USING XRF AND RAMAN FOR TARNISHING SILVER ARTIFACTS

ABSTRACT

In the article, the authors reviewed the methods and results of the study of tarnished silver and tested two spectrometers: a handheld XRF and a Raman microscope. XRF is rarely used for this kind of research, although it is widespread and suitable for sulfur and chlorine research. In contrast, a Raman microscope is often used to study tarnished silver, but usually only "in one direction": substance → spectrum. The capturing time for the existing handheld XRF (2015) spectrum is 20 seconds, the capturing time for the modern (2020) XRF is up to 999 seconds. The thickness of the tarnished layer cannot be calculated on the basis of the 20-second spectrum due to the low accuracy, unfortunately, but testing has shown that even the old XRF can distinguish silver from silvering. As for the Raman microscope, it is possible to establish a substance by its spectrum if all other spectra were somehow excluded.

АННОТАЦИЯ

В статье авторы рассмотрели методы и результаты исследований потускневшего серебра и протестировали два спектрометра: портативный РФА и Рамановский микроскоп. РФА не часто используется для такого рода исследований, хотя он широко распространен и подходит для изучения серы и хлора. Напротив, Рамановский микроскоп часто используется для изучения потускневшего серебра, но, как правило, только «в одном направлении»: вещество → спектр. Время снятия спектра у существующего РФА (2015 г.) спектра составляет 20 секунд, время снятия спектра у более современного (2020 г.) РФА - до 999 секунд. Толщина слоя потускнения не может быть рассчитана на основе 20-секундного спектра из-за его низкой точности, но, как показали  наши измерения, даже старый XRF может отличить серебро от серебрения. Что касается рамановского микроскопа, то можно определить вещество по его спектру, если как-то исключить все остальные спектры.

Keywords: tarnished silver, XRF.

Ключевые слова: потускневшее серебро, РФА.

Appearance, clearing and measurement errors.

There is silver in almost every museum in the world, and all curators of silver artifacts are concerned with the problem of how to preserve silver in its original form. Tarnishing is a thin layer of corrosion. Silver can tarnish over several months, especially if not properly cleaned.

There are two types of areas of silver: located in the light and located where there is no light at all. The first areas fade over time, the second ones do not, but, as a rule, do not shine like new metal. The tarnishing layer usually consists of three silver salts: Ag2S (black crystals if the layer is thick), Ag2O (also black crystals), and AgCl (white crystals that turn brownish in the light and then black). When the sulfide layer changes its thickness in the range of 0.01–0.1 µm, the color of the tarnished layer and the silver beneath it changes from yellow through red-brown to blue. It is a common mistake to think that silver sulfide (red-yellow) is silver chloride (brownish) and vice versa. The ratio of O, S and Cl in the tarnishing layer is still under discussion. Silver nanoparticles differ in shape and yellow color due to the plasmonic effect.

Table 1.

Three main silver tarnishing salts [1]

Light (sunlight, ultraviolet, etc.) is a source of additional energy for photosensitive reactions, but not a participant in them. However, some reactions are very slow in the absence of light. Visible light destroys organic matter, which increases the release of certain gases into the atmosphere. The temperature also rises in the light. Low light levels combined with other causes lead to the formation of chloride in the form of a layer of corrosion present on many coins and medals. Humidity also increases the rate of corrosion, but affects all areas at the same time. The effect of light illumination and relative humidity (RH) is shown in Figure 1a, 1b.

The metallic luster is a specular reflection of the incident light. Silver reflects almost 90% of visible light as a specular reflection (Figure 1, c) because 1) it has specific physical characteristics and 2) silver has a smooth surface. With an increase in the thickness of the tarnish layer, 90% decreases quickly. Also when silver is recovered from chloride, the surface is covered with 1-5-10 nm reduction pieces: the surface becomes unsmooth, and the reflection becomes diffuse (scattered in all directions). Hardness and ductility depend on the percentage of copper (Figure 1d).

Figure 1. All dependencies are redrawn from articles and books: a) Tarnish film thickness vs light or dark [2], b) Sulfide film thickness vs Relative Humidity [3], c) Visible light reflection vs layer thickness (3,9 nm and 10,9 nm) [4], d) Annealed silver alloy hardness vs copper percentage [5, p. 54]

Different cleaning techniques are known: mechanical, chemical,  electrochemical, others [6]. If the mechanical method restorer removes the tarnish layer with cotton wool and a special paste and polishes the artifact at the same time. Chalk (more precisely, calcium), which is found on some silver artifacts, is most likely a consequence of long-term cleaning of silver.

The State Hermitage Museum uses a chemical one mainly: the silver artifacts are cleaned in special local composition (Flurin), then rinsed with distilled water many-many times and wiped off with a towel. A white coating may remain on the silver — this is AgCl, which is removed with ammonia. Apparently, “Flurin” appeared according to patent # 2177053 [7] of Alexander Flate and Galina Yuretskaya. We do not know how “Flurin '' interacts with silver chloride: dissolves the chloride, rinses off the chloride (reducing adhesion), or nothing. What happens if, for example, horny silver is dipped in flurin? Most likely, nothing will happen, but this needs to be checked.

Any spectrometer, like other devices, is appropriate for capturing spectrum, if the spectrum error is less than the result itself. If the error is comparable to or greater than the result, the result is unreliable. The error δY is the distance between the true and measured values with a given probability. The dimension of the error is the same as the dimension of Y, the value of the error depends on the measurement method. As a rule, we either a) know δx, the derivative δY/δx and the probability distribution law or b) we can estimate δY from external considerations (the manufacturer does not help with this often). If the error is infinite (very large), then any measured value is correct and the probability is equal to one.

Accuracy is usually associated with a measurement scale if the device is professional, the accuracy is half a division of the scale. The error is also equal to half a division of the scale. The scale of any device is finite, and everything that is simply added to the measurements is not taken into account in the error. Precision depends on the distance between the experimental values. Resolution is the minimum distance that can be seen; this is a characteristic of optical instruments. If the relative error is 10%, then only 80% of the first digits are correct, if 30%, then 40%. If the relative error is 50% or more, the first digit is not true. Some special reason is required to measure with an accuracy of 30% -50%.

The result of portable XRF is the thickness and elemental composition of the layers, the result of IR or Raman spectrometers is the chemical formula. Both results have no mathematical meaning, these are spec characteristics, only written in symbols. Fitting means you find a spectrum that differs less from the experimental spectrum. The quality of the fit is ∑P_iN_i, where P_i is the probability that the experimental spectrum is similar to the spectrum i, N is the number of possible “formulas” for curve i.

Research methods: acceptable reactions, reaction rate, voltammetry, Auger spectrometer, computer simulator.

Since silver is available to everyone and tarnishes everywhere, many research methods have been applied to understand the nature of silver tarnish. Especially on what the ratio of oxide, chloride and sulfide depends and can this ratio be predicted? Also, what does the tarnishing layer depend on, why does the tarnishing layer grow slowly or quickly? The tarnished layer is the result of many reactions, often following one after another, the product of the first reaction is a reagent for the second, etc. Below are the methods used by many researchers, we will mention only a few of them [8, 9, 10].  Silver does not react with basic air components even at red heat, and thus was considered by alchemists as a noble metal like gold. Its reactivity is intermediate between that of copper and gold. But basic air contains various impurities and pollutants, such as ozone. Silver reacts with ozone at room temperature.

Ag + O3 = (room temperature) = Ag2O + O2

Ozone is light sensitive. And the oxide reacts with chlorine, forming a compound that darkens and decomposes if the light

Ag2O + 2HCl → 2AgCl + H2O

AgCl = (light, room temperature) = 2Ag + Cl2

Like copper, silver reacts with sulfur and its compounds; in their presence, silver tarnishes in air to form the black silver sulfide (copper forms the green sulfate instead, while gold does not react).

2Ag + H2S = (humidity) = Ag2S + H2       

4Ag + 2H2S + O2 = (air) = 2Ag2S + 2H2O

Unlike copper, silver will not react with the halogens, with the exception of fluorine gas, with which it forms the difluoride. While silver is not attacked by non-oxidizing acids, the metal dissolves readily in hot concentrated sulfuric acid, as well as dilute or concentrated nitric acid. In the presence of air, and especially in the presence of hydrogen peroxide, silver dissolves readily in aqueous solutions of cyanide.

AgCl = (light, room temperature) = 2Ag + Cl2

Like copper, silver reacts with sulfur and its compounds; in their presence, silver tarnishes in air to form the black silver sulfide (copper forms the green sulfate instead, while gold does not react).

2Ag + H2S = (humidity) = Ag2S + H2       

4Ag + 2H2S + O2 = (air) = 2Ag2S + 2H2O

Unlike copper, silver will not react with the halogens, with the exception of fluorine gas, with which it forms the difluoride. While silver is not attacked by non-oxidizing acids, the metal dissolves readily in hot concentrated sulfuric acid, as well as dilute or concentrated nitric acid. In the presence of air, and especially in the presence of hydrogen peroxide, silver dissolves readily in aqueous solutions of cyanide.

The reaction rate v is the rate at which concentration (for example [X]) changes over time.

v = k[X]^x[Y]^y, where k = A exp(-E_a/RT)

where gas constant R = 8.31 JK^-1mol^-1, Е_а is аctivation energy. The rate constant k depends on temperature (~√T), catalyst and inhibitor existion. The dimension of the reaction rate depends on the order of the reaction, which is understood as the sum of x + y. The formation of water H⁺ + OH^- → H2O is one of the fastest reactions. An example of a fast reaction is fire too, an example of a slow one is the chemical weathering of stones. 

Table 2.

Examples of the multiplier order A taken from [11]

We do not know the rate at which chlorine and sulfur reacts with silver, but apparently these rates are comparable because tarnishing contains oxide, sulfide, and chloride simultaneously. Oxygen can be from ozone (which is sensitive to light as well). Chloride decomposes in the light. Ozone and chlorine pollution is usually lower than sulfur pollution; light, humidity and air conditions affect the reaction rate too. The reaction rates are comparable with all of the above.

If a voltage is applied to the electrodes immersed in the solution, then an electric current will flow through the circuit. Electrochemical reactions take place: oxidation at the anode and reduction at the cathode. According to Faraday's laws of electrolysis, the transferred electric charge and mass are proportional to each other since, as we know today, they refer to one particle or ion. If we know this ratio, we can choose one reaction from a finite set of pre-selected ones.

The rate of the electrochemical process is determined by its slowest stage. If the speed of the slowest step is zero, the process does not take place. Voltammetry is an experimental and analytical method for analyzing the current-voltage (polarization) curve, which is built during electrolysis on a polarizable electrode. The method is an interpretation of the curves measured in a cell with a polarizable indicator electrode and a non-polarizable reference electrode (in our case, with a sulfate mercury electrode).

Typically, for films with a good surface, the voltage versus time plot should show a series of horizontal sections or steps, each corresponding to a specific recovery potential or voltage (Figure 3). The final potential stage corresponds to the reduction of hydrogen ions in solution (with the formation of gaseous hydrogen) and a limit arises. By the current/voltage ratio, it is possible to establish which substances participated in the reaction, and by other characteristics of the curve, even quantitative indicators of the reaction.

Figure 2. Examples of polarization curves

Figure 2 is taken from an article of Lin, Frankel and Abbot [12], where it can be seen that all silver salts are present at the same time during street corrosion. The region is distinguishable on the polarographic curve if the potentials of the substances differ by more than 0.1–0.2 V. If the potential difference is less than 0.1–0.2 V, the two substances merge into one. Various signal processing methods are used to increase the resolution of the polarographic method.

Auger spectroscopy (AES) is a method for studying the elemental composition of surfaces. The composition of several molecular layers is calculated using Auger spectra and then the surface is etched with ions to a certain depth. The composition calculation and etching is repeated over and over again until the specified analysis depth is reached.

The elemental composition of the layers and their condition when exposed to light were investigated using XPS and AES. First, the dependence of the effect on the wavelength of light is the opposite: the longer the length, the weaker the effect. As a result, the experimental wavelengths, according to the effect produced, are arranged in the following order: 1537 Å> 3650 Å> sunlight. Moreover, the longer the exposure time, the stronger the tarnish. The proportion of Ag2O was 42.9% of silver after exposure for 18 hours to ultraviolet light 1537 Å, but 17.6% if within 24 hours. The proportion of oxide is reduced by exposure to light. The XPS measurement showed that the surface, when exposed to light in the atmosphere, first oxidized and chlorinated, and then silver chloride could be converted to Ag2O, which eventually reduced to black superfine silver particles and again. (In the work [13] Fang Jingli and Yu Yaohua (1985) only the abstract was read)

Saleh, Xu, and Sanvito [14] used the ReaxFF [15] program to tarnish silver. Sulfur compounds, especially hydrogen sulfide (H2S), are believed to tarnish silver even at 1 ppb. Correct simulation can answer any question, but the power of the computer must be greater than the processes it counts. Or you have to use some tricks. The team used a method called ReaxFF (Reactive Force Fields). ReaxFF can "train" a computer to reproduce quantum chemical results and reveal the mechanisms underlying the silver-oxygen and silver-sulfur reactions. Simulation showed that the rate of sulfide formation is much lower than the rate of oxide formation. When the S8 molecules approach the silver, they quickly dissociate into individual atoms and react with the silver. Conversely, O2 dissociates slowly and has less "sticking" capacity. Moreover, it is generally expected that the growth  of the oxide or sulfide will slow down after the formation of the first tarnishing layer due to the oxygen or sulfur atoms having to diffuse through it in order to react with the silver atoms. The researchers found that silver atoms appear to diffuse upward towards the sulfur, allowing the silver sulfide to grow much faster. “Ag ion… highly mobile in Ag2S” as opposed to oxygen.

Examples

Example 1: "Biscuit Box"

The “Biscuit Box” is made of an alloy of silver (main) and copper with traces of impurities. The boxes had to be cleaned because they were blackish. Two XRF spectra were captured, one before and one after cleaning with “flurin”. XRF spectra had two tasks: the thickness of the tarnishing layer and the composition of the layers, which had to instrumentally decide whether it was necessary to clean and clean it well or not. The composition has been divided into two layers: tarnishing and alloy.

The depth of the surface analysis was about 15 μm, considering the K and L lines of silver. The average approximation error was 12% for the analyzed spectrum. Taking into account the estimated thickness of the upper layer of 0.01 μm, as well as its density, an error of 0.01 7/15 11.5 = 0.05% was needed and not 12%. As can be seen on the spectrum plot: the XRF spectra before and after cleaning practically do not differ (Figure 3).

Figure 3. Two experimental XRF spectra, one before and one after cleaning. Sulfur peak is 118 channel, chlorine peak is 134 channel. The two high peaks are the L line of silver

It is necessary to take more than 20 sec so that the spectrum becomes even, and not a herringbone. If we capture the spectrum not only for 20 sec, but much more, for example, 999 sec, the spectrum will rise up and become a smoother curve. The spectrum is simply a histogram of the registered impulses distributed over energies. We tested this statement on another XRF, Hitachi X-MET. We captured two spectra for 30 sec and for 600 sec, and compared them. Unfortunately for Hitachi there is no published calibration file for PiMca to use yet. When and if the calibration file becomes available it may be possible with the help of Hitachi to determine even the ratio of the content of radium to thorium and thereby establish the age of metal artifacts [16].

Example 2: "Reliquary"

The surface of the artifact was first divided into two layers: a tarnishing layer and an alloy layer. But all attempts to make the fit ended well, until we introduced a third layer: a layer of pure silver. The fit has become better, although not perfect, the approximation error has decreased. This can be seen on the plot (Figure 4).

Figure 4. Experimental XRF spectrum and two types of fit, where there are two and three layers, respectively

We have one explanation why the silvering was done with absolutely pure silver without any impurities in the 17th century. Even before Boulsover's invitation in 1742 (Sheffield plate), there were two methods [17] of industrial silvering: French coating [18] (or French Plaque) and close coating. The first method was invented in the early 1700s and used for electroplating later. Brass or copper, polished and degreased and heated almost to red hot before being placed in nitric acid. Silver plating was the application of four to six silver sheets. The sheet was also heated and then quickly and under pressure was applied to the surface. The sheets silvered by this method were specially marked indicating the thickness of the silver. The thickness was 6 to 33 μm minimum.

Example 3: "Modern jewelry"

The third example is related to the Raman spectrometer (Figure 5). We use the method of comparing spectra based on correlation coefficients (one of at least four methods). The correlation coefficient r is in the range -1 .. +1 where the correlation r = 0.95 corresponds to an error δ = 15%, according to pharmacologists [19]. Since a possible maximum δ = 30%, the correlation coefficient r can be 90%.

Figure 5. Raman spectra of Ag2S: a) experimental (more smooth) and b) Martina (less smooth) [20]

It is not enough to fit the spectrum with a given error, because people know tens of millions of organic and hundreds of thousands of inorganic substances, even if mixtures are not yet taken into account (Although it is the mixtures that are more common). No database includes that many. It is necessary to limit the number of possible formulas in any way. We used a portable XRF device. XRF completely eliminated any organic compound simply due to the sum of the number of pulses, еxcluded at least in a noticeable amount. The peak of Light Elements, generating organic matter (H; C; O; N) and air (mainly), is less than 1 keV and is poorly recorded by the built-in receiver. Also, C - O - F is more transparent to X-rays, and the resulting photon goes not only to the receiver, but in all directions. The distance increases as x, and the receiver area decreases as x². XRF allows determining the position of silver, for example, light elements (organic matter) are located before or after the silver foil.

Conclusion.

The test results have shown that portable XRF with a 20 second spectrum is not very suitable for measuring tarnish layer thickness. Perhaps, if the spectrum is recorded for 999 seconds, and not 20, it will become smoother and it will be possible to measure the layer thickness. The error is approximately equal to the ratio of the layer thickness to the depth of analysis in silver and, taking into account the density, which is approximately 0.04%, this is very small. To make a final conclusion, modern XRF and its software calibration are required. To identify a substance by its spectrum, an identification error is required that is less than the percentage of possible substances (now about 30 million are known). Or almost all substances should be excluded from consideration.

It seems that no one has tried to ground silver artifacts. Maybe grounding will reduce the growth rate of the tarnishing layer, but maybe not. The test is necessary.

Bibliography:

1. Pradyot Patnaik (2001). Handbook of inorganic chemical compounds. London: Mcgraw-Hill. ISBN 0070494398
2. Jiang Yang, C., Hao Liang, C., & Liu, X. (2007). Tarnishing of silver in environments with sulphur contamination. Anti-Corrosion Methods and Materials, 54(1), 21–26. https://doi.org/10.1108/00035590710717357
3. Reagor, B. T., & Sinclair, J. D. (1981). Tarnishing of Silver by Sulfur Vapor: Film Characteristics and Humidity Effects. Journal of The Electrochemical Society, 128(3), 701–705. https://doi.org/10.1149/1.2127485
4. Bennett, J. M., Stanford, J. L., & Ashley, E. J. (1970). Optical Constants of Silver Sulfide Tarnish Films. Journal of the Optical Society of America, 60(2), 224. https://doi.org/10.1364/josa.60.000224
5. Rhines, F.N. (1956). Phase diagrams in metallurgy : their development and application. New York: Mcgraw-Hill, ISBN 9780070520707
6. Palomar, T. & Ramirez Barat, B. & Cano, E. (2018). Evaluation of Cleaning Treatments For Tarnished Silver: The Conservator´s Perspective. International Journal of Conservation Science. 9. 81-90.
7. Patent, https://patents.google.com/patent/RU2177053C2/en,  Retrieved 01.01.2021
8. Burstein, G.T. (2019). The tarnishing conundrum of silver.
9. Selwyn, L., (2016) Understanding how silver objects tarnish, https://www.canada.ca/en/conservation-institute/services/training-learning/in-person-workshops/understanding-silver-tarnish.html, retrieved at 01.09.21
10. Inaba, M., (1996) Tarnishing of Silver: A Short Review, Conservation Journal January 1996 Issue 18, http://www.vam.ac.uk/content/journals/conservation-journal/issue-18/tarnishing-of-silver-a-short-review/, retrieved at 01.09.21
11. J. A. Manion, R. E. Huie, R. D. Levin, D. R. Burgess Jr., V. L. Orkin, W. Tsang, W. S. McGivern, J. W. Hudgens, V. D. Knyazev, D. B. Atkinson, E. Chai, A. M. Tereza, C.-Y. Lin, T. C. Allison, W. G. Mallard, F. Westley, J. T. Herron, R. F. Hampson, and D. H. Frizzell, NIST Chemical Kinetics Database, NIST Standard Reference Database 17, Version 7.0 (Web Version), Release 1.6.8, Data version 2015.09, National Institute of Standards and Technology, Gaithersburg, Maryland, 20899-8320.  Web address:  https://kinetics.nist.gov/
12. Lin, H., Frankel, G. S., & Abbott, W. H. (2013). Analysis of Ag Corrosion Products. Journal of The Electrochemical Society, 160(8), C345–C355. https://doi.org/10.1149/2.055308jes
13. 13. Fang Jingli and Yu Yaohua (1985) XPS and AES Study on the Tarnishing Mechanism of Silver-Electroplated Deposit(Ⅱ) Mechanism of Tarnishing Caused by Exposure to Light and Na_2S Treatment, CIESC journal
14. Saleh, G., Xu, C., & Sanvito, S. (2019). Silver Tarnishing Mechanism Revealed by Molecular Dynamics Simulations. Angewandte Chemie International Edition, 58(18), 6017–6021. https://doi.org/10.1002/anie.201901630
15. Senftle, T. P., Hong, S., Islam, M. M., Kylasa, S. B., Zheng, Y., Shin, Y. K., Junkermeier, C., Engel-Herbert, R., Janik, M. J., Aktulga, H. M., Verstraelen, T., Grama, A., & van Duin, A. C. T. (2016). The ReaxFF reactive force-field: development, applications and future directions. Npj Computational Materials, 2(1). https://doi.org/10.1038/npjcompumats.2015.11
16. Doménech-Carbó, A., Doménech-Carbó, M. T., Capelo, S., Pasíes, T., & Martínez-Lázaro, I. (2014). Dating Archaeological Copper/Bronze Artifacts by Using the Voltammetry of Microparticles. Angewandte Chemie International Edition, 53(35), 9262–9266. https://doi.org/10.1002/anie.201404522
17. Turner, E.,. (1993). Silver plating in the 18th century. In Metal Plating and Patination,
18.  Elsevier. https://doi.org/10.1016/b978-0-7506-1611-9.50024
19. Metropoliten Museum # 200377 and others
20. Gryniewicz-Ruzicka, C. M., Rodriguez, J. D., Arzhantsev, S., Buhse, L. F., & Kauffman, J. F. (2012). Libraries, classifiers, and quantifiers: A comparison of chemometric methods for the analysis of Raman spectra of contaminated pharmaceutical materials. Journal of Pharmaceutical and Biomedical Analysis, 61, 191–198. https://doi.org/10.1016/j.jpba.2011.12.002
21. Martina, Irene & Wiesinger, Rita & Jembrih-Simbuerger, Dubravka & Schreiner, Manfred. (2012). Micro-Raman characterisation of silver corrosion products. e-Preservation Science. 9. 1-8.