JOM • May 200638

ArchaeotechnologyResearch Summary

The “sintering” technique to produce Au-Pt alloys in pre-Hispanic times in the South American region Tumaco-La Tolita (North Ecuador, South Colombia) is reviewed in this paper. Two Au-Pt alloys were made using small pieces of pure gold and platinum in an attempt to simulate some pre-Hispanic artisan alloys. A “sintering” process (950°C, cold and hot hammering) was done with metallurgical considerations. The results were analyzed with optical and scan-ning-electron microscopy and compared to the available microstructures of some similar pre-Hispanic composition alloys.

INTRODUCTION

Pre-hispanic (300 B.C.–500 A.D.) objects made of platinum with gold (Figure 1), from the “Tumaco-La Tolita” area (North Ecuador, South Colombia)1 are delicate and beautiful art master-pieces whose purpose was undoubtedly religious, sacred. From a technological point of view, we know that those objects were made in processes that were unable to reach the platinum melting tempera-ture (1,769°C). Although it is known that the pre-Hispanic artisans could melt gold, actually an impure native gold (below 1,064°C), they must have worked at lower temperatures, producing what we see now as “macroscopically homo-geneous” Pt-Au alloys. They surely employed “sintering” or, more precisely, solid-state diffusion alloying with mechanical work. Several authors have simulated some Au-Pt pre-Hispanic alloys in different ways. Bergsoe2 imitated alloys without reporting structures. He observed semi-fi nished pieces from Tumaco-La Tolita, deducing the possible technique used: “the small grains of Pt were mixed with a little gold dust and small portions placed

About the Pre-Hispanic Au-Pt “Sintering” Technique for Making Alloys

M. Noguez, R. García, G. Salas, T. Robert, and J. Ramírez

Figure 1. (a) A feline mask with removable parts from Tumaco-La Tolita, 590 B.C.–350 A.D. (Banco Central del Ecuador Collection), and (b) a scheme of the removable diadem showing some of the compositions.1

b

a

2006 May • JOM 39

1 5 9 13

10

20

30

Pt (%

)

40

50

60

70

80

90

100

17 21 25 29Point

— Start— Cycle 1— Cycle 2

33 37 41 45 49 53 57

— Cycle 3— Cycle 4— Cycle 5

upon a piece of wood charcoal.” In this technique, gold melts and coats platinum grains; a pasty sintered mass is formed that withstands hammer blows, espe-cially when hot. “By alternately forging and heating it is possible gradually to build up a homogeneous mixture.”2

Handwerker et al.3 prepared samples of 15Pt-85Au (all compositions are in weight percent) with gold and platinum wire and heated them to 1,100°C and 1,200°C for a maximum of 7 min. They concluded that the Pt-Au objects were never heated to 1,100°C and probably never contained a liquid phase. Bustamante et al.4 made several Au-Pt alloys (50% platinum by weight) from alluvial gold and platinum samples, at 1,100°C and different times (10 min., 2 h, and 13 h). Some of the samples were mechanically worked and annealed at 800°C. When they compared the result-ing microstructures with some pre-Columbian pieces, Bustamante et al. hypothesized that artisans used the gold melting temperature to sinter, suggesting that they could use ceramic molds as dies to get the desired shape when ham-mering the pieces. In this work, metallurgical concepts are reviewed relative to the thermome-chanical treatment used by pre-Hispanic artisans. The focus is on the impressive solid-state diffusion and mechanical work or sintering involved when pre-Hispanic artisans made Au-Pt alloys. The Au-Pt solid diffusion process is followed up to equilibrium, and com-parisons with the microstructures reported by archaeologists are made. Hot and cold artisan hammering between thermocycles, diffusion below 1,100°C, and fabrication of a 60% platinum alloy, which have not been previously reported, are discussed in this work. The two chosen “equilibrium” alloys were pro-duced at 950°C with tiny pieces of pure gold and platinum. The results were analyzed with optical and scanning-electron microscopy.

SELECTING PARAMETERS TO SIMULATE THE

PRE-HISPANIC ALLOYS

Alloy Composition Selection

There is a wide range of Pt-Au com-positions reported in the pre-Hispanic pieces. Usually they include other impor-

Figure 5. The percentage platinum content of alloy 1 at some points during the processing cycles.

Figure 4. An optical micro-structure of alloy 1; gold dendrites are shown.

Figure 3. A macroscopic view of alloy 1: platinum strips within the gold matrix.

F igu re 2 . An Au-P t phase diagram showing the composit ions and temperatures used.

JOM • May 200640

tant metals but in small proportions, soluble in Au (Cu, Ag) or Pt (Fe, plati-noids). Attention in this study is focused on the Pt-Au percentage. In the corre-sponding equilibrium phase diagram (Figure 2) there are two alloy groups: one rich in gold, with monophasic alloys up to the partial solubility limit, and the other with compositions inside the biphasic immiscibility area. All the reported metallographic work done on the pre-Hispanic alloys from the Tumaco-La Tolita region1–5 reveals monophasic macrostructures and bipha-sic microstructures, which must be explained. Unfortunately, not all of the reported pieces have been fully chemi-cally analyzed. In this work, two percentages were used: 12Pt-88Au and 60Pt-40Au. The former is taken from some of the pieces reported by Bergsoe2 in his tables of gold objects (i.e., hook for pendant, 12.3Pt-71.9Au, also reported in Handwerker et al.3 as 12.7Pt-74.6Au). This composition is found in the richer gold area of the phase diagram. The other composition is taken from the data reported for the Bergsoe2 platinum alloys (piece of plate,

57Pt-38Au). The simulated alloys are marked on Figure 2, showing the working tem-perature (950°C). According to the binary phase diagram, the 12Pt-88Au equilibrium alloy is monophasic at all temperatures below the solidus line. The inhomogeneity could arise just for an incomplete alloying. This is not the case for the 60Pt-40Au equilibrium alloy, which will be biphasic at all temperatures below the maximum solvus (1,260°C). The equilibrium composition of the two phases changes, as the solvus lines indi-cate, depending on temperature. At

950°C, it has two equilibrium phases: gold-rich (30Pt-70Au) and platinum-rich (91Pt-8Au). The lever rule gives 49% of the former coexisting in equilibrium with 51% of the latter. It is expected that the fi nal sample will contain these two phases with the described compositions.

Particle Sizes

In sintering, particles are “powders” (approximately less than 100 µm) with high surface energy contributing to alloy-ing. Also, high pressures are used. Archaeologists mention “nuggets,” “sands,” or “fl akes” when referring to natural platinum and gold. In the pre-Hispanic technique, it could be visual-ized that artisans utilized not powder but sand sizes (0.023–2 mm). The particle dimension is the most important param-eter in the diffusion distance. Shapes and diffusion distances change during the process. It has to be remembered that the artisans used only the pressure of an arm and the blow of a stone to make the particles. At the beginning, the platinum width distance is the only diffusion distance to be considered because it is surrounded with molten gold, which holds all the pieces together. As the process continues, the particles fl atten with the blows, the platinum pieces are surrounded by gold on the sides not on the free surface, and some diffusion distances are bigger. In this work it was decided to cut platinum and gold sheets into small pieces within the size range of a grain of sand (approx-imately 2 mm square, 0.3 mm thick).

Thermomechanical Treatment

After the gold and platinum fl akes or nuggets were made, the fi rst step was joining them. It is reasonable to think that native gold, at its melting-point

Figure 7. An optical microstructure of alloy 1 at cycle 4.

Figure 6. An SEM analysis on alloy 1: (a) linear analysis, (b) x-ray mapping, and (c) microstructure of the analyzed region in a and b.

AuLa1, 4 PtLb1, 5

c

b

a

2006 May • JOM 41

temperature (lower than 1,064°C because of impurities), was used to conglomer-ate particles. Thereafter, the alloy could be subjected to a temperature below the melting point (~950°C), which is a reasonable normal fi re temperature. Handwerker et al.3 suggested that pre-Hispanic artisans might not have worked at melting-point temperatures. So, 950°C is perhaps enough to allow a substantial, rather slow, solid-state substitutional diffusion. Gold and platinum have big atoms, almost the same size; they need to fi ll in the vacancies left by one another. Gold is faster than platinum, so porosity will always be found on the gold side. Their diffusivities at 950°C, calculated after Brandes and Brook,6 are D

Au = 7.66*10–11

and DPt = 1.95*10–11 cm2/s. A very rough

approximation of the time needed to reach, for example, 88% gold at the center (or as an average composition), in a pure platinum sample of 0.015 cm of diffusion distance has been made by the authors with several conventional models. Results vary, depending on the diffusion coeffi cient and the model used. The ranges are from 200 h to 1,000 h. There is no way one can obtain a macro-scopically homogeneous alloy in 2 min., 7 min., 10 min., or even 2 h or 13 h. Undoubtedly, mechanical hammering enhances the diffusion because of the platinum cold-working substructure energy input. All metals, in general, work harden. Strain-hardening coeffi cients are particular for each metal and refl ect the atomic structural defect enhancement (mainly dislocations). Gold is the excep-tion because it can be mechanically worked at room temperature without hardening. Also, hammering and folding

several times brings the gold and plati-num phases in contact at several points during processing, helping diffusion. Annealing the pieces is necessary after some blows; cold work always leaves grains deformed. Just one of the archae-ological samples shows cold-working effects4 while the others evidence either annealing or hot working. There are several diffusion moments for these samples: fi rst, when gold melts, the liquid not only covers and sticks the pieces together superfi cially but actually penetrates or begins to penetrate the platinum; second, solid-state diffusion occurs when holding at temperature (~950°C); and third, diffusion occurs each time it is reheated or annealed between hammer blows or when it is struck. In this work, both types of mechanical work were done. To simulate having a hot metal piece on a hot stone, probably as a die,4 the authors used a gas torch. See the sidebar for experimental details.

EXPERIMENTAL WORK Pure gold strips, 0.5 mm thick, and pure platinum strips, 0.3 mm thick, were used to prepare the alloys: alloy 1: 60Pt-40Au (all compositions are in weight percent) and alloy 2: 12Pt-88Au. Strips were cut in approximate square or rectangular pieces of about 2 mm width for alloy 1 and small strips 10 mm long, 2 mm wide for alloy 2. They were placed in ceramic crucibles at 1,100°C in a furnace for melting the gold and bonding the platinum pieces. The pieces were air cooled. The samples were struck manually, using an iron hammer and a fl attened iron base. During the simulated work hammering, the sample was heated with a small gas torch (~950–1,000°C) and then hammered until it cooled down, or in some cases it was hammered within the fl ame. After 20 or more blows, it was necessary to anneal to continue hammering. So each hammering cycle represents several annealing events. Each sample is hit until fl at, then folded and hammered to preserve the original size. Table A shows the processing schedule, the diffusion time, and the number of blows for the two alloys in partial and cumulative records. The end point was set when the microstructure and scanning-electron microscopy (SEM) analysis almost completely revealed the desired monophasic 12Pt-88Au alloy and the biphasic equilibrium 60Pt-40Au alloy. The areas selected for SEM analysis were representative of the optical microscopy structure. Samples were etched with aqua regia (50% HCl-50% HNO

3). An Olympus PMG

microscope, and an SEM JEOL JSM–5900LV were used.

RESULTS AND DISCUSSION

Alloy 1: 60Au-40Pt

Alloy 1 required 304 h diffusion time and 1,410 hammer blows. This is too much time by 21st-century standards, although it is within the ranges calculated with the simple approximate models already mentioned. There is no intention to set this time and number of blows as a possible way to obtain the pieces, but in ancient times masterworks were created without regard for time. Figure 3 shows the macroscopic appearance of the fi rst small “conglom-erate” made after the melting time and some cold blows. Figure 4 shows gold dendrites, a platinum phase, and a third light gray phase in the grain borders, the beginning of alloying. Figure 5 displays some of the scanning-electron microscopy (SEM) analysis points. In the initial stage, the existence of micro-regions with 30% platinum and others with 95% platinum indicate that

1 cm

Figure 8. The fi nal macrostructure of alloy 1.

Table A. Alloy Processing Schedule

Partial Record Cumulative Record

Alloy 1* Alloy 2** Alloy 1* Alloy 2**

Diffusion Hammer Diffusion Hammer Diffusion Hammer Diffusion HammerOperation Time (h) Blows Time (h) Blows Time (h) Blows Time (h) Blows

Melting 2.65 202 2.22 55 2.65 202 2.22 55 and bondingCycle 1 0.93 66 17.73 70 3.58 268 19.95 125Cycle 2 25.42 96 32.33 60 29.00 364 52.28 185Cycle 3 219.40 523 140.08 55 248.40 887 192.36 240Cycle 4 54.30 430 — — 302.70 1,317 — —Cycle 5 1.30 93 — — 304.00 1,410 — —

* 60Pt–40Au; ** 88Au-12Pt

JOM • May 200642

Figure 11. The gold content (in weight percent) in alloy 2 at some points for each cycle.

equilibrium phases form rapidly even in some small micro-regions. Platinum contents on cycles 1, 2, and 3 reveal the spread of the compositions while the alloying is progressing; nevertheless, the equilibrium ones (30% platinum and 91% platinum) always exist. Cycle 4 and mainly cycle 5 manifest the biphasic equilibrium composition (approximately 30% platinum in equilibrium with 90% platinum). Figure 6 shows the SEM analysis and the microstructure of the Au-Pt diffusion in an equilibrium region in cycle 1. There, dendrites disappeared due to diffusion time at temperature (homogenization) and diffusion penetration of gold in platinum. Figure 6a exhibits a linear analysis of platinum and gold in the two regions. Figure 6b is an x-ray mapping where bright spots indicate greater con-centrations of either gold or platinum and Figure 6c is the SEM microstructure. The two different regions are two adja-cent grains in the alloy. Figure 7, from cycle 4, is the optical microstructure of the semi-fi nal alloy. The proportions of the two phases are near 50%-50%; Figure 8 is a macroscopic picture of the fi nal alloy. It looks to be homogeneous platinum to the eye. It is necessary to etch in order to reveal its biphasic, bicolor nature.

The breaking of platinum in small globules as gold penetrates, mentioned by Scott and Bray7 and Bustamante et al.,4 was not seen in this simulation. The work of pre-Hispanic artisans and the Bustamante et al. simulations4 used gold and platinum alluvial deposits. Scott and Bray7 give native platinum compositions, evidencing several impurities. Figure 9 shows one of their microstructures,7 where gold dendrites can be seen— although the authors never mentioned them—and the platinum globules. In the present study, pure platinum was used. This could be the reason for the different structure. Impurities could

give rise to platinum brittleness because they are usually located in grain bound-aries, limiting their ductility and fi nally breaking. These could be the origin of the broken or globular structure reported. Optical microstructures of four fi n-ished Au-Pt (31–47% platinum) pieces were reported by Bustamante et al.4 They exhibit large platinum grains surrounded by the golden matrix. They resemble the structure of this simulated alloy although their platinum grains are rounded, not striped as those in this study (shown in Figure 7).

Alloy 2: 88Au-12Pt

As shown in Table A, this alloy was fi nished after 192.36 h and 240 blows. The fi nal appearance is golden. The time required for fi nishing was less than in alloy 1 because there are fewer platinum particles. In the initial formation of the alloy, gold (a dark phase) penetrates around the white platinum grain border and forms the gray equilibrium alloy (Figure 10). Figure 11 gives the SEM analysis points in percentage gold after the cycles; from them the evolution of the alloying is readily seen. After 192 h the analyzed points have the same composition: around 88% gold. Figure 12 is an SEM image of the fi nal sample grains. The microstructures reported on Au-Pt pieces of similar composition by Handwerker et al.3 differ completely from the fi nal sample in this study. Their photomicrographs always show platinum particles embedded in a gold matrix. An equilibrium homogeneous composition was never attained in the pre-Hispanic

Figure 10. An optical microstructure of alloy 2 at the beginning of the process.

Figure 9. The microstructure of a gold-platinum object from Esmeraldas, Scott and Bray.7

3

30

20

10

0

40

50

60

70

80

90

100

1 5 7 9 11 13Point

Au (%

)

15 17 19 21 23

— Cycle 1— Cycle 2— Cycle 3

samples analyzed. Again, the globular breaking platinum was not observed.

GENERAL CONSIDERATIONS AND CONCLUSIONS

Pre-hispanic artisans did ingenious work when creating their alloys. They surely made trial-and-error mixtures of gold and silvery deposits to get the colors and shapes they wanted from natural raw materials. This is how metallurgy and technology, generally, have progressed. For a 21st century metallurgist, the consideration of all the processing details

for obtaining alloys is an intellectual task and time-consuming laboratory work, with the end point being just the alloy. In pre-Columbian times, the end point was not the alloy but a sacred masterpiece of art. We, as modern-day metallurgists, conducted this small study honoring them. In doing so we give a tribute to humanity of all cultures and races.

ACKNOWLEDGEMENT

Thanks to Guillermina Gonzalez Mancera for her scanning-electron microscopy work.

References

1. P. Estévez, “Platino en el Ecuador Precolombino,” Boletín del Museo del Oro (44-45) (1998), pp. 160–181.2. P. Bergsoe, The Metallurgy and Technology of Gold and Platinum among the Pre-Columbian Indians (Ingenior Videnskabelige Skifter No. A46: DNKS Copenhagen, 1937).3. C.A. Handwerker et al., “Fabrication of Platinum-Gold Alloys in Pre-Hispanic South America: Issues of Temperature and Microstructure Control,” Mat. Res. Soc. Symp. Proc. 185 (Pittsburgh, PA: Materials Research Society, 1991), pp. 649–664.4. N. Bustamante et al., “Tecnología del Platino, en la Fabricación de Piezas de Orfebrería Precolombina” (private communication, 2005).5. D.A. Scott and W. Bray, “Pre-Hispanic Platinum Alloys: Their Composition and Use in Ecuador and Colombia,” Archaeometry of the Pre-Columbian Sites, ed. D.A. Scott and P. Mayers (Marina del Rey, CA: Getty Conservation Institute, 1994), pp. 285–322.6. E.A. Brandes and G.B. Brook, editors. Smithells Metals Reference Book, 7th ed. (Great Britain: Butterworth-Heinemann, 1998), pp. 13–75.7. D.A. Scott and W. Bray, “Ancient Platinum Technology in South America,” Platinum Metals Review, 24 (4) (1980), pp. 147–157.

M. Noguez, G. Salas, T. Robert, and J. Ramírez are professors, and R. García is a student in the Materials Engineering Department at the Universidad Nacional Autónoma de Mexico.

For more information, contact María Noguez, Universidad Nacional Autónoma de Mexico, Department of Materials Engineering, Circuito Institutos s/n, Ciudad Universitaria, Mexico City, D.V. 04510, Mexico; e-mail nogueza@servidor.unam.mx.

Figure 12. The SEM micro-structure of alloy 2 at the process end.

17

18