Archaeometallurgical Study of a 16th–17th Century “Rapier” Sword Manufactured in Caino (Northern Italy)

Abstract

A metallurgical study was performed on a 16th–17th century “rapier” sword manufactured in Caino (northern Italy). Metallographic investigations and Vickers microhardness measurements indicate that the rapier was forged by assembling via hammer-welding different hypoeutectoid and near eutectoid steel bars. The rapier blade was heat treated by slack-quenching to increase its hardness, especially near the blade tip, improving the thrusting performance. The chemical composition of slag inclusions was analyzed by scanning electron microscopy coupled with X-ray dispersive spectroscopy. Compositional data of slag inclusions were analyzed by a multivariate statistical strategy aimed to distinguish and classify slag inclusions on the basis of their origin. It was estimated that the temperature reached during the finery and forging processes was at least 1270 °C and 1160 °C, respectively.

Introduction

During the Renaissance and Early Modern Age, some of the most important Italian armament production centers were placed in the district of Brescia (Lombardy, northern Italy), reaching their maximum expansion under Venetian domination (15th–18th centuries).^[1] As a manner of fact, a crucial role in sustaining this manufacturing growth was played by the iron ore extraction activities in the Brescian valleys mines, mainly located in Val Trompia and Val Camonica, along with the presence of an effective steelmaking production.^[2,3,4] In particular, iron and steel required for weapons fabrication were obtained by the indirect method, i.e., a two-step process based on the decarburization of cast iron produced in shaft furnaces named “a canecchio”.^[5] In fact, due to its poor workability, cast iron was decarburized to wrought iron or steel in open finery hearths named “large fires” where the metal mixed with charcoal was fired and exposed to an oxidizing environment.^[5,6] In this context, highly specialized sword-making workshops were developed in Caino, a small settlement located near Brescia. Based on historical documentation, it was estimated that the overall production capacity was up to a hundred blades per day.^[1] Proof-marked swords with the name of this village were even traded throughout Europe and perceived as high-quality weapons.^[2] Their great value can also be inferred from the many attempts to falsify their distinctive marks on blades manufactured elsewhere.^[1] Among the many blade types produced in Caino, falchions, and rapiers were the most common crab-claw hilt swords. Against this background, the present study is focused on a 16th–17th century rapier manufactured in Caino. In particular, the rapier is a one-handed thrusting civil sword especially suitable for dueling on foot or urban self-defense and even considered a status symbol component of the gentlemen's elegant attire.^[7] It is important to observe that the spread of this sword typology in Europe from the 16th to 18th centuries was contextual to the rise of new fighting techniques based on a predominant use of thrusting instead of slashing.^[8] This combat style shift stimulated the adoption of light and long swords whose functionality required suitable mechanical properties such as blade flexibility, high impact resistance, and fracture toughness. It is worth noting that proper manufacturing strategies must be adopted to attain these blade qualities. In the historical metallurgy field, the reconstruction of these forging techniques is still a crucial problem. To the author’s knowledge, only a few studies aimed to investigate the forging technology involved in the rapier fabrication process have been published so far.^[2,9,10,11] According to the ancient treatises, these studies point out the composite nature of the analyzed rapier blades, which were probably forged through shaping a layered lump constituted by low and high-carbon steel bars hammer-welded together and eventually subjected to faggoting with many folding cycles.^[12] In particular, metal bonding by hammer welding is a solid-state process. The metal surfaces to be joined were heated at 1100 °C and covered with silica flux. The chemical reaction between sand and iron oxide scale produces a low-melting point iron silicate (i.e., fayalite), which can be easily squeezed out by the hammer blows pressure allowing the surfaces to weld together through an interatomic diffusion mechanism.^[13,14,15] Moreover, the hardness and resistance of each analyzed rapier blade were enhanced by full or slack quenching and tempering heat treatments. Although the rapier, the object of this research work, was in the past subjected to a preliminary metallurgical examination by Beltrami,^[2] in this study further significant investigations have been carried out, by the addition of new deep insights into both smelting and forging processes. To achieve these objectives, a state-of-the-art reverse engineering approach based on the integration of metallographic and slag inclusions (SI) analysis was adopted.^{[1,16,17,18,19,20,21]}

Materials and Methods

The sword: Preliminary Description and Macro-observations

The rapier is without the guard and is characterized by a tapered double-edged blade with a rhomboidal cross-section. In Figure 1(a) is reported a photograph of the rapier as found before the samples taken in the work.^[2] The main rapier measurements, obtained by geometric reconstruction, are approximated as follows: total rapier length ≈ 880 mm, tang length ≈ 160 mm, ricasso length ≈ 58 mm, blade length ≈ 662 mm, and maximum blade width ≈ 18.8 mm. A proof mark “CAINO” and a crowded “S” letter were stamped on both ricasso sides (Figures 1(b) and (c), respectively). Based on these signatures, it was assumed that the rapier was produced by Sassi, an Italian sword-smith family who worked in Caino between the 16th and 17th centuries.^[2] It should be noted that the blade length is shorter than the usual length of this sword type which is approximately in the range of 1000–1100 mm.^[2] This evidence does not exclude the possibility that the length of the blade has been shortened due to damage limited to the final area of the rapier with subsequent repair. Five sections were taken along transverse axes from the tang (S1), ricasso (S2), and rapier blade (S4, S5, S6). In addition, a longitudinal section was sampled from the ricasso-tang transition zone (S3). The sampling scheme is shown in Figure 1(d).

Fig. 1

(a) Photographic image of the rapier with essential terminology, (b) “CAINO” proof mark with crowded “S” letter on the first ricasso side, (c) detail of the crowded “S” letter on the second ricasso side, and (d) sampling scheme

Experimental Methods

The experimental procedure employed in this research work was previously developed by the Authors to study other ancient artifacts.^{[1,16,17,20,22,23]} Each sample was ground with silicon carbide abrasive papers (grades from 80 to 1200) and polished with polycrystalline diamond pastes (3 µm and 1 µm). Metallographic observations of the as-polished sections were performed using a LEICA DMI 5000 M optical microscope to analyze both the morphology and distribution of slag inclusions. Then the samples were etched by immersion in 1 pct Nital solution (1 vol pct HNO₃ in ethanol) to reveal the microstructural constituents and their distribution and detect possible welding lines. This metallurgical information is crucial to reconstruct the artifact manufacturing process, especially in the case of composite objects. Vickers microhardness measurements were executed using a Micro Duromat 4000 Reichert Jung instrument, under a 300 g load applied for 15 seconds on the polished and etched surfaces to confirm the identification of the microstructural constituents. Later, the chemical composition of SI was investigated on each repolished sample to select a reliable number of non-metallic inclusions for each zone of interest, preventing any chemical measurements bias due to metallographic etching. LEO EVO-40 XVP Scanning Electron Microscope (SEM) in backscattering mode (BSD) was adopted to characterize SI microstructure in terms of mineralogical phases assemblage. Moreover, X-ray Energy Dispersive Spectroscopy (EDS) coupled with SEM was used to analyze in a semiquantitative way the SI chemistry. The chemistry of multi-phase SI was analyzed by examining all SI surfaces through area acquisition. The SI chemical composition is expressed by the following stoichiometric oxides: Na₂O, MgO, Al₂O₃, SiO₂, K₂O, CaO, TiO₂, MnO, FeO, SO₂, and P₂O₅. The surface of each SI was measured with the ImageJ software environment for image analysis.^[24]

Slag Inclusions Classification

SI entrapped in the metal matrix may have originated during the smelting and smithing processes. Key technological information about the reduction and smithing processes, e.g., steelmaking method identification and forging strategy reconstruction, can only be extracted from SI derived from the smelting and forging stages, respectively. For this reason, it is crucial to carefully examine their formation origin.^[17,20] The SI classification protocols introduced by Dillmann and L’Héritier^[25] and Charlton et al.^[26] were adopted to achieve this goal, as in previous Authors’ works.^[16,23] Both these approaches model SI as multicomponent oxide systems into which two main oxide types can be distinguished: reduced compounds (RCs) and non-reduced compounds (NRCs). It is worth noting that only NRCs (e.g., MgO, Al₂O₃, and K₂O) should be considered to classify the SI. RCs (e.g., FeO, and P₂O₅) can widely vary even inside SI with the same origin. As suggested by Dillmann and L’Héritier,^[25] smelting-related SI are featured by constant NRCs ratios (e.g., K₂O–CaO, MgO–Al₂O₃, Al₂O₃-SiO₂ ratios) in contrast to the SI originated during the forging process. Moreover, Charlton's method is based on multivariate statistical analysis, i.e., Principal Component Analysis (PCA) and Hierarchical Cluster Analysis (HCA). To use these models, the raw elemental chemical data of each analyzed SI were converted to oxides weight percentages (wt pct) by stoichiometric calculations. According to Charlton et al.,^[26] the main NRCs (i.e., MgO, Al₂O₃, SiO₂, K₂O, CaO, and MnO) were transformed into subcompositional ratios, dividing the composition of each NRCs by the sum of all the other major NRCs. In addition, before performing PCA, the subcompositional dataset was pre-processed by standardization through Formula (1) to equalize the NRCs weight magnitude and variance:

$${X^*}_{ij} = \left[ {{X_{ij}}-E\left( {X_i} \right)} \right]/{S_j},$$

(1)

where X^*_ij is the value for the jth NRC of the ith SI, E(X_i) is the mean value of the jth NRC and S_j is the sample standard deviation of the jth NRC. PCA was then performed on the standardized subcompositional SI dataset (X^*_ij). HCA based on the Ward method with Euclidean distance was adopted to discriminate the main SI clusters. The data treatment was performed using R 3.6.1, Excel, and OriginLab software environment for statistical computing and graphics.^[27]

Results and Discussion

Metallographic Examination and Microhardness Measurements

Six sections (Figure 1(d)) sampled from the rapier tang (S1), ricasso (S2), ricasso-tang transition zone (S3), and blade (S4, S5, S6) were observed by optical microscope. The analysis of as-polished S1, S2, and S3 sections allowed to detect several SI bands arranged alongside the main hot plastic deformation directions (Figures 2 and 3). In particular, SI stringers are well-visible on the longitudinal section S3 due to an extensive material drawing-out elongation and possible folding (Figures 2(a) and (b)). Moreover, the presence of truncated near-surface SI alignments is compatible with metal removal by a post-forging grinding treatment aimed at improving the ricasso-tang transition shape (Figure 2(c)). A crack was identified at one lower corner of the tang section S1 (Figures 2(d) and (e)).

Fig. 2

(a) As-polished longitudinal section S3 (collage of micrographs with 50X magnification) with (b) details of SI stringers and (c) truncated SI alignments near the outer surface, (d and e) as-polished S1 transverse section with a crack at the section lower corner

Fig. 3

(a) As-polished S2 transverse section (collage of micrographs with 50X magnification), (b) SI alignments near a sectioned maker’s mark

On the as-polished ricasso transverse section S2, the SI lines near the maker’s marks follow the metal deformation imposed during the hot punching process (Figures 3(a) and (b)).

The SI on the blade transverse sections S4, S5, and S6 are generally less oriented than on S1 and S2 sections, and their mean size decreases in sections S5 (Figures 4(a) and (b)) and S6, which are close to the blade tip. These features are probably due to an intense shaping process by hot hammering, which is more pronounced near the pointed blade tip. Moreover, the occurrence of interrupted near-surface SI alignments on the section S5 may indicate that a grinding treatment was carried out in the upper blade portion probably during reworking operations (Figure 4(c)). For further information on the as-polished sections S4 and S6, please see Figures S-1 and S-2 in the electronic supplementary material.

Fig. 4

(a) As-polished S5 transverse section (collage of micrographs with 50X magnification), (b) scattered SI, and (c) truncated SI alignments near the outer surface

The etched partial section S1 of the rapier tang shows a heterogeneous microstructure (Figure 5(a)). Small ferritic-pearlitic regions with Widmanstätten ferrite (F* + P) (Figure 5(b)) are surrounded by elongated zones where pearlite with some traces of allotriomorphic ferrite at prior austenitic grain boundaries was observed (P + F) (Figure 5(c)). Moreover, the left side of the section is mainly characterized by a fully pearlitic microstructure (P) (Figure 5(d)). The observed distribution of microstructural constituents is compatible with the assembly by hammer welding of different steel bars. In addition, the crack positioned at one lower corner of the section may have originated from metal layers separation.

Fig. 5

(a) Distribution map of the main microstructural constituents and features observed in the S1 section and (b through d) some representative 200X magnification optical micrographs (F ferrite, F* Widmanstätten ferrite, P pearlite, SI slag inclusions)

After chemical etching, section S2 of the rapier ricasso exhibits a heterogeneous banded microstructure (Figure 6(a)). In particular, the section is featured by an alternation of both continuous and discontinuous layers, which are constituted by pearlite (P) (Figure 6(b)), pearlite with traces of allotriomorphic ferrite (P + F) (Figure 6(c)), and ferrite with a small fraction of pearlite (F + P) (Figure 6(d)). In most of the section outer regions, the microstructure is fully pearlitic (P). It is worth noting that the described distribution pattern of microstructural constituents may derive from a billet composed of low and high-carbon steels.^[11]

Fig. 6

(a) Distribution map of the main microstructural constituents and features observed in the S2 section and (b through d) some representative 200X magnification optical micrographs (F ferrite, P pearlite, SI slag inclusions)

Consistently with the previous results, a banded distribution of microstructural constituents was also clearly detected on the etched longitudinal section S3 taken from the ricasso-tang transition zone (Figure 7(a)). Six almost entirely pearlitic layers with very small traces of ferrite (P) (Figure 7(b)) alternate with three mainly pearlitic layers with allotriomorphic ferrite (P + F) (Figure 7(c)) and two ferritic-pearlitic bands feature by allotriomorphic and Widmanstätten ferrite (F* + P) (Figure 7(d)). The interruption of the outer layers in conjunction with the section narrowing on its right side is probably caused by material removal by post-forging grinding operations (Figure 7(a)).

Fig. 7

(a) Distribution map of the main microstructural constituents observed in the S3 section and (b through d) some representative 200× magnification optical micrographs (F  ferrite, F* Widmanstätten ferrite, P  pearlite, SI slag inclusions)

The etched blade transverse section S4 is featured by a heterogeneous microstructure (Figure 8(a)). Discontinuous ferritic-pearlitic (F + P) (Figure 8(b)) and pearlitic-ferritic (P + F) (Figure 8(c)) bands were detected. In some regions, pearlite is partially spheroidized (P** + F) probably caused by intercritical annealing between Ac₁ and Ac₃ temperatures (Figure 8(d), high magnification of spheroidized pearlite in Figure 9). Small martensitic islands were also observed within the pearlitic-ferritic bands (P** + M + F) near the blade cutting edges (Figure 8(e)). This microstructure could be originated from rapid cooling after partial austenitization.

Fig. 8

(a) Distribution map of the main microstructural constituents and features observed in the S4 section and (b through e) some representative 200X magnification optical micrographs (F ferrite, F*  Widmanstätten ferrite, P pearlite, P** spheroidized pearlite, M  martensite, SI  slag inclusions)

Fig. 9

Spheroidized pearlite (500× magnification)

The distribution map of the main microstructural constituents for transverse section S5 taken near the blade tip is shown in Figure 10(a). The outer region is characterized by a martensitic microstructure (M) (Figure 10(b), high magnification in Figure 11). In addition, nodular pearlitic colonies at prior austenitic grain boundaries in a martensitic matrix (P* + M) (Figure 10(c), high magnification in Figure 12) were mainly detected in the inner part of the section.

Fig. 10

(a) Distribution map of the main microstructural constituents and features observed in the S5 section and (b through c) some representative 200X magnification optical micrographs (P* nodular pearlitic colonies, M  martensite, SI slag inclusions)

Fig. 11

Martensite (500× magnification)

Fig. 12

Nodular pearlitic colonies in a martensitic matrix (500× magnification)

The etched transverse section S6 exhibits similar microstructures (Figure 13(a)). Most of the section is marked by a fully martensitic microstructure (M) (Figure 13(b)). Martensite and nodular pearlitic colonies at prior austenitic grain boundaries (P* + M) (Figure 13(c)) were also detected mainly in the section core. In this zone, it was observed that the distribution pattern of these microstructural constituents exhibited a layered appearance (Figure 13(d)).

Fig. 13

(a) Distribution map of the main microstructural constituents and features observed in the S6 section, (b through c) some representative optical micrographs at 200×, and (d) 50X magnification optical micrographs (P* nodular pearlitic colonies, M  martensite, SI slag inclusions)

The microstructures observed in sections S5 and S6 taken from the blade tip may indicate that after austenitization the cooling rate was insufficient for a complete transformation of austenite to martensite. On the other hand, the presence of martensite in the outer zones indicates that a quenching heat treatment was executed. As previously mentioned in the Material and Methods section, the upper blade portion was probably subjected to reworking operations. Therefore, it is unclear if the microstructures observed in sections S5 and S6 derive from the original heat treatment or the subsequent reworking process.

Vickers microhardness measurements were performed for all six sections. The distribution map of microhardness indentations and measurement data for section S4 are shown in Figures 14(a) through (c), respectively. More details on the other sections investigated and an overall summary of the microhardness measurements are reported in the supplementary electronic material, in particular Figures S-3, S-4, S-5, S-6, S-7, and Table S-I.

Fig. 14

(a) Distribution map of Vickers microhardness indentations on S4 section and resulting measurements along (b) horizontal (L12) and (c) vertical (L13, L14, L15) lines (F  ferrite, F* Widmanstätten ferrite, P pearlite, P**  spheroidized pearlite, M martensite, SI slag inclusions)

Taking into consideration all the results achieved, the microhardness values in the fully pearlitic (P), pearlitic-ferritic (P + F), and ferritic-pearlitic (F + P) zones vary in the ranges 199-322 HV_0.3, 138-266 HV_0.3, and 113-217 HV_0.3, respectively. In the pearlitic-ferritic regions with partially spheroidized pearlite (P** + F), the microhardness measurements are in the interval 192–253 HV_0.3 and increase in the range 316–340 HV_0.3 where a small fraction of martensite is also present (P** + M + F). In the zones featured by a fully martensitic microstructure (M) and martensite with nodular pearlitic colonies (P* + M), the microhardness values vary in the ranges 530–635 HV_0.3 and 391–576 HV_0.3, respectively.

Metallographic analyses and Vickers microhardness measurements allowed to reconstruct the forging strategy adopted to produce the rapier sword (except for the guard which was not present).

The rapier blade, ricasso, and tang were shaped starting from a layered billet which was obtained by combining via hammer-welding different hypoeutectoid and near-eutectoid carbon steel bars. A near-eutectoid carbon steel bar was probably wrapped around and hammer-welded to a core consisting of five hypoeutectoid carbon steel bars alternating with four near-eutectoid carbon steel bars. This composite core could be produced by combining nine different bars or by a single folding of a billet consisting of five distinct bars (Figure 15(a)). The composite billet was subsequently drawn out by hot hammering (Figure 15(b)). After this stage, the rapier blade, ricasso, and tang were shaped and the Caino maker’s marks were hot punched on the ricasso (Figure 15(c)). The presence of martensite with a small fraction of nodular pearlite in the near-tip blade sections S5 and S6 suggest that a slack-quenching heat treatment was executed during the original forging process or subsequent reworking operations (Figure 15(d)). After slack-quenching, the rapier was ground to improve the shape of the ricasso-tang transition zone and it was finally polished (Figure 15(e)).

Fig. 15

Main steps of the rapier manufacturing process: (a) fabrication of a composite billet, (b) lump drawing-out, (c) tang, ricasso, and blade shaping by hot hammering, (d) blade slack-quenching, and (e) rapier polishing

SEM/EDS Analysis

The chemical composition of 96 SI was determined by SEM/EDS. Detailed chemical data for each SI, as numbered in Figures 16(f) and (g), are provided in the supplementary Tables S-II and S-III. The outcomes of PCA coupled with HCA for each SI are shown in Figure 16. The non-metallic inclusions were grouped into four main families on PC space (Figure 16(a)) by truncating the dendrogram plot at a height of 18 (supplementary Figure S-8).

Fig. 16

(a through e) Projection of SI chemistry on PC1–PC2 plane (solid and open versions of the same symbol refer to S3 and S4 sections, respectively) with loading vectors plot (the percentage of variation explained by each PC is indicated in brackets) and position of grouped and numbered SI on microstructural constituents maps of (f) longitudinal section S3 and (g) transverse section S4 (F ferrite, F* Widmanstätten ferrite, P pearlite, P** spheroidized pearlite, M martensite, SI slag inclusions)

The average oxides percentages weighted on the SI surfaces were computed for each SI group by Formula (2):

$${\rm{Pct}}\;{\rm{Oxide}}_j^* = \sum _{{\rm{i}} = 1}^{{n_j}}{\rm{Pct}}\;{\rm{Oxid}}{{\rm{e}}_{ij}} \times \frac{{{s_{ij}}}}{{{s_{Tj}}}}$$

(2)

where pct Oxide^*_j refers to the mean weighted percentage of a given oxide for the jth group, pct Oxide_ij is the oxide weight percentage for the ith SI in the jth cluster, S_ij correspond to the ith SI surface in the jth group and S_Tj is the total surface for the jth group. The average oxides weighted percentages and SI area for each SI group are listed in Table I.

Table I Average Weighted Oxides Percentages (Weight Percent) and SI Area for Each SI Group

Group 1 (green circles) comprises most of the analyzed SI and it is mainly located near the origin of the PC1–PC2 plane (Figures 16(a) and (b)). These SI are marked by a significant amount of MnO and probably derived from the finery process, i.e., decarburation by oxidation of Mn-rich cast iron (Table I). In fact, manganese is associated with the iron ore gangue and it can be reduced and dissolved as an alloying element into the cast iron produced in the shaft furnace.^[28] During the subsequent finery stage, most of the manganese dissolved in the cast iron is oxidized and can contribute to the slag formation as MnO.^[28] Group 2 (blue squares) is located at the upper extreme of the positively correlated K₂O–Al₂O₃ pair as graphically indicated by the small angle between these two loading vectors in the PC1–PC2 plane (Figures 16(a) and (c)). The main feature of group 2 chemistry is the high amount of Al₂O₃ (Table I). Following the classification model proposed by Charlton et al.,^[26] these results suggest that the chemical composition of group 2 SI was influenced by clay contaminations more than the other groups. Group 3 (red triangles) appears well-separated along the compositional variable SiO₂ in the PC1–PC2 plane (Figures 16(a) and (d)). Moreover, group 3 is characterized by a higher SiO₂ content and lower MgO and MnO amounts compared to the other groups (Table I). Accordingly to these observations, SI of group 3 can be classified as forging-related SI possibly formed after the chemical interaction between sand flux and iron oxide scale during hammer-welding operations. Group 4 (yellow inverted triangles) chemistry is marked by a high amount of MnO (Figures 16(a), (e), and Table I). Similarly to group 1, these SI probably originated during the finery stage. It should be noted that it is not possible to identify a clear distribution pattern for the SI groups on the microstructural constituents maps of sections S3 and S4 (Figures 16(f) and (g)). In particular, the finery-related SI of group 1 are scattered throughout all the zones of sections S3 and S4 (Figures 16(f) and (g)). This result is consistent with the hypothesis that the rapier was manufactured by a layered billet composed of low and high-carbon steels with a similar technological origin. A mineralogical investigation was carried out by SEM/EDS analysis to extend the SI characterization. Three main SI petrological classes were detected, i.e., two-phase Fe-rich mangano-wüstite-olivine SI (Figure 17(a)), two-phase Mn-rich mangano-wüstite-olivine SI (Figure 17(b)), and single-phase glassy SI (Figure 17(c)). Example chemical data for the main mineralogical phases is shown in Figure 17 and Table II. The Fe-rich mangano-wüstite-olivine SI are mainly included in groups 1 and 4. The Mn-rich mangano-wüstite-olivine SI only belongs to group 4. Moreover, the single-phase glassy I is generally characterized by a low iron oxide content and is greatly predominant in groups 2 and 3.

Fig. 17

SEM micrographs of the three main SI mineralogical types: (a) two-phase Fe-rich mangano-wüstite (grey)-olivine (dark grey) SI, (b) two-phase Mn-rich mangano-wüstite (grey)-olivine (dark grey) SI, and (c) single-phase glassy SI. The letters A, B, C, D, and E indicate the zones chemically analyzed by SEM/EDS

Table II SEM/EDS Microchemical Analyses of the Areas Highlighted in Figure 17

FeO–SiO₂ and NRCs bivariate scatterplots were also examined (Figures 18(a) and (b)) to deepen the SI classification. As an example, the SiO₂–Al₂O₃ biplot was shown in Figure 18(b). It was noted that the point clouds corresponding to groups 1 and 2, approximately follow the same linear trend because of similar NRC ratios (Figures 18 (a) and (b)). This feature confirms that these groups include finery-derived SI with the same origin. On the other hand, group 3 follows a linear trend with a different slope with respect to groups 1 and 2 (Figures 18(a) and (b)). This difference is mainly due to the higher SiO₂ content observed in the SI of group 2 compared to the other groups. These results confirm that group 2 has a different origin than the finery-related groups 1 and 2. In particular, group 2 probably contains SI formed during the forging process. The linear trend observed in the FeO–SiO₂ bivariate scatterplot for group 4 significantly deviates from that of the other groups because of the high FeO–SiO₂ ratio (Figure 18(a)). In fact, the SI of group 4 is mainly constituted by Fe-rich mangano-wüstite and Mn-rich mangano-wustite phases with low amounts of SiO₂ and Al₂O₃ (Figure 18(b)).

Fig. 18

Scatterplots: (a) FeO–SiO₂, (b) Al₂O₃–SiO₂, and (c) Al₂O₃– FeO–SiO₂ (solid and open versions of the same symbol refer to S3 and S4 section, respectively)

To summarize the results illustrated in the bivariate scatterplots FeO–SiO₂ and Al₂O₃–SiO₂, each SI group was also projected in the ternary plot Al₂O₃–FeO–SiO₂ (Figure 18(c)).

Estimation of Finery and Forging Temperatures

The finery and forging temperatures were assessed by estimating the liquidus temperature of finery and forging-related SI. The liquidus temperature was computed by MELTS-rhyolite v.1.0.2 software package starting from the average chemical composition of SI groups 1 and 3 (Table I), which include finery and forging-derived SI, respectively.^[29,30] The thermodynamic modeling of liquidus temperature was performed assuming a pressure of 1 bar (0.1 MPa) and iron-wüstite (IW) oxygen fugacity buffer (input files are provided in the electronic supplementary material). The results for SI groups 1 and 3 are 1270 °C and 1160 °C. These values can be considered reasonable estimations of the lower limits of the maximum temperatures reached during the finery and forging processes, respectively.

Comparison with a “Falchion” Sword Manufactured in Caino in the Same Historical Period

The compatibility between the finery-related SI of the rapier and a previously studied falchion sword also produced in Caino in the 17th century was assessed by focusing on their NRCs ratios.^[1] Compositional data of finery-related SI from the falchion sword blade were taken from Tonelli et al. ^[1] Finery-derived SI of the two objects was visualized in the SiO₂–K₂O, SiO₂–CaO, and SiO₂–Al₂O₃ bivariate scatterplots (Figure 19).

Fig. 19

Bivariate scatterplots: (a) SiO₂– K₂O, (b) SiO₂–CaO, and (c) SiO₂ –Al₂O₃ (green circles and grey diamond symbols correspond to SI group 1 of the rapier and finery-related SI of the falchion (F), respectively) (Color figure online)

The average oxides weighted percentages for SI group 1 of the rapier and finery-related SI of the falchion (F) are summarized in Table III.

Table III Average Weighted Oxides Percentages (Weight Percent) of Finery-related SI of Rapier (Group 1) and Falchion swords

In the bivariate scatterplots SiO₂–K₂O (Figure 19(a)), SiO₂–CaO (Figure 19(b)), and SiO₂–Al₂O₃ (Figure 19(c)) point clouds corresponding to finery-related SI of rapier (green circles) and falchion (grey diamond symbols) roughly follow linear trends and overlap each other due to similar NRCs ratios. It is worth noting that NRCs ratios of finery-derived SI strictly depend on both the raw materials used during the smelting and finery process (e.g., iron ore, charcoal, and refractory materials) and the steelmaking technology adopted.^[25] Therefore, the observed compatibility supports the hypothesis that the rapier and falchion swords were produced in the same technological context, i.e., the Caino sword-making workshops. Furthermore, NRCs ratios measured for finery-derived SI of the rapier and the falchion provide a specific chemical signature for Caino swords that can be used to test provenance hypotheses for unknown-origin swords.

Conclusions

The focus of this research is the archaeometallurgical study of a 16th–17th century rapier sword (missing the guard) manufactured in Caino (Brescia, northern Italy). An inverse engineering approach based on metallography and slag inclusions (SI) analysis was carried out to deeply investigate the manufacturing strategy adopted and extract information about smelting and forging processes. The results of microstructural analysis and Vickers microhardness measurements point out that the rapier was forged by assembling via hammer welding different steel bars. A near-eutectoid carbon steel bar was wrapped around a billet constituted by five hypoeutectoid steel bars alternated with four near-eutectoid carbon steel bars. This banded core could be produced by combining nine different bars or by a single folding of a billet consisting of three hypoeutectoid and two near-eutectoid steel bars. Both these forging strategies reflect the intention to produce a highly functional composite blade characterized by good flexibility and fracture toughness. The rapier was forged by drawing out and shaping the composite billet through a hot hammering process. Subsequently, the rapier blade was heat treated by slack-quenching to increase its hardness, especially near the tip improving the thrusting performance. After the slack-quenching, the geometry of the ricasso-tang transition zone was improved by grinding. The SEM/EDS analysis of SI compositional data via multivariate statistical techniques, i.e., PCA and HCA allowed to distinguish and classify four SI groups. Finery-derived SI are greatly spread throughout the analyzed blade sections suggesting that the rapier was forged with steel bars characterized by similar technological origin. The temperatures reached during both the finery and forging processes were estimated by thermodynamic modeling of the liquidus temperatures of finery and forging-derived SI with the software MELTS. The lower limits of the maximum finery and forging temperatures are 1270 °C and 1160 °C, respectively. The chemical composition of SI from the rapier and a falchion sword manufactured in Caino in the same historical period was compared focusing on NRCs ratios. It was proved that the two swords featured compatible chemical signatures (i.e., similar NRCs ratios) which are probably specific to the Caino sword-making workshops and can be used to test provenance hypotheses for unknown-origin swords in future works.