|
|
|
Trace Element Constraints on the Source of Silica Sand Used by the Boston and Sandwich Glass Co. (c. 1826–1888), Massachusetts
J. Victor Owen, Katherine L. Irwin, Charles L. Flint, and John D. Greenough
Glass manufactured by the Boston and Sandwich Glass Company (BSGCo) is compositionally diverse. Most analyzed samples have alkali-lead (Na-K-Pb) compositions, but some have low Na contents (i.e., are potash-lead glass), whereas others have ternary Ca-Na-K compositions. BSGCo glass was made from silica that originated from (1) silica sand from southern New Jersey (NJ) and, starting in the 1850s, (2) Cheshire quartzite from western Massachusetts. Although ultrapure (>98 wt.% SiO2), these silica sources differ in their concentrations of niobium, a trace element concentrated in Ti and Fe-Ti oxide minerals. Most samples of NJ sand have no Nb anomaly on mantle-normalized diagrams, or they have a weak positive or negative one. In contrast, Cheshire quartzite has a pronounced negative Nb anomaly. So too do alkali-lead glass samples for which trace element data are available. These samples are interpreted to have been made using Cheshire quartzite or magnetically treated NJ sand from which Fe-Ti oxides have been removed. In contrast, selected samples of potassic soda-lime and potash-lead glass have only weak negative Nb anomalies, thereby indicating the use of untreated NJ sand in their manufacture. These interpretations are supported by the ratios of other trace elements (e.g., La/Ce) determined for the sand and glass. Chromium-bearing Fe oxides were detected in the quartzite but not in NJ sand, so potentially Cr can also be used to recognize glass made from these different sources of silica. In the case of black glass, however, the trace element signature of the silica has been masked by the Fe-Mn ore used to darken this medium. The source of sand that BSGCo used to make black glass remains undetermined.
|
|
|
| |
|
|
Introduction
|
|
One of the objectives of industrial archaeology is to gain insight into the technology of emerging industries. This not only includes the manner in which various products were manufactured but also includes the nature of raw materials and their sources. Such information is often found in factory records and, piecemeal, in contemporary print media. Questions can still persist. For example, factory records are likely to be incomplete, if they are preserved at all, and print media, then as now, can provide an even more distorted view of an industry's operation. For example, the manufacturers of mid-19th-century glass in the Como-Hudson area west of Montreal, Quebec, claimed that they located to that area because local sand was "better than the best that could be had in England."1 This sand, however, contains only about 71–77 wt.% silica (as opposed to 100% SiO2 in pure quartz sand). Mass-balance calculations using analytical data for the sand and for pale green and colorless Como-Hudson glass show that their compositions are inconsistent with one another.2 The reason for the apparent deception promulgated by the manufacturers of this glass is not known, but it does serve to caution researchers that the historical record is not always valid. Fortunately, in this instance, the veracity of the claims can readily be tested using analytical data for the glass and sand.
|
1
|
| Additional concerns can arise in situations where enterprises obtained the same raw material from multiple sources. Does this reflect economic factors, as various suppliers periodically underpriced their competitors or went out of business? Did technological reasons come into play? For example, in the glass industry, different types of glass can require different grades of silica sand. Fortunately, the compositional signatures of this raw material's trace elements are commonly inherited by the glass of which it is a part. Once identified, these peculiarities can potentially be used to authenticate finds; to better understand technical difficulties faced by historical manufacturers (e.g., how much and what type of decolorant was needed to produce transparent, colorless glass); and to constrain the age of artifacts from a particular glassworks' production line if the dates when particular raw materials were used are known. Consequently, analytical data for selected raw materials and the objects they were used to produce can be used to answer some though not all of the questions commonly posed by archaeologists.3 This study uses compositional data to distinguish between the two sources of silica sand known to have been used in the manufacture of a remarkably wide range of types of glass produced at the Boston and Sandwich Glass Company (BSGCo), a mid-19th-century glassworks in Cape Cod, Massachusetts. |
2
|
| |
|
|
Historical Background
|
|
In what is now the United States, the manufacture of glass was first attempted at Jamestown, Virginia, in the early-17th century, but commercial success was not achieved until the establishment of the Wistarburgh glassworks in southern New Jersey c. 1739. By the 19th century, many dozens of glassworks were established in this area, principally because of the abundance of fuel (firewood) and highly silicious glass-grade sand. Fewer glassworks were established in New England, but one of these, BSGCo, survived for several decades (c. 1826–1888).
|
3
|
| Deming Jarves, a Bostonian, incorporated BSGCo. He located the glassworks at Sandwich because of the suitability of its harbor and the local supply of wood for firing furnaces. Local sand, however, was insufficiently silicious for producing colorless glass, although it was used for initial experiments at the glassworks.4 Instead, Jarves imported silica sand from the Maurice River area of southern New Jersey (see figure 1A) and from the Berkshire Hills of western Massachusetts (figure 1B) by the 1850s, where both silica sand and highly silicious quartzite had been discovered.5 Although the expansion of the railway system made the Massachusetts material somewhat more affordable, it was still more expensive than New Jersey sand (i.e., $6.75/ton vs. $1.50/ton). For this reason, BSGCo continued to import New Jersey sand. Peak production at BSGCo was achieved in the 1860s, but once the workers organized, their union's demands could not be met, and the glassworks closed on 1 Jan. 1888. |
4
|
|
| |
|
Figure 1. Location of (A) the Maurice River area of southern New Jersey, and (B) the historic silica sand producing areas near Cheshire and Washington in the Berkshire Hills of western Massachusetts.
|
|
|
|
|
|
| |
|
|
Production Methods
|
|
The name Sandwich conjures up an image of lacy pressed glass, but BSGCo used many methods to shape its wares. The earliest products were free-blown and mold-blown objects, but soon Jarves had patented several improvements to a lever-operated machine for pressing glass into metal molds. This more efficient and less demanding approach lowered production costs and ensured the financial success of the operation. Mold pressing of glass was not without its difficulties, however, because the process tended to create surface imperfections as a result of the different cooling rates of glass and metal molds. Consequently, molds were made with a stippled pattern to help mask these flaws. These dots and circles lend a "lacy" aspect to the glass that has become the signature of BSGCo.
|
5
|
| BSGCo specialized in lead-rich glass, a medium patented in 1674 by Englishman George Ravenscroft (1618–1681). This glass is generally made from silica sand, lead oxide, and potash, with minor amounts of other ingredients. Lead not only acts (as does the potash) as a flux, lowering the melting temperature of the silica sand, but it also gives glass a high refractive index. The resulting brilliance and high luster of this glass make it particularly well suited for cutting and engraving. This glass is often, but not exclusively, used to produce lead crystal. Most of the lead-rich glass produced by BSGCo contains substantial amounts of soda (Na2O), so it is best described as an alkali-lead (Na-K-Pb) glass. Moreover, it has a very wide range of compositions, and its trace element signature, along with that of other types of glass the company manufactured, mirrors the source of silica sand used in the glass batches. |
6
|
| |
|
|
Geological Background
|
|
| |
|
Cheshire Quartzite
|
|
Cheshire quartzite formed when the Cambro-Ordovician (~542–444 million years [Ma] old) quartz sands deposited on the eastern edge of the Laurentian continental shelf were metamorphosed during the Taconic Orogeny (~450 Ma). It overlies Grenvillian gneisses.6 Accessory minerals can include iron oxides, rutile, tourmaline, and zircon. Cheshire quartzite's purity indicates that its source was far removed from the underlying rocks.7
|
7
|
Orthoquartzite was quarried sporadically in western Massachusetts during the 19th century. The quartzite quarry near Cheshire commenced operation in 1812 (possibly earlier). The nearby (20 km distant) quarry at Washington, located ~100 m NE of Sandwash Reservoir (see figure 2) at N42º22'17", W73º9'35" at an elevation of about 610 m, started up in 1845. Both quarries exploited Cheshire quartzite.
|
8
|
|
| |
|
Figure 2. The quartzite quarry and crushing facility near the town of Washington, Berkshire Hills, western Massachusetts.
|
|
|
|
|
|
Artifacts recovered from the Washington site provide evidence for the manner in which workers quarried and processed this resource. These artifacts include oxen and horse shoes, sledge hammers, crowbars, steel wedges, and feathers, the latter being flat pieces of steel that fit on each side of a wedge to keep it from getting stuck in the fissures during the quarrying operation. The miners likely freed blocks from the bedrock using black powder. The quarry consists of a large, hillside outcrop of quartzite. Nearby are one large and one small cellar hole, remnants of the buildings where the workers evidently lived. Other evidence of habitation connected to the quarrying operation is provided by numerous stone walls, which apparently outline the foundations of large outbuildings, including barns used for storage as well as to shelter oxen and horses along with the fodder they required (figure 2). There is also evidence of a blacksmith shop, no doubt required to furnish the miners with the tools of their trade. The miners probably excavated headstone-size rocks and loaded them onto wagons pulled by teams of oxen downhill to the washhouse (elevation 570 m) straddling nearby Roaring Brook, which drains Sandwash Reservoir; the name of the latter testifies to one of the objectives of this operation.
|
9
|
The washhouse foundation consists of large stone walls that straddled Roaring Brook, which evidently was dammed with very large blocks. The foundation has a small spillway on one side with a water chase for a waterwheel, possibly to power a crushing facility (see figure 3). Adjacent to the washhouse, mounds of quartzite blocks remain where the workers stockpiled quarried material destined for crushing. After being crushed and washed, the silica sand was taken by wheelbarrow to drying tables. The clean white sand, spilled as it was being moved, occurs below the black loam in the soil profile. After the sand was sun dried, it was shoveled into barrels and loaded onto wagons. The sand was then taken either to a glassworks in Lenox Dale village or to the Lenox Dale Railroad Station.
|
10
|
|
| |
|
Figure 3. The remains of the streamside facility where Cheshire quartzite was crushed. Photo by Charles L. Flint.
|
|
|
|
|
|
| The Roaring Brook site was active for approximately 24 years (c. 1854–1878). The earlier Cheshire quarry has not been located, but it is known to have been near Ashley Lake, a small pond just northwest of Cheshire. The Cheshire quarry supplied silica sand to two early, short-lived glass houses (1812–1814) in the area: Farmer Glass Works in Clarksburg and Cheshire Glass Works. Later during the 19th century, as quarrying activity expanded, silica sand derived from the Cheshire quartzite was sent not only to BSGCo but also to other glassworks, including those in Trenton, Nova Scotia.8 Despite all efforts to remove impurities, glass made from this sand nonetheless required the addition of decolorants (Mn or As oxides) to avoid a pale green or aqua hue. |
11
|
| |
|
New Jersey Silica Sand
|
|
In southern New Jersey, glass-grade silica sand occurs in the pre-Holocene, Cenozoic, Cohansey, and Cape May formations. Based on its mineralogy, the Cohansey has been subdivided into three provinces. From north to south, these are (1) ilmenite-bearing sand, (2) clayey sand, and (3) the Maurice River area high-silica sand in southernmost New Jersey (figure 1A).9 All are friable, poorly consolidated quartz sandstones that are easily reduced to loose sand. The Cohansey has a maximum thickness of ~30 m. Accessory minerals can include zircon, ilmenite, rutile, tourmaline, sillimanite, kyanite, and staurolite, with subordinate hornblende, chloritoid, garnet, epidote, monazite, hypersthene, and rare glauconite.10 These minerals originated in Precambrian gneisses and granite, with contributions from Triassic rocks.
|
12
|
The Cape May Formation overlies the Cohansey. Due to fluvial and marine reworking of underlying rocks, it contains similar heavy minerals as does the Cohansey. The Cape May Formation is up to about 40 m thick. Although highly silicious, sand from the Cohansey Formation it can contain high concentrations of dark, organic-rich, muddy material, which can be easily washed out, leaving behind the silica sand.
|
13
|
Two methods were used to quarry NJ silica sand. One approach simply involved shoveling the sand into carts or cars, whereupon it was hauled to the washer.11 This method worked best in areas where the sand had many impurities, as it allowed the impure layers to be selectively removed before washing and produced higher quality sand. The second method involved suspending the sand in water directly in the pit, where it was then drawn by pipelines to the washer. This method could only be used where sand was relatively free of impurities. The sand, with the accompanying clays and fine sands, was sieved into a 30-mesh screen, allowing grains too large for glassmaking to be separated from grains that were more suitable. The latter material was then washed with water to remove unduly fine-grained sand and clay. This method was much cheaper than the first and became widely used as time went on.
|
14
|
To help ensure that glass made from this sand would be nearly colorless, attempts were made to remove magnetite and Fe-bearing Ti oxides (notably ilmenite) magnetically, a process not known to have been undertaken in the Berkshire Hills. It is unknown how efficient this process was or if it was routinely done during the time when BSGCo was in production.
|
15
|
| One would expect that glass made from the cleansed NJ material would be significantly depleted in trace elements associated with these minerals (notably Nb in rutile and Ti-rich Fe oxides) compared with untreated samples. As will be seen, only some BSGCo glass shows this feature; other samples appear to have been made from untreated (or inefficiently treated) NJ sand. |
16
|
| |
|
|
Sample Descriptions
|
|
| |
|
Cheshire Quartzite
|
|
Five samples of quartzite (CQ1 to CQ5) and three samples of sand (Berksand 1, CQ 6WS, and CQ 7WS) were taken from the site of the old quarry and the nearby streamside crushing facility, respectively. Quartzite samples CQ1, CQ3, CQ4, and CQ5 contain visible dark minerals that define thin laminations in the rock. In contrast, CQ2 is an extremely clean, white quartzite and has no visible bedding or dark minerals. Based on electron probe microanalysis and SEM compositional maps of heavy mineral separates, accessory minerals in the quartzite include, in approximate descending order of modal (vol.%) abundance, zircon, ilmenite, magnetite, Cr-bearing Fe oxide (probably a spinel), and staurolite.12 Heavy liquid separation of these minerals indicates that they constitute 0.02 to 0.17 wt.% of the quartzite.
|
17
|
| The three sand samples were obtained from beneath the loam by the remains of the crushing facility on Roaring Brook (figure 2). Evidently, the samples represent material that had been spilled after the quartzite had been crushed and washed. In stark contrast to the black soil and underlying bedrock, it is gleaming white. |
18
|
| |
|
New Jersey Sand
|
|
| The eight samples (NJ1 to NJ8) analyzed here were collected along a south-to-north traverse from East Point at the mouth of the Maurice River (figure 1A) to north of Vineland (see appendix 1). Samples NJ1 and NJ2 are Holocene beach sands derived from the underlying Cohansey Formation; the other samples were taken from the Cohansey itself. Accessory minerals in the sand include, in approximate descending order of modal abundance, zircon, ilmenite, magnetite, rutile, and staurolite. Heavy liquid separation of these minerals indicates that they constitute 0.11 to 0.65 wt.% of the sand. |
19
|
| |
|
Sandwich Glass
|
|
| The analyzed glass samples are all fragments of shaped (worked) objects (see appendix 2). The glass can be transparent to nearly opaque and can be colorless, white, black, red, blue, or green. As will be seen, the range of colors shown by the glass is mirrored by a remarkable range in their chemical compositions. |
20
|
| |
|
|
Analytical Results
|
|
Major elements and selected trace elements in the sand and quartzite samples were determined by X-ray fluorescence. Other trace elements were determined by inductively coupled plasma-mass spectrometry (ICP-MS). It should be noted that the list of elements determined is not all-inclusive and so omits some of the components used to color glass (e.g., gold and copper). It does, however, include the range of trace elements judged to be of potential use in characterizing the sources of silica used in the manufacture of BSGCo glass.
|
21
|
| Major and selected trace element glass compositions were determined by electron microprobe. Other trace elements were determined in selected samples by ICP-MS. Analytical methods are described in appendix 3. |
22
|
| |
|
Cheshire Quartzite and New Jersey Sand
|
|
Although both are remarkably silicious, the average composition of Cheshire quartzite and NJ sand show some differences in their major and trace element signatures. NJ sand averages 99.6% SiO2, 1% higher than the quartzite, and has lower alumina (0.12 vs. 0.64% Al2O3) and potash (0.01 vs. 0.13% K2O) contents, indicating the presence of a higher proportion of alkali feldspar components in the latter (see table 1).13 Although they have similar concentrations (~0.1%) of both TiO2 and Fe2O3 (i.e., Fe-Ti oxide components), they differ in their Nb contents, the NJ sand being enriched in this component by, on average, a factor of 2.3 (i.e., 1.83 vs. 0.80 ppm) (see table 2). Both have similar average Zr contents (~90 ppm), but the quartzite shows a nearly threefold enrichment in Y (2.52 vs. 0.94 ppm) (see figure 4) and most of the rare earth elements (REEs) (see figure 5). Although differing in the concentration of the REEs, both NJ sand and quartzite have similar REE patterns (figure 5) with pronounced negative Eu anomalies and flat to shallowly positive sloping heavy REE signatures. In addition to the REEs, the quartzite is also enriched in Rb (1.87 vs. 0.56 ppm), consistent with its higher potassium content. Consequently, it lacks the pronounced negative Rb anomaly that characterizes the pattern for NJ sand shown on a mantle-normalized multi-element ("spider") diagram (figure 4).
|
23
|
|
| |
|
Figure 4. Primitive mantle-normalized multi-element diagram for (A) NJ sand and (B) Berkshire quartzite.
|
|
|
|
|
|
|
| |
|
Figure 5. Chondrite-normalized REE diagram for NJ sand and Berkshire quartzite. Chondrite data are after N. Nakamura, "Determination of REE Ba, Fe, Mg, Na, and K in Carbonaceous and Ordinary Chondrites," Geochim, Cosmochim Acta 38(1974): 757–75.
|
|
|
|
|
|
Table 1
Major and selected trace element geochemistry of Cheshire quarzite and New Jersey silica sand |
| SiO2 (wt%) |
TiO2 |
Al2O3 |
Fe2O3 |
MnO |
MgO |
CaO |
Na2O |
K2O |
P2O5 |
L.O.I. |
V (ppm) |
Zr |
Ba |
Zn |
Nb |
| Cheshire quartzite |
| CQ1 |
98.3 |
0.088 |
0.52 |
0.09 |
0.001 |
<0.01 |
0.04 |
<0.01 |
0.16 |
0.008 |
0.75 |
12 |
97 |
165 |
13 |
1 |
| CQ2 |
98.7 |
0.059 |
0.37 |
0.01 |
0.001 |
<0.01 |
0.04 |
<0.01 |
0.04 |
0.004 |
0.77 |
7 |
16 |
155 |
11 |
<1 |
| CQ3 |
98.9 |
0.084 |
0.49 |
0.12 |
0.001 |
<0.01 |
0.04 |
<0.01 |
0.13 |
0.007 |
0.21 |
10 |
78 |
147 |
7 |
2 |
| CQ4 |
98.3 |
0.081 |
0.70 |
0.36 |
0.001 |
<0.01 |
0.04 |
<0.01 |
0.11 |
0.011 |
0.41 |
12 |
71 |
134 |
7 |
3 |
| CQ5 |
99.0 |
0.093 |
0.57 |
0.07 |
0.001 |
<0.01 |
0.04 |
<0.01 |
0.14 |
0.007 |
0.10 |
13 |
99 |
113 |
14 |
2 |
| CQ6 |
98.2 |
0.105 |
1.19 |
0.05 |
0.002 |
<0.01 |
0.05 |
<0.01 |
0.16 |
0.008 |
0.24 |
14 |
125 |
146 |
11 |
3 |
| CQ7 |
98.6 |
0.096 |
0.63 |
0.09 |
0.001 |
<0.01 |
0.05 |
<0.01 |
0.14 |
0.007 |
0.37 |
13 |
93 |
149 |
11 |
3 |
| New Jersey sand |
| NJ-1 |
98.4 |
0.398 |
0.58 |
0.49 |
0.003 |
<0.01 |
0.02 |
<0.01 |
0.05 |
0.014 |
0.00 |
34 |
339 |
35 |
17 |
1 |
| NJ-2 |
99.4 |
0.044 |
0.20 |
0.07 |
<0.001 |
<0.01 |
0.01 |
<0.01 |
0.02 |
0.010 |
0.29 |
5 |
39 |
<25 |
7 |
<1 |
| NJ-3 |
99.3 |
0.104 |
0.19 |
0.14 |
<0.001 |
<0.01 |
0.01 |
<0.01 |
0.01 |
0.007 |
0.20 |
6 |
71 |
<25 |
8 |
<1 |
| NJ-4 |
98.5 |
0.326 |
0.55 |
0.27 |
0.002 |
<0.01 |
0.01 |
<0.01 |
0.03 |
0.011 |
0.30 |
25 |
244 |
<25 |
9 |
2 |
| NJ-5 |
99.4 |
0.135 |
0.18 |
0.15 |
<0.001 |
<0.01 |
0.01 |
<0.01 |
0.02 |
0.008 |
0.10 |
8 |
94 |
<25 |
7 |
<1 |
| NJ-6 |
98.5 |
0.454 |
0.40 |
0.39 |
0.004 |
<0.01 |
0.01 |
<0.01 |
0.04 |
0.011 |
0.20 |
34 |
322 |
<25 |
11 |
1 |
| NJ-7 |
99.3 |
0.168 |
0.22 |
0.23 |
<0.001 |
<0.01 |
0.01 |
<0.01 |
0.03 |
0.009 |
0.00 |
14 |
131 |
27 |
8 |
<1 |
| NJ-8 |
99.6 |
0.074 |
0.12 |
0.10 |
<0.001 |
<0.01 |
0.01 |
<0.01 |
0.01 |
0.007 |
0.10 |
<4 |
46 |
<25 |
8 |
<1 |
| LOI: loss on ignition |
| Note: XRF data for selected trace elements were below detection limits (Cr < 4 ppm, Ga < 5 ppm, Ni < 3 ppm, Sr < 5 ppm). |
|
|
Table 2
Trace element geochemistry of Cheshire quartzite, crushed quartzite, and New Jersey silica sand |
| |
Rb (ppm) |
Sr |
Y |
Zr |
Nb |
Cs |
La |
Hf |
Ta |
Ce |
Pr |
Nd |
Sm |
Eu |
Gd |
Tb |
Dy |
Ho |
Er |
Tm |
Yb |
Lu |
Th |
U |
| Detection Limit: |
0.05 |
0.5 |
0.02 |
4.0 |
0.2 |
0.007 |
0.02 |
0.1 |
0.2 |
0.07 |
0.006 |
0.03 |
0.01 |
0.005 |
0.009 |
0.003 |
0.008 |
0.003 |
0.008 |
0.003 |
0.01 |
0.003 |
0.06 |
0.007 |
| Cheshire quartzite |
| CQ1 |
1.80 |
3.4 |
2.40 |
85.4 |
0.7 |
0.034 |
3.64 |
2.3 |
N.D. |
8.39 |
0.804 |
2.98 |
0.49 |
0.082 |
0.365 |
0.059 |
0.408 |
0.091 |
0.295 |
0.051 |
0.35 |
0.059 |
0.77 |
0.307 |
| CQ3 |
1.32 |
3.3 |
2.33 |
80.9 |
0.6 |
0.028 |
3.87 |
2.1 |
N.D. |
9.50 |
0.937 |
3.57 |
0.60 |
0.091 |
0.440 |
0.064 |
0.401 |
0.090 |
0.271 |
0.045 |
0.33 |
0.055 |
0.78 |
0.339 |
| CQ4 |
1.55 |
1.9 |
3.57 |
75.7 |
0.6 |
0.029 |
3.70 |
2.0 |
N.D. |
8.66 |
0.954 |
3.86 |
0.76 |
0.162 |
0.688 |
0.109 |
0.661 |
0.130 |
0.380 |
0.056 |
0.40 |
0.060 |
0.81 |
0.350 |
| CQ5 |
1.91 |
4.3 |
2.45 |
105.9 |
0.9 |
0.036 |
3.71 |
2.7 |
N.D. |
8.34 |
0.833 |
3.00 |
0.51 |
0.077 |
0.396 |
0.066 |
0.417 |
0.087 |
0.276 |
0.046 |
0.34 |
0.055 |
0.79 |
0.363 |
| CQ5 dupl* |
1.88 |
4.3 |
2.44 |
88.1 |
0.8 |
0.031 |
3.56 |
2.3 |
N.D. |
8.00 |
0.799 |
2.99 |
0.51 |
0.081 |
0.387 |
0.064 |
0.409 |
0.087 |
0.279 |
0.045 |
0.32 |
0.049 |
0.75 |
0.325 |
| Crushed quartzite from 19th century crushing mill |
| Berksand1 |
1.79 |
1.9 |
2.37 |
70.0 |
1.0 |
0.044 |
3.06 |
1.9 |
N.D. |
7.04 |
0.706 |
2.78 |
0.46 |
0.104 |
0.379 |
0.062 |
0.411 |
0.090 |
0.275 |
0.041 |
0.30 |
0.047 |
0.72 |
0.286 |
| CQ6WS |
1.77 |
1.5 |
2.38 |
90.9 |
0.8 |
0.041 |
3.66 |
2.3 |
N.D. |
7.74 |
0.858 |
3.18 |
0.60 |
0.096 |
0.511 |
0.075 |
0.442 |
0.093 |
0.297 |
0.046 |
0.32 |
0.050 |
0.73 |
0.278 |
| CQ7WS |
2.34 |
1.5 |
2.29 |
81.2 |
0.9 |
0.049 |
3.67 |
2.1 |
N.D. |
8.22 |
0.835 |
3.15 |
0.55 |
0.093 |
0.463 |
0.072 |
0.434 |
0.088 |
0.288 |
0.044 |
0.30 |
0.049 |
0.76 |
0.296 |
| CQ7WS dupl* |
2.44 |
1.5 |
2.42 |
88.0 |
0.9 |
0.049 |
3.33 |
2.3 |
N.D. |
7.50 |
0.806 |
2.96 |
0.53 |
0.099 |
0.431 |
0.067 |
0.445 |
0.093 |
0.301 |
0.045 |
0.32 |
0.054 |
0.71 |
0.322 |
| New Jersey silica sand |
| NJ1 |
1.66 |
4.5 |
1.69 |
175.2 |
3.7 |
0.129 |
2.60 |
4.2 |
0.21 |
5.10 |
0.590 |
2.26 |
0.40 |
0.055 |
0.313 |
0.047 |
0.298 |
0.064 |
0.211 |
0.036 |
0.26 |
0.047 |
0.82 |
0.382 |
| NJ2 |
0.46 |
2.1 |
0.71 |
44.6 |
0.5 |
0.087 |
1.10 |
1.2 |
N.D. |
2.10 |
0.233 |
0.88 |
0.17 |
0.021 |
0.150 |
0.021 |
0.122 |
0.026 |
0.087 |
0.014 |
0.10 |
0.016 |
0.29 |
0.162 |
| NJ3 |
0.47 |
2.4 |
0.71 |
46.9 |
0.5 |
0.096 |
1.02 |
1.2 |
N.D. |
1.90 |
0.227 |
0.80 |
0.14 |
0.022 |
0.140 |
0.019 |
0.125 |
0.026 |
0.085 |
0.013 |
0.10 |
0.016 |
0.27 |
0.142 |
| NJ4 |
0.23 |
1.7 |
0.61 |
65.6 |
0.9 |
0.095 |
1.08 |
1.5 |
N.D. |
2.05 |
0.224 |
0.78 |
0.14 |
0.016 |
0.107 |
0.017 |
0.098 |
0.022 |
0.077 |
0.013 |
0.10 |
0.018 |
0.29 |
0.210 |
| NJ5 |
0.51 |
2.4 |
1.05 |
122.9 |
2.0 |
0.104 |
1.54 |
2.8 |
N.D. |
2.89 |
0.318 |
1.17 |
0.19 |
0.025 |
0.167 |
0.026 |
0.158 |
0.036 |
0.124 |
0.020 |
0.16 |
0.029 |
0.49 |
0.207 |
| NJ6 |
0.24 |
2.0 |
0.72 |
66.9 |
1.6 |
0.091 |
0.99 |
1.6 |
N.D. |
1.84 |
0.194 |
0.72 |
0.13 |
0.020 |
0.112 |
0.017 |
0.118 |
0.028 |
0.092 |
0.015 |
0.11 |
0.020 |
0.26 |
0.157 |
| NJ7 |
0.73 |
2.2 |
0.99 |
127.0 |
2.1 |
0.100 |
1.28 |
3.0 |
N.D. |
2.36 |
0.278 |
1.06 |
0.18 |
0.027 |
0.155 |
0.025 |
0.167 |
0.037 |
0.123 |
0.021 |
0.16 |
0.029 |
0.37 |
0.228 |
| NJ8 |
0.20 |
1.3 |
0.78 |
51.7 |
1.1 |
0.078 |
0.78 |
1.3 |
N.D. |
1.46 |
0.164 |
0.62 |
0.11 |
0.016 |
0.110 |
0.019 |
0.118 |
0.029 |
0.091 |
0.015 |
0.10 |
0.017 |
0.22 |
0.134 |
| NJ9 |
0.56 |
2.6 |
1.20 |
155.0 |
4.1 |
0.115 |
1.71 |
3.6 |
0.24 |
3.44 |
0.375 |
1.38 |
0.24 |
0.028 |
0.195 |
0.032 |
0.185 |
0.046 |
0.146 |
0.026 |
0.20 |
0.036 |
0.62 |
0.274 |
ND-not detected
dupl*: duplicate analysis of sample |
|
|
NJ sand has higher La/Ce ratios than the quartzite (i.e., La/Ce = 0.50–0.54 vs. 0.41–0.47, respectively). Consequently, data for both sources of silica cluster in separate fields on plots that compare La/Ce with their Tb/Lu, Nb/Ce, Nb/Y and Y/Ho ratios (see figure 6).
|
24
|
|
| |
|
Figure 6. Distinction between Cheshire quartzite and NJ sand and the glass made there on the basis of their La/Ce vs. (A) Tb/Lu, (B) Nb/Ce, (C) Nb/Y, and (D) Y/Ho ratios. Note that (A), (B), and (C) are semi-log plots and that black glass sample BS11 plots as an outlier in each diagram.
|
|
|
|
|
|
| The quartzite and NJ sand contain similar accessory minerals, but Cr-bearing Fe-rich oxide was identified only in the quartzite. Ilmenite in the NJ sand can contain up to 2.8% MnO, whereas no analyzed ilmenites from the quartzite were seen to contain more than 1% of this component. Like manganese and niobium, chromium can substitute for Fe and Ti in ilmenite and can be incorporated into this mineral during chemical weathering.14 This component was not detected in the ilmenites from either locality. |
25
|
| |
|
Glass
|
|
| Sandwich glass is unusual because the samples described here span a wide range of compositions. Indeed, a bewildering variety of glass batch recipes used by BSGCo were reported in a series of monthly articles published by A. Silverman, and transcriptions of original BSGCo notebooks have been compiled more recently by Pamela Vandiver.15 Most glass factories produced either soda-lime or potash-lead glass with a restricted range of compositions. Sandwich glass, in contrast, comprises variations of alkali-lead (Na-K-Pb), potash-lead (K-Pb), and lead-poor, variably potassic soda-lime (Na-Ca) glass. The latter is represented by white (milk) glass (BS5) (see appendix 2) and parts of the red-colored bands in two samples (BS20, BS21) (see figure 7) of Vasa Murrhina, a multicolored glass often containing metallic chips. |
26
|
|
| |
|
Figure 7. Backscattered electron image of Vasa Murrhina sample BS20, showing potassic soda lime glass (dark grey band) associated with lead-rich (42% PbO) red glass (narrow white band), enclosed by less lead-rich (28% PbO), clear and blue-green glass (medium grey, upper left and lower right, respectively).
|
|
|
|
|
|
Twenty-one samples of Sandwich glass were analyzed for major elements by electron microprobe (see table 3). Trace elements were determined by ICP-MS for six samples (see table 4) spanning the range of major element compositions identified by microprobe. To facilitate systematic classification of this diverse glass, two adjoining ternary diagrams (Ca-Na-K, Pb-Na-K) have been devised to distinguish between variably sodic potash-lead glass and its calcic counterpart (see figure 8). Including colored bands present on some samples, the glass shows a remarkable range of compositions, particularly in terms of silica (43–72% SiO2) and lead (0.2–49% PbO). Soda-lime (Na2O/CaO) and potash-soda (K2O/Na2O) ratios range between 0.3–47.0 and 0.1–16.0, respectively. The sample suite is dominated by various types of alkali-lead glass, so most analyses plot in the Na-K-Pb field as shown in figure 8. Several samples, however, have sufficiently low Na contents so that they can be classified as potash-lead glass. Analyses of milk glass and part of the red bands in Vasa Murrhina glass (figure 7) with low lead contents (i.e., Pb/[Ca+Na+K+Pb] < 0.01) plot in the interior of the Ca-Na-K field in figure 8.
|
27
|
|
| |
|
Figure 8. Compositions of BSGCo glass plotted on Pb-Na-K and Ca-Na-K diagrams. XPb=Pb/(Pb+Ca+Na+K). Black dot is black glass sample BS11.
|
|
|
|
|
|
Table 3
Composition of Sandwich glass |
| |
SiO2 (wt%) |
TiO2 |
Al2O3 |
FeO |
MnO |
MgO |
CaO |
Na2O |
K2O |
PbO |
P2O5 |
Cl |
CoO |
As2O5 |
BaO |
SO3 |
Sb2O5 |
SeO2 |
Total |
| BS1* |
50.80 |
0.05 |
0.21 |
0.08 |
0.01 |
|
|
0.28 |
11.11 |
36.82 |
|
0.23 |
|
0.02 |
0.08 |
0.31 |
|
|
100.0 |
| BS2 |
59.65 |
0.14 |
0.30 |
0.19 |
0.13 |
0.09 |
0.32 |
2.49 |
10.27 |
25.33 |
|
0.23 |
|
0.27 |
0.41 |
0.18 |
|
|
100.0 |
| BS3 blue |
63.31 |
0.14 |
1.84 |
0.23 |
0.17 |
0.12 |
4.32 |
3.77 |
7.09 |
17.33 |
0.11 |
0.18 |
0.19 |
0.65 |
0.20 |
0.22 |
|
0.14 |
100.0 |
| BS4 green |
66.88 |
0.07 |
0.50 |
0.23 |
0.14 |
0.13 |
2.12 |
12.08 |
3.09 |
13.67 |
|
0.27 |
0.12 |
0.27 |
0.13 |
0.31 |
|
0.03 |
100.0 |
| BS5 white |
71.92 |
0.05 |
4.82 |
|
0.09 |
0.06 |
8.72 |
12.22 |
1.47 |
0.17 |
|
0.09 |
|
0.31 |
|
0.11 |
|
|
100.0 |
| BS6 white |
67.83 |
|
1.75 |
|
0.01 |
|
5.65 |
1.66 |
9.28 |
13.73 |
|
0.09 |
|
|
|
|
0.02 |
|
100.0 |
| BS6 red |
49.81 |
0.12 |
1.25 |
0.46 |
0.43 |
0.13 |
0.41 |
3.79 |
2.97 |
39.06 |
|
0.14 |
0.15 |
|
0.19 |
0.09 |
0.89 |
0.14 |
100.0 |
| BS7 |
54.07 |
0.06 |
0.45 |
0.07 |
0.06 |
0.01 |
0.20 |
2.50 |
6.85 |
35.42 |
|
0.14 |
|
|
0.01 |
0.06 |
0.08 |
0.05 |
100.0 |
| BS7red |
49.53 |
0.08 |
0.42 |
0.11 |
0.28 |
0.02 |
0.09 |
2.01 |
6.30 |
39.87 |
|
0.20 |
0.97 |
|
|
|
0.10 |
0.05 |
100.0 |
| BS8 |
57.34 |
0.22 |
0.48 |
0.23 |
0.30 |
0.04 |
0.13 |
4.71 |
6.66 |
28.08 |
0.19 |
0.10 |
0.34 |
0.28 |
0.40 |
0.24 |
0.02 |
0.28 |
100.0 |
| BS9 |
56.71 |
0.02 |
0.22 |
0.02 |
0.12 |
|
0.05 |
9.72 |
4.03 |
28.66 |
|
0.28 |
0.02 |
|
0.04 |
0.02 |
0.01 |
0.04 |
100.0 |
| BS9 blue |
56.59 |
0.02 |
0.22 |
0.10 |
0.12 |
0.01 |
0.06 |
9.72 |
3.70 |
28.89 |
|
0.29 |
0.04 |
|
0.05 |
0.03 |
0.09 |
0.05 |
100.0 |
| BS10 |
61.02 |
|
0.40 |
|
0.03 |
|
0.75 |
3.98 |
7.61 |
25.99 |
|
0.17 |
|
|
|
|
|
|
100.0 |
| BS10 blue |
46.37 |
0.07 |
0.45 |
0.17 |
0.15 |
0.03 |
0.09 |
0.44 |
6.29 |
45.45 |
|
0.10 |
0.06 |
|
0.03 |
0.03 |
0.20 |
0.06 |
100.0 |
| BS11 black |
56.97 |
0.08 |
0.44 |
8.52 |
6.14 |
0.07 |
4.79 |
8.10 |
10.65 |
2.90 |
|
0.19 |
0.10 |
0.66 |
0.12 |
0.31 |
|
|
100.0 |
| BS12 |
67.65 |
|
0.30 |
|
|
0.03 |
2.58 |
13.17 |
6.20 |
9.38 |
|
0.40 |
|
|
|
0.37 |
|
|
100.1 |
| BS13 green-blue |
62.27 |
0.12 |
0.46 |
0.33 |
0.11 |
0.11 |
1.12 |
5.51 |
10.58 |
18.19 |
|
0.19 |
0.17 |
0.32 |
0.20 |
0.21 |
|
0.16 |
100.1 |
| BS14 |
60.62 |
|
0.40 |
0.09 |
0.10 |
0.04 |
0.73 |
4.36 |
8.09 |
25.26 |
|
0.14 |
0.02 |
|
0.02 |
0.07 |
0.09 |
0.02 |
100.1 |
| BS14 red |
46.48 |
0.15 |
0.45 |
0.20 |
0.28 |
0.07 |
0.09 |
0.34 |
6.32 |
44.46 |
|
0.25 |
0.08 |
|
0.28 |
0.09 |
0.29 |
0.24 |
100.1 |
| BS15 blue |
72.34 |
0.03 |
0.93 |
0.13 |
0.14 |
0.05 |
2.46 |
14.77 |
2.11 |
6.48 |
|
0.16 |
0.01 |
0.23 |
0.02 |
0.18 |
|
|
100.0 |
| BS16 blue-green |
50.91 |
0.24 |
2.40 |
0.46 |
0.27 |
0.12 |
4.37 |
2.34 |
5.79 |
30.96 |
0.35 |
0.08 |
0.33 |
0.49 |
0.38 |
0.26 |
|
0.27 |
100.0 |
| BS17 green |
59.65 |
0.14 |
0.45 |
0.19 |
0.16 |
0.07 |
1.93 |
6.11 |
5.43 |
24.75 |
0.01 |
0.19 |
0.19 |
0.26 |
0.28 |
0.18 |
0.04 |
0.02 |
100.1 |
| BS18 green-blue |
58.85 |
0.17 |
0.43 |
0.39 |
0.20 |
0.08 |
1.72 |
5.25 |
5.32 |
26.24 |
|
0.18 |
0.26 |
0.31 |
0.27 |
0.15 |
|
0.21 |
100.0 |
| Vasa Murrhina |
| BS19 |
53.71 |
0.04 |
0.16 |
0.13 |
0.21 |
0.02 |
0.04 |
0.71 |
10.43 |
33.82 |
|
0.15 |
0.05 |
|
0.12 |
0.05 |
0.25 |
0.15 |
100.0 |
| BS19 red |
39.21 |
0.14 |
0.16 |
0.30 |
0.21 |
0.05 |
0.54 |
1.84 |
7.38 |
46.34 |
|
0.17 |
0.24 |
1.96 |
0.31 |
0.17 |
0.59 |
0.36 |
100.0 |
| BS19 blue |
53.75 |
0.05 |
0.15 |
1.29 |
0.07 |
0.01 |
0.06 |
0.79 |
11.44 |
31.75 |
|
0.14 |
0.11 |
|
0.11 |
0.03 |
0.22 |
0.09 |
100.1 |
| BS20 |
58.72 |
0.01 |
0.13 |
0.13 |
0.08 |
|
0.02 |
0.90 |
10.56 |
28.97 |
|
0.13 |
0.04 |
|
0.05 |
0.05 |
0.20 |
0.04 |
100.0 |
| BS20 red |
40.24 |
0.13 |
0.20 |
0.29 |
0.17 |
0.08 |
0.61 |
5.26 |
4.61 |
42.01 |
|
0.24 |
0.21 |
4.70 |
0.26 |
0.18 |
0.39 |
0.39 |
100.0 |
| BS20 red |
61.04 |
|
1.51 |
2.33 |
0.30 |
1.79 |
9.22 |
12.95 |
8.25 |
1.12 |
0.22 |
0.99 |
|
|
|
0.10 |
0.20 |
|
100.0 |
| BS20 blue |
56.54 |
0.02 |
0.16 |
0.93 |
0.11 |
0.01 |
0.04 |
0.92 |
10.99 |
29.64 |
|
0.11 |
0.02 |
0.02 |
0.08 |
0.02 |
0.45 |
0.02 |
100.1 |
| BS21 |
56.30 |
0.02 |
0.15 |
0.11 |
0.09 |
|
0.02 |
0.94 |
11.44 |
30.36 |
|
0.13 |
0.01 |
|
0.07 |
0.02 |
0.29 |
0.09 |
100.0 |
| BS21 red |
43.39 |
0.14 |
0.15 |
0.21 |
0.28 |
0.06 |
0.16 |
1.68 |
5.55 |
44.91 |
0.01 |
0.22 |
0.37 |
1.09 |
0.31 |
0.13 |
1.01 |
0.38 |
100.1 |
| BS21 red |
61.11 |
|
1.55 |
2.34 |
0.33 |
1.77 |
9.33 |
11.61 |
9.70 |
1.10 |
0.27 |
0.99 |
|
|
|
0.12 |
|
|
100.2 |
| BS21 blue |
56.85 |
0.01 |
0.20 |
0.89 |
0.13 |
0.01 |
0.04 |
0.90 |
10.92 |
29.34 |
|
0.12 |
0.02 |
0.03 |
0.10 |
0.04 |
0.39 |
0.04 |
100.0 |
| *Note: Unless otherwise indicated, the analyzed glass is colorless. |
|
|
Table 4
Trace element composition of selected samples of Sandwich glass |
| |
Rb (ppm) |
Sr |
Y |
Zr |
Nb |
Cs |
La |
Hf |
Ta |
Ce |
Pr |
Nd |
Sm |
Eu |
Gd |
Tb |
Dy |
Ho |
Er |
Tm |
Yb |
Lu |
Th |
U |
| BS1 |
>150 |
4.20 |
0.83 |
61.0 |
1.00 |
0.912 |
1.28 |
1.40 |
0.24 |
2.32 |
0.251 |
0.90 |
0.17 |
0.029 |
0.130 |
0.022 |
0.134 |
0.029 |
0.100 |
0.016 |
0.11 |
0.018 |
0.34 |
0.231 |
| BS5 |
65.7 |
34.10 |
4.96 |
48.5 |
2.70 |
4.120 |
3.42 |
1.30 |
0.65 |
6.32 |
0.746 |
2.74 |
0.62 |
0.196 |
0.632 |
0.106 |
0.674 |
0.132 |
0.394 |
0.058 |
0.38 |
0.058 |
1.77 |
2.310 |
| BS11 |
>150 |
129.00 |
35.50 |
92.0 |
1.40 |
0.302 |
13.9 |
2.10 |
0.24 |
21.10 |
4.600 |
21.3 |
5.79 |
1.400 |
7.020 |
1.100 |
6.390 |
1.290 |
3.580 |
0.497 |
2.98 |
0.431 |
0.69 |
0.906 |
| BS18 |
45.7 |
29.90 |
2.14 |
102.0 |
0.80 |
0.100 |
2.77 |
2.50 |
0.19 |
6.61 |
0.711 |
2.69 |
0.52 |
0.093 |
0.419 |
0.064 |
0.383 |
0.080 |
0.257 |
0.040 |
0.29 |
0.046 |
0.68 |
2.270 |
| BS7 |
37.6 |
4.30 |
2.22 |
110.0 |
0.80 |
0.140 |
3.18 |
2.70 |
0.20 |
7.34 |
0.791 |
2.92 |
0.60 |
0.108 |
0.445 |
0.068 |
0.406 |
0.083 |
0.265 |
0.043 |
0.30 |
0.050 |
0.72 |
0.764 |
| BS17 |
50.6 |
23.30 |
2.32 |
104.0 |
0.80 |
0.133 |
2.97 |
2.60 |
0.20 |
6.73 |
0.727 |
2.75 |
0.53 |
0.099 |
0.427 |
0.066 |
0.404 |
0.086 |
0.270 |
0.043 |
0.31 |
0.051 |
0.70 |
>20 |
|
|
The six samples chosen for trace element analysis represent the main compositional groupings of Sandwich glass. Sample BS1 is a colorless, potash-lead glass, containing about 51% SiO2, 11% K2O, and 37% PbO (table 3). Sample BS5 is a slightly potassic soda-lime milk glass with one of the highest amounts of SiO2 (~72%). It is also strongly enriched in alumina (4.8% Al2O3). Samples BS7, BS17, and BS18 have alkali-lead compositions with about 25–35% PbO (table 3). Black glass sample BS11 is a lead-bearing, potash-soda glass.
|
28
|
The three ternary Na-K-Pb samples (BS7, BS17, and BS18) for which high precision trace-element data are available show good correspondence in their trace element signatures, with nearly coincident REE patterns (see figure 9A) as well as other trace elements, although their uranium contents can vary (see figure 10A). Each shows a pronounced negative Nb anomaly (figure 10A). They also cluster in a tight field among data for the quartzite on plots comparing La/Ce with the ratios of other trace elements (figures 6A to 6D).
|
29
|
|
| |
|
Figure 9. Chondrite-normalized REE diagram for BSGCo glass.
|
|
|
|
|
|
|
| |
|
Figure 10. Primitive mantle-normalized multi-element diagram for BSGCo glass. Mantle data are after S. S. Sun and W. F. McDonough, "Chemical and Isotopic Systematics of Oceanic Basalts: Implications for Mantle Composition and Processes," Magmatism in the Ocean Basins, ed. A. D. Saunders and M. J. Norry (London: Geological Society of London, 1989), 313–45.
|
|
|
|
|
|
Potash-lead sample BS1 is depleted in trace elements compared with the other types of glass, but its REE pattern (figure 9B) mirrors those of the alkali-lead glass samples. In contrast, the REE pattern of the potassic soda-lime glass (BS5), unlike all the other samples, has no Eu anomaly. Moreover, it has a relatively weak negative Nb anomaly (figure 10B). Sample BS1 plots among data for NJ sand in figures 6A to 6D, but BS5 plots either in or at the end of the linear array of the sand data (figures 6B, 6C) along an extension of this array (figure 6A) or as an outlier (figure 6D).
|
30
|
| The most distinctive sample is BS11. This sample of black glass has a slightly plumbian (2.9% PbO), potash-soda composition (figure 8). Like many black glasses, it is enriched in both iron and manganese, but unusually so (~8% FeO, ~6% MnO). The latter component is a decolorant which, when used in excess as here, renders glass very dark (as does the iron). It also accounts for the negative Ce anomaly on a chondrite-normalized REE diagram (figure 9A), this signature no doubt being inherited from the original iron-manganese ore.16 Given the amount of Fe-Mn ore used in its manufacture, not surprisingly this sample is also unusually enriched in the REEs and in high field strength elements (e.g., Ti, Y) (figure 10B). Consequently, it plots as an outlier in each diagram in figure 6. Moreover, compared with the other glass samples, its REE pattern (figure 9A) is relatively unfractionated. These data suggest that the original trace element signature imparted by silica sand has been masked by the Fe-Mn ore used to darken this sample. |
31
|
| |
|
|
Discussion
|
|
There is a long tradition of using compositional data to trace ancient lithic artifacts (e.g., obsidian, chert, dacite, jadeite) to their points of origin.17 Indeed, the compositions of specific minerals can be sufficiently diagnostic so that they can be used to trace artifacts (or even loose sand) back to their geological sources.18 In the case of archaeological glass, of course, raw materials are completely melted, so direct comparison with suspected sources is not possible. Consequently, some workers have relied on indirect means (e.g., experimental reproduction of early glass and philological studies) to determine the source of silica sand used to manufacture ancient glass.19
|
32
|
The compositional signatures of accessory minerals present in silica sand are carried through to finished glass objects. Since accessory minerals host many (although certainly not all) of the trace elements found in geological materials, then the concentrations of these components can be used to constrain the source of silica sand used in historical glass batches. The admixture of silica sand to other batch ingredients requires that analytical data be interpreted judiciously because of the influence that other ingredients (e.g., alkalis) can have on the trace element signature of glass (see Shortland and Eremin).20 Spider diagrams for BSGCo glass (figure 10), sand (figure 5A), and quartzite (figure 5B) show some commonalities, notably a positive Zr anomaly. This clearly records the presence of zircon in these silica sources and the inheritance of this signature in the resultant glass. Any differences in the compositional signatures of the glass and silica sources must be interpreted cautiously because not all trace elements are governed by sand (or quartzite). For example, in geological materials, Rb and Sr are affiliated with feldspars; whereas in glass, they can originate in fluxes (potash) and stabilizers and hardeners (lime), respectively, that are added to glass batches. In contrast, Nb is associated with Ti and Fe-Ti oxides, in much the same way as Zr is controlled by zircon in rocks. Most samples of NJ sand have no Nb anomaly or only a weak negative one. In contrast, the quartzite has a pronounced negative Nb anomaly, as do the alkali-lead glass samples (BS7, BS17, BS18) and the plumbian potash-soda black glass sample (BS11). In contrast, the slightly potassic soda-lime milk glass (BS5) and potash-lead glass (BS1) have only weak negative Nb anomalies.
|
33
|
Because Nb is affiliated with common Ti-bearing oxide minerals (rutile, ilmenite, titaniferous magnetite), some of which can be magnetically separated from sand, glass with indistinct negative Nb anomalies cannot have been made from a silica source (such as the quartzite) with strong anomalies. Rutile (TiO2), of course, would be unaffected by this process. Any magnetic separation of iron-bearing titanium oxides would only intensify this anomaly. The glass samples with pronounced Nb anomalies have compositions consistent with the use of quartzite in their batches, whereas untreated NJ sand is the likely source of silica used in the manufacture of most of the remaining samples. The alkali-lead samples BS7, BS17, and BS18 have compositions consistent with the use of Cheshire quartzite in their batches, whereas BS1 and BS5 were more likely to have been made using NJ sand. These interpretations are consistent with the clustering of trace element ratios for the glass in the La/Ce plots in figure 6. The possibility cannot be excluded that the alkali-lead glass samples with strong negative Nb anomalies made use of NJ sand that had been magnetically treated to remove oxide minerals. The alkali-lead glass analyses, however, plot with the quartzite, even on diagrams that do not portray this component (e.g., Tb/Lu vs. La/Ce) (figure 6A).
|
34
|
Despite its strong negative Nb anomaly, black glass sample BS11 plots as an outlier in figures 6A to 6D and other trace element diagrams (figures 9A, 10A), so the source of silica used in its manufacture remains problematic. Many black glasses contain elevated alumina contents and are unusually enriched in the REEs and other trace elements, indicating that their batches included relatively impure sand with abundant feldspars and accessory minerals.21 BS11 has a low alumina content (0.4% Al2O3) but is, nonetheless, enriched in the REEs and HFSE (e.g., Ti, Y) compared with the other glass samples. This suggests that the silica sand used to make this black glass had low feldspar contents, but no inference can be made about any enrichment in accessory minerals in this sand, given its very high concentrations of iron and manganese.
|
35
|
Glass made from sand from which magnetic oxide minerals have been removed should have different ratios of ilmenite and magnetite components to trace elements associated with nonmagnetic accessory minerals in untreated sand. By comparing the ratios of these components in the sand and glass, it should be possible to distinguish glass made from magnetically treated sand from that made from sand or silicious rocks that are naturally impoverished in these minerals (e.g., Cheshire quartzite). This approach will only be successful if the ratios are based on components confined to sand sources and are not associated with other glass batch ingredients.
|
36
|
Potentially, zirconium (Zr) and yttrium (Y) are suitable components in this regard. The Ti/Zr ratios of NJ sand are much more variable, however, than those of quartzite (i.e., 5.9–40.7 vs. 4.4–7.1). The same holds true for their Ti/Y ratios (372–3783 vs. 141–252). This is hardly surprising given the much larger area over which the sand was sampled, but it seriously compromises the usefulness of these data as a means of distinguishing magnetically treated NJ sand from untreated sand derived from Cheshire quartzite.
|
37
|
Another strategy would involve distinguishing these two sources of silica based on possible differences in their chromium (Cr) contents, Cr-bearing Fe oxides having been identified only in the Massachusetts samples. There is no evidence that sand rendered from quartzite was ever magnetically treated. Apparently, it was only washed, so this geochemical signature should be preserved in glass made from this sand. High precision Cr data, however, are lacking for this material.
|
38
|
| Distinctions between the sources of silica sand used by BSGCo provide a loose constraint on the relative ages of some of the analyzed glass samples. For example, alkali-lead glass samples with strong negative Nb anomalies indicative of Cheshire quartzite used in their batches likely postdate those lacking this geochemical signature, since this source of silica sand was not exploited by BSGCo until the 1850s. Approximate ages are known for only five of the analyzed samples (Appendix 2), and of these, ICP-MS trace element data are available for only one (BS1). It is nonetheless noteworthy that the composition of this artifact is consistent with the early date (c. 1835–45) assigned to it. |
39
|
| |
|
|
Conclusions
|
|
Historical records indicate that BSGCo used silica sand derived from the Cheshire quartzite and the Cohansey and Cape May formations of southern New Jersey. Given its purity, distinguishing the source of the sand used to make particular sample glass objects is challenging and must rely on comparing their trace element signatures with those of glass recovered from the Sandwich glass factory site.
|
40
|
In terms of their trace elements, the silica sources used by BSGCo differ mostly in their niobium signatures: quartzite (and sand derived therefrom) consistently having strong negative Nb anomalies and NJ sand having an indistinct Nb anomaly or no anomaly whatsoever. This component is governed by Ti-bearing oxides in silica sands and silicate rocks and is not affiliated with any other ingredients commonly added to glass batches. It can be used to identify the likely source of silica used to make different types of BSGCo glass, the only complicating factor being that some NJ sand might have been magnetically treated to remove oxide minerals. On this basis, it appears that the alkali-lead glass samples for which trace element data are available (i.e., NS7, BS17, BS18) were made from Cheshire quartzite or magnetically treated NJ sand. The alkali-lead glass compositions were found, however, to plot with the quartzite, even on diagrams that do not portray Nb (e.g., on a Tb/Lu vs. La/Ce plot), thereby supporting the former (quartzite) interpretation. Soda-lime glass and potash-lead glass with negligible or weak negative Nb anomalies evidently was made from (untreated or inefficiently treated) NJ sand.
|
41
|
| Chromium-bearing Fe spinel has been identified only in the Cheshire quartzite, so potentially Cr can also be used to distinguish between NJ and Massachusetts sources of silica used by BSGCo. The trace element signature of the black glass (BS11), however, appears to have been modified by the addition of Fe-Mn ore, obscuring recognition of the sand used in its manufacture. |
42
|
| |
|
|
Acknowledgments
|
|
| We thank the curator of the Sandwich Glass Museum for making available the samples described here. Analytical expenses were funded by Natural Sciences and Engineering Research Council of Canada Discovery grants to J. Victor Owen and John D. Greenough. We are especially grateful to Ed Kirby who accompanied Charles L. Flint on several arduous treks through the Berkshire Hills, searching for abandoned quartzite quarries, and who also participated in digging test pits and interpreting artifacts recovered from the Washington quarry and Roaring Brook sites. We thank three anonymous IA referees for their comments on the manuscript and Patrick Martin for editorial assistance. |
43
|
Notes
1. Olive I. Wilson, Les Verreries de Vaudreuil, Bas-Canada, 1845–1877 (Glass in Vaudreuil, Lower Canada, 1845–1877) (Vaudreuil, Quebec: Musée Régional du Vaudreuil, 1996), 28.
2. J. Victor Owen, "The Como-Hudson Factories (c. 1845–77): Results of Geochemical Analyses for Quebec's First Known Glassworks." Canadian Journal of Archaeology 25(2001): 74–97.
3. Robert C. Dunnell, "Why Archaeologists Don't Care about Archaeometry," Archaeomaterials 7, no. 1 (1993): 161–65.
4. A. Silverman, "Sandwich Glass," The Glass Industry 17, no. 2 (1936): 50.
5. Raymond E. Barlow and Joan E. Kaiser, The Glass Industry in Sandwich, Vol. 2 (Lancaster, Penn.: Schiffer Publishing Ltd., 1989), 267–82.
6. J. W. Skehan, Roadside Geology of Massachusetts (Missoula, Mont.: Mountain Press Publishing Co., 2001).
7. Norman Herz, "Geologic Map of the Cheshire Quadrangle, Massachusetts Bedrock Geology," U.S. Geological Survey Geologic Quadrangle Map GQ-108 (1958).
8. J. Victor Owen, Jaroslav Dostal, Lane Humphreys, Brenda Orr, and Stephen Powell, "Nova Scotia Glass c. 1881–1917," submitted to Journal of Glass Studies.
9. F. Markewicz, "Ilmenite Deposits of the New Jersey Coastal Plain," Geology of Selected Areas in New Jersey and Eastern Pennsylvania and Guidebook of Excursions, ed. S. Subitzky (New Brunswick, N.J.: Rutgers Univ. Press, 1969), 369.
10. Accessory minerals in NJ sands are described in two articles in Subitzky, Geology of Selected Areas (see n. 9). These articles are J. P. Owens and N. F. Stohl, "Shelf and Deltaic Paleoenvironments in the Cretaceous-Tertiary Formations of the New Jersey Coastal Plain," 235–78, and F. Markewicz, "Ilmenite Deposits of the New Jersey Coastal Plain," 363–82.
11. H. Kümmel and R. B. Gage, "The Glass-Sand Industry of New Jersey," New Jersey Geological Survey, Annual Reports of the State Geologist (Trenton, N.J.: Geological Survey of New Jersey, 1904), 84–86.
12. The compositions of these minerals are as follows: zircon ZrSiO4; ilmenite Fe2+TiO3; magnetite Fe3+2Fe2+O4; Fe-Cr spinel FeCr2O4; staurolite (Fe2+,Mg)2Al9(Si,Al)4O20(O,OH)4. The spinel analyzed here was not chromite but an Fe spinel that contained about 10 wt.% Cr2O3.
13. The XRF calibration curve allows the reliable determination of silica up to concentrations of about 97 wt.%. We have taken silica contents of NJ sand and Cheshire quartzite to equal the difference between 100 and the sum of other major element components, calculated volatile free.
14. I. Grey, C. MacRae, E. Silverster, E. Sasini, and J. Sasini, "Behaviour of Impurity Elements during the Weathering of Ilmenite," Mineralogical Magazine 69 (2005): 437–46.
15. Silverman, "Sandwich Glass," 50 (see n. 4); Pamela Vandiver, "Transcription of Batchbooks from the Collection of the Sandwich Glass Museum, Sandwich, Massachusetts" [unpublished] (1969).
16. See discussion in J. Victor Owen, "Geochemistry of Wistarburgh Glass (ca. 1739–1777): Implications for Batch Recipes and Distinction from Other South Jersey Wares," Canadian Journal of Earth Sciences 41 (2004): 691.
17. Agazi Negash and M. S. Shackley, "Geochemical Provenance of Obsidian Artefacts from the MSA Site of Porc Epic, Ethiopia," Archaeometry 48, no. 1 (2006): 1–12; N. Malyk-Selivanova, G. M. Ashley, R. Gal, M. D. Glascock, and H. Neff, "Geological-Geochemical Approach to 'Sourcing' of Prehistoric Chert Artifacts, Northwestern Alaska," Geoarchaeology 13, no. 7 (1998): 673–708; J. D. Greenough, L. M. Mallory-Greenough, and J. Baker, "Orthopyroxene, Augite, and Plagioclase Compositions in Dacite: Application to Bedrock Sourcing of Lithic Artefacts in Southern British Columbia," Canadian Journal of Earth Sciences 41(2004): 711–23; Jan Spisiak and Dusan Hovorka, "Jadeite and Eclogite: Peculiar Raw Materials of Neolithic Stone Implements in Slovakia and Their Possible Sources," Geoarchaeology 20, no. 3 (2005): 229–242.
18. See, for example, Leanne M. Mallory-Greenough, John D. Greenough, and J. Victor Owen, "The Stone Source of Predynastic Basalt Vessels: Mineralogical Evidence for Quarries in Northern Egypt," Journal of Archaeological Science 26 (1999): 1261–72, and William R. Dickinson, "Petrography and Geologic Provenance of Sand Tempers in Prehistoric Potsherds from Fiji and Vanuatu, South Pacific," Geoarchaeology 16, no. 3 (2001): 275–322; Dennis A. Darby, "Trace Elements in Ilmenite: A Way to Discriminate Provenance or Age in Coastal Sands," Geological Society of America Bulletin 95 (1984): 1208–18.
19. A. Silvestri, G. Golin, G. Salviulo, and R. Schievenin, "Sand for Roman Glass Production: An Experimental and Philological Study on Source of Supply," Archaeometry 48, no. 3 (2006): 413–32.
20. A. J. Shortland and K. Eremin, "The Analysis of Second Millennium Glass from Egypt and Mesopotamia, Part 1: New WDS Analyses," Archaeometry 48, no. 4 (2006): 581–604.
21. J. Victor Owen, "The Como-Hudson Factories (c. 1845–77): Results of Geochemical Analyses for Quebec's First Known Glassworks, Canadian Journal of Archaeology 25 (2001): 74–97.
APPENDIX 1: New Jersey (NJ) Sand Sample Locations and Description
NJ1: Silica sand from the eastern bank of the Maurice River at Dorchester (N39º15'11.0" W74º59'42.7").
NJ2: Silica sand from the beach at East Point lighthouse, very clean sand with no pebbles (N39º14'36.9" W75º00'06.6").
NJ3: White sand from a meter-scale lens that occurs in beige, pebbly sand at Menantico Ponds wildlife management area, Millville; these white sand lenses are very white on the surface but below a few mm in the ground are dark grey due to organic-rich, muddy material (N39º22'02.6" W74º59'51.8").
NJ4: Beige, pebbly sand that hosts the white sand lenses from the same site as NJ3.
NJ5: White sand from a lens in beige pebbly sand (N39º22.01.2" W74º59'41.7").
NJ6: Pale grey sand from a lens in brown sand in Union Lake Preserve, near Vineland (N39º23'37.6" W75º01'03.6").
NJ7: More quartz-rich sand from the location as NJ6 (N39º26'47.7" W75º04'53.0").
NJ8, NJ9: White sand from near the junction of routes 55N and 352 (N39º26'10.3" W75º03'31.4").
APPENDIX 2: Boston and Sandwich Glass Co. Glass Sample Descriptions
BS 1: colorless Eagle salt cellar (c. 1835–45)
BS 2: colorless Peacock Eye dish fragment (c. 1835–45)
BS 3: opaque blue candlestick socket (c. 1840–60)
BS 4: colorless/green candlestick socket (c. 1840–60)
BS 5: opaque white (milk glass) bottle fragment
BS 6: etched milk glass/red glass overlay shade with floral and dragonfly design (c. 1870–87)
BS 7: colorless/light ruby fragment
BS 8: colorless/light ruby fragment
BS 9: colorless/light ruby fragment
BS 10: colorless/light blue fragment
BS 11: opaque black dish fragment
BS 12: colorless dish fragment
BS 13: colorless/green-blue d | |