Insight into the structure of black coatings of ancient Egyptian mummies by advanced EPR spectroscopy of vanadyl complexes.

: Advanced EPR techniques such as ENDOR and pulsed EPR are used to investigate the enigmatic black coatings of ancient Egyptian mummies, consisting in a complex and heterogeneous mixtures of conifer resins, wax, fat and oil with variable amounts of bitumen. Natural bitumen always contains traces of vanadyl porphyrin complexes that we used here as internal probes to explore the nanoscale environment of V 4+ ions in these black coatings by hyperfine spectroscopy. Four types of vanadyl porphyrins were identified from the analysis of 14 N hyperfine interactions. Three types (referred to as VO-P1, VO-P2 and VO-P3) are present in natural bitumen from the Dead Sea, among which VO-P1 and VO-P2 are also present in black coatings of mummies. The absence of VO-P3 in mummies, which is replaced by another complex VO-P4, may be due to its transformation during preparation of the black matter for embalming. Analysis of 1 H hyperfine interaction shows that bitumen and other natural substances are intimately mixed in these black coatings, with bitumen aggregate sizes not larger than a few nanometres.

Mummies and wooden coffins, funerary artifacts and panel paintings in ancient Egypt were often covered with enigmatic organic black materials, made of a heterogeneous mixture of natural substances such as fat, oil, wax, conifer or mastic tree resin, pitch, animal glue, plant gum and bitumen) in variable proportions. [1][2][3][4] Those are characterized by a variety of molecular biomarkers identified mainly by gas chromatography -mass spectrometry (GC-MS). However, the presence or not of bitumen in these black materials has been the subject of a long controversy due to the fact that the analytical protocols used were often not well-adapted to the detection of bitumen, so that two opposite opinions have emerged among researchers analyzing these black coatings: those who did not believe in the presence of bitumen, [5][6][7] and those who claimed its presence. 1,[8][9][10][11][12] With the detection of specific biomarkers (hopanes, steranes) and radiocarbon analyses (bitumen has lost its 14 C), a consensus has recently emerged on the increasing presence of bitumen in embalming materials from the New Kingdom (ca 1550-1070 BC) to the Ptolemaïc/Roman period ending in the 4th century AD. 3 Despite the inestimable contribution of GC-MS for revealing the composition of these black coatings, this microdestructive technique requires preliminary steps of fractionation and separation, which exclude any direct, non-destructive identification of bitumen, so that structural information on this black material cannot be obtained at the nanometer scale. Such kind of information can only be obtained with non-destructive analysis techniques, i.e. which leave the samples intact. This is the case with magnetic resonance techniques because the low frequency electromagnetic fields (radiofrequency for NMR and microwave frequency for EPR) penetrate the whole sample and deposit a totally negligible energy in the material compared to the other spectroscopic techniques. Multinuclear magnetic resonance ( 1 H-and 23 Na-NMR) of mummified tissues is mainly used in the imaging mode 13,14 rather than the spectroscopic mode, 15 owing to the rather low sensitivity and spectral resolution of NMR for these highly disordered solid materials.
Electron paramagnetic resonance (EPR) is the electronic equivalent of NMR, and applies in the presence of unpaired electron spin density, i.e with electron spin S ≥ 1/2. The spectroscopic resolution of EPR is optimal when the paramagnetic entities (transition metal ions, radicals, …) are magnetically diluted, which correspond to defects and impurities in material. 16 It is well known that oil and bitumen contain organic radicals and porphyrinic complexes of vanadyl VO 2+ ions (V 4+ ion, 3d 1 configuration). These very stable paramagnetic complexes are present mainly in asphaltene -the most refractory fraction of oil and bitumen -and can be considered as molecular markers of bitumen which can be detected with high sensitivity by EPR spectroscopy. [17][18][19][20] Generally speaking, vanadyl porphyrins complexes (hereafter referred to as VO-P) are specific of oils and bitumen of marine origin, [21][22][23] while carbonaceous radicals (hereafter referred to as C 0 ) are present in all fossilized organic matters, whether of marine or terrestrial origin, [24][25][26] and even in the extraterrestrial organic matters of the most primitive carbonaceous meteorites. 27 Recently, we showed that EPR analysis of VO-P complexes and C 0 radicals is a simple and nondestructive way (no sample preparation) to reveal the presence of bitumen in black coatings of Egyptian mummies, even in small amount. 28 In addition to VO-P, the latter contain also non-porphyrinic VO 2+ complexes (hereafter referred to as VO-nP), with four oxygen ligands in nearly square planar configuration. 28 These VO-nP complexes are absent in Dead Sea bitumen used by Egyptians, and present in black coatings containing natural substances in addition to bitumen. We hypothesized that these VO-nP could result from the demetallation of VO-P of bitumen followed by the complexation of VO 2+ by oxygenated functions of other components of the black matters.
However the resolution of EPR spectra of VO-P in such disordered materials is limited by the fact that the weak hyperfine (hf) interactions with other nuclei, namely 1 H (I=1/2, 100% abundance), 14 N (I = 1, 99.6 % abundance) and 13 C (I = 1/2, 1.1 % natural abundance) are unresolved. These hindered hf interactions contain precious information about the structure of VO-P complexes, their environment and possibly on the degree of alteration of the black matters. This information can be recovered by the indirect detection of NMR transitions of magnetic nuclei in the environment of the unpaired electron spin. By this means, VO 2+ of bitumen can be considered as internal probes which "see" their nuclear spin environment in a non-destructive manner.
Here we used Electron Nuclear Double Resonance spectroscopy in continuous-wave mode (cw-ENDOR) for 1 H nuclei, and HYperfine Sublevel CORrelation (HYSCORE) spectroscopy for 14 N nuclei to study the same corpus of Egyptian black coatings than in the preliminary cw-EPR study. 28 We found that proton ENDOR is sensitive to the bitumen content of the black matters and to the size of bitumen aggregates in the mixture, while HYSCORE of nitrogen reveals the presence of different types of VO-P complexes in the material. Special attention was given to the black coating of a human mummy of unknown origin (Hum 3), but whose EPR characteristics differ clearly from coatings of the other studied mummies. 28

Results and discussion
The origins of black coatings are summarized in Table 1, and samples studied by EPR are described in more details in Table S1 and Fig. S1. Three samples were taken from the coating of an anthropomorphic coffin (Hum 1) dated from the Ptolemaïc period, and two human mummies (Hum 2 and Hum 3) dated from the third Intermediate period. Four samples were taken from animal mummies dated from the same periods as human mummies, three rams (An

1, An 2, An 3) and one crocodile (An 4).
Their spectra were compared with those of two pure bitumen samples: a fragment of natural asphalt from the Dead Sea (Ref 1), and a commercial powder of bitumen of Judea (Ref 2).

Figure 1. Origin of some samples of black matters studied by EPR; Hum 1)
Coffin of Irethorerou (Ptolemaïc dynasty), The Art and History Museum, Narbonne, France, the black matter was sampled at the bottom of the coffin (arrow), reproduced with permission of C2RMF/Anne Chauvet; Hum 2) Human mummy of the Late Period, The Hieron museum, Paray-le-Monial, France, the sample is a fragment of black matter covering the mummy, reproduced with permission of C2RMF/Hélène Guichard; Hum 3) Head of the mummy from the Late period, Chateau-musée, Boulogne, France, the sample was taken from the mummy's neck, reproduced with permission from Frédérique Vincent; An 2/An 3) Ram mummy of the Late Period, The Thomas Dobrée museum, Nantes, France, An2 and An3 samples are a fragment of black matter and a fragment of tissue with brown matter, respectively, reproduced with permission of C2RMF.  . We concluded that mummy Hum 3 was covered with pure bitumen, which was confirmed by GC-MS analysis. 28 In addition, the spectrum of Hum 3, An 1 (Fig. S3) and, to a lesser extent An 2, show a broad baseline distortion due to a ferromagnetic resonance (FMR) signal of iron oxide microparticles. 28 This FMR signal, which does not give electron spin echo, can be eliminated by recording the echo-detected EPR spectrum (ED-EPR), as clearly shown for An 2 and Hum cw-EPR spectra (in red) and pseudo-modulated ED-EPR spectra (in black) at X-band; the blue area represents the position of the strong C 0 signal, that has been suppressed for the sake of clarity; b) cw-EPR spectra at Q-band; some EPR lines of VO-nP complexes in An 2 are represented by green circles. The magnetic field settings for ENDOR and HYSCORE experiments correspond to 51 V hyperfine lines marked by arrows.
The EPR parameters of VO-P complexes were deduced from the fitting of spectra at both X and Q-bands ( Table 2). The slight differences of EPR parameters between the three samples fall within error bars of the simulations, except for parameter AꞱ, for which the differences are clearly visible on the spectra at Q-band ( Fig. 2b). showing the electron spin S on vanadium and two nuclear spins I on nitrogen and on the bridging hydrogen Hmeso; b) EPR with a non-saturating microwave field, the dashed circle representing the limited resolution of EPR, which does not reach neighboring ligand nuclei; c) cw-ENDOR: a saturating microwave field modifies the populations of the nuclear spin state (step 1), while a strong rf field at nuclear frequency restores the nuclear populations, which in turn desaturates the EPR transition (step 2); d) HYSCORE spectroscopy: a sequence of two /2 microwave pulses separated by time  induces a nuclear coherence in each electron spin state ms (step 1), a  pulse after time t1 produces a transfer of nuclear coherence between the two electron spin states ms (step 2); after an evolution time t2, the nuclear coherences are transferred back to the electron coherence by a /2 pulse, giving an electron spin echo at time  (step 3).
The shape of EPR spectra of VO-P complexes (Fig.2) is entirely controlled by the anisotropy of the g-factor and of the strong hf interaction with the central 51 V nucleus, which reflect the electronic structure and the geometry of the complex. For this reason, the weak unresolved hf interactions with 1 H, 14 N and 13 C nuclei of the porphyrin ligand can only be revealed by ENDOR and HYSCORE spectroscopy.

H-ENDOR analysis.
In a cw-ENDOR experiment, a specific EPR transition is partially saturated at high microwave power and at fixed magnetic field, which modify the population of the nuclear spin states. A saturating radiofrequency (rf) field of frequency  is then swept through the NMR frequencies of 1 H nuclei. The populations of nuclear spin states are modified at each nuclear resonance frequency, which are detected by a change in the EPR intensity  It must be noticed that additional ENDOR features that would be expected at A// = 1.3 MHz and AꞱ = 0.4 MHz for hydrogen of pyrrole groups, 34,35 are not observed in our ENDOR spectra.
This indicates that all pyrrolic hydrogen atoms are substituted with alkyl groups, as known in the case of vanadyl geoporphyrins in oil and bitumen, such as vanadyl etioporphyrin (VO-EP) and vanadyl deoxophylloerythroetioporphyrin (VO-DPEP) for example (Fig.S6). 36 The narrow signal centered at H, referred to as the "matrix" line, 37 Fig.4c. However, since this Judea bitumen is commercial, it may have undergone some undocumented treatment that could have modified some of its characteristics.
It is important to note that a matrix ENDOR line can occur only if the vanadium-proton distances R are sufficiently small to give non-zero electron-proton dipolar interactions (i.e R < 5-6 nm). 37, 38 We may conclude that only protons at distance R > 0.6-0.7 nm from the vanadium atom (~ half the size of the porphyrin ligand) and R < 5-6 nm (limit for non-zero dipolar interaction) can contribute to the 1 H matrix line. This limited distance range for matrix protons has two consequences: (i) for a given proportion of bitumen and natural substances, the amplitude Y of the matrix line should increase (X/Y decrease) upon decreasing the radius RA of bitumen aggregates in the black matters (see Fig.S7 for a schematic representation), and reach a maximum amplitude for RA < 5-6 nm (i.e. all VO-P complexes "see" protons of natural substances); (ii) for bitumen aggregates with mean RA value, X/Y should decrease upon increasing content of natural substances, as more and more bioorganic hydrogen atoms are present in the vicinity of VO-P complexes. According to (i), X/Y should be nearly independent of VO-P content of the black matter if RA lies in the micrometer range or larger, because in this case, only the small fraction of VO-P close to the surface of bitumen aggregates "see" the bioorganic protons. According to (i) and (ii), the fact that the decrease of X/Y with decreasing VO-P content is regular (Fig. 4c) for our corpus of black coatings (if we except the commercial bitumen Ref 2) suggests that RA < 6-7 nm in all cases, and that X/Y depends only on the ratio bitumen/natural substances of the black matters.
A lower limit of the size RA of bitumen aggregates can also be estimated from the fact that VO-P and radicals C 0 are spatially connected in asphaltene, with (VO-P)-C 0 distances not larger than 1-3 nm. 39 We previously showed that such spatial connection is conserved in bitumen of Egyptian black coatings. 28 Consequently, we may roughly estimate that the sizes of bitumen aggregates in the studied black coatings lie in the range ~1 nm < RA < 6-7 nm. Three scales of asphaltene aggregation were proposed in the Yen-Mullins model of asphaltene hierarchical structure: 40 the molecular (~1.5 nm), the nanocluster (~2.0 nm) and the Cluster (~5.0 nm) scales.
It appears that the sizes of bitumen aggregates in black coatings correspond to the cluster scale of the Yen-Mullins model. 40 Whatever the actual distribution size of asphaltene aggregates be, this ENDOR analysis shows that the bitumen and natural substances are intimately mixed in black coatings. For such small sizes of bitumen aggregates, where the surface/volume ratio is high, this could also explain why a significant fraction of VO-P are transformed into oxygenated VO-nP complexes at interfaces between bitumen aggregates and natural substances. 28 The regular variation of the ENDOR shape factor X/Y with VO-P content (Fig.4c) can be reproduced with a simple model considering that X and Y amplitudes are both sums of contributions from protons of VO-P complexes and of the matrix layer around the bitumen aggregates, respectively, so that the ratio X/Y is given by (see SI for demonstration): Experimental data were nicely fitted to Eq.1 with b = 15 (Fig.4c). This good agreement shows that a is a constant almost independent of sample. As developed in SI, this means that the larger the concentration of bitumen in the black matters, the larger the mean size of bitumen aggregates. Consequently, this result shows that a simple measurement of the amplitude of ENDOR lines can give direct information on the bitumen content of a black coating.
14 N-HYSCORE analysis. The very small variations of EPR parameters g and A of VO-P from one sample to another ( Table 2) may suggest that several types of slightly different VO-P complexes are present in variable proportions in the bitumen component of black coatings. We used HYSCORE spectroscopy at X-band to discriminate different types of VO-P by their 14 N hf interaction, with the perspective to use in the future these metallic complexes for getting information on the geographical origin of the bitumen and on its chemical or thermal treatment during the preparation of the mummy. In pulsed EPR spectroscopy, 29 a spin echo is generated by a series of /2 and  microwave pulses ( and /2 represent the rotation angles of the electron spin magnetization) separated by controlled time delays (Fig.3d). By varying these time delays, the echo intensity is modulated at frequencies of the hf interactions. HYSCORE spectroscopy is based on the pulse sequence /2--/2-t1--t2-/2--echo, where  is the delay between the first and second /2 pulses. The first /2 pulse generates an electronic coherence (a mixing of the two ms = ± 1/2 states), and the second /2 pulse after time  transfers the electronic coherence to nuclear coherences (mixing of mI states). After an evolution time t1, a  pulse transfers the nuclear coherence from one ms state to the nuclear coherences of the other ms state, which creates correlations between nuclear transitions of these two ms states. After another evolution time t2, a third /2 pulse transfers the nuclear coherence back to the electronic coherence for detection, which generates an electron spin echo after time . The echo intensity is measured for the two times t1 and t2, which are varied stepwise at constant  value. The 2Dfrequency plot (HYSCORE) is obtained by 2D-Fourier transformation of the data set in time domain.
The correlations between nuclear transitions in the two ms = ±1/2 states appear as cross peaks in the 2D frequency plot, which are distributed in two different quadrants (+,+) and (+,-) corresponding to ω2  0, ω1  0 and ω2  0, ω1  0, respectively. 29 For an electron spin S=1/2 interacting with an I = 1/2 nuclear spin such as 13 C and 1 H, we expect two cross peaks in each quadrant, which take the shape of ridges perpendicular to the diagonal ω1 = ω2 for anisotropic hf interactions in disordered materials. 29 The spectrum is more complicated for a nuclear spin I = 1 such as 14 N, which can give up to 18 cross peaks and ridges in each quadrant. 41 The situation is even more puzzling if the electron spin is coupled with several nuclear spins (which indeed is the case of VO-P) because such multi-spin systems can give additional zero-and multi-quanta coherences, as well as suppression effects. 38 Fortunately, many of these spectral features are too weak to appear in the 2D-plot, so that the HYSCORE spectra remain interpretable. [42][43][44] Representative HYSCORE spectra for samples The energy level diagram in Fig.6 describes the spin states and the corresponding nuclear transitions for an S = 1/2, I = 1 system. The frequencies for the single quantum (mI = 1) and double quantum (mI = 2) transitions, referred to as sq and dq transitions, respectively, are given by: 42 where N is the nuclear Zeeman frequency, and A and Q the hf interaction and the quadrupolar interaction, respectively, of 14   , all sq and dq frequencies of the spin system in Fig.5 can be simply deduced. The results for VO-P1 and VO-P2 are shown in Fig.S9. Neglecting again 2 nd order terms, the quadrupolar parameter Q can be estimated from Eqs.2 by: . The values for VO-P1 and VO-P2 are Q ≈ 0.47 ± 0.03 MHz and Q ≈ 0.55 ± 0.02 MHz, respectively (see SI for the estimation of the second order terms).
The 14 N-hf interaction is anisotropic, with two components A// and AꞱ, corresponding to B0 parallel and perpendicular to the V-N bond. For HYSCORE spectra recorded from the mI = +3/2Ʇ field setting, which span all orientations of B0 in the porphyrin plane, the measured value of A is the average  Table 3. All this interpretation was based on the analysis of only a small portion of each HYSCORE spectrum (rectangular boxes in Fig.5). This procedure raises the question of the interpretation of all other correlation peaks and ridges present in the HYSCORE spectra (Fig. 5). To test the validity of the proposed analysis, the whole 14 Table 3 calls for several comments: (ii) All the other correlation peaks and ridges visible in the (+,-) quadrant are clearly due to the same porphyrinic nitrogen atoms that give rise to the observed dq-dq peaks. Thus it is not necessary to invoke hf interactions with other nuclei ( 14 N, 13 C or other).
(iii) Unfortunately the (+,+) quadrant does not give any information on the 13 C hf interaction because the corresponding peaks are hindered under 14 N correlations which come out in the same frequency range as 13 C (C = 3.7 MHz), as shown by the simulation in Fig. 5. Also, in a multi-spin system such as VO-P complexes, nuclei with weak modulations (such as 13  century. This beautiful mummy is covered with a solid, black and shiny substance, and has therefore an unknown origin (Fig.1). Cw-EPR and GC-MS analysis showed that this black coating is made of pure bitumen. 28 Contrary to other black coatings studied in this work (animal and human mummies, coffin), Hum 3 contains no VO-nP (non-porphyrinic vanadyl complexes), and its EPR spectrum is very similar to that of pure bitumen (Ref 1 and Ref 2). 28 This similarity with native bitumen is confirmed by the almost identical ENDOR spectra of are characterized by the presence of VO-P3 complex. As the preparation of the embalming coating by ancient Egyptian implies that bitumen was heated to the liquid state in order to be mixed with the other ingredients and spread on the mummy, we may hypothesize that VO-P4 originates from the thermal transformation of an unstable complex VO-P3 initially present in bitumen. Laboratory experiments will be necessary to test this hypothesis.
In summary, vanadyl porphyrin (VO-P) complexes commonly found at trace level in bitumen can be used as intrinsic paramagnetic probes for a non-destructive analysis of the black matter covering ancient Egyptian mummies. Four types of VO-P complexes were identified in HYSCORE spectra by the double-quantum (dq-dq) correlation peaks of 14  with hyperfine spectroscopy (ENDOR, HYSCORE) is a promising tool for a non-destructive exploration of the nanostructure and composition of black coatings of ancient Egyptian mummies and funerary artifacts. The same strategy could be applied to other historical collections such as Islamic metal works for which this black matter could have been widely used as adhesive for metal carving. 47 EPR tools could also permit to better apprehend local economies, workshop practices and recipes, supply areas as well as trade routes of bituminous materials in the past.

Methods
All samples (10-20 mg) were inserted into quartz Suprasil EPR tubes. They are shown in Fig.S1.
Continuous wave Electron Paramagnetic Resonance (cw-EPR) measurements were performed at room temperature and at 100 K with a Bruker Elexsys E500 EPR-ENDOR spectrometer operating at about 9.6 GHz (X-band) and 34 GHz (Q-band), equipped with a high sensitivity X-band 4122SHQE/0111 EPR cavity and a Q-band ER5106QTE resonator for both EPR and ENDOR. Cw-ENDOR at Q-band was used to measure the hyperfine interaction with 1 H nuclei of porphyrin ligands and their molecular environment (see Fig. 2a). The ENDOR spectra were recorded at 100K by using a CF935 helium flow cryostat from Oxford Instruments. The radiowaves were amplified by an ENI3100L amplifier, and the ENDOR signals were detected by a 25 kHz frequency modulation of the sweeping rf field, with a modulation depth of 100 kHz.
The rf was swept in the range 45-60 MHz, centered at the proton Larmor frequency.
EPR spectra at X-band were also recorded by 2-pulse echo field sweep experiment, using the standard Hahn echo sequence /2----echo. The resulting echo-detected absorption EPR spectrum (ED-EPR) was pseudo-modulated to give a first derivative ED-EPR spectrum similar to the cw-EPR spectrum.
HYSCORE experiments were performed at 6K with the pulse sequence /2--/2-t1--t2-/2--echo, with pulse lengths of 22 ns and 44 ns for /2 and  pulses, respectively, and the delay  = 200 ns was chosen as an optimum to prevent blind spot effects. 20 The spectra were recorded with 256×256 data points for t1 and t2 time domains. The unmodulated part of the echo was removed by second-order polynomial subtraction. Final HYSCORE spectra were obtained by 2D-Fourier transformation of the data set, using a Hamming apodization window function.
EPR and HYSCORE spectra were simulated with the EasySpin toolbox for Matlab (version

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.