Bull. Mater. Sci., Vol. 24, No. 3, June 2001, pp. 317–322. © Indian Academy of Sciences.

Characterization of ancient Indian iron and entrapped slag inclusions

using electron, photon and nuclear microprobes
P DILLMANN† and R BALASUBRAMANIAM*
Department of Materials and Metallurgical Engineering, Indian Institute of Technology, Kanpur 208 016, India
†
Laboratoire Pierre Süe CEA/CNRS, CE Saclay 91191, Gif sur Yvette Cedex, France

MS received 8 February 2001; revised 29 March 2001

Abstract. Compositional and structural information were obtained from an ancient 1600-year old Indian iron
using microprobe techniques (EDS, µXRD and µPIXE). Several different local locations in the iron matrix and
in the entrapped slag inclusions were analyzed. The P content of the metallic iron matrix was very hetero-
geneous. Lower P contents were observed in the regions near slag inclusions. This was correlated to the
dephosphorization capacity of the slag. The crystallized phases identified in the slag inclusions were wüstite
and fayalite. The compositions of the slag inclusions were relatively homogeneous.

Keywords. Ancient iron; slag inclusions; characterization; electron, photon and nuclear microprobe
techniques.

1. Introduction eventual presence of CaO in the charge must be linked to

the ore composition. The efficient dephosphorization
The corrosion resistance of ancient Indian iron has been capacity of CaO is well recognized (Turkdogan 1996).
debated for several decades (Ghosh 1963; Lahiri et al Therefore, in the absence of CaO in the slag, a greater
1963; Prakash and Tripathi 1986; Prakash 1991; Bala- amount of P can be left behind in the metal. In order to
subramaniam 1995; Anantharaman 1997). A shining gain further insights into the operating conditions of
example of the excellent corrosion resistance of ancient ancient furnaces, it becomes important to obtain accurate
Indian iron (dating to period before the tenth century AD) chemical compositions and structural analyses (both from
is the Delhi iron pillar, a 1600-year old structure that has the metal and the entrapped slag inclusions) from ancient
withstood significant corrosion damage. A detailed cha- Indian iron.
racterization of the oldest Delhi iron pillar (DIP) rust As the ancient iron-making process did not utilize
(Balasubramaniam and Ramesh Kumar 2000) has esta- smelting of the metal, ancient irons possess very hetero-
blished that DIP obtains its excellent corrosion resistance geneous compositions (Ghosh 1963; Lahiri et al 1963;
due to the formation of a protective crystalline iron Prakash 1991). The minor element (C, P, Si, etc) content
hydrogen phosphate hydrate at the interface between the can vary depending on the location in the sample (Ghosh
metal and the atmospheric rust (Balasubramaniam 2000). 1963). Moreover, ancient irons invariably contain
One of the reasons for the formation of this phosphate is entrapped slag inclusions. These entrapped slag inclusions
the availability of a significant amount of P in the under- result from the process by which iron was made (Lahiri
lying metal. Based on several reported analyses, the ave- et al 1963; Prakash and Tripathi 1986; Prakash 1991). In
rage P content in the Delhi pillar iron has been reported to order to understand all the local compositions and struc-
be 0⋅24% (Balasubramaniam 2000). Moreover, available tures, it is necessary to use characterization techniques
compositions from ancient Indian iron objects indicate that are capable of sampling very localized locations. The
relatively high P contents (Ghosh 1963; Lahiri et al 1963; present paper outlines the application of some micro-
Prakash and Tripathi 1986; Prakash 1991). The high P diffraction techniques (µXRD performed with a photonic
content could originate either from the ore or from the microprobe and µPIXE performed with a nuclear micro-
charcoal ash. The former appears likely as Indian iron probe) to determine localized metal (and slag) composi-
ores are relatively rich in P2O5. However, the P content of tions and structures in an ancient Indian iron. Indeed, the
the metallic matrix is determined by the slag composition strong heterogeneity of these kinds of samples needs the
in the shaft furnace. The use of limestone in the charge of use of microbeams. Moreover, to understand the complex
ancient Indian shaft furnaces is not attested. Thus, the mechanisms linked to the production of ancient iron and
to understand the influence of the structure and composi-
tion on properties (like, say, corrosion behaviour), it is
*Author for correspondence necessary to obtain very fine local information about
317
318 P Dillmann and R Balasubramaniam

structure and composition. Therefore, a combination of coupled to the SEM operating at 20 keV, using a Si(Li)
photonic, electronic and nuclear microprobe techni- detector with a thin beryllium window.
ques (like µXRD, EDS and µPIXE) provide valuable To detect the minor and trace elements in the material,
insights. the micro particles induced X-ray emission (µPIXE)
The iron (from an iron clamp) utilized in the present method was utilized. A detailed description of the PIXE
analysis was obtained from an ancient Gupta temple (now method is described elsewhere (Johansson and Campbell
in ruins) at a site called Eran in Madhya Pradesh in Cen- 1988). The PIXE method is a convenient method to che-
tral India. Inscriptions in the temple site indicate that the mically analyze the sample because it quantifies relatively
temple was constructed in the later half of the fifth cen- low amounts of elements (Z higher than 12) in the meta-
tury AD (Tripathi 1992). The corrosion behaviour of this llic matrix and allows simultaneous multi-elemental ana-
iron has been investigated (Puri et al 1997) and the lyses. This aspect is important specially when studying
atmospheric corrosion products characterized (Ramesh archaeological iron samples because they usually contain
Kumar and Balasubramaniam 1998). As these two earlier various concentrations of minor elements (phosphorus and
studies did not consider microscopic compositional and silicon, for example). The 1⋅5 MeV proton beam of the
structural variations of the metal and the slag inclu- Pierre Süe Laboratory (at CEA Saclay, France) nuclear
sions, this aspect will be highlighted in the present microprobe was used. This energy was selected to favour
communication. the X-ray emission corresponding to the lighter elements.
The beam current was about 150 pA for a 2 × 2 µm 2 beam
spot size. The distance between the sample and the X-ray
2. Experimental Si(Li) detector (positioned at 45° from the incident beam)
was 6 cm. A 15 µm Mylar filter was used to reduce Fe
2.1 Sample preparation pill-up peaks. The data collected were analyzed with the
GUPIX 99 program. The relative error obtained on the
Two different sample preparation methods were utilized, quantitative results is about 5%. The profile data were
depending on the analysis technique. In the first method, a computerized with the RISMIN program. The detailed
polished section of the material (ancient Indian iron with experimental procedure is provided elsewhere (Neff and
entrapped slag inclusions) was prepared after mounting Dillmann 2001). A reference sample of low-alloyed steel
the specimen in epoxy resin and polished with SiC paper containing 450 ppm phosphorus was utilized.
(grades 180 through 2500) and diamond paste (3 and The X-ray microdiffraction (µXRD) experiments were
1 µm). This mounted specimen was observed in Cam- conducted on the D15 beamline at Laboratoire pour l’Utili-
bridge Stereoscan 120 scanning electron microscope (SEM). sation du Rayonnement Electromagnetique (LURE) at
The same sample was also utilized for analysis with the Orsay, France. The white X-ray beam delivered by the
nuclear microprobe of the Pierre Süe Laboratory for PIXE DCI ring was focused and monochromatized by a carbon/
analyses. However, the X-ray microdiffraction analysis, tungsten(C/W) Bragg Fresnel Multilayer Lens (BFML)
performed at LURE, requires to be conducted in the (Chevallier et al 1996). The BFML is a wide band pass
transmission mode. Therefore, after mounting the speci- monochromator resulting in broadening of the diffraction
men in the resin, a thin film of 20 µm thickness was pre- peaks. Photons centred around 14 keV (λ = 0⋅8857Å) were
pared. In order to accomplish this, first, a slide of the selected and focused down to a 10 × 10 µm 2 beam. The
mounted sample was cut with a diamond cutting wheel. diffraction patterns were collected with an image plate
This slide was later mechanically trimmed and polished downstream the sample. A detailed description of the
with 1200 grade SiC paper. One of the faces of the thin experimental set-up is provided in Dillmann et al (1997,
film was fine polished with 3 µm diamond paste so that it 2000). 1-D diffraction patterns were obtained by circu-
could be observed with an Olympus optical microscope larly integrating diffraction rings using the FIT2D soft-
(OM). The thin film was deposited on a 30 µm kapton ware developed at the European Synchrotron Radiation
scotch tape and the final sample was carbon coated by Facility (ESRF) (Hammersley et al 1996). The reference
sputtering. In this way, the thin film could be observed in sample used was silicon powder. The spectra were
the OM and SEM. Specific localized regions of interest compared with the JCPDF database using the Diffrac+
were then investigated by X-ray microdiffraction. program.
Simultaneous with the X-ray microdiffraction mea-
surements, the fluorescence signal emitted by the sample
2.2 Techniques was detected at 90° from the direct beam. Due to the
absorption of the low energy photons by air present
The metallographically polished section and the thin film between the sample and the detector, only elements heavier
were both examined by optical microscopy and SEM. The than potassium could be detected. The micro X-ray fluo-
major elements in the inclusion and in the oxide scales rescence under focalized synchrotron radiation (µSXRF)
were detected by energy dispersive spectrometry (EDS), data were analyzed with the WAPI software (Wang 1995).
 Characterization of ancient Indian iron 319

3. Results and discussion iron matrix at different locations in order to examine the
variation of the minor elements’ compositions. In the pre-
Two distinctly different localized zones of the ancient sent sample, the phosphorus content was examined at
Indian iron sample were analyzed: the iron matrix and the different points of the matrix using the µPIXE method. A
entrapped slag inclusions (particularly, the ones that were typical µPIXE spectrum is provided in figure 1. Several
surrounded by an oxide scale). different points were chosen in the iron matrix and the P
contents determined at these locations are provided in
table 1. Some details about the P contents determined
3.1 Iron matrix
should be noted. First, the data shows a wide variation in
the P content thereby indicating the heterogeneous nature
The composition of ancient iron is heterogeneous in
of the iron matrix. The global P content of the same mate-
nature because of the nature of the ancient iron making
rial (including the P in metal as well as in the slag) was
processes (Prakash and Tripathi 1986; Prakash 1991).
determined by wet chemical analysis and this was found
Therefore, it was important to compositionally analyze the
to be 0⋅25% (Ramesh Kumar and Balasubramaniam
1998). However, the P content in the metal can fluctuate
from 570 to 2300 ppm depending on the location. Points 7
Table 1. Phosphorus con- and 9 were both collected near the metal/oxide interface
tent in the Fe matrix (wt. but at different locations of the sample. The enhancement
ppm). Points 7 and 9 were of P near the metal surface observed by an electron probe
collected from near metal–
microanalyzer (EPMA) (Ramesh Kumar and Balasubra-
oxide interfaces at two diffe-
rent locations. maniam 1998) is partially verified, although the EPMA
study did not report the regions of lower P content.
Point P (wt. ppm) Secondly, the relatively higher amount of P content in the
1 1371 Gupta period ancient Indian iron has to be noted (when
2 1462 compared with modern steels).
3 1151 A phosphorus concentration profile was obtained with
4 2179 the µPIXE method in the metallic matrix near a slag
5 1531 inclusion (see figure 2). As observed in other phosphoric
6 570
7 2300 ancient irons (Neff and Dillmann 2001), the profile
8 1309 revealed decreasing P content near the slag inclusion (for
9 750 distances less than 10 µm). This effect must result due
to the dephosphorizing power of the slag. At locations

Figure 1. A typical µPIXE spectrum obtained from the iron matrix in ancient Indian
iron. Notice the distinct presence of P in the metal. The composition of P was obtained
from the spectrum.
320 P Dillmann and R Balasubramaniam

further away from the inclusions (about 40–50 µm), the the inclusions. A typical example is presented below to
variation in the phosphorus content were similar to that provide the insights given by the µXRD analysis tech-
typically observed in the metallic matrix. nique in understanding local nature of an entrapped inclu-
sion in the ancient Indian iron.
The SEM metallograph of an entrapped slag inclusion
3.2 Entrapped inclusion that was sampled by µXRD is presented in figure 2. The
OM and EDS observations revealed that the inclusion was
made of wüstite dendrites entrapped in a polyphased
The global composition of several inclusions was deter-
matrix. One part of this slag appeared to be made of
mined by EDS coupled to SEM (see table 2). These
glassy phases but some other areas were crystallized. The
results indicate the relative homogeneity of compositions
µXRD analyses reveal that these crystallized areas were
of the inclusions. Quantitative µPIXE analyses were not
made of fayalite Fe2SiO4 (figure 3). The presence of
performed in the slag inclusion because the interpretation
wüstite dendrites in glassy and fayalitic slags in ancient
of µPIXE spectra is very complex in multiphase samples.
Indian iron is usually the case as the main constituents of
The compositions determined by the EDS alone does
ancient Indian iron-making slags were SiO2 and FeO. No
not provide any information about the nature (i.e. struc-
other crystallized phase could be detected in the inclu-
ture, phase composition, etc) of the entrapped inclusions.
sions. The fact that the diffraction peaks of wüstite do not
It is in this regard that the µXRD analyses are very useful
appear on the spectrum can be easily explained by the fact
in determining local structure of the phases that constitute
that the dendrite is monophased. Thus the FeO crystal is
not well oriented to diffract under the monochromatic
beam. The wüstite dendrites are monocrystals. If these are
Table 2. Typical composition of inclusions sampled with a monochromatic X-ray beam, there is no
(at %) that were analyzed by EDS. chance to be in the right position to obtain diffraction
peaks (Bragg position). Diffraction is obtained from poly-
Inclusion Si P Fe
crystals. EDS analyses reveal that in addition to Fe, Si and
1 30 6 64 O, the inclusions also contained P. The absence of phos-
2 26 8 65 phates diffraction peaks on the µXRD spectrum revealed
3 25 11 64 that the phosphates contained in the inclusions were
4 24 7 68
glassy in nature.

Figure 2. SEM micrograph of an entrapped slag inclusion in the iron matrix.

 Characterization of ancient Indian iron 321

Several inclusions, like the one presented in figure 2, can be observed on the spectrum are due to the iron ma-
were surrounded by an oxide phase near the metal/oxide trix. The surrounding oxide appears most likely to be a
interface. Microdiffraction experiments were also con- corrosion product, as the material had been handled in the
ducted to examine the oxide surrounding the inclusion of atmosphere for some time. This could be either from the
figure 3. The spectrum obtained is shown in figure 4. This slag (wüstite being later converted to goethite) or metal
oxide is made of goethite (α-FeOOH). The iron peaks that (oxidation of the metallic matrix in the zone surrounding

Figure 3. µXRD pattern from the interior of the inclusion presented in figure 2, identifying the presence of fayalite.

Figure 4. Microdiffraction spectrum from the region in the inclusion of figure 2 obtained near the metal-inclusion interface. The
identified phase is goethite (α-FeOOH). The iron peaks that can be observed on the spectrum are due to the iron matrix.
322 P Dillmann and R Balasubramaniam

the inclusion). The latter appears reasonable because Balasubramaniam R 2000 Corr. Sci. 42 2103
wüstite is entrapped in a (relatively stable) fayalite Balasubramaniam R and Ramesh Kumar A V 2000 Corr. Sci.
matrix. 42 2085
Chevallier P, Dhez P, Legrand F, Erko A, Agafonov Y,
Panchenko L A and Yakshin A 1996 J. Trace and Micro-
4. Conclusions probe Techniques 14 517
Dillmann P, Populus P, Chevallier P, Fluzin P, Béranger G and
Local compositions from the iron matrix and entrapped Firsov A 1997 J. Trace and Microprobe Techniques 13 251
slag inclusions in an ancient 1500-year old Indian iron Dillmann P, Regad B and Moulin G 2000 J. Mater. Sci. Lett. 19
were determined using microprobe techniques (EDS, 907
µXRD and µPIXE). The P content of the metallic matrix Ghosh M K 1963 NML Tech. J. 5 31
was very heterogeneous (varying between 570 ppm and Hammersley A P, Svensson S, Hanfland M, Fitch A N and
2300 ppm). These observations are in agreement with Hausermann D 1996 High Press. Res. 14 235
earlier studies on ancient phosphorus-containing irons. Johansson S A E and Campbell J L 1988 PIXE, a novel tech-
The crystallized phases identified in the slag inclusions nique for elemental analysis (New York: John Wiley and
were wüstite and fayalite. The P content in the iron matrix Sons)
was lower near the slag inclusion due to the dephosphori- Lahiri A K, Banerjee T and Nijhawan B R 1963 NML Tech. J. 5
46
zation capacity of the slag. The advantage of utilizing
Neff D and Dillmann P 2001 Nuclear Instrum. & Meth. B
different microdiffraction techniques to understand local
(accepted)
compositions in ancient Indian iron has been highlighted. Prakash B 1991 Ind. J. Hist. Sci. 261 351
Prakash B and Tripathi V 1986 Met. & Mater. 2 568
Acknowledgement Puri V, Balasubramaniam R and Ramesh Kumar A V 1997 Bull.
Metals Museum 28 1
The cooperation of the Archaeological Survey of India is Ramesh Kumar A V and Balasubramaniam R 1998 Corr. Sci.
gratefully acknowledged. 40 1169
Tripathi R S 1992 History of ancient India (New Delhi: Motilal
Banarsidass Publishers) p. 252
References Turkdogan E T 1996 Fundamentals of steel making (London:
Institute of Materials) p. 186
Anantharaman T R 1997 The rustless wonder – a study of the Wang J X 1995 Utilisation du rayonnement synchrotron pour
Delhi iron pillar (New Delhi: Vigyan Prasar) la détermination d’éléments traces par fluorescence, Ph D
Balasubramaniam R 1995 NML Tech. J. 37 123 Thesis, Orsay University, Paris