Results from our experimental and methodological investigation into the physical characteristics and remaining trace evidence of manufacture, treatment, corrosion and archaeology are presented below.
On mirrors 1,2 and 3, polishing marks were seen on both sides. On mirror 1 and 2, by using the polarized light microscope (PLM) with reflected light, parallel lines with sharp angles were observed. This type of line is likely caused by mechanical polishing of the surface . For mirror 1 the marks can be seen on-top of some protruding corrosion rather than passing beneath them (Fig. 2a & 2d). This indicates that this polishing was done post formation of these corrosion and that an aggressive cleaning of the surface was undertaken quite recently, perhaps to remove significant corrosion to get to the base metal surface in attempts to improve the mirror “aesthetics”. For mirror 2 it is clear that newer corrosion components have formed on top of some of the presumably older tool marks (see arrowed examples in the figure) while the surface has some inclusions that are many times (10x) larger than those small surface defects seen on mirror 1 (Fig. 2b & 2e). Mirror 3 is similar to mirror 2 but has a factor of 3-5x more surface defects with a mixture of broader and finer tool marks. In some cases the cleaning tracks can be seen going through these surface defects. In another case (top left of Fig. 2c) these tracks are under the corrosion. These tracks also have a broader spread of angles across the surface indicating several different cleaning episodes. Due to the highly corroded surface, there are no polishing marks that can be observed on mirror 4.
Mirrors 1 and 2 have a black, shiny patina on their fronts (see Fig.1b) common to mirrors from the Warring States period. The areas without pitted corrosion still present a reasonably reflective surface. Mirror 3 has as a variable, very dark to lighter green patina. From a region towards the edge of mirror 3 where the overlaying shiny patina has become detached, the division of surface patina and further corroded metal underneath is distinctive. In this small area, where the lighter shiny olive green patina has come away, a lower level “crinkly” sub-surface is evident Fig.3 (left). This is seen in many other bronzes from antiquity. An example is shown in Fig.3 (right) of a 2nd century AD Roman brooch from Cirencester museum in the UK (photo taken by the 2nd author in-situ in December 2017). This is shown next to mirror 3 together with arrowed inserts showing the common patination features. Both mirror and brooch are from approximately the same time period, both are small in size and have similar light to dark olive green shiny patinas. However, in both examples the top level patina has flaked off in some places to reveal a crinkly, sub-surface condition. Such features are, to our knowledge, not present in modern replicas.
Mirror 1 does not have any obvious archaeological evidence such as any encrusted earth deposits or any other phenomenon related to a burial history such as attached organic material, including textile pseudomorph marks from wrapping in cloth prior to burial. Mirror 2 does have some apparent encrustation remaining on the mirrors’ 3mm wide edge that extends for about 2cm. Images of this attached encrustation are shown in Fig. 4a at 40x magnification. A fine implement can remove this material quite easily. Both mirrors 1 and 2 appear to have been heavily cleaned (see Figs. 2a-b). For mirror 3, within the incisions of the inscribed pattern of the jade annular attachment, some traces of earthen encrustations were also observed. By touching by scalpel, these encrustations are very hard to remove. The formation of this type of crust on a jade surface is usually caused by a prolonged burial history. The jade annulus itself is 12mm wide with an outer diameter of 52 mm and inner diameter of 28mm with a typical late-Warring States simple, scroll-type, panchi pattern. From microscopic examination, it is clear this piece was not worked with modern machine tools but betrays the tell-tale marks and scratches typical for jade hand-worked from this period and evident in Fig. 4b. The jade annulus itself is at least partially recessed into the bronze mirror being almost flush with the surface on one side with a protruding thickness of ~1.9mm for the most part but closer to 3mm for the inner edge closer to the central boss.
On mirror 4, the overall surfaces are quite heavily corroded and covered with deep greenish blue corrosions raised from the base mirror surface together with some white crust along one edge. Textile pseudomorph marks were observed on this particular corrosion layer (see Fig. 5a,b). This area also fluoresces bright yellow under UV light indicating some non metallic possibly organic material is present in this region (see Fig.13). The impressed textile marks look natural. From the woven pattern (Fig. 5b), it is likely from a linin fabric. It is not rare to see such textile pseudomorph marks on bronze mirrors and other bronzes. As an important and elegant personal belonging, mirrors were usually buried with the owner, and sometimes wrapped with a textile [15, Fig.10]. After a long time underground, the textile, which is an organic material, deteriorates away. Each textile fibre has been replaced by mineral salts preserving the pattern but the actual organic component has gone. This is true especially in the regions where underground water levels fluctuate actively. Several insect carcasses can also be seen embedded in the thick corrosion and some charcoal (see Figs. 6 & 12).
The surfaces of mirror 1 have been well preserved but extensive abrasive cleaning seems also to have been undertaken perhaps in the recent past. On the front surface, towards the mirror edges severe pitted corrosion can be seen in several places that has eaten into the smooth, light green surface patina. By looking at these areas where the surface patina has been broken-up, 3 distinct layers can be distinguished (see Figs. 7 & 8). The top layer is a light green, ~200 micron thick shiny patina, underneath is a black layer and under this is a layer of greenish powdery, pitted depressions. It is hard to tell if the corrosion starts from inside the metal or has ingressed from outside. We believe the black middle layer is the original mirror surface akin to what is seen for the other “warring states” mirror 2 and resulting from application of the “xuan xi” technique referred to earlier.
However, given the absence of metal surface where the pits are exposed and the trend of the increasing density of number of pits around the powdery area, it is very likely that the corrosion starts from each single pit where moisture has gotten into the metal to stimulate the process of decuprification to form green corrosion and then turn into powdered corrosion. In certain regions the original metal surface is gradually replaced by this green corrosion. Along with forming increasing green powder corrosion, the green surface has been eaten away. Therefore, the green surface is the middle phase of the whole deterioration process. On the back surface, towards the mirror edges, the thick green layer is absent where a black rough metal underneath is exposed (Fig. 8).
For mirror 2 both front and back surfaces are well preserved with the surfaces dark, almost black in colour. A similar corrosion phenomenon (Fig. 9) to mirror 1 can also be seen on parts of the front surface of mirror 2 although in this case there is no additional green surface patina on top of the dark layer. It is possible this has been removed by aggressive cleaning. Both mirrors one and two were acquired at the same time from the same vendor. On the back surface towards the edge in one area in particular, a rougher surface is observed. It was likely formed during the manufacturing process since part of the relief is disturbed (Fig. 10).
For mirror 3 the front surface is mainly smooth and dark. However, there are regions of bright green malachite corrosion along with small regions of red corrosion which is assumed to be cuprite, as shown in Fig. 11. On the front surface there is also a meandering hairline crack within the body of the mirror (refer Fig.14) that does not extend to the edges. It has an overall extent of ~5cm. Mirror 4 is covered with thick green corrosion. In some regions, small, black “charcoal” like inclusions can be seen within the corrosion (Fig.12 right).
Small samples of the greenish white powdery corrosion from the pitted areas on mirrors 1 and 2; the green and red corrosion from mirror 3; and the white encrustation, green corrosion and black inclusions from mirror 4 were taken. These samples were mounted with MeltmoundÒ 1.662 resin on glass slides for PLM. The result are given in Table 3.
On mirror 1 and 2, the greenish white powdery corrosions were identified to be copper hydroxide (spertinite). Copper hydroxide is regarded as an unstable and an intermediate product that will transform to a more stable mineral such as copper carbonate during the corrosion process [23, p.98]. There is no indication as to when and how this deterioration process started on mirror 1 & 2, especially with traces of the modern intervention found.
On mirror 3, the green corrosion on the reflective surface is confirmed to be malachite by PLM and confirmed by infrared spectroscopy (see later). Due to the limited sample obtained from the red in round-shape corroded area (Fig.11), only one particle was successfully mounted on a glass slide. The particle looks quite dark in both plane and cross-polarized light. PLM did not offer strong diagnostic information to this red corrosion. However, from the formation shape and colour, it is very likely cuprite [32, Fig. 240]. Scott [23, p.106] indicated that a natural transaction from cuprite to malachite is very hard to duplicate in the laboratory. In other words, the presence of cuprite covered with malachite, as seems the case here, is strong proof of the truth that the formation of malachite is natural on this mirror.
For mirror 4, the known process of the migration of lead within the alloy affects the composition of the distinct but localised white crust that bears the textile marks (pseuodmorphs). The pXRF result on this area (see later) shows that the whitish crust on mirror 4 has a very high lead value (up to 93%). When the content of lead exceeds tin in a bronze it would form thick and crusty corrosion products mostly of carbonates and oxides of lead . PLM proves the white mineral is lead white (lead carbonate). The black inclusion is confirmed to be carbon based black. The bond of this top black inclusion and the green corrosion underneath is strong which does not allow easy sampling. This indicates the carbon black existed on the surface along with the growth of the copper corrosion. The green corrosion on mirror 4 was identified to be atacamite.
UV fluorescence Imaging
UV fluorescence imaging has been a valuable diagnostic tool in the art and archaeology fields since the 1920s . The four mirrors were subjected to UV illumination to detect if any modern restorations or applications (e.g. using adhesive/paint to attach a false surface patina), have been done. The results show that no restorations have been found on any of these mirrors. Mirror 3 had evidence for a brown “glue” under the jade annulus that is fixing it onto the back surface of the mirror. UV imagery did not reveal any fluorescence indicating that whatever was used to fix the jade annulus in place was not any modern resin or adhesive. Furthermore, the strongly attached encrustations here and there within the jade carvings did not fluoresce either so these are not modern “glued-on” material that is sometimes used to create an impression of authenticity by the unscrupulous. UV imagery of Mirror 4 did reveal in stark relief the grey-white areas of corrosion associated with the textile pseudomorph regions. This indicates that some organic components may remain in the fluorescent region.
Surface Metallurgy from pXRF measurements
For each mirror, 3-4 test spots were chosen for analysis using pXRF. Here it is crucial to appreciate that the pXRF measurements only penetrate a small distance into the surface. Hence, where the corrosion is thick, only these products can be assessed. All data was therefore from the surface but the cleanest regions where chosen wherever possible. For mirror 4, the data was collected from the area with the least green corrosion. The “vacuum” attachment and lower kV and high current was used for the corroded area in order to detect chloride and lighter elements. Bruker® Artax analytical software was used to analyse the data. By analysing the net count rates, which are the number of photons recorded for each element after removing other elemental interference and background, a percentage by abundance for each recognized element was calculated. Given the data was taken from the surface layers, where any corrosion or residues would affect the absolute value, the averaged results from the combined 4 spot measurements should be regarded as quantitative of the mirror surfaces only.
The pXRF data was taken at 40kV, 20 µmA, with the yellow filter for metal analysis with 120 second exposures for each spot. The Table 4 results in Appendix I are the averages from the 3-4 spot measurements. The K12, L1 and M1 designations refer to electron shells where K denotes the first shell (or energy level), L the second shell and M, the third shell.
From the pXRF analysis of the alloy composition of the four mirrors they are all bronze, i.e. a copper-tin-lead ternary alloy with variation of minor and trace constituents. Those additional trace elements apparent in the mirrors are also seen in other ancient Chinese bronze artifacts. Iron, silver, gold, nickel, cobalt, zinc,bismuth and antimony are regarded as impurities from the unrefined base ores accessed in antiquity . Manganese, barium, calcium are from the earth encrustation or contamination on the surface of the mirrors. Silicon was not detected because the pXRF measurement parameters are set for metals. Zinc is not over 10% in any of these mirrors which indicates that none are poorly made, easily acquired modern fakes where high levels of zinc are common [25,26].
As seen in Table 1. the observed fraction of copper in provenanced Chinese bronze mirrors is quite wide, ranging typically from 65% to 80%. In the case of a heavily corroded surface or a surface of qigu, the copper content can be quite low. Hence copper content is not a top diagnostic factor in bronze mirror appraisal based on elemental composition. He [15,p.76] states that mirrors dated to the Warring States typically have a tin fraction that is much higher than lead. The range in the amounts of tin and lead found in mirrors 1 and 2 does not match the opinion of He. Here it is only 17-24% tin c.f. 8-20% lead. He & Song  also indicated that mirrors dated to the Warring States with low fraction of iron (Fe) were probably unearthed from central China Shaanxi, Shanxi province or from inner Mongolia. For mirror 2 in particular, the observed 20.2% lead is much higher than the current reference record.
For mirror 3 the surface copper content is extremely low while the tin content is 50%. The likely formation of green qigu (Fig. 3) contributes to this outcome. Similar data on Cu-Sn ratios were achieved on mirror surfaces by He  and given in a scientific report  by the lab in University of Science and Technology of China in 1988. Although 50% tin was achieved from surface measurements, the presence of a crack indicates the brittleness of the alloy in this mirror (Fig.14). Indeed, the drawback of high tin content is that it can lead to a brittle copper alloy. Most published data on the lead content on the surface of mirrors dated to the Han dynasty are less than 10%. Therefore, the lead in mirror 3 is in the reasonable range for provenanced Han bronze mirrors.
For mirror 4, due to the highly corroded surface, the ratio of major elements found should not be considered as representative of the metal surface. Even so, the ratios of each element fall into the range of published data on archaeological provenanced mirrors dated to the same period. According to He & Song , a Song mirror unearthed from Linxi, Shandong province (northern China) present a high lead content (35.11-35.75%) on its surfaces. Conversely, mirrors unearthed from Echeng City in Hubei province (southern China) show a relatively low lead component (average 20.220% for 12 mirrors). An interesting phenomenon for mirror 4 is the much more highly corroded surface compared to the other three mirrors and that the measured lead fraction is high, averaging ~38% on the front surface. There are two possible explanations and sources for the lead. 1). It is migration to the surface from within the alloy itself; 2). It arises from the lead mercury finishing paste used to keep the reflective surface shiny . This could at least also partially explain why the front surfaces for mirrors 2, 3 and 4 also have a higher lead content. No matter what the source is, lead is an active factor in causing more corrosion. This is also true for other types of bona-fide archaeologically excavated bronzes with high lead content .
pXRF and the Presence of mercury (Hg)
Trace but reliable levels of mercury (Hg) were detected on both mirrors 3 & 4, indicating that these two mirrors at least were likely treated with the quicksilver polishing technique described in the historical records. No mercury was detected on mirrors 1 and 2. It is hard to conclude that no mercury was used to treat these mirrors since if heat was involved during the process, the mercury may evaporate .
pXRF and the Presence of chromium (Cr)
There is chromium (Cr) present on both mirrors 1 and 2 as a trace element. A few ancient bronzes have been discovered bearing chromium that have been studied in the past 30 years. Scholars are interested in whether such chromium was added deliberately by the ancients for anti-rust purposes. However, most investigations [38-40] pointed out that the chromium was not an intentional addition but a contamination from the natural environment, except perhaps for one bronze Pan dated to the Warring States Chu Culture. The content of chromium in the sample surface-film by Luo  for this bronze Pan is 35.61%, much higher than the average content in any other publication and much higher than found here for mirrors 1 and 2 (less than1% chromium). As chromium (III) oxide (Cr2O3) is a modern polishing product used on metal we conclude that the detected chromium is more likely a result of cleaning with chromium oxide more recently, especially considering the sharp angled modern polishing tool marks observed on the reflective surfaces on mirrors 1 and 2. Interestingly, these two mirrors came from the same source. This may also explain the absence or very low levels of mercury on those two mirrors, where a heavy, abrasive polish could have removed any remaining trace of mercury.
pXRF of the corroded areas
The pXRF with settings of 15Kev, 36 µmA, vacuum implemented and with 120 second exposure testing times were used to analyse chosen corroded areas on pitted areas of mirror 1 and 2 to detect chloride (Cl) which commonly causes ‘bronze disease’, seen as greenish white powder on bronzes . However, no chloride was detected. For mirror 3, data was collected on the green corrosion on the reflective surface as in Fig. 11. The green corrosion and thick crust on mirror 4 were also analysed. Chart 1 shows the elemental distribution on the targeted areas on the four mirrors.
A MARS tomographic 3-D X-ray spectral scanner was used to produce saggital X-ray scans of these 4 mirrors. This is new technology designed for bio-imaging applications that generates multi-energy images with high spatial resolution and low noise. The use of this scanner for bronze artifacts is a world first that required extensive testing and characterisation before useful data could be obtained for these mirrors. The best scanning characteristics established through trial and error are given in Table 5. below.
Table 5. MARS best-estimate scanning characteristics determined for these mirrors.Notes to Table 5: The ASN# is just the identification number for a given scan run. FOV stand for field of view (or sample diameter), SOD is the source-to-object distance, and ODD is the object-to-detector distance. As these are 3-D tomorgraphic scans there is a volumetric pixel or “voxel” which here was set at a resolution of 0.09mm. To reduce artifacts in the scans a 7mm copper filter was inserted between the artifact and the X-ray beam in conjunction with long exposure times and high beam currents to maximise transmitted X-ray photons through each mirror.
The most interesting X-ray results comes for the Han mirror 3 which shows that the metal surface under the jade annulus (not registered in these higher energy X-ray images) is flat and unadorned as shown in Fig.15a. Most Han mirrors have a pattern here but the lack of one indicates the high likelihood that the mirror was made for the jade annulus (which may itself date from the Warring States) rather than being added later. The hole through the central knob is also shown to be non-tangential. As saggital scans delve deeper into the mirror irregularities in the bronze smelt are also evident (Fig. 15b) giving further credence to imperfections in the smelting and alloy mixing process. This process is much more homogeneous in modern replicas.
For mirror 2 no irregularities in the saggital slices through the mirror were seen implying a much more homogeneous alloy mix – so no imagery is presented.
Due to its thickness mirror 4 was very difficult to scan properly with MARS but from what was possible it is clear there are significant variations in the internal alloy density distribution and structure as shown in the two examples below for the same mirror segment.
An ADS portable, high-resolution, near-infrared (NIR) spectroradiometer (model A122320) from the PSML at HKU was used in its 3 mm fibre-optic probe configuration to perform contact spectroscopy of the green corrosion region on the front surface of mirror 3 as shown in Fig.11 over the wavelength range 1.8 to 2.4 microns at a reslotion of ~6 nanometres. The resultant NIR spectroscopic results from this test are shown in Fig.18. The radiance data were calibrated against a white reflectance standard to produce calibrated reflectance values. The reflectance spectrum is compared to a calibrated reflectance spectrum of malachite from the United States Geological Survey spectral library. There is excellent agreement between these spectra. The base bronze alloy and red cuprite corrision have no spectral features in this wavelength range allowing purer identification. The malachite is distinct from other carbonate minerals by its inidcative green colour and the clear C-O absorptions located near 2.27 microns. This NIR analysis confirms the identification of this green corrosion product formally as malachite, as already inferred from simple visual inspection and from the separate Polarised Light Microscopy.
Chemical Solubility Spot Testing
Solubility spot testing across each mirror was the last process applied as part of this forensic investigation. Three chemical solvents were used: ethanol, acetone and turpentine (paint thinner). Here the intention was to see if there was any evidence for any acrylic or oil based paint or adhesive application of “false” patina to any of these mirrors. If so then application of these chemicals would dissolve any such constituents. Small cotton swabs were used to apply each solvent in turn to each mirror. Several regions were chosen on each mirror targetting the various corrosion products and surfaces evident. In no case was any obvious dissolving of any surface noticed and the cotton swabs remained clean with no discolouration.