Amend 2 Attach 7 SAFA Report 1618-Metals Analysis-1.pdf

Page 1 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Scientific Analysis of Fine Art Report 1618

Identification of Alloys, Corrosion Products, and Corrosion Mechanisms in the Harbor Light, Dry Tortugas National Park (1876)

December 11, 2017

Executive Summary

I. Wrought iron substrate

The 1876 Harbor Light structure, which is comprised of iron alloy plates and rivets, consists of several varieties of wrought iron. Some of the locations reveal a traditional wrought iron microstructure with silicate slag stringers running throughout, indicating the working direction of the plate (samples 1, 2, and 3). The microstructure of the iron alloy of the plate that was riveted

(sample 4) suggests puddled wrought iron rather than simple wrought iron (puddled iron is wrought iron that has been decarburized from pig iron) based on the presence of inclusions of MnS (manganese sulfide) rather than slag. The one rivet studied appears to have been hot rivetted based upon the widmenstätten ferrite observed (see discussion below). Wrought iron was a common 19th century construction material, used on the Eiffel tower and the framework for the Statue of Liberty. However, historic wrought iron structures require continuous maintenance, and the prior corrosion protection layers should ideally be removed before new ones are applied.1 This procedure was not followed for the maintenance of the harbor light. The corrosion protection technologies applied to the Harbor Light over time and the reason for their ultimate failure are discussed below.

II. Corrosion phenomena identified

The corrosion products identified using Raman spectroscopy include lepidocrocite [γ-FeO(OH)], hematite (Fe2O3), and goethite [FeO(OH)]. Electron microscopy shows the presence of chloride ions in the corroded wrought iron layers of the structure, in addition to the presence of chloride ions in the anti-corrosion coatings. However, no iron chlorides or iron chloride hydroxides have been identified in the wrought iron. For example, the akageneite form of iron oxide hydroxide

[FeO(OH,Cl)], which commonly denotes high concentrations of chloride in an environment, was

1 Tolboom, Hendrik Jan, “Instandhouding van smeedijzer in het exterieur. (Maintenance of wrought iron in open air)”RDMZ info no. 32 (2003 Jan), pp. 1-8 Netherlands. Rijksdienst voor de Monumentenzorg, Zeist, Netherlands

Page 2 of 67 © 2017 Scientific Analysis of Fine Art, LLC not observed.2 Likewise, limonite [FeO.OH.nH2O], which can form on steels in the presence of marine aerosols, was not observed.3 Other possible phases that can be involved in historic and/or archaeological iron corrosion include the iron hydroxychloride β-Fe2(OH)3Cl, an alpha form of this same mineral, and a gamma phase known as hibbingite.

Since the iron corrosion products were identified using micro-Raman spectroscopy, akageneite if present may have been converted into hematite by the power of the incident laser.4 It is also notable that the Raman spectrum of akageneite will depend upon the amount of chloride substitution for hydroxide, and so this can also make its identification challenging. Finally, any lack of crystallinity for this compound can hinder its ability to be detected by Raman spectroscopy (as well as x-ray diffraction, in both cases the characteristic peaks will be broader and less intense if the compounds are not crystalline).5

III. Corrosion protection technologies employed and their performance on samples 1–4.

The most important data collected relating to the preservation of the Harbor Light include:

1) The microstructures of the wrought iron plates and the rivet provided, revealing their state of preservation

2) The history of the anti-corrosion coatings applied to the Harbor Light

3) The performance of these anti-corrosion coatings

4) The failure mechanisms of the anti-corrosion coatings

This was studied using visible light photomicroscopy, ultraviolet light photomicroscopy, backscattered electron imaging in the scanning electron microscope, and x-ray imaging in the scanning electron microscope.

Sample 1. Overview

Three campaigns of anti-corrosion coatings are present. The first (earliest) campaign is made up of one discrete layer of relatively monodisperse small (average of ~10 µm) grey and white particles, with the grey metallic particles predominately at the bottom and the white particles

2 L. Bellot-Gurlet et al., “Raman studies of corrosion layers formed on archaeological irons in various media”

Journal of Nano Research (2009) 8, 147 – 156.

3 J. Aramendia et al., “Portable Raman study on the conservation state of four CorTen steel-based sculptures by

Eduardo Chillida impacted by urban atmospheres”, Journal of Raman Spectroscopy, 2012, 43, 1111-1117.

4 M. Hanesch, “Raman spectroscopy of iron oxides and (oxy)hydroxides at lower laser power and possible applications in environmental magnetic studies”, Geophysical Journal International (2009), 177, 941-948.

5 S. Reguer et al., “Deterioration of iron archaeological artefacts” micro-Raman investigation on Cl-containing corrosion products”, Journal of Raman Spectroscopy, 2007, 38, 389-397.

Page 3 of 67 © 2017 Scientific Analysis of Fine Art, LLC predominately toward the top of the layer (see Figure 5, Figure 6, Figure 8, Figure 18, and Figure

31). There is iron corrosion on both sides of this layer. The second campaign consists of white paint layer with large (up to 100 µm) metal spheres (prills). These layers are separated by a layer of iron corrosion, indicating the failure of the first layer and that the two layers were applied years apart. There is a thin and inconsistent layer of iron corrosion between the second and third anti-corrosion campaigns (see, for example, Figure 12 and Figure 16). The third campaign is made up of five discrete layers (two repeating pairs separated by a translucent coating). This campaign can be seen to consist of (from first to last application):

1. A white paint layer

2. A dark grey (almost black) paint layer

3. A transparent layer that appears free of particulates (see Figure 8, Figure 12, and Figure

35).

4. A second white paint layer

5. A final dark grey paint layer

That all of these five layers make up one coating campaign (i.e., that they were applied at the same time) is indicated by the lack of any surface soil or other discontinuity between the layers.

The blue fluorescence of coatings 1, 3, and 4 of this campaign in the ultraviolet suggests a zinc component to these layers (either as a dissolved zinc carboxylate drier or a dispersed zinc oxide)

(see Figure 14-Figure 17).

Sample 1. Anti-corrosion campaign 1

The three corrosion protection campaigns for Sample 1 can be observed most clearly in the backscattered electron image in Figure 40. The original corrosion protection layer is filled with metallic zinc spheres with a predominately oxide corrosion layer surrounding them (a small amount of chlorides is also present here). Some inclusions of silica can also be observed as well as iron oxide and iron chloride corrosion products (see electron microscopy images in Figure 40, Figure 41, and Figure 43-Figure 47). This corrosion protection strategy is sometimes referred to as ‘cold galvanizing’, and is essentially zinc dust mixed with inorganic fillers and organic binders. This mixture is then sprayed onto the surface of the wrought iron substrate.6 Surface preparation of the underlying metal surface is critical to achieve adhesion with this technology, and it is recommended that the iron alloy substrate be sandblasted prior to its application. This also includes removing all “loose, flaking, and deteriorated paint from the iron, as well as dirt

6 https://www.galvanizeit.org/corrosion/corrosion-protection/zinc-coatings/zinc-rich-paint

Page 4 of 67 © 2017 Scientific Analysis of Fine Art, LLC and mud, water-soluble salts, oil, and grease.”7 Furthermore, it is recommended that the zinc-rich paint be applied within four hours after sandblasting.8 In aggressive environments such as the

Dry Tortugas, it is further recommended that the zinc-rich paint act as a primer to be covered with a top coating and enhance the lifetime of the system. No evidence of such a top coating is observed here. One failure mechanism of a zinc prill paint with an organic binder is its susceptibility to damage upon exposure to ultraviolet light. Voids in the organic binder are observed for this coating, indicating this mechanism of failure has occurred as well.

One late 20th century procedure for a corrosion protection coating applied to architectural wrought iron involved one alkyd undercoat, two coats of zinc phosphate epoxy primer, and then two alkyd topcoats.9 Another recommended corrosion protection system for freshly sand-blasted historic iron surfaces is an epoxy base coat, a zinc-rich primer, and then two urethane finish coats.10 While the Harbor Light was initially protected with a thickly applied zinc-based corrosion protection campaign, it consists of a single thick (~400 mm) layer only – it does not contain a carefully engineered stratigraphy with undercoats and topcoats to protect it. It is notable that the service life of such a coating in a coastal environment is estimated at ~16 years.

It is also notable that, in addition to having been applied incorrectly (without any barrier coatings) this first corrosion protection campaign, and the manganese-silicon-iron one that followed it, were not fully removed before the third corrosion protection campaign was applied.

While total removal of earlier campaigns is not always possible, this can lead to more adhesion issues with later campaigns.

Ideally the concentration of the zinc particles in the paint will be sufficient to provide cathodic protection to the wrought iron layer underneath. However, for this to take place, the concentration of the zinc prills in the paint needs to be high enough for electrical conductivity between the zinc and iron alloy layers. Since the paint binder is organic and non-conductive, cathodic protection is challenging to achieve with this technology. Note that the late 20th century zinc-based corrosion protection formulations are relying on cathodic protection as a secondary measure, since there is an undercoat applied beneath the zinc-containing layer.

7 M. Gayle, D. W. Look, and J. G. Waite, Metals in America’s Historic Buildings, Uses and Preservation

Treatments, U.S. Department of the Interior National Park Service Cultural Resources Preservation Assistance, Washington D.C. 1992, p.137.

8 Hot-dip galvanized steel vs. zinc spray metallizing, American Galvanizers Association, undated.

9 N. Ashurst, “Conserving corroded ironwork”, Conservation bulletin (English Heritage) no. 6 (1988 Oct) pp. 4-5.

10 M. Gayle, D. W. Look, and J. G. Waite, Metals in America’s Historic Buildings, Uses and Preservation

Treatments, U.S. Department of the Interior National Park Service Cultural Resources Preservation Assistance, Washington D.C. 1992, p.140.

Page 5 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Zinc is commonly applied over iron alloys (during the galvanization process, for example) for two reasons. The first is its corrosion resistant nature, which protects the iron alloy from agents of degradation including water, salt spray, and oxygen. The second is that as it does corrode, it produces corrosion byproducts that will be dense and strongly adhere to the zinc. These zinc oxides will form around the zinc prills in the paint and provide a barrier layer protecting the wrought iron structure beneath. This process can be observed in Figure 41. Zinc-rich paints, however, suffer from a number of limitations, including inconsistent and incomplete coverage during application. This limits the efficacy of the coating, particularly at the corners and edges of the wrought iron plates.11 This type of coating is also soft and prone to scratching, exposing the wrought iron substrate to the elements over time.

The zinc-rich paint used here and the extent of its corrosion can be observed by comparing

Figure 43 and Figure 44. It is notable that most of the zinc prills have been converted to zinc oxide (the zinc particles appear brighter in the backscattered image), which should add bulk and protection to this film. However, one can also conclude that this corrosion protection layer has far exceeded its approximately 16-year lifetime, and it is possible to observe that the binding medium is missing from this paint layer in some areas, likely due to UV exposure. In other words, there is in places a breakdown of the organic binder that should be holding the zinc particles together, which will allow agents of degradation such as oxygen, water, and chloride-containing aerosols to penetrate through to the wrought iron. This is evidenced by the presence of iron corrosion both above and below the zinc-rich paint. Thus, the zinc-rich paint had failed before the next corrosion protection layer was applied. It is also notable that the presence of extant porous corrosion layers beneath later-applied coatings will act as a reservoir to trap water against the surface of the monument and accelerate corrosion.12 The presence of chloride acting as an electrolyte for the corrosion process will further worsen this problem.

The failure of the first zinc-based anti-corrosion coating and its incomplete removal (along with the porous iron corrosion above it), has allowed active chlorides to enter the wrought iron plates comprising the Harbor Light’s structure. In the case of the corrosion layer it has also provided a reservoir of electrolyte-rich moisture that will accelerate future corrosion.

Sample 1. Anti-corrosion campaign 2

Directly above the zinc-rich corrosion protection layer is campaign 2, a unique layer comprised of large (over 100 µm) spheres/prills of metallic iron enriched with silicon and manganese (see

11 https://www.galvanizeit.org/corrosion/corrosion-protection/zinc-coatings 12 M. Gayle, D. W. Look, and J. G. Waite, Metals in America’s Historic Buildings, Uses and Preservation

Treatments, U.S. Department of the Interior National park Service Cultural Resources Preservation Assistance, Washington D.C. 1992, p.131.

Page 6 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 49). This coating also appears to have been engineered as a corrosion-protection layer, since the silicon and manganese in the iron prills will form passivating oxide surfaces on them, although not with the volume expected for the zinc prills. The distribution of these iron prills observed in Figure 16 suggests that they are an intentional addition to this layer, rather than an impurity. This coating layer appears to be poorly formulated however, and is predominately comprised of other inert filler materials that have no corrosion protection action. These include magnesium silicates. The magnesium silicates in this layer have an acicular or their needle-like morphology, as is commonly observed for talc, [Mg3Si4O10(OH)2] a common architectural paint filler. Equiaxial calcium-rich particles can be observed in this layer too, these are a calcium-rich mineral because they also contain aluminum, silicon, and oxygen. The potential for cathodic protection from this layer is thus minimal, and it functions predominately as a protective paint film. The absence of any base coat or top coats, however, represent another serious flaw in campaign 2 in terms of its corrosion protection efficacy.

A thin iron oxide and chloride-rich corrosion layer is also visible here. Chlorides are, in fact, distributed throughout the entire layer system.

Sample 1. Anti-corrosion campaign 3

Layers 1 and 4 of the most recently applied (uppermost) anti-corrosion campaign have the same composition. They are comprised of silica (SiO2) particles in an organic matrix (see Figure 49).

The organic matrix was confirmed by infrared spectroscopy to be an epoxy resin. This layer has no active anti-corrosion activity, and is acting as a barrier layer or topcoat. Layers 2 and 5 have a mixture of barium sulfate (BaSO4), magnesium silicate, and calcium silicate particles. The presence of barium sulfate was confirmed by Raman spectroscopy. Note that the magnesium silicates are likely talc [Mg3Si4O10(OH)2], a common architectural paint filler. The calcium silicate particles are also acting as a colorless paint filler.

Sample 2. Discussion

Sample 2 is also from the first panel and the proper left of the structure. The microstructure revealed in the bright field images is characteristic of a wrought iron, with slag stringers throughout that reveal the working direction of the metal used to make the plate (see Figure 19-

Figure 24 for visible light cross-section photomicrographs of this sample, and Figure 53 for the secondary electron and backscattered electron images of this sample). Figure 54, at a higher magnification, reveals that the sample is mostly mineralized (chloride-containing corrosion layers), with only a small amount of metallic iron left. The slag stringers present in this sample can be identified by their silica and phosphorus content. The iron oxide in the slag appears to be wüstite (FeO) based on its shape. X-ray microanalysis data showing the composition of the slag

Page 7 of 67 © 2017 Scientific Analysis of Fine Art, LLC stringers are shown in Figure 55. Note the presence of one metallic tin inclusion – possibly a remnant of a tinned iron fragment?

Sample 3. Wrought iron plate from proper left

The initial sectioning of the plate showed that no metal remained. The layers of corrosion protection paint were intact, and the banding seen in the corroded metal suggested that the plate was heavily worked into its final shape (see Figure 59 and Figure 60).

Under the microscope, the layer structure of the coatings can clearly be seen. An iron corrosion layer is followed by a second corrosion layer prior to the application of a failed paint coating

(possibly a primer) and then a metallic zinc rich corrosion protection (cathodic) paint layer. The corrosion protection layer is then covered by two additional coatings, in this case paint layers with no specific corrosion protection properties. These are clearly seen in the brightfield and darkfield images (see Figure 60). Electron microscopy revealed more details of layering structure

(Figure 61 and Figure 62). The lowest level near the metal is a magnesium silicate – judging by the elongated crystal structure likely talc, a common architectural paint filler Mg3Si4O10(OH)2.

This is followed by a layer rich in metallic zinc particles, meant to act as a sacrificial corrosion protection layer (in the manner of galvanized steel). While the zinc particles are reactive with the environment, their corrosion products form a dense protective layer which blocks agents of degradation from reaching the iron alloy below. Note however that chloride corrosion has penetrated this layer and reached all the way down to the iron alloy substrate. The next corrosion protection layer repeats the first magnesium silicate layer, but is mixed with a calcium silicate-containing filler and also barium sulphate. The thin topmost layer is mainly composed of a silicon dioxide (quartz or sand) filler. This is similar to the stratigraphy observed in Sample 1 but the layer with the silicon-manganese-iron inclusions (campaign 2) is missing. This stratigraphy is essentially the layer structures recommended in the 20th century historic preservation literature for wrought iron – the application of a base coat followed by a corrosion protection layer and then the application two topcoats. However, the coatings have failed, with chloride corrosion visible throughout the thickness of the iron substrate, and no metallic iron remaining.

Sample 4

The cross-section of the rivet (sample 4) revealed that some plate metal has remained allowing for a more complete investigation of the two metal structures (see Figure 63). Focusing first on the rivet itself, it is made from wrought bloomery iron as evidenced by the heavy remaining slag stringers (Figure 64). The long directional nature of the slag down the length of the rivet suggest that it was heavily worked prior to being used. The slag is a fayalitic slag (Fe2SiO4) with wüstite dendrite growth (FeO).

Page 8 of 67 © 2017 Scientific Analysis of Fine Art, LLC

The makeup of the slag is more clearly seen under the electron microscope (Figure 65). Spot analysis shows the general composition of the dendrites to be approximately 79% Fe and 20% O and the general composition of the fayalite (Fe2SiO4) to be approximately 56% Fe, 13.5% Si and

30% O.

Upon etching, a low carbon ferrite grain structure was revealed (this is the body centered cubic α form of iron, see Figure 66Figure 7). This is also consistent with a bloomery iron. The grains at the center of the rivet are uniform in size and show no-to very little evidence of any eutectoid structure. This indicates a base carbon content very close to 0.008 weight percent.

An interesting observation was noticed at the rivet interface with the top of the metal plate. This is the location of the largest deformation of the rivet shape when being inserted. There is growth of widmenstätten ferrite extending from the grain boundaries (Figure 67). This formation tends to occur in low carbon iron at temperatures close to the austenite (γ-Fe) transformation (~900oC for pure iron). It is apparent in ‘finger like’ structures extending out from the grain boundaries

(as seen in meteorites). This type of structure suggests that the joining was hot riveted.

In addition, near the edges of the rivet, there is some pearlitic growth (eutectoid) indicating an enriched carbon region which may be associated with the heating of the rivet prior to insertion and working [note that pearlite is a lamellar structure composed of alternating layers of ferrite (Fe3O4) and cementite (Fe3C)]. The plate attached to the rivet conforms with what was seen previously in the fully corroded sample, namely heavy working as evidenced by the elongated inclusions. The inclusions themselves are not silicate slag and suggest an alternative method of manufacture. Electron microscopy analysis of the inclusions revealed that they are

MnS and suggest that the iron used for the plates was decarburized through puddling (therefore, puddled wrought iron, see Figure 68).

Etching reveals that this iron is a hypo-eutectoid steel with a low carbon concentration of around

0.1%. This is evident in the pearlitic structure found at most grain boundaries as seen in Figure

69Figure 10.

Interface between Rivet and Plate

The interface between the rivet and the plate shows that a coating may have applied to the plate surface prior to the joining (Figure 70). Electron microscopy analysis shows that the interface coating layer is mainly comprised of a calcium-rich compound such as chalk (Figure 71). On either side of the paint layer are layers of iron oxide in various oxidation states. A thin silicon rich layer resides to the left of the calcium rich section and is coated with the zinc sacrificial/cathodic paint. The silicon rich layer is likely paint that has been filled/bulked with

SiO2, and may represent the initial coating layer on the artifact.

Page 9 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Additional x-ray maps of the rivet head show a layering structure similar to that which was seen on the plate samples alone. However, in this situation, the layering is as follows: the first layer on the rivet is a Mg, Si, S, Ba, and Ca layer (likely a paint bulked with talc, barium sulfate, and chalk again). This is followed by a SiO2 layer with the zinc rich layer residing on top (Figure 72).

This sample reveals a 'bad' weld joint, and likely rivets that are slightly different in their mechanical properties than the wrought iron plate. Variable expansion of the metal at the weld edges and at rivet heads likely caused initial cracking and allowed corrosion to penetrate below the protective layer.

Page 10 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Object Description and Reason for Analysis

Figure 1. The 1876 Dry Tortugas harbor light

Several layers of corrosion protection paint have been applied to this iron alloy structure.

However, failure of those layers is observed, and molecular, microscopic, and elemental analysis information is required to understand the mechanisms of failure present and the corrosion phenomena taking place (see figures below). One proposed mechanism for the corrosion protection challenges of this structure is the use of electrochemically incompatible metals such as copper, whose presence would accelerate the corrosion of the iron alloy. As a result, alloy analysis is accompanied by corrosion product analysis and also corrosion protection surfaces.

The composition and microstructures of the corrosion protection surfaces were used to reveal their mechanisms of corrosion protection and also their limitations.

Page 11 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Results and Discussion

Data were collected from the following provided samples.

Sample Description

1 first panel, proper left of structure

2 first panel, proper left of structure

3 lower plate, proper left

4 rivet from lantern base

Figure 2. Location of Sample 1. First panel, proper left of structure.

Page 12 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 3. Location of Sample 2. First panel, proper left of structure.

Figure 4. Location of Sample 3. Lower plate, proper left.

Page 13 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Sample 4, a rivet from the lantern base, was collected by Al Hodson, Resurgence Engineering, and does not have an accompanying photograph.

Page 14 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Visible and Ultraviolet Light Microscopy

Figure 5. Sample 1. Visible microscopy, 100× magnification, 189 ms exposure, left region.

Figure 6. Sample 1. Visible microscopy, 100× magnification, 189 ms exposure, right region.

Figure 7. Sample 1. Ultraviolet microscopy, 100× magnification, 1000 ms exposure, left region.

Campaign 3

Campaign 2

Campaign 1

Page 15 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 8. Sample 1. Visible microscopy, 250× magnification, 200 ms exposure, region 1.

Figure 9. Sample 1. Visible microscopy, 250× magnification, 200 ms exposure, region 2.

Figure 10. Sample 1. Visible microscopy, 250× magnification, 200 ms exposure, region 3.

Page 16 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 11. Sample 1. Visible microscopy, 250× magnification, 200 ms exposure, region 4.

Coating campaigns 2 and 3 only.

Figure 12. Sample 1. Visible microscopy, 250× magnification, 200 ms exposure, region 5.

Coating campaigns 2 and 3 only.

Page 17 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 13. Sample 1. Visible microscopy, 250× magnification, 200 ms exposure, region 6.

Coating campaign 2 and 3 only.

Figure 14. Sample 1. Ultraviolet microscopy, 250× magnification, 1000 ms, region 1.

Figure 15. Sample 1. Ultraviolet microscopy, 250× magnification, 1000 ms, region 3.

Page 18 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 16. Sample 1. Ultraviolet microscopy, 250× magnification, 1000 ms, region 4, campaigns

2 and 3 only.

Figure 17. Sample 1. Ultraviolet microscopy, 250× magnification, 1000 ms, region 5, campaigns

Figure 18. Sample 1. Ultraviolet microscopy, 250× magnification, 1000 ms, region 6, campaigns

Page 19 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 19. Sample 2. Visible light microscopy, 100× magnification, 700 ms exposure, region 1, bright field.

Figure 20. Sample 2. Visible light microscopy, 100× magnification, 700 ms exposure, region 2, Figure 21. Sample 2. Visible light microscopy, 100× magnification, 700 ms exposure, region 3, Page 20 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 22. Sample 2. Visible light microscopy, 100× magnification, 700 ms exposure, region 4, Figure 23. Sample 2. Visible light microscopy, 250× magnification, 500 ms exposure, detail 1, Figure 24. Sample 2. Visible light microscopy, 250× magnification, 500 ms exposure, detail 2, Page 21 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 25. Sample 3. Visible light microscopy, 100× magnification, 700 ms exposure, top layer left, bright field.

Figure 26. Sample 3. Visible light microscopy, 100× magnification, 700 ms exposure, top layer right, bright field.

Figure 27. Sample 4. Visible light microscopy, 100× magnification, 1000 ms, left, bright field.

Page 22 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 28. Sample 4. Visible light microscopy, 100× magnification, 1000 ms, right, bright field.

Figure 29. Sample 4. Visible light microscopy, 100× magnification, 600 ms exposure, left, Figure 30. Sample 4. Visible light microscopy, 100× magnification, 600 ms exposure, right, Page 23 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Brightfield and Darkfield Optical Microscopy and Electron Microscopy.

Figure 31. Sample 1, bright field.

Figure 32. Sample 1, dark field.

Page 24 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 33. Sample 1, fully-crossed polarizing filters.

Page 25 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 34. Sample 1, region 1 at higher magnification; (top) bright field, (middle) dark field, (bottom) fully-crossed polarizing filters.

Page 26 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 35. Sample 1, region 2 at higher magnification; (top) bright field, (middle) dark field, (bottom) fully-crossed polarizing filters.

Page 27 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 36. Sample 2, bright field. The bottom surface in these images are towards the exterior, and the top surface is towards the interior of the sample flake.

Figure 37. Sample 2, bright field, higher magnification.

Page 28 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 38. Sample 2, dark field, higher magnification.

Figure 39. Sample 2, fully crossed polarizing filters, higher magnification.

Page 29 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 40. Sample 1, backscattered electron image.

Page 30 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 41. Backscattered electron image showing the zinc spheres in the initial anti-corrosion layer (campaign 1) and their partial oxidation to help protect the wrought iron layer beneath them.

Page 31 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 42. Backscattered electron image of the two most recent corrosion protection campaigns.

Page 32 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 43. Backscattered electron image of zinc-based anti-corrosion layer (campaign 1).

Page 33 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 44. Backscattered electron image showing the remaining metallic zinc in the initial anti-corrosion layer (campaign 1).

Page 34 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 45. X-ray maps of the region shown in Figure 44, first anti-corrosion layer (campaign 1).

Page 35 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 46. X-ray spectroscopy of the region shown in Figure 44, first anti-corrosion layer

(campaign 1).

Page 36 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 47. x-ray spectroscopy of the region shown in Figure 44, first anti-corrosion layer

(campaign 1).

Page 37 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 48. Backscattered electron image of the top two corrosion layers (campaigns 2 and 3).

Page 38 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 49. X-ray maps of top two corrosion layers (campaigns 2 and 3). Note that the map for titanium is, because of an overlap of x-ray energies, actually also a map for barium.

Page 39 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 50. Overview of Sample 1 corrosion protection layers. Layer 6 is campaign 1, layer 5 is campaign 2, and layers 1- 4 are the inorganic-containing layers of campaign 3 (the middle layer in campaign 3 has no inorganic component and so shows up black on this x-ray map).

• SiO2 particles in an organic matrix

• Layer 2 + 4

• barium sulphate, magnesium silicates, and calcium rich particles – likely chalk

• Layer 5

• Thin iron oxide

(chloride) corrosion layer

• Large balls of metallic iron enriched with silicon and manganese

• Smaller spherical balls of calcium with

Sample 1

Page 40 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 51. X-ray spectra for components of campaigns 2 and 3. Iron silicates are identified in campaign 2, and also magnesium carbonate. Silica is identified in campaign 3, and also barium sulfate.

Page 41 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 52. x-ray microanalysis of campaigns 2 and 3. Talc is identified in campaign 2. Barium sulfate is identified in campaign 3, along with silica and talc.

Page 42 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 53. Sample 2, (left) secondary electron image and (right) backscattered electron image, 282X magnification.

Page 43 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 54. Sample 2. Pure wrought iron sample. A small amount of metallic iron (with traces of phosphorus) remain. Majority is corrosion with layers of chlorides. Some slag stringers are present with a higher silica and phosphorus content. The iron oxide in the slag is reminiscent of wüstite based on shape.

Page 44 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 55. X-ray maps showing elemental composition of corrosion layers and slag stringers in sample 2.

Page 45 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 56. Composition of slag and slag inclusions in Sample 2. The slag contains phosphorus and silica, and the inclusion is consistent with wüstite.

Page 46 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 57. Secondary electron image and backscattered electron image of sample 2, 644X magnification.

Page 47 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 58. Metallic iron is identified in Sample 2, along with a tin inclusion and iron oxide corrosion.

Page 48 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Sample 3 – plate

Figure 59. One sample of the plate metal showing the banded structure indicative of heavy working. All metal has corroded away. The layer of paint remains attached as seen at the top of the sample.

Figure 60. Bright field and dark field images of the paint layering on the plate sample.

Page 49 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 61. Scanning electron microscopy backscattered electron image showing the region mapped.

Page 50 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 62. X-ray maps showing the distribution of elements across the paint layers.

Page 51 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Rivet

Figure 63. Section of the rivet revealing remnant of plate metal as well as the painted coating layers.

Figure 64. Elongated large section of slag indicative of wrought bloomery iron. On closer inspection of the slag itself, one can see a base of fayalite with wüstite dendrite overgrowth.

Page 52 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 65. SEM image (both secondary electron and backscattered electron) showing the slag makeup of a fayalitic base with wüstite overgrowth.

Figure 66. Etched sample showing the regular-sized ferrite grains intermixed between the slag stringers.

Page 53 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 67. Widmenstätten ferrite growth extending from the grain boundaries. Darkened color is partially due to over etching of sample in this region and the presence of pearlite (eutectoid).

Page 54 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Plate.

Figure 68. Micrograph of the plate section attached to the rivet showing MnS inclusions closely associated with puddle iron.

Figure 69. Etched plate metal showing ferrite grains surrounded by pockets of pearlite.

Page 55 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Interface between Rivet and Plate

The interface between the rivet and the plate show that a coating may have applied to the plate surface prior to the joining.

Figure 70. Interface between the plate and the rivet in brightfield and darkfield showing some base iron oxidation on the plate covered with a ‘paint’ coating. To the right is a coating of the zinc rich layer, covered by additional layers of paint.

Figure 71. Elemental maps of the distribution of elements at the interface between the rivet and the plate.

Page 56 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 72. Elemental maps of the element distribution of the coatings on the rivet head.

Page 57 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Micro-Raman Spectroscopy

Figure 73. Sample 1. Raman spectrum (upper panel) of layer 2 (lower panel). The peaks at 451, 615, and 985 indicate the presence of barium sulfate. Conichalcite (hibbingite) was also identified. 785 nm, 10% laser power, 60 seconds/acquisition, 5 acquisitions. Peaks above 1000 are due to the epoxy resin.

Page 58 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 74. Sample 1. Raman spectrum (upper panel) of layer 5 (lower panel). The peaks at 449, 612, 637, and 982 indicate the presence of barium sulfate. Conichalcite (hibbingite) was identified. 785 nm, 10% laser power, 60 seconds/acquisition, 5 acquisitions. Peaks above 1000 are due to the epoxy resin.

Page 59 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 75. Sample 1. Raman spectrum (upper panel) of layer 7 spot 4 (lower panel). The peaks at 225, 241, 291, 408, 495, 608, and 1316 indicate the presence of hematite. 785 nm, 10% laser power, 30 seconds/acquisition, 5 acquisitions.

Page 60 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 76. Sample 1. Raman spectrum (upper panel) of layer 8 white bloom (lower panel). The peaks at 224, 291, and 408 indicate the presence of hematite. The peak at 435 indicates the presence of zinc oxide. 785 nm, 10% laser power, 60 seconds/acquisition, 10 acquisitions.

Page 61 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 77. Sample 2. Raman spectrum (upper panel) of greenish (lower panel). The peaks at

247, 299, and 383 indicate the presence of goethite. 785 nm, 1% laser power, 30 seconds/acquisition, 10 acquisitions, 1200/mm grating.

Page 62 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 78. Sample 2. Raman spectrum (upper panel) of red degradation interior (lower panel).

The peaks at 215, 249, 305, 341, 375, 525, and 650 indicate the presence of lepidocrocite. 785 nm, 10% laser power, 30 seconds/acquisition, 3 acquisitions, 1200/mm grating.

Page 63 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 79. Sample 2. Raman spectrum (upper panel) of red on metal 2 (lower panel). The peaks at 248, 297, and 386 indicate the presence of goethite. 785 nm, 1% laser power, 30

Page 64 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 80. Sample 4. Raman spectrum (upper panel) of red edge 3 (lower panel). The peaks at

215, 249, 309, 345, 378, 527, and 644 indicate the presence of lepidocrocite. 785 nm, 1% laser power, 30 seconds/acquisition, 10 acquisitions, 1200/mm grating.

Page 65 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Figure 81. Sample 4. Raman spectrum (upper panel) of red interior (lower panel). The peaks at

249, 299, and 383 indicate the presence of goethite. 785 nm, 1% laser power, 30

Page 66 of 67 © 2017 Scientific Analysis of Fine Art, LLC

Analytical Methodology

Visible and Ultraviolet Light Photomicroscopy

Samples were collected and mounted in styrene resin and cured overnight. Cross-sections were made and polished with Micro-Mesh silicon carbide in successive steps to 3600 grit (8 microns).

Visible and ultraviolet light microscopy was performed with a Ni-U microscope (Nikon

Instruments) with 10X oculars, 4X, 10X, and 20X Plan Fluor objectives, and a 2.5X F-mount adapter for a total magnification of 100X, 250X, and 500X. Samples were illuminated by a

SOLA SM II light engine (Lumencor) in visible or blue/violet illumination (395 nm). Images were acquired with a DS-Ri2 16.25MP scientific CMOS camera (Nikon Instruments) and NIS-

Elements F4.30 (Nikon Instruments).

Fourier Transform Infrared Spectroscopy

A size 11 steel blade scalpel was used to remove small (millimeter sized) samples for molecular analysis by FTIR (Fourier transform infrared spectroscopy). Samples were mounted on a diamond half-cell support for transmission mode. Data were acquired and analyzed with a

Thermo Scientific Nicolet iN-10 FTIR microscope controlled by OMNIC Picta software. A total of 128 scans were collected over a spectral range of 4000 – 650 cm-1 with a spectral resolution of

4 cm-1. A set of databases, including the IRUG (Infrared and Raman Users Group) database and a number of commercial databases, were used for identification.

Micro-Raman Spectroscopy

A Horiba XploRA PLUS Raman spectrometer was used with a 785 nm (red) diode laser operating at a laser power of 1% or 10% (300 mW total power) as indicated in each figure. The instrument was frequency calibrated with a silicon reference standard. Data were collected over a range of 100 cm-1 to 1500 cm-1 or 100 cm-1 to 2400 cm-1, and the spectral resolution of the instrument is 3 cm-1. A 50× objective was used for data collection.

Scanning Electron Microscopy – Energy Dispersive Spectroscopy

Samples were cut on an Isomet slow speed saw. They were then mounted in epoxy resin and ground and polished to a 1 µm finish. For a more detailed investigation of the microstructure of the rivet, the sample was etched by swabbing in Nital for 30 seconds.

A Zeiss Axio Imager A1m equipped with illuminators for reflected light, transmitted light and fluorescent microscopy was used for microscopic analysis. Images were taken using Zen 2

Page 67 of 67 © 2017 Scientific Analysis of Fine Art, LLC software. Samples were viewed under brightfield and darkfield. Objective magnifications range between 50x and 1000x for reflected light microscopy.

Secondary electron and backscatter electron images were obtained using a Tescan Vega3 XMU tungsten variable pressure scanning electron microscope. Samples were analyzed under high vacuum. The accelerating voltage of 15 kV was used. X-ray spectra and maps were collected and processed using an Oxford Instruments X-Maxn Silicon Drift Detector (SDD), energy dispersive spectroscopy with a 50 mm2 detector and AZtecEnergy analysis software. Samples were carbon coated to reduce surface charging.