technical study of photographic glass

history of flat glass

The industrial revolution led to mass-production of many materials, a good number of which found their way into 19th-century cultural heritage. One such material was flat glass, the demand for which was driven by the window glass and mirrored glass markets. Flat glass was usually of an alkali silicate composition. With the invention of photography, flat glass, also known as sheet glass or plate glass, was quickly adopted by photographers for both picture glazing and image substrates. Continuing developments in the flat glass industry influenced the quality of glass that eventually found its way into 19th-century photographs, which now make up an important body of cultural heritage collections worldwide. The history of flat glass development and its effect on photographic advancements in the 19th century are crucial to understanding and assessing photographic glass stability.

highlights of findings

test samples (“Flea market glass”)

A small group of ambrotypes and tintypes, i.e., objects containing cover or image glass, were purchased from the open market and analyzed to get a sense of typical photographic glass compositions. Seven samples of glass from the objects were ground into powder for fully quantitative XRF analysis. Results show that among these random samples, six are high lime soda glass (also known as soda lime glass) with:

~13-15% CaO (lime)
~10-14% Na2O (soda)
~0-500 ppm MnO (manganese oxide)
~0-250 ppm As2O3 (arsenic oxide)

Although these results do not disqualify any of the cover glasses as being modern replacements (as often occurs), no boron or other elements that are typical of modern glass were detected.  

The remaining sample, an uncased ambrotype (pictured above as FG-5I), is a higher lime potash glass, with a composition of:

19% K2O (from potash)
8% CaO (lime)
1.5% Na2O (soda)
467 ppm Sb2O3 (antimony oxide)
155 ppm MnO (manganese oxide)

UV examination of historical photographic glass is often difficult to determine and can be influenced by the light source and viewing set up, but is best assessed on the glass edges. UVA examination technique is only appropriate for cover glasses, since image glass normally has a dip-coating and/or emulsion, which itself will fluoresce. Nevertheless, UV examination of photographic glass reveals some possible patterns related to compositional types. Fairly pure soda lime glass usually has only weak fluorescence in UVA (365 nm). However, the presence of MnO, a decolorizer, is quite common and causes distinct fluorescence in UVA of both soda lime and potash glass, producing a range of yellow colors---from orangish to greenish. This variation in color depends on the concentration of Mn, and more particularly on the presence of other trace elements which can act as fluorescence co-activators or quenchers. For example, the presence of As2O3 seems to cause a more lemony yellow fluorescence, while Fe2O3 may quench and mute fluorescence. In UVC illumination (254 nm), coatings normally do not fluoresce, and soda lime glass has very weak to strong fluorescence, sometimes indistinct and sometimes in a similar range of yellow colors, again depending on the elements present. UVC illumination of potash glass that contains Sb2O3 is interesting and yields a strong milky light blue in historical glass, as viewed as part of this study. The latter visible fluorescence color can, however, be confused with clear light blue fluorescence caused by PbO, which is normally only present in very trace amounts in photographic glass.

Two encased daguerreotypes from a private collection were examined without the opening of the encasement using non-invasive XRF, light microscopy, and OCT. XRF shows that the cover glass in one of the two objects is a potash glass composition containing an antimony oxide fining agent, while the other is soda lime glass.

Both objects display significant deterioration on the underside of their cover glasses, including a high concentration of liquid droplets and precipitated salts. OCT, a medical imaging technique that has become standard in the field of ophthalmology, was used in situ to non-invasively view and document this deterioration by imaging through the glass without opening the encasement. This allowed the droplets on the underside of the cover glass to be examined without danger of drying out, and is the first time OCT has been used for this purpose as far as we know. OCT images show that the liquid droplets extend about 100 µm below the glass surface. Images taken of the potash cover glass also allowed distinction and measurement of dry particulates.

historical societies

Several institutions were visited to survey a randomly sampled portion of the photographic collections, including: The Virgina Museum of History and Culture, The Historical Society of Washington DC, The Historical Society of Pennsylvania, and The Special Collections Research Center at The George Washington University. Three of these collections contain early photographic materials that previously had poor storage conditions and have never been treated by a conservator, as is typical of such under-served institutions.

Approximately 145 objects, including daguerreotypes, ambrotypes, tin types, glass plate negatives and glass lantern slides were examined with the non-invasive, portable instrumentation shown in this image. No sealed enclosures were opened during the surveys. XRF analysis in general showed a range of glass types similar to that represented in the model glass study. Unfortunately, signs of deterioration among the objects examined were far more common than predicted from composition alone, particularly in encased or enclosed objects. For example, about 90% of one historical collection’s encased photographs display liquid droplets and/or crystalline particles on the underside of the cover glass. Most surprisingly, these deteriorated cover glasses include both unstable glass formulations, where stabilizers are insufficient in quantity compared to alkalis, and what are normally considered to be stable formulations of soda lime glass. Glass deterioration symptoms that we observed are mostly on the underside of the cover glasses in the form of tiny precipitated-looking particles spread evenly around the surface, liquid droplets, and/or liquid droplets with a solid center. Contrary to deterioration symptoms predicted by model studies, no cracking in glass was observed, but some cracked emulsions of glass plate negatives was found, signaling possible glass deterioration underneath the emulsion. Biological growth on some cover glass was also observed, suggesting exposure to humidity. For more discussion of glass deterioration symptoms, see visual vocabulary.

Historical collection surveys determined that roughly 25-30% of the overall collections examined may show condition issues, either as liquid droplets or crystalline deposits, which is far greater than expected. Materials at the greatest risk for deterioration include encased photographs as well as glass lantern slides, which are usually sandwiched between plates of glass. Tests performed on liquid droplets on a soda lime glass lantern slide from about 1858, in which the tape seal had detached, showed that the droplets have a pH of about 9.0-9.5. This high alkalinity confirms their identification as symptoms of severe glass deterioration. These results also underscore the potential for any type of 19th-century glass to undergo deterioration in adverse environments.

Decision Tree

The results highlighted above have been used to construct a decision tree for assessing at-risk photographic glass collections. The decision tree is meant to be used alongside our visual vocabulary for deterioration. The tree guides museum professionals through simple tools, followed by XRF analysis, if available, for better confidence in assessment of preservation priority.

19th Century Vessel Glass: Decision Tree Guide for Use with Claude Laurent Flutes

Step 1. This condition assessment in the form of a decision tree is specific to glass flutes made by Claude Laurent. However, the tree may be helpful for other types of 19th century glass WITH CAUTION.

Step 2 uses a simple, inexpensive ultraviolet light source, preferably with both long-wave (UVA, 365 nm) and short-wave (UVC, 254 nm), viewed in the dark. Protective UV goggles must be worn for safety to your eyes. UV fluorescence behaviour as shown indicates the presence of manganese (Mn) or lead (Pb) in Laurent’s formulations; this step may reveal different behaviour in other types of glass. X-ray fluorescence spectroscopy (XRF) is the best non-invasive method for determining glass type. See Appendix I.

1.      UVA:

a.      Yellowish-green fluorescence means the glass contains Mn. For Laurent flutes this indicates the glass is likely to be a high potassium, low calcium glass (potash glass). Go to UVC examination.

b.      Pinkish or indeterminate. This indicates the glass is not Laurent’s potash formulation.

2.      UVC:

a.      Pinkish or indeterminant. This indicates the glass is not high-leaded (“crystal glass”).

b.      Light, clear blue. This indicates the glass is headed (“crystal glass”), and is necessary to confirm composition category. Note that the amount of lead necessary to produce this fluorescence has not been determined.

Step 3. Viewed with the naked eye, glass may appear somewhat opaque when it becomes severely degraded due to light scattering from cracks, losses, and surface accretions. However, glass finishes, including rough polishing and etching/frosting, also cause light scattering, so that opaqueness itself may be part of the original design. In this step, judge the presence of any opacity only as relative to different areas/joints of the glass object.

Step 4. Refines condition assessment using simple to complex tools, including examination with 10x or greater magnification in normal light, plus other analytical methods, as possible.:

1.      Determine the presence of cracking patterns, using light at different angles if possible. Patterns include:

a.      Parallel or arcing cracks are often seen in the fluting depressions of Laurent flutes.

b.      T-joint cracks that appear to join the previous types of cracks at a sharp angle.

c.      Polygonal cracking, with areas defined by cracks roughly 100-200 microns in diameter.

d.      Losses between cracks.

2.      Working from left to right, roughly categorize level of deterioration:

a.      Advanced, presence of:

   i.           polygonal cracking.

   ii.           losses related to cracks (i.e., not isolated chips).

   iii.           significant amounts of accretions evenly spread across surface (note that this is not often seen on Laurent flutes).

  iv.           Liquid droplets or what appear to be dried films of liquid.

b.      Intermediate: presence of very fine cracks that are only visible with magnification, including all types of shapes

c.      Not apparent: None of other symptoms; surfaces appear to be pristine, with only polishing marks and debris visible. Note: if glass has been determined to be potash, deterioration level should be considered as incipient, i.e., in initial stages but not visible

3.      Supplemental analysis:

a.      pH measurement of the surface is subject to variability in application and should be interpreted with caution. See Appendix II. If pH measures > 8.0, the object should be assumed to be undergoing at least incipient deterioration.

b.      Fiber optic reflectance spectroscopy (FORS), using a light source in the visible to near infrared range, is a good non-invasive method for verifying whether the glass is hydrated, which part of the alkali silicate degradation process. See publications.

c.      Optical coherence spectroscopy (OCT) is a specialized, non-invasive method that can give quantitative information about the structure and depth of an existing hydration layer on the glass surface, if it is about 10 microns or thicker. This is a powerful tool, especially when combined with FORS. See publications.

d.      X-ray diffraction (XRD) and µRaman spectroscopy are excellent methods for analysis of surface accretions or particulates that can be removed for analysis. Identification of samples suggests the deterioration mechanism, and most importantly distinguishes debris, including polishing residues, that can easily be mistaken for evidence of deterioration.

Step 5 includes currently recommended actions for best preservation actions. Note that all potash glass should be considered inherently unstable. High-leaded glass is often observed to be relatively stable, as is soda lime glass with formulations that include high calcium content. Glass flutes being used for performance need extra care. Cleaning methods are currently under study, but need to be undertaken if the flute is being handled and played. Dry wiping with a 100% cotton soft cloth is appropriate.

Appendix I. Method for characterizing glass type from XRF measurement

1.    Objects can be examined non-invasively. For use with a Bruker Tracer instrument,  conditions should include He flow at two power settings to optimize separately for detection of lighter or heavier metals: a) light element mode, 15kV, 50 µA, no filter; and b) heavy element mode, 50 kV, 35 µA, Cu/Ti/Al filter.

2.    For application of semi-quantitative ratios of alkali:Ca greater or less than 1.5, external linear regression calibrations must be constructed for the instrument in use at each set of operating conditions using glass standard reference materials. These can be obtained from NIST, and should include Brill standards if possible. Any glass analysed by inductively coupled plasma spectroscopy can be used as a standard as well.

3.    If this method is not possible, very rough estimates of “high Ca” or “low Ca” alkali silicate glass can be made, since the sensitivity of K and Ca are not that different in the light element mode conditions (due to their close molar mass). In addition, neither Na nor Mg will be detected. Qualitative ratios of raw intensity for K and Ca can be calculated in Bruker Artax software after processing data with the evaluation function. Spectra can than be sorted as follows:

a.    Prominent Ca peak and relatively smaller K peak (medium-high alkali:Ca):

  i.         UVC illumination: milkly blue or white: this is likely to be a potash type glass; priority should be high.

  ii.         UVC illumination: indeterminant: this is likely to be a “high Ca” alkali silicate glass, although the amount of Na is unknown, so the glass should remain in the high priority category IF it is encased. If it is a single plate, the priority is medium to low, depending on evidence of deterioration.

b.    Prominent Ca peak and very weak K peak (low alkali:Ca): this is likely to be an inherently stable soda lime glass (and should have exhibited pinkish/orangeish UVA fluorescence). If the glass is part of encasement, the priority remains medium - high, depending on evidence for deterioration. If the glass is a single plate, the priority should be low – medium, depending on whether any deterioration in the emulsion or coating is evident.

c.    Prominent Ca peak and relatively equal or greater K peak (high alkali:Ca): this is potash glass and is inherently unstable; the UVC fluorescence was likely milky blue; the priority is high.

d.    Ca and other lighter element peaks dwarfed by Pb: this is leaded glass, which is unusual, but should be fairly stable unless it is part of an encasement. In the latter case, priority is low – medium depending on any evidence for deterioration.

Appendix II. Method for approximating pH of a glass surface

1.      Surface pH measurements may be conducted using:

a.     Deionized water and colorpHast pH indicator strips (not expired), where the pH strip is wet with deionized water, blotted of excess moisture, then held to the surface of the glass sample for two minutes. The pH strips should be cut into small slivers in order to test small spots. The test should be done at least 3 times in different areas. The pH of the deionized should be measured to make sure it is not too acidic. Note that near neutral pH is difficult to measure, and strong responses are easier to judge.

 b.      A pH meter measurement can be used on a 1 micoliter droplet of deionized water applied to the flute. This results takes special equipment and is not necessarily any more accurate than the pH strip method.