Journal of Archaeology in the Low Countries 3-1 (November 2011)Janneke Nienhuis; Jilt Sietsma; Stijn Arnoldussen: The Production Process and Potential Usage of Bronze Geistingen Axes

TOC | article | PDF

3 Results and discussion

3.1 Composition

next section

The results concerning the composition of the two Geistingen axes, determined with NRCA (Postma et al. 2005a; 2011/in press; personal communication 2010), XRF and EPMA are summarised in table 1 and illustrated in figure 4.

Table 1 Average compositions (in at%) of two Geistingen axes determined with NRCA, XRF and EPMA.

Axe		Pb	Sn	Sb	As	Fe	Ni	S	Ag	Cu (by balance)

AC20	Mean	0.21	9	1.4	0.72	0.4	2	0.95	0.378	85
	error	0.01	2	0.2	0.01	0.2	1	0.06	0.003	3

BH76	Mean	0.4	0.62	4	2.9	0.2	6	1.26	0.605	84
	error	0.3	0.08	1	0.1	0.1	2	0.09	0.004	4

Figure 4 Compositions (in at%) of two Geistingen axes determined with NRCA, XRF and EPMA.

Table 1 and figure 4 show the relevance of using three complementary techniques to determine the chemical compositions. NRCA is able to quantify silver while EPMA is in this case the only technique that measures sulphur. Both elements are important in deducing the raw material used for production (see sections 3.2 and 3.3). NRCA, XRF and EPMA all confirm the presence of the main alloying elements tin, antimony and nickel. The main differences are seen in the number of elements and their percentages, which originate from the technical specifications and limitations of the techniques used (see section 2.2).

Due to systematic instrumental uncertainties for each technique that cannot be accurately determined, a larger variation in the concentration values is found than expected on the basis of the statistical errors per technique. Therefore, for every element detected with more than one technique, the arithmetic mean is calculated and the corresponding error is calculated by using the least squares method. The use of these complementary analytical methods thus provides a more accurate determination of the composition than by using only one.

The results lead to the conclusion that AC20 can be identified as a binary copper-tin bronze with 9 at% tin, while BH76 is essentially a ternary copper-antimony-nickel alloy with 4 at% antimony and 6 at% nickel.

Comparing these numbers with known compositions from functional and contemporary axes can highlight the differences and help form a hypothesis on the function of Geistingen axes. Only a modest corpus of metal analyses on Late Bronze Age axe types is available for the area under study (table 2). The corpus mostly consists of axes of the Plainseau and Geistingen types and very few other types (Niedermaas, Armorican, Wesseling; Butler 2003/2003) are represented. Moreover, the quality of the analytical techniques is varied (chemical extraction versus XRF/EPMA). Nonetheless, the data allow characterising the composition of certain axe types. The (functional) axes of the Plainseau-type can be roughly characterised as copper containing tin (2-10 at%) and lead (2-7 at%) with virtually no other elements present (Van Impe 1994; Wouters 1994, 42). The Geistingen axes studied as part of the SAM series can be characterised as copper with 0.5-5.8 at% tin, 1.8-2.9 at% arsenic, 0.2-2.3 at% antimony and low (

Table 2 Metal composition of Late Bronze Age socketed axe types, measured using different analytical techniques.

As		Sb		Ni		Fe		References		Remarks
0,036		0,001		0,001		0,001		Wouters 1994, 41 tab. 1;		Pres. Niedermaas-type axe; average of 5 measurements
								Van Impe & Creemers 1993;
								Butler & Steegstra 2002/2003, 269-271.
0,031		0,001		0,08		0,001		Wouters 1994, 41 tab. 1;		Pres. Niedermaas-type axe; average of 6 measurements
								Van Impe & Creemers 1993;
								Butler & Steegstra 2002/2003, 269-271.
0,054		0,001		0,001		0,001		Wouters 1994, 41 tab. 1;		Pres. Niedermaas-type axe; average of 5 measurements
								Van Impe & Creemers 1993;
								Butler & Steegstra 2002/2003, 269-271.
0,009		0,001		0,002		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,007		0,001		0,057		0,082		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,007		0,001		0,001		0,89		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,01		0,001		0,022		1,51		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,009		0,001		0,02		0,35		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,009		0,001		0,035		0,19		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,001		0,001		0,006		0,16		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,009		0,001		0,006		0,089		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,008		0,001		0,012		0,007		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,008		0,001		0,001		0,39		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,001		0,001		0,06		0,31		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,01		0,001		0,02		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,01		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,011		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,003		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,011		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,011		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,014		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,01		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,01		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,01		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,012		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,01		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,011		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,009		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,01		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,01		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,01		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,01		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,01		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,01		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,011		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,01		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,01		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,01		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,011		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,001		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,012		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,012		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,011		0,001		0,001		0,43		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,009		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type; average of 2 measurements
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,011		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,01		0,001		0,001		0,56		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,009		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,008		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,008		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,009		0,001		0,001		0,001		Wouters 1994, 42 tab. 2;		Axe of 'Plainseau' type
								Van Impe 1994;
								Butler & Steegstra 2002/2003, 280-282
0,01		0,001		0,001		0,001		Wouters 1994, 42 tab. 2		Axe of 'Plainseau' type; pres. part of Heppeneert hoard
0,01		0,001		0,001		0,001		Wouters 1994, 42 tab. 2		Axe of 'Plainseau' type; pres. part of Heppeneert hoard
n.a.		n.a.		n.a.		n.a.		Van Impe 1995/1996, 31		Axe of 'Plainseau' type
n.a.		n.a.		n.a.		n.a.		Van Impe 1995/1996, 31		Axe of 'Plainseau' type
0,08		0,08		0,05		n.a.		Verlaeckt 1996, 88-89		Axe of 'Plainseau' type; Ag 0,07
1,18		0,56		0,07		n.a.		Verlaeckt 1996, 90		Axe of 'Plainseau' type; Bi 0,1
0,18		n.a.		0,25		0,18		Verlaeckt 1996, 90		Axe of 'Plainseau' type; Bi 0,1
0,06		0,09		0,02		n.a.		Verlaeckt 1996, 109		Axe of 'Plainseau' type; Bi 0,002; Ag 0,05
0,35		<0,62		0,12		0,39		Verlaeckt 1996, 119		Axe of 'Plainseau' type?; Bi 0,06
0,316		0,596		0,161		0,96		Verlaeckt 1996, 119		Axe of 'Plainseau' type?; Bi 0,046
0,07		0,15		0,1		0,08		Verlaeckt 1996, 120		Axe of 'Plainseau' type?; Bi 0,004
1,5		0,7		0,25		0,01		Verlaeckt 1996, 120		Axe of 'Plainseau' type?; Bi 0,015
1		0,8		0,25		0,002		Verlaeckt 1996, 103		Local type (cast in Heusden mould?)
0,535		1,26		1		0,072		Verlaeckt 1996, 90		Axe of type Sompting
2		1		0,25		0,035		Verlaeckt 1996, 106		Axe related to type Sompting ('Atlantic')
1		1		0,35		0,1		Verlaeckt 1996, 106		Axe related to type Sompting ('Atlantic')
0,55		0,15		0,06		0,03		Verlaeckt 1996, 120		Axe related to type Sompting ('Atlantic'); Bi 0,01
0,56		n.a.		0,66		0,23		Verlaeckt 1996, 101		Atlantic'?
3,3		4,1		1,9		n.a.		Kibbert 1984, 214;		Axe of type Geistingen; Co 0,06
								Junghans et al.1960-1974 (SAM 16559)
0,55		2,6		0,5		n.a.		Kibbert 1984, 214;		Axe of type Geistingen, Co 0,04
								Junghans et al.1960-1974 (SAM 16552)
2,1		0,046		0,03		n.a.		Kibbert 1984, 214;		Axe of type Geistingen; Co 0,08
								Junghans et al.1960-1974 (SAM 21612)
1,7		2,7		1,2		n.a.		Kibbert 1984, 214;		Axe of type Geistingen, Co 0,26
								Junghans et al.1960-1974 (SAM 16560)
1,07		2,17		1,97		n.a.		Postma et al. in press table 1;		Axe of type Geistingen
								Butler & Steegstra 2002/2003, 305
1		3,62		4,47		n.a.		Postma et al. in press table 1;		Axe of type Geistingen
								Butler & Steegstra 2002/2003, 305
2,36		4,27		7,58		n.a.		Postma et al. in press table 1;		Axe of type Geistingen
								Wielockx 1986
2,69		4,47		7,73		n.a.		Postma et al. in press table 1;		Axe of type Geistingen
								Wielockx 1986
3,09		4,65		6,59		n.a.		Postma et al. in press table 1;		Axe of type Geistingen; XRF&EPMA
								Wielockx 1986		measurements done (this study)
3,05		5,54		7,83		n.a.		Postma et al. in press table 1;		Axe of type Geistingen
								Wielockx 1986
1,97		3,21		4,3		n.a.		Postma et al. in press table 1		Axe of type Geistingen
2,39		2,66		3,95		n.a.		Postma et al. in press table 1		Axe of type Geistingen
1,24		2,4		6,53		n.a.		Postma et al. in press table 1;		Axe of type Geistingen
								Butler & Steegstra 2002/2003, 309
1,19		1,76		1,5		n.a.		Postma et al. in press table 1;		Axe of type Geistingen
								Butler & Steegstra 2002/2003, 309
0,62		0,87		<0.90		n.a.		Postma et al. in press table 1;		Axe of type Geistingen
								Butler & Steegstra 2002/2003, 309
0,82		2,21		<2.5		n.a.		Postma et al. in press table 1;		Axe of type Geistingen; XRF&EPMA
								Butler & Steegstra 2002/2003, 305		measurements done (this study)
traces		n.a.		n.a.		n.a.		Jacobsen 1904, 18		Axe of 'Plainseau' type; Ag 0,17
traces		n.a.		n.a.		1,5		Jacobsen 1904, 18		Axe of unclear (Niedermaas) type; Ag 0,19
n.a.		n.a.		n.a.		0,77		Jacobsen 1904, 20		Axe of 'Plainseau' type
n.a.		n.a.		n.a.		1,05		Jacobsen 1904, 24		Axe of Armorican type?
n.a.		n.a.		n.a.		1,05		Jacobsen 1904, 39		Axe of Armorican type?; S 0,275
2,35		3,4		0,19		1,99		Verlaeckt 1996, 109		Axe of 'Plainseau' / 'Atlantic' type?; Bi 0,11 (Armorican?)
0,15		0,09		0,02		0,08		Verlaeckt 1996, 87-88		Axe of Armorican type (Tréhou); Ag 0,06
0,46		0,11		0,02		0,16		Verlaeckt 1996, 93		Axe of Armorican type (Tréhou); Bi 0,71
0,39		0,37		0,04		n.a.		Verlaeckt 1996, 89		axe of Armorican type (Plurien); Bi 0,32
0,48		0,29		0,03		0,07		Verlaeckt 1996, 89		Axe of Armorican type (Plurien); Bi 0,35
0,59		<0,17		0,02		0,54		Verlaeckt 1996, 97		Axe of Armorican type (Plurien); Bi 0,22
0,53		0,23		0,06		0,11		Verlaeckt 1996, 119		Axe of Armorican type (Plurien); Bi 0,19
0,29		n.a.		0,03		0,11		Verlaeckt 1996, 101		Axe of Armorican type (Plurien); Bi 1,01
0,31		n.a.		0,05		0,18		Verlaeckt 1996, 89		Axe of Armorican type (Couville); Bi 0,32
traces		n.a.		n.a.		0,35		Jacobsen 1904, 43		Axe of type 'winged with biconical collar'
n.a.		n.a.		n.a.		0,315		Jacobsen 1904, 44		Axe of linear faceted type
n.a.		n.a.		n.a.		0,507		Jacobsen 1904, 44-45		Axe of type Wesseling
1,1		1,9		0,55		n.a.		Kibbert 1984, 168-170; 214		Axe of type Amelsbüren; Bi 0,027; Ag 0, 58
0,29		0,3		0,14		n.a.		Kibbert 1984, 168-170; 214		Axe of type Amelsbüren; Bi 0,033; Ag 0,22
0,42		0,46		0,17		n.a.		Kibbert 1984, 168-170; 214		Axe of type Amelsbüren; Bi 0,025; Ag 0,3
0,33		0,53		0,13		n.a.		Kibbert 1984, 168-170; 214		Axe of type Amelsbüren; Bi 0,022; Ag 0,36
0,95	.	6,4	.	0,65	.	0,01	.	Craddock 1979, 380 table 2	.	Socketed looped axe of continental type

When drawing comparisons with the data available for axe (type)s known to be functional (see table 2), Geistingen axes differ in a number of ways, although there are also some similarities. First, the tin content of the Geistingen axes falls in the range present in several Plainseau and Armorican axes, but since there is such a wide margin (5 at% antimony or >4 at% arsenic, is the result of the re-melting of scrap bronze (Curtis & Kruszyński 2002, 91). Axe BH76 from the Geistingen depot shows such a combination (see table 1). However, sometimes antimony is seen as an intentional replacement for tin (Craddock 1979, 380). This indicates that the combination of tin and antimony seen in axe BH76 is the result of intentional addition of only antimony. This however does not explain the high nickel and arsenic content found, so it is argued that re-melting in this case seems more likely because of the additional elements present. When looking at the arsenic content of the axes shown in table 2, a dichotomy is seen: the Geistingen axes contain more than 1 at% (up to 3 at%) arsenic, while the content of all other axes measured is (much) less than 1 at%. Next to making the copper appear more silvery, addition of more than 2 at% arsenic will make the object harder without loss of integrity (Lechtman 1996, 506; 509; Junk 2003, 23; 24). The same arguments apply to nickel. Furthermore, the Geistingen axes contain 1-7 at% nickel, while other axes contain (much) less than 1 at%. The effect of nickel as an alloying element is comparable to that of arsenic, resulting in good cold and hot working properties of the alloy (Cheng & Schwitter 1957, 351). The amount of lead found in the Geistingen axes is lower than 1 at% and the majority of the axes listed in table 2 contain more than 1 at%. The addition of less than 1 at% lead to copper will lead to increased fluidity of the melt, making the casting of hollow shapes more easy (Craddock 1979, 383). Adding more than 1 at% will significantly reduce the melting point of copper, with a reduction of almost 200 °C for a bronze containing 18 at% lead (Craddock 1979, 383). However, upon solidification of a leaded copper alloy, the lead will form insoluble globules that are dispersed throughout the microstructures. Increasing the lead content will increase the size of these globules, which will cause major areas of weakness. Therefore, bronze objects containing large amounts (>1 at%) of lead might be cast as ingots and not intended to be functional objects (Craddock 1979, 383). In case of the Geistingen axes, lead is not intentionally added to make the axe unusable since its concentration is below 1 at%. Its presence can however be attributed to the reduction of the melting point and the increase of fluidity. Another element found in more than 1 at% in Geistingen axes but not, or in less than 1 at%, is antimony (see table 2). One exception is the socketed looped axe of continental type (Craddock 1979, 380). In general, the addition of antimony to copper will lower the melting temperature and in amounts over 0.5 at% will slightly increase toughness and ductility, as well as the hardness of copper (Junk 2003, 26; 28). However, in percentages over approximately 4% the result is a brittle alloy (Charles 1980 in Moorey 1999, 241). The Geistingen axes analysed in this research contain respectively 1.4 at% and 4 at% (see table 1), which means one of them is on the verge of being unusable because of brittleness. These numbers indicate that the antimony was either added on purpose (perhaps to make the axes less suitable as tools or weapons), or the presence of the element could not be controlled, i.e. the ore or scrap contained large amounts of antimony. In both cases, the result is that the finished Geistingen axe could not have been functional on compositional grounds as well as the low weight of the objects.

Therefore, the combination of >1 at% antimony, >1 at% Ni and >1 at% As in the Geistingen axes has no counterpart in the composition of any of the contemporary axes in table 2. This sets the Geistingen axes even further apart from other socketed axes.

3.2 Microstructure

The microstructure is the assembly of microscale crystals and inclusions in a material with characteristic features like grain dimensions, morphology and composition. This microstructure is the result of specific steps of the production process. Consequently, by studying the features of the two Geistingen axes, this process can be reconstructed.

3.2.1 Dendrites

The axes from Nijmegen and Tongeren both exhibit a microstructure (fig. 5) that contains tree-like structures with a rounded outline, called dendrites, which is typical for cast structures (Allen et al. 1970). The dendrites are intact and the intradendritic pores are spherical, suggesting that no mechanical working of the upper part of the axes has taken place during the production process of the axes.

Figure 5 Microstructures of AC20 (top) and BH76 (bottom), Scanning Electron Micrographs. The distance indicated as d2 is the secondary dendrite arm spacing.

The dendritic phase in axe AC20 is the α-phase of copper, with tin and nickel in solid solution. In axe BH76 the dendritic α-phase contains antimony, arsenic and nickel in a solid solution with copper.

Dendrites consist of main (or primary) branches which at a certain stage of solidification develop secondary side-branches. The number density of these secondary arms, quantified by the secondary dendrite arm spacing (d2, see fig. 4), depends on the thermal and compositional conditions during cooling. Dendrites in copper alloys typically form in the temperature range 800-500 °C, in which both the undercooling below the melting temperature and the atomic mobility are sufficiently high to form the solid-state microstructure (Porter & Easterling 2001). The cooling rate in this temperature range therefore governs the characteristics of the resulting microstructure. This means that d2 provides an indication for the average cooling rate that was applied and d2 can thus be used to determine the cooling rate during the production of the Geistingen axes. The values for d2 in this research have been determined using detailed scanning electron micrographs of the microstructure (fig. 5). For the AC20 axe, the average secondary dendrite arm spacing is determined as 9±1 µm and for BH76 it is 14±2 µm. The relation of this spacing with the cooling rate can be expressed in a general empirical form as (Miettinen 2001, Kumoto et al. 2002, Frame & Vandiver 2008):

where d2 is the secondary dendrite arm spacing in µm, α a factor dependent on the alloy composition, R the cooling rate in °C/s and n a constant with an empirical value of 0.3. Using the average calculated values α = 30 µm °Cⁿ/sⁿ for AC20 and 39 µm °Cⁿ/sⁿ for BH76 (using equations for α from Frame & Vandiver 2008), the average cooling rate is calculated to be approximately 39 °C/s for the Nijmegen axe and 22 °C/s for the Tongeren axe. The relative uncertainty in R is estimated at 30%. Since for both axes the temperature range of 800-500 °C is the range at which the secondary dendrite arm spacing evolves, this result implies that most features of the microstructure are formed in approximately 10 seconds, which is only a part of the time in which the whole axe is solidified.

3.2.2 Inclusions

Several types of inclusions can be found in the microstructures of the Geistingen axes.

Small silver and lead-antimony particles are sporadically seen throughout the structure of AC20. The most plausible explanation for their presence is that they are remnants of the ore since their presence is not uncommon in copper ores (Lindgren 1933). The melting point of pure silver (962 °C) is higher than that of bronze with a composition like AC20 (~950 °C) and of BH76 (~800 °C), and the silver particles thus remain solid in the liquid bronze. Lead has a very poor solid solubility in copper. Both cases result in these elements (either solid silver or liquid lead) being dragged along with the solidification front in the melt subsequently to be found in the interdendritic phase between the solid dendrites.

Both AC20 and BH76 display another distinct type of inclusion, namely copper-sulphide particles. Almost all of these particles in the two axes contain the same elements but their concentration differs: copper, sulphur, iron, oxygen and sometimes also tin, nickel and antimony. The majority of the particles can be identified as Cu₂ S with the aforementioned elements in solution. A difference seen between the two axes lies in the morphology of these inclusions (see fig. 5). They are spherical and star-shaped in AC20, while BH76 only contains star-shaped particles. Substructures have been qualitatively identified in the spherical particles, but further research is needed to provide more information about the origin of these structures. The different inclusion morphologies indicate that for both axes the temperature of the melt has been around the melting temperature (1130 °C) of copper-sulphide. The spherical particles have been completely molten and solidified into spheres, while the star-shaped particles represent particles of which only the outside has been molten. This implies that the temperature of the melt of both axes ranges between 1100 °C and 1150 °C, and the absence of spherical Cu₂ S-inclusions in BH76 indicates a somewhat lower temperature than for AC20. For both axes the applied temperature is higher than the melting temperatures of the main phases.

The copper-sulphide particles solidify during cooling at a higher temperature than the bronze does and can therefore end up between the dendrites, within the interdendritic phase as described above. This type of inclusion represents an important clue to the production of Geistingen axes since they can be identified as matte, a by-product of the smelting of copper-sulphide ore. It is therefore assumed that the presence of silver, lead-antimony and matte particles with their different morphologies can be attributed to the imperfect smelting of sulphidous copper ore, used for the initial melt that eventually formed the Geistingen axes after re-melting.

3.3 Production

When looking at macroscopic aspects of the two axes from Museum Het Valkhof (AC20) and the Gallo-Romeins Museum (BH76), it is evident that their size and socketed shape are almost the same. Casting seams are also present on both. These features suggest a casting technique which is the same for both axes and for which most probably a bi-valve mould has been used. The microstructural research presented in this paper provides valuable additional information.

A decisive indication for the casting process is the porous dendritic microstructure with different types of inclusions. It indicates that the bronze has been molten and relatively rapidly solidified. The average cooling rate is approximately 30 °C/s for both axes, which suggests that they have been cooled using the same method. This rapid cooling can be attained by quenching with water (Frame & Vandiver 2008).

Several materials can be used as mould material, of which clay (cf. Van Impe 1995/96, 20) and bronze are frequently used for bi-valve moulds (Kuijpers 2008). Clay has a very low thermal conductivity, which means that several steps need to be undertaken to cast a sound and strong bronze object. After casting the bronze in the pre-heated clay mould, the ensemble needs to be cooled with water for a few minutes. As soon as the bronze is solidified, the clay mould should be removed to allow a faster cooling of the object by submerging it in (cold) water again. This is a practice that calls for considerable skill and experience since the more rapid cooling should be applied during the formation of the microstructure, i.e. during the solidification process. Using a bronze mould (cf. Butler 1973, 322; 338; Butler & Steegstra 2005/06, 209) will eliminate the extra step of removing the mould and is therefore more likely to be applied. In addition, the identical shape of the various Geistingen axes supports the use of a re-usable bronze mould.

There are two options for the type of raw material used; either ore is smelted, or bronze scrap or exchanged ingots are re-melted. One indication for smelting lies in the presence of matte particles in the microstructure, which are remains of the roasting step in the smelting process. Since other impurities like titanium, silver, cobalt and zinc are present as well, no further refining steps have then been undertaken (Lindgren 1933, Craddock 1995, Figueiredo et al. 2009). The composition of the two bronzes indicates it is a possibility that both axes are smelted using the same type of ore with different ratios and additions of tin and/or nickel. Fahlerz is a sulphidic ore known to have been used during the Copper Age and the Early Bronze Age (Biek 1957, Tylecote et al. 1977) and contains elements like copper, arsenic and antimony (Gainov et al. 2008). If ore from the same source is used for both Geistingen axes, the Sb/As ratio should be the same in both cases. However, the Sb/As ratio is different for both objects (2.0 for AC20 and 1.5 for BH76). If nickel would have been added, AC20 and BH76 should contain at least 4 times the arsenic amount than they do now since nickel is usually associated with arsenic (Lindgren 1933). These two features indicate that it is unlikely that the same type of ore from the same source has been used for both axes. So either different types of ore have been used during smelting or (different pieces of) bronze scrap formed the raw material of these Geistingen axes. The lack of data in the literature on microstructures of re-melted bronze makes it difficult to reach a firm conclusion on the raw materials used. However, some information on the change in composition during re-melting is available. Two cycles of re-melting and hot-working in air can result in the loss of antimony and arsenic content until less than one per cent is left (Junk 2003, 26; 29; Tylecote 1977, 329). Since these two Geistingen axes still contain relatively high amounts of both elements (1-4 at%) it is assumed that extensive re-melting and hot-working have not taken place. This assumption is supported by the relatively large amount of copper sulphide particles in the microstructure.

From an archaeological perspective, it is very plausible that ingots or, even more likely, scrap was transported to areas themselves lacking metal ores. In the Netherlands, the famous Drouwen hoard (Butler 1986) found in Drenthe yielded 1.1 kg of bronze scrap of non-local (southern Scandinavian and central European) origin (Butler 1986, 138). The equally well-known Voorhout hoard (traditionally interpreted as a trader’s stock), was recently convincingly shown to have contained not pristine axes (as is to be expected for a ‘merchant’s hoard’) but old, worn and no longer functional axes (i.e. scrap) of French and English origin instead (Fontijn 2008, cf. Kuijpers 2008, 43; 74). Finds like Drouwen and Voorhout, together with the unlikelihood of transporting large quantities of ore, suggest that the Geistingen axes have probably been re-melted from different scrap resources.