Appendix I
A Scanning Electron Microscope Looks at a Daguerreotype
Surviving Daguerreotypes exhibit several kinds of surface deterioration. The appearance of mechanical scratches and large area abrasions can effectively be eliminated by retouching a copy (not the original). Stains may be cancelled by copying with colored filters, and weak images can be improved by high-contrast copying. However, large area silver tarnishing that obscures image detail on many Daguerreotypes cannot be compensated by optical copying methods. For this reason, chemical removal of tarnish on the original plates was a common practise for many years.

Potassium cyanide was first used to remove tarnish and, inevitably, some of the image information, since it dissolves silver, but in the early 1970's a "new and improved" formula was published that utilized acidified thiourea. It became widely used because, besides being less toxic than cyanide, it produced bright clean surfaces that appeared not to have sustained noticeable damage or loss of image.

Of course it was realized that tarnish returns quickly to clean silver unless the storage environment is completely free of sulfur. But some cleaned Daguerreotypes soon developed unsightly blemish spots that were dubbed "measles", rather than the expected hazy film of tarnish.

In March 1973 the author, at the request of colleague Leon Jacobson, examined corrosion spots on a sixth plate Daguerreotype of an unknown subject using a scanning electron microscope (SEM). The results were published in a short article [80] in 1974. Following is a more complete discussion of the technique and results, including previously unpublished SEM micrographs from that work.

Figure 9 shows the appearance of the test picture chosen for analysis, after it had been cleaned in the thiourea solution. The "measles" spots are hardly visible in this specimen, but they were sufficient for analysis. They were of much greater concern on other historically valuable Daguerreotypes.

  
Figure 9

Because of the vacuum environment in electron microscopes, it was necessary to remove the Daguerreotype plate from its case and from its binding tape and cover glass. Thus prepared, the bare Daguerreotype plate is better able to withstand a vacuum environment than any other photographic image. The plate was larger than our available SEM specimen stages so a holder was improvised that, unfortunately, did not allow optimum tilt angles, but the resolution was not seriously degraded at magnifications less than about 10,000.

In the years since this work, many other SEM analyses have been reported, notably by M. Susan Barger and coworkers, and by Swan et al. But the earlier work still usefully illustrates the nature of a corrosion problem and one of the pitfalls of restoration. It also reveals details of the Daguerreotype microstructure that a light microscope cannot achieve.

Microstructure of a Daguerreotype Image
Fig. 10 is a low magnification (about 15x) SEM micrograph of a portion of the white shirt chosen for its sharp edge contrast. Various kinds of blemishes are visible, some of which are nearly invisible by light microscopy. The SEM image depends on the secondary electron emission properties of surfaces rather than on light reflection. This fundamental difference between the two imaging processes often reveals organic and inorganic thin film contaminants not visible in light, even though the concept of color is inapplicable to electron images as it is to light images.

Figure 10 
 Figure 11

Figure 11 shows the light/dark boundary at about 200 magnification; the particulates in the white region are becoming visible. Figure 12 clearly shows the amalgam particles in the white area, as well as buffing scratches in the silver-plated base metal. Figure 13 shows details of the amalgam particles at about 5000 magnification. This sequence of pictures shows that the "white" expanse of the shirt contains many more amalgam particles than the dark regions. The particles are silver-white in visible light, and their shape scatters incident light so that the viewer's eye has an appreciable acceptance angle for this reflected light. Light that is reflected from the highly polished areas where there are no particles is efficiently reflected, but in a narrow angle that depends precisely on the angle of incidence. This has the effect of sharply reducing the eye's acceptance angle. Thus a viewing angle can be found where the contrast is at a maximum, within perhaps twenty degrees on either side of the perpendicular. The actual dependence of contrast on viewing angle depends on several factors; it has been studied by Barger et al [12]. If the eye is far off to the side, contrast is nearly zero, and the image vanishes.

Figure 12
Figure 13

Daguerreotypes have been described as "grainless", but from these pictures this is obviously in error. The grain of the particles is apparent in a light microscope at 300x. They appear textureless in comparison with salt prints, their contemporary competitors, which had a visible paper texture.

The mechanism of particle nucleation and growth which accounts for the range of particle sizes is discussed more fully by Barger [8, 12] and by Pobboravsky [117]. Since we examined only one specimen, we have no information on the original effects of process and materials variations.

The fact that the particles are bright by reflected light and also bright in secondary electron images does not have an intuitively obvious explanation. It has been said that the earliest secondary electron images surprised the pioneering workers because of their unpredicted resemblance to light microscope images. SEM images, besides being capable of more than fifty times greater magnification, have some five hundred times greater depth of field than light micrographs. It is convenient that the two imaging technologies complement each other so well.

The width of the black band at the bottom of some of the pictures is a micrometer marker (not all the pictures have a marker because one of the SEMs we used lacked a marker mechanism). The band marked '100 microns' thus represents about 0.004"; '4 microns' represents 0.00016". One micrometer, or its formerly-used synonym "micron", equals one thousandth of a millimeter or about 4 one-hundred thousandth of an inch; the wavelength of green light is half a micrometer. It is more accurate to refer to these internal markers, because apparent magnification may change during subsequent reproduction. The maximum magnification of which most light microscopes are capable is less than 2000x.

Corrosion Analysis
Figure 14 shows one of the "measle" spots near the left side at about 850x; it consists of a dark center surrounded first by a narrow white ring, then a broader dark ring. This specimen had been cleaned in the acidified thiourea solution. The corrosion site is approximately twenty-six times larger than a typical amalgam particle, making it visible to the unaided eye.

  
Figure 14

We performed X-ray fluorescence analysis in the SEM by focusing a stationary electron beam with an energy of 9 kilovolts on the center of the corrosion site and on other selected sites for comparison. The energy spectrum of the X-rays emitted at the site of electron bombardment was analyzed by a solid state energy dispersive detector. X-rays are emitted from a pear-shaped volume substantially smaller than the overall corrosion site but large enough to include several of the crystalline "petals".

Analytical results were as follows:

1. Center of corrosion site: strong silver and sulfur, trace chlorine and mercury.
2. Small white particles surrounding corrosion "petals": strong silver and sulfur.
3. Dark zone - Fig. 10: strong silver and sulfur.
4. Amalgam particle: silver, mercury, trace chlorine.
5. Clean base metal between particles: silver, mercury. Gold was not detected; not all Daguerreotypes were toned.

Towler [145] listed the following materials used to make Daguerreotypes:
Jeweler's rouge (iron sesquioxide).
Iodine, and sometimes bromine, sensitizer.
Mercury.
Sodium hyposulfite ("hypo").
Gold chloride toner, not always used.
However, this list is oversimplified: there were many variations. Other polishing compounds such as pumice were used, and combinations of sensitizers were used, including chlorine, as discussed by Swan et al [138].

Both the narrow light ring and the broad dark ring showed strong silver and sulfur. The results indicate that these collars are spreading contaminants that hide the normal composition of the clean surface, and that they are largely responsible for the expanded visibility of the corrosion sites. Neither the amalgam particles nor the base metal between particles contained detectable sulfur.

The plate was then cleaned again in thiourea solution, followed immediately by several distilled water washes and an ultrasonic wash in distilled water. A second SEM analysis showed essentially no change in the appearance of the crystalline corrosion, but the sulfur and chlorine peaks in the X-ray spectrum were almost undetectable. The measles were much less apparent to the eye, and did not change over a storage period of six months.

Conclusions and discussion:
The crystalline corrosion spots act like tiny sponges that retain traces of the thiourea cleaning solution. This thiourea, which contains sulfur, effused outward over a period of days, forming a collar of increasing visibility around each corrosion site. It was the thiourea residue that was largely responsible for the visibility of the measle spots: the original crystalline centers were much smaller and relatively obscure. The ultrasonic wash was vigorous enough to remove the residual traces from the interstices of the microcrystalline "sponges".

Thiourea is the active ingredient in most commercial silver cleaners. It is an organic compound containing sulfur, nitrogen, carbon, and hydrogen (H2NCSNH2). It had been recommended by reputable restorers in the 1970's for removing tarnish on Daguerreotypes (eg Weinstein & Booth [148], and the 1979 edition of Eastman Kodak Publication F-30.) But there is no cleaning process that removes chemically bound corrosion without also losing some picture information. Chemical cleaners cannot convert silver corrosion compounds back to metallic silver and redeposit it precisely in its original sites. Cleaners convert the corrosion products (usually sulfides) to a soluble organo-metallic complex that can be washed away. This selectivity is useful: the silver in the corrosion is lost but not the uncorroded silver. Dirt and inactive foreign substances, if they are not chemically bound to silver, may be removed by solvents or detergents.

The cause of the crystalline form of corrosion is unknown. The fact that no copper was detected was interpreted to mean that there was no pinhole in the silver plating to expose the base copper. This is not conclusive: the crystalline structure may have grown in several phases, effectively concealing the original defect. We believe that the most effective means of analysis would be to remove the corrosion by argon or krypton focused ion bombardment in the SEM; Barger et al [11] discusses this technique. This would permit SEM inspection during the dissection process and eliminate exposure to other chemical reagents that would confuse interpretation. At the time of our original work (1973) this technique was being explored but was not then operational. It has become a recognized tool in recent years.

Neither bromine nor iodine (the usual sensitizers) were detected in our analyses. Pobboravsky [117] has measured typical silver iodide film thicknesses of the order of 30 nanometers, or about 300 atomic diameters. Because of the unfavorable placement of the specimen plate in our SEM, it is likely that this was below our detection limit. The presence of these materials was not of particular interest unless they were concentrated in the corrosion sites, which was not the case.

The origin of the chlorine traces is not certain. It may have been added as an accelerator during sensitizing. It may also have been a trace impurity in the original process (before the days of 'Chemically Pure' reagents), or simply have come from recent handling or during more than a century of storage.

Particles of the original polishing compound may have been left on the surface, which could have served as corrosion nucleation sites. Our SEM had a substantial iron background X-ray peak caused by wall scattering and aggravated by the unfavorable specimen position. Therefore no conclusion was justified on this question.
Other limitations of the SEM analysis:
The X-ray spectrum at the time of this analysis detected chemical elements but did not yield information on the chemical compounds or on the quantitative amounts. In a heterogeneous surface such as this specimen, quantitative information would be meaningless unless the analyzed microvolume could be defined.

There are detection problems with elements whose atomic numbers are below about fluorine, which includes carbon, oxygen, and nitrogen. The X-ray yield is small at low atomic numbers, and the escape probability of the low energy X-rays also decreases, especially in heavy matrices such as silver. Instrumentation is continually improving, and many new analytical techniques are emerging that are capable of identifying organic compounds in microstructures.

Our results, like those of many other investigations, leave unresolved a number of questions. They did lead to a conclusion regarding a cleaning process that was experimentally verified, which is a useful outcome for a small volunteer effort. The study has been discussed in detail to show the power of the scanning electron microscope, a modern analytical tool in common use in many fields. Hopefully this experience may encourage other workers to make similar efforts.