in all forged blades. In the first two experiments the standard flux of clear bottle glass, oyster shells, and green leaves was used. In the next five experiments the oyster shells were removed and in the final experiment the green leaves were removed. These results indicated that a simple bottleglass slag was sufficient for successful band formation. After the second, third, and seventh successful ingots were prepared, ingots were produced using identical procedures but the Sore1 iron was removed and the C was added by addition of charcoal mixed in to the iron charge in the crucible. These three ingots were forged to blades at the same time as one of the successful ingots and in all three cases no surface pattern and no ce-

FIG.4. Appearance of ingot 12193 following the TC treatment. Boiling picric etch. (a) low magnification. (X 25) and (b) High magnification of region in box on (a). (X200).

COMPOSITION

DEPENDENCE

A major

problem in doing systematic research on GF ingots has been the lack of reproducibility in pattern formation in the forged blades. Recently, however, the process has been developed to the point where the success rate is 100% provided: (1) only those ingots are forged which are not domed, (2) the initial high-temperature forging is used to avoid graphite formation, (3) the optimum time/temperature process control is used, and (4) the metal charge material is restricted to a mixture of two commercially available irons, Armco iron and Sore1 iron. Sore1 iron is a highpurity iron containing around 4”/,C and by mixing it with the high-purity Armco iron one may match the C levels of wootz Damascus blades. A series of eight experiments has recently been completed in

(b) FIG. 5. Picric + boiling picric etch on: (a) longitudinal section of blade from Armco-only ingot 6794. (b) longitudinal section of blade from Armco+Sorel ingot 6694. X 100.

1. D. Verkoeven et al.

16 Table 3

Chemical Analyses of Two Types of Ingots, Armco + Sore1 and Armco Only: All Analyses by Weight and Are in Parts per Million (ppmw), Except for C

M.

Armco + Sore1 (6694) Nucor

Armco only (6794) Nucor

Armco + Sore1 (6694) Mass spec.

Armco only (6794) Mass spec.

Armco + Sore1 Range 8 samples Nucor

Armco only Range 3 samples Nucor

1.35-1.73%

1.24-1.52%

100 130


22 100

6 20



S Si

110 1700

70 1100

24 1100

16 360

loo-250 700-1700

70-90 1000-1200

Ni Cr

100 100

25 73

13 22

C Mn P

1.35%

1.41%


MO



CU

Cl00

100

4


<4 48

Al

10

V

90

Nb Pb Sn Ti Zr B

Ca

Cl00 20


9.3


10
2

1

0.17


<1

co.3

K co Zn Mg

Na W 0 N

1.3

5.2

5.1 4.7 8.2 co.4

co.4 36 0.7

CO.008 0.37 22

35

<4 0.66 co.2 5.8 1.8 1.4 co.4 0.061

0.56


Cl00


Cl00







10-20



1


co.3 10 1.1

CO.016 0.40 <3

Combustion analysis 46 45 39 44

The data of the cations of interest are in bold

mentite bands on metallographic sections were observed. Figure 5 presents a comparison of the longitudinal sections of blades from an Armco-only ingot (no. 6794) and an Armco + Sore1 ingot (no. 6694) and one sees the remarkably distinct difference which is responsible for the lack of a surface pattern in the Armco-only blades. The chemical compositions of the initial four blades of this study were analyzed using a combination of combustion and emission spectrograph methods at both Chicago Spectra Laboratories and Nucor Steel and excellent agreement was obtained. Subsequent ingots were analyzed only at Nucor Steel and

the ranges of compositions found for the major expected impurity elements for all eight ingots when using Armco + Sore1 and the three ingots when using Armcoonly are presented in the right-most two columns of Table 3. One sees that the impurity levels are very small in both cases, with only three elements having levels >lOOppmw (O.Ol”/J: Si, P, and S. Analysis of the Sore1 and Armco irons revealed that the only element clearly being introduced by the ingot making is Si, which was present in the Sore1 at 800ppmw and in Armco at
17

Wootz Damascus Steel Blades

the Armco + Sore1 ingots a mass spectrographic analysis was carried out on one Armco + Sore1 ingot (no. 6694) and one Armco-only ingot (no. 6794) and the results of both the emission spec. and the mass spec. analyses are presented in Table 3. The accuracy of the mass spec. analyses is not expected to be as good as that of the emission spec. because the latter has been carefully calibrated with standards. However, the emission spec. analysis is not sensitive to the low levels possible with the mass spec. analysis. Therefore the emission spec. analysis is expected to be more accurate for those elements present at levels high enough for it to detect and the mass spec. analysis provides a good measure of the relative compositions between the Armco + Sore1 versus the Armco-only ingots for the remaining elements. The mass spec. technique was able to analyze for all remaining elements, except H, and it shows that there is no rare element present in the Armco + Sore1 ingots. The major impurity element is Si and the emission spec. analyses show that its level is statistically the same in the two types of ingots. The Armco-only ingots are seen to be purer with respect to a few elements as is summarized in Table 4. One sees that the Armco + Sore1 ingots contain more of fivecarbide-segregating elements V, Ti, Cr, Co, and W, and two non-carbide-forming elements I’ and S, but in all cases the levels are quite low. At this point in the research it seemed more likely that P and S were playing the key role in the banding mechanism because of their higher level in the steels than the carbide formers as well as their well-established surface active character, known to produce large physical effects at low composition levels. Table

4

Lists of Elements with Significantly Ingots

TYPC

P

Armco + Sore1 Armco only

130-280 30-60

Units by weight, parts per million (ppmw).

S

loo-250 70-90

ROLE OF THE SLAG A major difference between the VIM ingots and the GF ingots was the use of a slag with the GF ingots and it was initially thought that the slag played a key role in the band formation. The initial effort to study this problem involved making a VIM ingot with a slag and the final column of Table 1 presents the details of the experiments. Note that the charge used highpurity additions to the Armco iron rather than the Sore1 addition. Two experiments were done with one showing no banding and the other showing a limited amount of banding, which indicated that the slag might be playing a key role. Therefore the slag composition was determined on several of the ingots utilizing energy-dispersive spectroscopy (EDS) analysis in the scanning electron microscope. A tempered glass of known composition was tested with the EDS method and it was determined that the analysis was accurate to about 22% of the quoted values. The results for five experiments are presented in Table 5. The top two pairs of ingots both used oyster shells and clear bottle glass in the slag and it is seen that the CaO content increased as might be expected. The next two ingot pairs used clear bottle glass for the slag. The final ingot was the VIM ingot and it used tempered glass for the slag which was higher in MgO than the other slags. (Note: There may have been B203 present in the bottle glass, as EDS could not detect B, but there was a nil level in the tempered glass.) It is likely that the slag is controlling the oxygen activity in the crucible melts. The Fe0 = Fe + [0] reaction can be used to characterize the oxygen activity since one expects the activity of dissolved oxygen to

Different Levels in Armco + Sore1 versus Armco-Only

co

V

Ti

Cr

W

36 10

60-90 Cl0

17-34 9-14

73 22

<3

22

18

1. D. Verhoeven et al.

Table 5

Compositions of Slags: Percent by Weight

ExperimenP

NazO

SiOz

CaO

MgO

Al203

Fe0

Bands

GF:A+S(OyS) GF: A (Oy S)

10.3 9.9

65.3 67.1

20.6 18.9

0.6 0.8

2.2 2.1

0.7 1

Yes No

GF:A+S GF: A

11.2 10.9

75.4 75.7

10.1 9.9

0 0

1.5 1.5

1.6 1.3

Yes No

VIM: A + (S, I’, Si, Mn)

10.9

72.9

8.1

5.3

0.7

1.7

Weak

“A, Armco iron; S, Sore1 iron; Oy S, oyster shells; and VIM, vacuum induction melting.

be given by the product of the equilibrium constant for this reaction and the activity of the FeO. The data in Table 5 show that the slag is picking up FeO, but there is no systematic change in Fe0 levels between ingots giving bands and those not giving bands. Hence, although the data of Table 1 indicate an increased 0 level in blades made from slagged ingots over nonslagged ingots, the slag analysis indicates that this difference was probably not present in the melt but resulted from oxide inclusions or changes in 0 level on cooling or handling. Turkdogan [ 111 has shown that the total oxygen content determined by chemical analysis of steel bears little relationship to the state of oxygen in the liquid steel, and studies of 0 levels in high-C melts [12] show that at the temperature and compositions employed here the expected 0 levels are less than around 10ppmw. The slag might also be introducing some of its cations into the liquid metal which could affect the subsequent cementite band formation. The cations of interest would be Na, Ca, Mg, Al, and Si. There is clearly some pickup of Si from the slag but the data of Table 3 show that the Si level is essentially the same in the banding and nonbanding ingots, and in the nonslagged VIM ingots Si had been added to match these levels. The data of Table 3 show that the other four cations of interest (bold in the table) are present at extremely low levels and that their levels appear to be equal or higher in the ingot which did not produce bands (see 6794 ingot versus 6694 ingot). Hence the chemical analyses data indicate that pickup of cations from the slag does not play a role in the banding process.

At this point in the research two experiments were done in the GF furnace using no slag. The charge consisted of the Armco and Sore1 irons (1900g and 94Og, respectively) and the green leaves with a charcoal chunk above the leaves in the sealed clay graphite crucible. The leaves and charcoal were found to provide an adequately reducing atmosphere to avoid oxidation and both ingots were successfully forged to blade shapes, although this was difficult because both ingots contained a layer of cavity porosity at their bottom surface, which was not encountered with slagged ingots. Both blades displayed good surface patterns and strong banding, similar to Fig. 5(b). The range of the chemical composition of the two blades is presented in the first data column of Table 6. It was found that the P and S levels were below the range of the eight successful slagged ingots of Table 3, but the V, Cr, and Ti levels were within or above the range of these ingots. These experiments coupled with the slag analysis just discussed provided strong evidence that the slag does not play a key role in the banding formation and they suggested that the carbide-forming elements V, Cr, and Ti might be responsible for the band formation. CONTROL

COMPOSITION

EXPERIMENTS

IN GF INGOTS

A final series of experiments was carried out in the GF furnace in which controlled alloy additions were made to the Armco iron. Small 5Og buttons of four master alloys (Fe + S), (Fe + I’), (Fe + P + S + V + Cr + Ti), and (Fe + V + Cr + Ti) were prepared by arc melting in an inert atmo-

Wootz

Damascus

19

Steel Blades

Table 6 Compositions of Blades Made from Gas-Fired Ingots, (Same Units as Table 3): Alloys of First Column Used No Slag and Those of Remaining Columns Were Made from Armco Iron + Additions from Master Alloys Avmco+P&S + v, Cr, 7-i

M.

Armco + Sore1 No slag

Bands

Yes

No

Yes

1.8% 100 70-110 70-90 700
1.3221.50% Cl00 100-180 90-290 1000-1500 Cl00 200-300
1.62%
C Mn P S Si Ni Cr MO cu Al V Nb Pb Sn Ti Zr B Ca

Armco+P&S


sphere. Pieces of these master alloys of the calculated weights necessary to match the composition levels found in successful Armco + Sore1 experiments were placed in the bottom of the crucible. A mixture of Armco iron + powdered charcoal was added and covered with green leaves and clear green glass. The glass slag was added because it avoids the problem with cavity porosity and the above analysis indicated it was not adding impurity atoms, other than Si, to the melt. The first two experiments added only P and S to the Armco iron and neither one produced bands. The chemical analysis range on these two blades is presented in the second data column of Table 6 and it is seen that the major difference with the steels which formed bands is the lower levels of V and, to a lesser extent, Ti. The final four columns of Table 6 list the chemical composition of the blades of ingots made with the carbide-forming elements V, Cr, and Ti added from the master alloys. All four of these blades had an ex-

Armco + v, Cr, Ti

Aimco + v, Cr, Ti

Armco + v, Cr, Ti

Yes

Yes

Yes

1.58% Cl00 20 90 900 Cl00 100 Cl00 100
1.43% Cl00 10 30 900 100 200 Cl00
1.75%
cellent surface pattern and displayed good internal band formation, similar to that shown in Fig. 5(b). The data show clearly that the bands form with or without the P and S additions and that one or more of the three carbide-forming elements V, Cr and Ti appears to be the critical elements needed to cause the banding to form.

DISCUSSION The initial phases of this work established that the rim heat treatment was not a necessary part of the reconstructed process for producing wootz Damascus blades, because good-quality blades were formed in ingots which had the P level lowered to the point where the ingots were not hot short. It also established that the graphite problem could be elminated by doing the initial forging treatments at temperatures above the A,, temperature. There is good evidence [lo] that the graphite is nucleated by

20

the microporosity in the ingots which is closed up by the high-temperature forging. Here are the additional conclusions of this experimental work that relate to the cause of the cementite band formation: 1. Both slagged and nonslagged ingots made from Armco + Sore1 irons in the GF furnace produce good surface patterns and internal bands of clustered sheets of cementite particles. Nonslagged VIM ingots made from Armco + I’, Mn, S, and Si additions and having the same dendrite spacing, macrostructure, and apparent composition will not produce cementite bands and corresponding surface patterns when given the same forging process. 2. The TC treatment applied to banded blades reforms the bands of cementite particles and they are found to lie in the IRS. The TC treatment of nonbanded blades from VIM ingots produces random cementite arrays. 3. The TC treatment applied to slagged ingots forms clusters of cementite particles in the IRS, but when applied to the VIM ingots the cementite particles form random arrays. 4. GF ingots made from Armco + Sore1 iron produce good patterns whereas those made from Armco-only do not form the bands of the cementite particles and the corresponding surface patterns. The impurity-element level in both types of ingots is extremely low. All impurities are present at levels
J. D. Verkoeven et al.

iron at the levels found in the successful Armco + Sore1 runs. These experiments found that banding occurred with either the V, Cr, or Ti addition alone or the V, Cr, and Ti plus S and I’ addition. However, the addition of only S and P did not produce banding. These results present strong evidence that the banding is induced by addition of the carbide-forming elements, and apparently I’ and S do not play a role in the banding. As discussed in the introduction, previous experiments with the TC treatment [6] have presented strong evidence that formation of the aligned bands in wootz Damascus blades is produced by a carbidebanding mechanism. Results of the present TC experiments, such as that shown in Fig. 3, show that the carbide bands form in the IRS of the steels. The TC results of conclusions 2 and 3, taken together, provide strong confirming evidence that the aligned sheets of cementite particles are being produced by a type of carbide-banding mechanism in the GF ingots which does not operate in the VIM ingots. The carbide banding observed in these steels is slightly different from the classic ferrite banding which occurs in hypoeutectoid steels. Ferrite banding [13-151 is known to result from microsegregated minor elements, such as I’, which form arrays in the IRS of the cast metal that are aligned into bands during hot rolling and are effective in causing ferrite to nucleate preferentially on slow cooling from the austenite range. Whereas ferrite banding is complete after just one thermal cycle, the carbide banding found here is not apparent after one cycle but slowly develops over several cycles [4, 61. The slow evolution of the banded structure during the cycling indicates that some type of a preferred coarsening mechanism is involved. The TC provides an additional mechanism for coarsening over the familiar Ostwald ripening mechanism. As the temperature rises on the heat-up part of a cycle the volume fraction of cementite drops and the smallest particles disappear first. Therefore, on each heat up the smallest particles

Wootz Damascus Steel Blades dissolve preferentially. There are two possible mechanisms for formation of more cementite during the cooling cycle: (1) the existing cementite grows larger and (2) new cementite particles nucleate. In general it is difficult to nucleate cementite without supercooling. Therefore it is unlikely that many new cementite particles will nucleate because cementite particles are present nearby. Hence, the thermal cycling will tend to promote the growth of the larger particles so that once larger particles have formed in band regions they will tend to grow at the expense of the smaller interband particles due to both Ostwald ripening and the thermal cycling. Selective coarsening produced by the TC would be influenced by effects that reduced the mobility of the cementite/austenite interfaces. To illustrate, suppose that microsegregated elements reduced the growth rate of the cementite/austenite interfaces selectively in the IRS. During heat up there will be less dissolution of cementite in the IRS so its distribution will be biased to larger sizes there. During cool down, however, the cementite in the IRS would grow more slowly. Hence considering growth only it appears that the bias in the growth rate would not produce a net difference in particle size between the IRS and the dendritic regions (DRs). However, the nucleation problem on cool down must be considered. After heat up there will be a reduced density of cementite particles in the DRs because the faster interface growth there will have dissolved more particles. The larger number of particles that have dissolved in the DRs on heat up will not come back on cool down because of the nucleation problem. This means that the particle density will be reduced in the DRs relative to the IRS. Hence microsegregation effects which reduce the cementite/austenite interface growth rate in the IRS should promote cementite band formation during TC. The experimental results summarized as points 4 and 5 show that the microsegregated elements responsible for the selected formation of the cementite particles in deformed IRS of the blades are some combina-

21 tion of the three carbide-forming elements V, Cr, and Ti and not the elements I’ and S as had been our initial guess. Examination of the data of Tables 3 and 6 indicate that the V level is consistently higher in blades giving patterns, while the Ti level is lower in four of the five blades that did not form patterns, and Cr is lower in only three of the five blades that did not form patterns. It is suspected that the variation of the Cr level is due to calibration drift in the emission spectrographic analysis. More experimental work is needed to determine clearly which of the three carbide elements is most important in causing the bands to form, but the present data indicate that it is probably V. The experimental data show that the carbide-forming elements are causing the cementite particles to form preferentially in the IRS during the TC. The mechanism producing this action is unknown and different theoretical models can be postulated. At present the following model seems most reasonable. During solidification the V, Ti, and Cr atoms microsegregate to the IRS. When the cementite grows these elements become incorporated into the cementite forming M& where M is Fe + the carbideforming elements. Growth or dissolution of the M&Z carbide will require diffusion of the carbide-forming elements and since they are substitutional solutes in Fe the growth rate will be reduced relative to FeK. The higher concentration of these elements in the IRS will cause a reduced growth rate of cementite there and as discussed this will result in selective coarsening of the cementite in the IRS thereby producing the band structure. If this type of mechanism is correct, then one would expect the effects of carbide-forming elements to be cumulative since most all carbide-forming elements in steel are substitutional solutes that would have low diffusion coefficients in austenite.

CONCLUSIONS

Thermal cycling experiments applied to both reconstructed Damascus steel blades and the ingots from which they were made of-

1. D. Verhoeven et al.

22

fer very strong evidence that the formation of the clustered cementite particles into the banded array that gives the famous damask pattern is produced by a carbide-banding mechanism, with the carbides forming in the interdendritic regions of the steels. The carbide-banding mechanism is similar to the ferrite-banding mechanism which occurs in hypoeutectoid steels in that the cementite bands result from microsegregated elements causing selective formation of cementite in the intermediate regions that have been deformed into sheet geometries. It differs from the ferrite banding in that the cementite bands require several thermal cycles to form compared to a single cycle for ferrite banding. The cementite bands form by a selective coarsening process during the thermal cycling done at temperatures below the A,, temperature. Chemical analysis of blades that produce bands versus those that do not have shown the following: (1) Bands can be produced in steels that are extremely clean, having S and P levels as low as 0.003 and O.OOlwt.%, respectively, with the total remaining impurity level
Sciences, through the Ames Laboratory, Iowa State University, Contract W-7405-ENG-82.

References 1. J. D. Verhoeven and D. T. Peterson, What is a Damascus steel? Muter. Chav. 29:335-341 (1993). 2. M. Sache, Damascus Steel, Myth, History, Technology App/ications, Stahleisen, Dusseldorf, pp. 67-68 (1994). 3. J. D. Verhoeven and A. H. Pendray, Experiments to reproduce the pattern of Damascus steel blades, Mater. Char. 29:195-212 (1992). 4. J. D. Verheoven and A. H. Pendray, A reconstructed method for making Damascus steel blades, Metals Mater. Proc. 4:93-106 (1992). 5. J. D. Verhoeven and A. H. Pendray, Studies of Damascus steel blades, I: Experiments on reconstructed blades, Mater. Char. 30:175-186 (1993). 6. J. D. Verhoeven and A. H. Pendray, Studies of Damascus steel blades, II: Destruction and reformation of the pattern, Mafer. Char. 30:184-200 (1993). 7. P. M. Berge, A study of Damascus steel, MS thesis, Iowa State University, Ames, IA (1994). 8. J. D. Verhoeven and L. L. Jones, Damascus steel, II: Origin of the damask pattern, Mefnl/ography 20: 153-180 (1987). 9. D. T. Peterson, H. H. Baker, and J. D. Verhoeven, Damascus steel: characterization of one Damascus steel sword, Mater. Char. 24:355-374 (1990). 10. P. M. Berge, J. D. Verhoeven, D. T. Peterson, and A. H. Pendray, Graphitization of plain carbon steels, Iron Steelmaker 22:67-72 (1995). 11. E. T. Turkdogan, Nucleation, growth and flotation of oxide inclusions in liquid steel, J. Iron Steel Inst. 204:91&949 (1966). 12. T. Fuwa and J. Chipman, The carbon oxygen equilibria in liquid iron, Trans. Met. Sot. AlME 218: 887-891 (1960). 13. L. E. Samuels, Ferrite-pearlite banding in steel, in Optical Microscopy of Carbon Steels, American Society for Metals, Metals Park, OH, pp. 127-161 (1980). 14. R. A. Grange, Effect of microstructural steel, Met. Trans. 2:417426 (1971).

banding in

15. I’. G. Bastien, The mechanism of formation of banded structures, I. Iron Steel Inst. (London), 193: 281-290 (1957). Received June 1995; accepted January 1996.