Archaeological context and samples
Tomb M54 was discovered and excavated in 2000 and 2001. More than 210 pieces of jade were excavated in this tomb, which is the second largest quantity after the tomb of Fu Hao. The inscription “Ya Chang” on most of the bronze ritual wares indicates that the occupant of tomb M54 must be a leader of the “Chang” clan, and it is dated to late phase II of the Yinxu period [27].
Three jade artifacts were selected from tomb M54 (Figs. 2, 3), and their material was determined by infrared spectroscopy to be D-type nephrite, i.e. dolomitic marble-related nephrite [28]. They are translucent, as they are hardly affected by weathering. They have a bright yellow‒green hue, show a very good sense of warmth and fineness, and are of high quality, usually without blemishes. Because of their similar colour and feel, it is often intuitively assumed that they have the same provenance.
The handle-shaped artifact (M54:398) is a particular type of jade object from the Xia, Shang, and Zhou dynasties. It is particularly abundant in the late Shang period, and was usually found in the tombs of high-ranking nobles in the capital of the kingdom. They usually consist of a narrow, rectangular body and a thin, short, sometimes pieced tang at the bottom; some even have a scabbard, and most are of small weight and size [29]. Arrowheads belong to the category of weapons, and this type of arrowhead made of nephrite is also a characteristic feature of the Yinxu culture, which, although not practical, reflects the dignified status of the tomb owner. Samples M54:364 and M54:366 represent exactly two different types of arrowheads. The former has an elongated shaft at the lower end that can be inserted into a groove, while the latter has a recessed notch in the axis of symmetry and has a perforation that allows the rope to anchor and secure itself. From the type and style of these objects, it can be concluded that they were processed in the late Shang period.
Based on several reasonable assumptions about the availability of jade and considerations of its quality, nephrite from Xinjiang, Qinghai, and Liaoning has always been considered the most likely material for the production of ancient jade from Yinxu, which is why we made it the focus of comparative study. Modern samples closer in color to the yellow-green tones of Yinxu jade are selected for comparative study as much as possible. To ensure that they are all D-type like the ancient jade, the selected samples were tested with pXRF for a composition of Mg/(Mg + Fe) well above 0.93 [2] (Additional file 1: Table S1).
Method
It is important to emphasize that multispectral imaging is a method for detecting features of nephrite used by traditional visual methods. So far, the structural features of nephrite have been commonly used in the market, but the terminology used by different people is inconsistent, and some connoisseurs or folk experts borrow a lot of metaphorical rhetoric to describe the characteristics of nephrite as they see them, and these experiences can lead to some misunderstandings or errors when they are disseminated. As a result, many visual studies of nephrite have taken on a strong occult flavour and are often popular among a large group of Chinese folk connoisseurs, which in turn can cause many researchers to deliberately avoid them. Although these subjective experiences have actually had a positive effect on advancing jade research, they have not yet been rationally interpreted, much less confronted by scientific research. The scientific recording of images of nephrite features is sufficient to break through this limitation and closure, providing a common ground and basis for discussion for all and another way to solve the problem of determining provenance.
In this study, the CRi Nuance imaging system (USA) was used. The images are acquired in the wavelength range of 450–740 nm with a scanning step of 10 nm, covering the visible (VIS) and near-infrared (NIR) regions. The general experimental procedure is as described in our previous articles [25, 26], with some exceptions highlighted here for archaeological samples. The main operational improvement is to place the light source under the sample, i.e., to illuminate the sample transmissively instead of picking up scattered or reflected light (Fig. 4). Archaeological samples must be non-destructive, and no risky tests should be performed. In our multispectral experiments, although there is no direct contact with the sample, too much light can lead to high temperatures that can change the properties of the sample due to the light source that must pass through the sample. For this reason, transmitted light produced by light emitting diodes (LEDs) is used to illuminate the sample instead of halogen lamps and other variable light sources. The sample was placed on the support platform at the correct height and angle to ensure that it is within the field of view of the objective. To avoid stray light, an enclosed space is constructed around the specimen to serve as a darkroom. The distance between the flat surface for all of samples and the lens is the same, the depth of field of focus will be consequently easier to be determined. The focal length of the lens was then adjusted until the image of the sample was clearly shown. The exposure parameters were preset and self-adjusted and the images will be recorded and saved by software.
It is apparent from a series of multispectral images that each image receives a different amount of light energy. With the increase of wavelength, the image gradually changes from dark to light and then from light to dark. When the light wavelength is about 600 nm, the brightness of the image will reach the highest. In most cases, the area with 0 grey level is the background which need to be discarded, whereas the brightest area is also invalid due to the high intensity of illumination. If there is an area with grey level higher than 230, the observation will be affected. Hence, if the overall grey levels were set up more balanced, the original image will contain more details and gradations. Based on the above reasons, the moderate image of band 660 nm is selected as representatives in each series, and at the same time, the band of 510 nm could also be chosen due to symmetry presented by the multispectrum [26].
Furthermore, to make the target features clearer, an idea of subtracting the fitted light distribution image from the original image (660 nm band) was adopted. Here a smoothing linear filter was applied to simulate the light distribution. As a low-pass filter, it has the net effect of blurring or smoothing the image, e.g. bridging small gaps in lines or curves. The value of each pixel in the image is replaced by the weighted average value which is the intensity levels in the neighborhood defined by the filter mask, this process will reduce abrupt transitions in an image [30]. After many trials, a filter mask of 40 × 40 was chosen for the linear filtering, which greatly enhances the visual effect without deviating from the original image features [26].