Practices and Rules of 16th Century Genoese Gilding: Exploring Gold Leaf Thickness and Caratage through X-ray and Ion Beam Techniques

Abstract

This study investigates the practices and rules of Genoese gilding, drawing insights from a 16th-century manuscript containing regulations for gold leaf production. Employing X-ray and ion beam techniques, we quantitatively assess the manuscript’s gold leaf thickness without destructive sampling. Artisanal goldbeater-produced leaves of different thicknesses, applied with a guazzo or mordant technique, served as standards. Further analysis of samples with unknown thickness from the furniture of Palazzo Spinola di Pellicceria in Genoa (Italy) has confirmed the method’s applicability to practical cases. External beam Rutherford backscattering spectrometry (RBS) and particle-induced X-ray emission (PIXE) analyses were carried out using 3 MeV protons at the LABEC accelerator laboratory in Florence. A linear relationship between Gold Lα peak yield and leaf thickness, as measured by RBS, has been established for optimal calibration of portable or hand-held X-Ray fluorescence (XRF) instrumentation for in situ measurements. Moreover, the caratage of the gold foil preserved in the manuscript has been assessed.

1. Introduction

In Genoa, decorating with thin gold foils—called gold leaf gilding—has been regulated since the medieval age by strict rules. The Genoese panorama represents a unique case due to the vast archival documentation available, among which the Giovanni Battista Paggi 1590 querelle (trial), an ideological debate, stands out [1,2]. The painter launched a legal battle to liberate the art of painting from the corporative system, in which craft tasks, such as gilding, were still practiced. The consequence was that gilding emerged as an autonomous discipline and the gilders founded their own guild [3]. Therefore, since the 16th century, numerous regulations governing the application of gold leaf in Genoa have existed [4,5]. Trade in ‘gilded tin’ was expressly forbidden and gilding was to be done only with aurum bonum (“good/pure gold”—predominantly 23 ¾ and 24 carat—i.e., 100% pure gold). Moreover, in numerous documents, we find sample leaves of 9 × 9 cm², meaning that in Genoa, a larger foil size than the ones produced by Venetian and Florentine guilds (7.2 × 7.2 or 7.3 × 7.3 cm²) was allowed [6,7,8].

The discovery of a manuscript that preserves the prototype of the gold leaf intended for trade in Genoa, signed by the producers (i.e., the goldbeaters) and the customers (i.e., the painters) who used it in their gilding, represents a turning point in the research in this field (Figure 1a).

It is known that the censors, with whom the gold leaf prototype was deposited, regulated the production of goods in the city and had the duty of guaranteeing their quality [9,10,11]. It is, therefore, a true commercial agreement that regulates the shape, thickness, and carat of the gold leaf in commerce. Thickness is quite important to characterize gilding: the progress made by goldbeater artisans through the ages can be traced thanks to the evolution of the thickness of the gold foil product [8,12,13,14]. Moreover, assessing the caratage of the leaf can provide insight into the actual compliance with the guild’s aurum bonum rule. Unfortunately, the integrity of the leaf in the document is quite compromised both because of the normal wear and tear derived from the page consultation and because of intentional engravings (Figure 1b), but an estimation of the original features, like composition, thickness, size, and type of substrate employed, can be obtained by means of X-ray fluorescence (XRF) analysis, which usually does not require invasive sampling.

XRF is a versatile technique that combines high analytical sensitivity with low cost and simple instrumentation and it can be used to assess the gold foil thickness [15,16,17,18,19,20,21,22,23,24], appropriately calibrating the instrument with the support of ion beam analysis (IBA) techniques, namely Rutherford backscattering spectrometry (RBS) and the particle-induced X-ray emission (PIXE) technique [25]. Caratage (K) is, instead, estimated by calculating the ratio of pure gold to the total weight of the alloy, determining the percentage of gold in the gold leaf, with higher caratages indicating a higher proportion of gold over copper or silver [26,27]. Moreover, thanks to the possibility of scanning large areas with an XRF spectrometer (Macro Area XRF or MA-XRF), we can determine the foil size, which is assessed thanks to the increased thickness derived from the overlapping areas [15,28,29,30].

The aim of this paper is to set up a method for inferring gilded artifact properties through XRF analysis, and fully characterize the preserved gold leaf prototype, determining its caratage and thickness. The approach lays a foundation for future comparisons of the manuscript between 16th-century gilded Genoese artifacts, further verifying if the gold foil used is actually in compliance with the censor rules, in terms of composition and size. To this purpose, gold leaves of different thicknesses and caratages were employed to build a set of matrix standards and calibrate the XRF quantitative analysis (Figure 1c). The foils were produced by a certified artisanal goldbeater and the models were prepared on different substrates. Initially, the gold foil thickness was assessed by means of RBS, and the linear relation between the gold foil thickness and the PIXE response was established. The Au standards were analyzed along with a set of gilded samples of unknown thickness and composition coming from the furniture of Palazzo Spinola di Pellicceria, hereafter called “historic samples”, which can be dated to the middle decades of the 18th century and are used as the validation set (Figure 1d). Consequently, the XRF instrument has been calibrated and the information collected has been used to fully characterize the manuscript, assessing the composition, size, caratage, and thickness of the gold foil.

2. Materials and Methods

2.1. Gold Leaf Reference Sets

The gold leaf was handcrafted by a Venetian goldbeater who maintained the steps of production used in ancient goldbeating. Leaves were mounted by a professional restorer on wood cubes either using a mordant or a guazzo technique. The former refers to a gilding technique in which the gold leaf adheres directly to the surface using an oil–resin adhesive, to which drying pigments may be added, while the latter involves the application of the gold leaf over a layer of dampened bole on a surface that has been previously prepared with gypsum. The term bole, or bolus, describes a clay material malleable enough to allow burnishing of the superimposed gold leaf, conferring a smooth and shining effect to the gilded surface. It is mainly composed of Al, Si, and K, with Fe, which provides the typical red color. Red (or Armenian) bole began to be recommended as a base for gilding around the end of the 14th century and it can be found on paper, parchment, or wood supports [31,32]. Moreover, gold leaves have also been overlapped to achieve larger thicknesses. A summary of the standard samples is given in

Table 1. Gold leaf reference sets. Each sample is identified by a letter and a number as follows: C = 24 K single leaves mounted either with mordant (1–4) or guazzo (5–8) technique; M = 24 K multiple overlapping leaves on guazzo; K = single leaves of different caratages mounted on guazzo. Leaves are differentiated by thickness according to the information provided by the manufacturer.

Dataset	Caratage (K)	Substrate	Sample Name	Thickness (g) 1	Nominal Thickness (nm)
Thickness calibration	24 K	mordant	C1	20	160
			C2	22	180
			C3	32	260
			C4	40	320
	24 K	guazzo	C5	20	160
			C6	22	180
			C7	32	260
			C8	40	320
			M1	20 + 22	340
			M2	22 + 22	360
			M3	32 + 40	580
			M4	40 + 40	640
			M5	32 + 40 + 40	900
Caratage calibration	18 K	guazzo	K1	32	260
	22 K		K2	32	260
	23 3/4 K		K3	32	260
	24 K		K4	32	260

¹ Value provided by the goldbeater to distinguish between different thicknesses, refers to the initial weight of the unbeaten speck of gold used to produce 1000 8 × 8 cm² leaves.

.

2.2. X-ray Fluorescence (XRF) Analysis

XRF was performed with a portable energy-dispersive spectrometer ELIO (XG Lab, Milan, Italy; Bruker Optics, Billerica, MA, USA). The XRF employs a 25 mm² Silicon Drift Detector with CUBE technology and a 50 kV–4 kW X-ray tube generator based on an Rh anode. The excitation X-ray beam is collimated to a 1.2 mm spot diameter on the sample surface. The typical energy resolution is below 140 eV FWHM for the Mn Kα line and elements from Na to U can be detected. Elemental 2D mapping is achieved through automatic XY raster scanning with a total travel area of 10 × 10 cm². XRF point measurements were performed at a 40 kV tube voltage and 30 µA tube current for 120 s. Mapping measurements were performed employing a tube voltage of 40 kV and a tube anode current of 30 μA, with an acquisition time of 5 s for each point. Data was processed using ELIO 1.6.0.29 software.

2.3. Particle-Induced X-ray Emission (PIXE) and Rutherford Backscattering Spectrometry (RBS) Analyses

PIXE and RBS analysis were carried out simultaneously using the 3 MV Tandetron electrostatic accelerator (by High Voltage Engineering Europa) at the INFN LABEC ion beam laboratory in Florence, Italy [33]. The protons produced by a 358 Duoplasmatron source were accelerated up to 3 MeV and delivered to the sample via the +45° beamline specifically designed for measurements on cultural heritage objects. The beam was extracted in air through a Si₃N₄ window (energy on target is 2.935 MeV) and the beam size was about 0.5 mm in diameter on the target surface; the measurements were conducted by moving the sample over a 1 cm long line scan in order to average out possible local inhomogeneities in the layer thickness. The instrumental set-up, extensively described elsewhere, includes, among others, detectors for PIXE and RBS, as well as a rotating chopper for the beam charge monitor [34]. The two X-ray detectors for the PIXE technique were a 10 mm² Ketek SDD with a 125 eV energy resolution FWHM (for the light element (Na-Ca) analysis) and a 150 mm² Ketek SDD with a 135 eV energy resolution FWHM, equipped with a 450 μm thick Mylar absorber for mid-heavy elements (Fe-Pb). Both detectors were at 45° with respect to the beam direction. The particle detector for the RBS technique was a 10 × 10 mm² Hamamatsu Si pin diode, placed at a 135° scattering angle, mounted in an aluminum case, and kept at 10−1 mbar pressure. The overall energy resolution, including detector resolution, extracted beam energy straggling, and kinematic straggling, was about 30 keV FWHM. A helium flow was maintained in front of the target in order to reduce the attenuation of low-energy X-rays and the stopping of the extracted proton beam. The measurement time was 300 s and beam currents between 40 and 500 pA were used. Some standards (metallic and not) were also measured to check the data reduction procedures.

3. Results and Discussion

3.1. Measurement of Gold Leaf Thickness

Artistic gold leaves are traditionally manufactured with thicknesses of the order of a few hundred nanometers (e.g., Table 1). In the thin target hypothesis, while analyzing gold leaves, we expect that a gold characteristic X-ray yield (e.g., Lα yield at 9.713 keV), produced by a leaf (either ancient or modern) both in X-ray fluorescence and in PIXE, depends linearly on the thickness D_Au:

$$Y_{XRF}\left(Au{L}_{\alpha }\right)=K\left(Au{L}_{\alpha }\right)·D_{Au}$$

(1)

Details are given in the Supplementary Material. The coefficient K is an instrumental constant that is determined by the experimental setup and by the specific X-ray line observed. It can be computed through calibration with gold standards of known thickness (in the range expected for leaves), in fixed experimental conditions. The factor can be used in the same instrumental conditions to determine the thickness of unknown samples from the X-ray yield. This is the experimental approach we employed—using the standards presented in Table 1 and the unknown samples of this study (Figure 1b,d). First, the thickness of the reference standard leaves was measured by RBS. The sample thickness can be calculated either using a simple equation in the thin target hypothesis or by a simulation of the RBS spectra. In the thin target hypothesis, the proton energy loss in the leaf is negligible, thus the backscattering cross-section can be assumed constant and the gold leaf thickness D_Au is given by the following equation:

$$D_{Au}\left[\mathrm{nm}\right]=\left(\frac{Y·e\left[\mathrm{C}\right]·A\left[\mathrm{g}\right]}{Q\left[\mathrm{C}\right]·N_0·\rho \left[{\mathrm{gcm}}^{-3}\left]·d\Omega \left[\mathrm{sr}\right]·\frac{d\sigma }{d\mathsf{\Omega }}\right[{\mathrm{cm}}^2{\mathrm{sr}}^{-1}\right]} \right)\times {10}^7$$

(2)

where Y is the yield of the protons backscattered on Au, e is the elementary charge, A is the atomic weight, Q is the collected beam charge, N₀ is Avogadro’s number, dΩ is the detector solid angle, and dσ/dΩ is the Rutherford elastic scattering cross-section. A more accurate result is obtained by the simulation of RBS spectra with the SIMNRA code [35] in which experimental parameters and microphysical data (cross-sections and stopping powers) are introduced, and the sample thickness is adjusted to obtain the best fit to the shape of the backscattered proton peak. An example of an acquired RBS spectrum and its SIMNRA simulation is given in Figure 2.

The PIXE spectra acquired on reference standards have been analyzed by the GUPIX code [36]. Among the various options that the code offers to describe the sample, gold has been considered as a thin surface layer sitting in front of a bulk containing elements expected in bole or from the mordant. In this configuration, by introducing the beam and instrumental parameters, the code iteratively adjusts the gold layer’s thickness, to finally obtain—by considering absorption in the gold and the bulk—the correct relative height of the peaks in each group of characteristic X-rays (K, L, M). It calculates the bulk concentration from the fundamental parameters. From the PIXE analysis, we extracted the gold thickness, the 2-FWHM area of the gold Lα peak, and the concentration of the bulk (substrate) elements. The results of the investigation of reference samples are given in

Table 2. Results from the reference sample investigation.

Sample	RBS				PIXE
	Thickness from Au RBS Peak Area		Thickness from SIMNRA Fit		Thickness from Au Lα Peak Area	Thickness from GUPIX Fit
	DAu (nm)	Err [%]	DAu (nm)	Err [%]	DAu (nm) *	DAu (nm)	Err [%]
C1	304.9	4.51	299.5	0.02	332.1	305.1	1.1
C2	217.4	2.96	219.7	0.76	237.5	225.7	1.98
C3	143.3	2.41	140	1.09	154.7	142.2	2.73
C4	174.7	2.91	169.4	1.07	190.7	174.5	3.25
C5	260.9	6.46	254.4	0.95	271.0	242.6	0.81
C6	217.3	1.74	219.1	0.1	227.9	202.9	0.94
C7	154.7	0.04	154.1	0.55	164.4	147.5	1.17
C8	158.2	3.16	161.9	0.74	170.3	153.6	1.08
M1	322.9	7.97	311.9	1.47	343.7	304.1	0.58
M2	503.7	8.38	499.7	0.67	493.3	438.2	0.82
M3	540.9	4.47	534.7	0.35	527.1	475.5	0.92
M4	590.9	4.8	578.9	0.28	585.9	525.3	0.74
M5	769.6	5.47	789.7	1.99	724.9	657.2	0.46

* Error on the coefficient derived from the linear regression fit in Figure 4 = 1.85% (1σ).

. As shown in Figure 3, the thicknesses calculated from Equation (2) and SIMNRA differ by less than 1% on average while those calculated in the GUPIX fitting of PIXE spectra are, on average, almost 10% lower. Using the thicknesses calculated from the RBS data, we demonstrate (Figure 4) the linearity existing between the yield of the gold Lα peak and the leaf thickness, as from Equation (1). It is worth repeating that the coefficient of linearity that we have deduced is only valid for the specific experimental setup since the K factor in the equation depends on the beam and detector parameters. However, the calibration procedure can be repeated with the same set of reference samples in any other set-up since their thicknesses are now assessed through the analysis of the RBS data. For double-checking the procedure, we calculated the reference leaf thicknesses starting from the Lα line and using the calibration coefficient 0.0058, obtained from Figure 4. In Figure 5, the PIXE-computed thicknesses are plotted as functions of those calculated from the RBS peak area (Equation (2)). In this case, there is very good linearity between the two sets of data; on average, their ratio is practically 1. The slight deviations observed for thicknesses below and over 400 nm can be due to the thin target approximation assumed. This approximation for PIXE, including both variations in X-ray production energy loss due to finite proton energy loss and self-absorption of the emitted X-ray, can be considered valid within 5% for thicknesses up to 500 nm and within 10% for thicknesses around 800 nm. As discussed in the Supplementary Materials, for XRF, we expect a validity of the thin target approximation around 7% once thicknesses exceed 600 nm.

3.2. Calibration of the ELIO Setup and XRF Analysis of Historic Samples

The calibration procedure was repeated using the XRF ELIO setup. Figure 6 shows the linear relation between the RBS thickness and the XRF Au Lα area net yield. Once determined, the linearity coefficient and the XRF yield were transformed into thicknesses for the historic samples from Palazzo Spinola di Pellicceria, and analyzed in the same instrumental conditions as the standards. Even though these gildings are not all coeval, they represent suitable empirical validation sets because they show a variety of thicknesses, compositions, caratages, and shapes. They also differ greatly in the state of conservation. After careful inspection, some areas of gilding appear quite damaged, while others are well preserved. A minimum of five XRF point measurements have been taken for each historic sample.

Table 3. Results of the thickness calculation from XRF data collected for the historic samples from Palazzo Spinola di Pellicceria.

Historic Samples	XRF Au Lα Area Net Count			XRF Thickness (nm)
	Min	Average	Max	Min	Average	St. Dev	Max
S1	4910	8815	12,501	124	222	78	315
S2	5826	8429	14,395	146	212	70	362
S3	1071	5635	11,534	26	142	98	290
S4	2147	4280	5689	54	107	34	143
S5	5498	9345	14,074	138	235	68	354
S6	5611	8461	17,043	141	213	94	429
S7	4373	8284	14,979	110	208	108	377
S8	6091	10,285	19,430	153	259	129	489
S9	3797	5293	8060	95	133	43	203
S10	5535	10,458	16,443	139	263	103	414

shows the Au Lα area net counts of the eleven historic samples with the associated calculated thicknesses. Minimum, maximum, and average values are reported. Assuming that the highest net count corresponds to a pristine gold leaf, this value should be indicative of the original thickness of the employed foils, while the minimum net area count might be linked to a damaged leaf. The original thickness spans from around 143 nm for S5 to around 489 nm for sample S9. Additional XRF point analyses were performed in the same points as the RBS/PIXE measurements and the relative thickness values were compared. The XRF results agree with RBS/PIXE results and the techniques provide compatible data.

Moreover, to test the capability of the instrument to provide useful thickness information in the mapping mode, a 10 × 10 cm² sample made from overlapping foils of different thicknesses applied in various combinations was analyzed. This is useful because, during the gilding process, a portion of the gold leaf overlaps with the previously applied one (sormonto); with the XRF mapping function, it is possible to visualize these overlapped zones and assess the original sizes of the leaves. In Figure 7, the XRF map is shown. By extracting values from a specific pixel column, a thickness profile can be generated to assess the gilding pattern. From a compositional point of view, both PIXE and XRF analyses have revealed the presence of Al, Si, S, K, Ca, Ti, and Fe, demonstrating the constant use of the guazzo technique in the artistic samples of Palazzo Spinola. The guazzo technique, in fact, involves applying gold foil over a mirror-polished gypsum (CaSO₄) ground, which is covered by a thin bole layer—a red ochre composed of an aluminosilicate fraction, plus Fe and Ti [31]. Furthermore, the presence of Sr has been associated with the natural provenance of gypsum, while impurities of V, Cr, and Ni are commonly associated with ochres [37].

3.3. Characterization of the Manuscript

The trade agreement governing the sale of gold leaf in the city of Genoa was signed during three different periods: 1520, 1550, and 1641. The original copy of the stipulations is preserved within a manuscript drafted annually by the magistracy of the censors in order to regulate what was produced in the city. The peculiarity of the first agreement (1520) is that a prototype of gold leaf was included next to the text, to prove what finished product had to be used and traded. To fulfill the covenant, we find the commitment of the goldbeater and the painter on the sides of the prototype leaves. The signature of the painter was presumably affixed with a carbonaceous ink or a lacquer (no XRF signals), while the goldbeater engraved the inscription “hec est forma auri” (“this is the shape of gold”). The size of the leaf is confirmed to be 9 × 9 cm², as is reported in other documents. XRF point analyses were performed on the manuscript and three main areas of interest were identified: parchment, ink, and gold leaf (Figure 8). The page composition slightly differs between the central area (where the words were meant to be written) and the borders. The hypothesis is that the page was prepared with a Ca-based primer in written parts [38,39,40]. An iron-based ink, presumably an iron gall ink, was uniformly used. Analysis of the gold leaf shows the presence of Cu as an alloying element. In combination with the gilded area, Ca, Fe, and Pb are also detected. Their presence may suggest the use of a mordant prepared with a Pb-based pigment or adhesive, along with the addition of bole as a tinting element [41,42,43,44]. The mapping function of the XRF allowed us to verify how plausible the assumptions were so far about the stratigraphy. It can be observed that Au-Cu-Pb has the same distribution, while Ca results are complementary. This, coupled with the substantial presence of Ca in the written areas, suggests that it was used only as a page primer and not as the gold foil substrate. Moreover, the XRF mapping enabled the assessment of the state of conservation of the leaf from the perspective of abrasion damage. The maximum thickness, calculated in the best-preserved areas, is equal to 493 nm in the center and 767 nm on the border, i.e., the less worn portion. To determine the caratage of the unknown gold foil, we used a reference set, including 24 K, 23 ¾K, 22 K, and 18 K gold leaves, i.e., gold alloys with 99.9%, 99.1%, 91.7%, and 79.2% gold content, with the rest being copper. The procedure consisted of relating the known caratage to the net Cu Kα to Au Lα ratio and using the curve as a calibration for unknown samples, similar to what was done for thickness assessment. In Figure 9, the calibration curve is presented. Assuming a carat value of 19.5 K, the corresponding Cu/Au ratios are equivalent to the experimental data measured on the manuscript leaf, with an error of 0.006.

4. Conclusions

The Genoese gilding context is unique due to the discovery of a prototype gold leaf that is supposed to have been used as a reference for all the gildings made in the city, starting from the 1520s. To gain insight into the formulations used by goldbeaters, who typically did not reveal (in writing) their methods, and into the actual guild rules, understanding the exact features of this foil is very important. This study offers a comprehensive characterization of the gilding technique and materials employed in the prototype leaf. Through the integration of X-ray fluorescence analysis (XRF) and ion beam analysis (IBA) techniques, such as particle-induced X-ray emission (PIXE) and Rutherford backscattering spectrometry (RBS), various aspects of gilding, including thickness, caratage, and composition, were analyzed. After calibration against laboratory-based IBA techniques, XRF proved to be a valid method for measuring in situ the thickness of gilding, as attested by the results obtained in measurements on historical samples from Palazzo Spinola di Pellicceria, which exhibited diverse thicknesses, compositions, and states of conservation. Furthermore, the examination of caratage using reference sets of known compositions allowed for accurate assessments of gold purity, shedding light on the adherence to guild regulations governing the use of aurum bonum. The manuscript leaf was found to have an Au-Cu composition, a maximum thickness of 767 nm, and a caratage of 19.5 K. This latter value is significant as it deviates considerably from the expected standard of aurum bonum.

In conclusion, this manuscript provides unique insights into historical trade agreements and gilding practices in Genoa, serving as a solid basis for future comparative analyses of Genoese artistic gilding. Through this comparison, it will be possible to understand if painters and gilders adhered, as promised, to the norms in force in the city regarding leaf carat and thickness. One might wonder whether Genoese production is distinguished by a low carat, or paradoxically, whether the leaf used as a model was of much lower quality than the gilding carried out daily in the workshops of Genoese masters.