Research article

Received: 2 November 2012 Revised: 23 May 2013 Accepted: 31 May 2013 Published online in Wiley Online Library: 9 July 2013

(wileyonlinelibrary.com) DOI 10.1002/xrs.2504

Application of a portable XRF analyzer to

investigate the medieval wall paintings of
Yemrehanna Krestos Church, Ethiopia
Kidane Fanta Gebremariam,* Lise Kvittingen and Florinel-Gabriel Banica
Despite the large body of ancient Ethiopian works of art in the form of murals, icons, and illuminated manuscripts, the phys-
icochemical examinations carried out on them are few. This study is an in situ investigation of the wall paintings in the early
12th century Yemrehanna Krestos Church, Ethiopia. Fast, onsite, nondestructive analysis of the painting materials was carried
out using a portable X-ray energy fluorescence dispersive spectrometer. It is believed to be the first onsite technical examina-
tion of Ethiopian mural paintings. This work resulted in information about the painting materials and the existence of differ-
ent painting programs. The analysis revealed that the main pigments were red and yellow ochre, minium, cinnabar, orpiment,
gypsum, lead, and white and carbon black; those typically employed in medieval times with no indication of later conserva-
tion–restoration intervention. Correlation between concentration of elements and multivariate statistical analysis was used
to identify the most probable compounds and to classify sets of pigments used in two painting programs. The portable
X-ray energy dispersive fluorescence spectrometer analyzer is found valuable to guide the in situ analysis and assess potential
sites for microsampling for further investigations with complementary analytical techniques. However, in this expedition, we
were unable to collect sufficient microsamples to warrant adequate complementary analyses. Characterization and documen-
tation of the materials of the church murals support art historical studies and eventually conservation intervention plans.
Copyright © 2013 John Wiley & Sons, Ltd.

Introduction chemical perspective is scant. No onsite investigations, as far as

This is an in situ investigation using a portable X-ray energy dis-
persive fluorescence spectrometer (pXRF) to study paintings be-
longing to the 12th century Yemrehanna Krestos Church, which
was built in a cave at an altitude of 2650 m on the west side of
Mount Abuna Yosef, 12 km slightly northeast of Lalibela, a United
Nations Educational, Scientific and Cultural Organization cultural
heritage site. The church is one of the oldest surviving examples
of Christian architecture in medieval Ethiopia decorated with
paintings. It is constructed in the form of a basilica with three
aisles, a cupola, and a tall saddle roof. The church's architecture
is derived from an earlier Aksumite building tradition of alternat-
ing layers of wood and stone.[1] It was, and still is, an important
site of pilgrimage used by the local people. The church was built
by the priest–king of the Zagwe Dynasty who ruled at the end of
the 12th century. Several of the Zagwe rulers were great church
builders with the tradition claimed to have begun during the
reign of Yemrehanna Krestos.[2–4] There are both figural and geo-
metrical decorations in the church. These are located on walls,
arches, capitals, columns, and ceilings. Part of the ceiling is also
carved into medallions and painted. The wall paintings represent
various religious themes.[1] The paintings are suffering from con-
tinued deterioration and disfiguration, as well as being covered
with dust and soot. They have apparently never been restored
but are clearly in need of a conservation intervention. This inves-
tigation is a step on the ladder to well-informed restoration and
conservation. It could also add to the knowledge we have about
the materials and techniques of the paintings for art historical
studies.
Despite the large body of extant Ethiopian art in the forms of
462
we know, have previously been carried out on Ethiopian wall
paintings to identify the chemical composition of the pigments
or artistic techniques. This is not surprising as undertaking inves-
tigations onsite in Ethiopia is challenging. These sites, which are
still sacred, are located in remote areas, some accessible only
on foot or even by climbing in addition to being at high altitudes,
e.g. in the Lalibela area between 2600 and 3700 m. Time for an
investigation may also be limited, due to various concerns and
activities. Thus, a range of ad hoc changes have to be accommo-
dated in the field work.
X-ray fluorescence spectrometry, XRF, as a fast and nonde-
structive method, has in recent decades proven versatile for
cultural heritage studies.[5] Examples of both portable and sta-
tionary XRF systems used for in situ and ex situ studies are numer-
ous and diverse in the literature. Some recent applications are for
paintings,[6–8] investigation of paintings and other cultural heritage
objects along with complementary analytical techniques,[9–15] a
remote-controlled system used for field investigation,[16] and
analysis of murals using confocal synchrotron micro-XRF.[17] The
typical characteristics of a handheld XRF were also deemed
appropriate for investigations of Ethiopian artifacts in the field. In
particular, the extreme portability of the analyzer (weight ~ 1.3 kg),
the short analysis time, and the nondestructive nature of the
* Correspondence to: Gebremariam, Kidane Fanta, Norwegian University of
Science & Technology (NTNU), Department of Chemistry, Trondheim, Norway.
E-mail: fanta@ntnu.no
Department of Chemistry, Norwegian University of Science & Technology

murals, icons, and illuminated manuscripts, their study from the (NTNU), Trondheim, Norway

X-Ray Spectrom. 2013, 42, 462–469 Copyright © 2013 John Wiley & Sons, Ltd.
Portable XRF for the study of murals of Yemrehanna Krestos Church
measurement along with the wide range of elements that can be
detected have prompted the selection of the method.
The general objective of the study is to determine the nature
of materials used in the paintings of Yemrehanna Krestos Church.
The sets of pigments used in the different locations of the church
are compared to indicate whether different programs of
paintings existed. The application of pXRF beyond identifying
pigments can guide microsampling for further analyses. How-
ever, on this expedition, we were not able to collect sufficient
samples to warrant adequate complementary investigations. This
investigation is part of a broader project on physicochemical
characterization of the medieval paintings in Ethiopian churches
and monasteries.
Experimental
Energy dispersive X-ray fluorescence analysis: instrument
description and experimental setups
A pXRF analyzer, NITON XL3t 900 (Thermo Fisher Scientific),
was used for the investigation. The analyzer uses a miniaturized
X-ray tube coupled with a high-performance, thermoelectrically
cooled Si-PIN diode detector. The X-ray tube has a silver
anode (50 kV, 40 μA, 2W maximum) and a Kapton window. The
high X-ray flux permits a lower detection limit for many elements
and reduces the analysis time significantly. The analyzer is fitted
with an internal charge-coupled device (ccd) camera, allowing
closer visualization of the measurement area. Depending on the
details to be analyzed, either 8 or 3 mm diameter collimation of
the primary beam was used. The figural paintings on the lower
walls of the northern isle were analyzed with 3 mm collimation,
whereas the geometric decorations on different arches and
capitals with 8 mm.
In this investigation, Cu\Zn mining mode was used because
most of the pigments suspected are inorganic minerals as
suggested by visual examination. Compton normalization and
fundamental parameters are the algorithms used to compute
concentrations of each element within the range of elements
detected. In the case of the fundamental parameters algorithm
used for determination of elemental concentration, factory set
‘internal’ calibrations are used. As the mining mode is intended
for quantitative analysis of metals in mineral matrices, e.g. in
ore prospecting, the assumed matrix can only be an approximate
for the heterogeneous and multilayered painting. Although use-
ful information can be extracted from the quantitative data, the
analyzer may underestimate or overestimate the elemental
composition. For this reason, the numerical results are taken as
semiquantitative, and the qualitative information is generally
emphasized in this study. The light elements not detected by
the analyzer are estimated with reference to the Compton peak
in the fluorescence spectra. The measured value from the
detected elements and the value for the nondetected light
elements, labeled as ‘balance’ by the analyzer, are normalized
to 100%.
Silver is used for calibration of the detector. Various filters are
used for specific ranges; molybdenum in the high range
(covering from barium to silver), aluminum\iron in the main
range (from manganese to bismuth), and copper in the low range
(from potassium to chromium). In the main range, even lighter el-
ements can be detected; however, sensitivity to these elements
increases when light range is selected. In the light range,
elements from magnesium to phosphorous can be detected with
X-Ray Spectrom. 2013, 42, 462–469 Copyright © 2013 John Wiley & Sons, Ltd.
blank filter and helium purge. No measurement under He purge
was affected as gas cylinders were for obvious reasons not possi-
ble to bring on this expedition. The excitation energy levels are
controlled by the mechanical rotation of a wheel of the primary
beam filters between the X-ray tube and the emission window.
For the main range, the tube voltage is 40 kV, and the current is
up to 50 μA. For the light range, the values are 6.2 kV and
199 μA. The energies in the low and high ranges are 20 kVp/
100 μA and 50 kVp/40 μA, respectively. The main and the light
options were selected to cover a wide range of elements with
30 s set for each, making a total counting time of 60 s.
PC-compatible NITON Data Transfer (NDT©) software was used
to customize the instrument, for file transfer, data management,
and data analysis. The numerical results and spectra that can be
displayed on screen in real time allow a quick survey of the ele-
mental composition of the different colors and tones in the paint-
ings and to select the measurement areas for analysis. The
system electronics used in real-time signal processing, computa-
tion and communication include a 533 MHz ARM 11 CPU, a
300 MHz dedicated digital signal processor and a 4096 channel
multichannel analyzer.
Sampling sites
The selection of the measurement areas (spots) was based on the
visual examination of the colors and tones, although somewhat
constrained by access restrictions to certain locations in the
church. Comparison between the geometric and the figurative
paintings' component pigments was addressed, and underpaint-
ing materials exposed in damaged parts of the paintings were in-
cluded. Examples of the figurative and the geometric paintings
are shown in Fig. 1. Parts of the geometric figures and especially
the painted ceiling with diverse figures and curved medallions
are not covered in this study.
The aperture size was selected after monitoring the details on
the instrument's screen and visual examination of the homoge-
neity in color at the specific analysis area. The selected measure-
ment areas were not prepared by dusting off the surface or by
any other means. Examples of measurement areas captured by
the inbuilt camera are shown in Fig. 2. Locations of the measure-
ment areas are described in the columns of Tables 1 and 2,
labeled ‘origin’.
Results and discussion
Two groups of paintings have been investigated; figural paintings
from the lower walls of the northern isle and geometrical decora-
tions on the capitals of columns and arches at various locations in
the church. Both groups appear to have been made using mostly
mineral-based pigments; however, they are quite distinct from
one another. The figural paintings on the walls are executed with
a limited palette, mostly using earth pigments (red and yellow
ochre color), whereas in the geometrical paintings, additional
pigments have been used.
The chemical compounds associated with different colors,
tones, and discolorations are summarized in Tables 1 and 2. Cer-
tain limitations have complicated the interpretation of the results:
the inability to determine specific molecular and structural fea-
tures of the pigment compounds from the elemental informa-
tion; the challenge to detect lighter elements due to X-ray
463
absorption by air; and the nature of the painting structure, which
wileyonlinelibrary.com/journal/xrs
 K. F. Gebremariam, L. Kvittingen, and F. G. Banica

Figure 1. Some of the investigated sites. (a) One of the three figurative paintings that was investigated depicting the story of Jesus Christ washing the
Apostles' feet. (b) A column of the church shown along with the capital, the arches, and the part of the ceiling decorations.

Figure 2. Image taken by the inbuilt camera of the analyzer at two measurement areas. Close-up views of the measurement areas from (a) a figurative
painting and (b) a geometric pattern. The red circle in (a) represents the 3 mm collimation; the larger illuminated circles in (a) and (b) show the 8 mm.

can result in X-ray fluorescence signal originating from multiple
layers. In addition, there are effects of lack of ideal homogeneity
and flat surface, chemical matrix, and spectral interferences.
Preparatory layer analysis of both painting programs
Both visual examination and identification of gypsum, as the
material used for the ground preparation in all parts investigated,
indicate a fresco secco technique. The ground layer in the figura-
tive paintings appears to be mainly composed of gypsum, as con-
firmed by the relatively high XRF peaks for sulfur and calcium in
most cases. The peaks of calcium and sulfur in the spectrum of
gypsum obtained during the analysis of an exposed preparatory
layer (spot 12 in Table 1) are shown in Fig. 3. A good correlation
between calcium and sulfur concentrations in percentage by mass
is observed in the XRF data collected from all parts analyzed, as
displayed in Fig. 4(a). The correlation based on the elemental
information suggests the use of gypsum even though sulfur
can be incorporated in other pigments such as orpiment and
vermillion. Moreover, exposed underlying layers (e.g. spot 12) are
found to be rich in calcium and sulfur, supporting the suggestion
of gypsum as used in the preparation of the ground layer. Sulfur
was measured in the light range. Its signal is more intense in this
464
The onboard software takes into account chemical matrix effects
and artifact peaks when computing the concentration from the
fluorescence signal. Normally, the escape peak of calcium, which
can potentially interfere with the quantitative determination of
sulfur, is small in magnitude, comprising less than 1% of the
parent peak. The intensity of the sulfur peak is detected and thus
the computed concentration is generally high as shown in Fig. 3
as well as Tables 1 and 2. The good correlation obtained between
calcium and strontium could indicate the same origin for the
gypsum used in the ground layer preparation in the figurative
and geometric figures of the church.
Painting layers analysis
Figural paintings on the lower wall in the northern isle
The XRF analysis detected iron in parts colored red and brownish
red indicating red ochre. In spots 11, 14, and 15, lead was found
associated with iron and aluminosilicates, which implies admix-
tures used to impart a desired color tone. High lead content and
relatively moderate concentration of iron with aluminosilicates indi-
cate the use of minium along with red ochre. A variety of the red
and brown colorations can be produced by a combination of lead,
iron-based red pigments (Pb3O4 and Fe2O3), and carbon black. The

range than in the main range. lead peak obtained could also be a result of lead white applied

wileyonlinelibrary.com/journal/xrs Copyright © 2013 John Wiley & Sons, Ltd. X-Ray Spectrom. 2013, 42, 462–469
 Table 1. Results of XRF analyses of figural paintings located in the northern walls of Yemrehanna Krestos Church, indicating the current color of the measurement areas, the inferred pigments along with
elements detected, and the elemental concentration. The figures shown are concentrations computed as weight percentage by an algorithm optimized for the mining mode as described in the exper-
imental section. Note that the light elements not detected are reported as ‘balance’ and this constitutes the major portion of the total elemental composition. The balance is not included in the tables
Spot Color Origin Elements detected Pigments identified
nr

1 Darkish gray Shoe of one of the apostles Ca (6.3) S (2.9 ) Si (7.4) P (1.3) Fe (2.0) Al (1.9) Ti (0.6) — Gypsum and carbon black
2 Yellowish Area below the shoe of one of the apostles Ca (3.4) S (1.1 ) Si (12.8) Fe (2.2) Al (3.2) Ti (0.9) P (0.5) Cl (0.4) Gypsum, goethite and FeO(OH)
3 Orange Edge of the foot of one of the apostles Ca (2.1) Si (3.6) Pb (1.3) P (1.0) Fe (0.6) As (0.7) — — Gypsum and lead white
4 Dark Dark disfigured part of the cloth under cloak Ca (2.4) S (2.7) Si (3.7) P(1.1) Pb (1.8) P (1.1) Cl (1.5) Fe (0.8) Lead white and gypsum covered

X-Ray Spectrom. 2013, 42, 462–469

by carbonaceous matter
5 Yellow Yellowish ground close to a geometrical pattern on Ca (5.2) S (4.0) Si (5.4) Fe (2.3) Al (1.2) Ti (0.7) P (0.5) — Yellow ochre and gypsum
an arch to the left of the painting in Fig. 1(a)
6 Yellowish-white Yellowish ground close to the washing bath Ca (4.6) S (2.3) Si (9.2) Fe (2.6) Ti (1.0) Al (1.9) P (0.6) Cl (0.4) Gypsum and yellow earths
7 Reddish line Edge of the foot of one of the apostles Ca (2.1) S (1.5) Si (4.0) P (1.4) Pb (1.2) P (1.3) As (0.7) Fe (0.6) Gypsum and lead white
8 Red Reddish pattern on an arch to the left of the wall Ca (2.7) S (8.2 ) Si (1.7) P (2.1) Hg (7.0) P (2.1) Ti (0.2) — Cinnabar/vermillion
painting in Fig. 1(a)
9 Yellow Yellow pattern on an arch to the left of Ca (5.3) S (8.2 ) Si (2.0) As (1.5) Fe (0.5) Ti (0.2) — — Orpiment
the wall painting in Fig. 1(a)
10 Yellow Yellow pattern on an arch to the left of Ca (9.1) S (14.5 ) As (1.1) Fe (0.4) — — — — Orpiment
the wall painting in Fig. 1(a)
11 Reddish-brown From smaller wall painting located to Ca (1.2) Si (2.0) Pb (1.9) As (0.7) Fe (0.7) P (0.9) — — Lead white or minium and red ochre
the right of the figure in 1a. The cross of the
Portable XRF for the study of murals of Yemrehanna Krestos Church

one crucified to the right of Jesus Christ

12 White White exposed part above the reddish-brown, spot 11 Ca (9.9) S (13.0) Si (5.5) Fe (1.6) Ti (0.3) — — — Gypsum
13 Yellowish-white Yellowish toned unpainted part of the Ca (5.4) S (3.6) Si (9.1) Fe (2.7) Al (1.8) Ti (0.5) P (0.6) — Gypsum and ochres
Crucifixion painting
14 Reddish-brown From the cloth of the one crucified to the Ca (2.2) Si (3.7) Pb (3.1) Fe (1.1) P (1.2) Cl (1.2) As (0.5) — Lead white or minium and red ochre

Copyright © 2013 John Wiley & Sons, Ltd.

right of Jesus Christ
15 Light pinkish Left hand of the one crucified to the right of Ca (2.0) S (1.7 ) Si (7.8) Fe (1.8) Al(2.5) Pb (0.8) Ti (0.9) — Lead white or minium and red ochre
Jesus Christ in the Crucifixion scene
16 Reddish-brown Dark brown reddish boarder used at Ca (3.6) S (2.0) Si (7.8) Fe (2.0) Al (1.3) Ti (0.7) Cl (0.5) — Hematite and red ochre
the Crucifixion scene
17 Black Darkened scaly part on the cloth of an angel Ca (5.9) S (2.7) Si (9.4) P (1.1) Fe (1.3) Al (1.8) Ti (0.5) — Gypsum covered with carbonaceous
and earth materials
18 Black Dark part above a camel figure Ca (3.0) Si (5.2) Pb (2.5) Fe (2.0) Cl (1.4) As (0.6) Ti (0.4) — Lead white
19 Gray Grayish color of a camel depicted laying Ca (3.5) S (3.1 ) Si (3.3) P (1.0) Pb (1.1) As (0.6) Fe (0.9) Cl (0.7) Lead white and gypsum
under a figure of a saint mounted on a horse
20 Yellowish-white A lightly colored part in front of a camel Ca (1.7) S (1.7) Si (2.4) P (1.0) Pb (1.4) Cl (1.0) As (0.7), Fe (0.6) Lead white and gypsum
mentioned in the case of spot 19.
21 Reddish-brown Background used in the horse riding saints area Ca (3.3) S (1.8) Si (8.2) Fe (1.6) Al (1.3) Ti (0.9) P (0.8) Cl (0.4) Red ochre
22 White White exposed part close to a camel figure Ca (1.7) Si (2.4) P(1.3) Pb (3.1) As (0.9) Fe (0.5) Cl (0.8) — Lead white
23 Black Horse hoof Ca (3.3) S (1.6 ) Si (9.2) Fe (3.5) Al (2.2) As (0.3) Ti (0.6) P (0.9) Gypsum and carbon black
24 White White exposed part close to an angel figure Ca (12.4) S (21.0) Si (2.1) Fe (0.5) Ti (0.2) — — — Gypsum and ochres

wileyonlinelibrary.com/journal/xrs
465
 K. F. Gebremariam, L. Kvittingen, and F. G. Banica

Table 2. Results of the XRF analyses of geometrical interlacings on capitals of columns and various arches indicating their colors, elemental compositions, and inferred pigments. The figures shown are

Gypsum applied with carbon black

Pigments identified

Cinnabar/vermillion (HgS)

Orpiment (As2S3)

Orpiment (As2S3)
Gypsum
Minium
Figure 3. Spectrum indicating gypsum in a ground layer found exposed
because of damage. The spectrum is obtained in the light range of the
‘mining mode’ without He purge.
Ti (0.2)

—
—
—
—
as underpainting; however, this is difficult to confirm with the
Ti (0.1)
Fe (0.2)

pXRF as the relative concentration of oxygen and carbon cannot

—
—
—
—

be compared. In an exposed white area without painting layer

(spot 22), a high level of lead was detected. This reinforces the
claim that lead is more likely linked with lead white than with
concentration computed as weight percentage by an algorithm optimized for the mining mode as described in the experimental section

Fe (0.3)

Ti (0.2)

Fe (0.2)
Fe (0.2)
Al (0.3)

minium. Along with the iron signal detected, titanium was found
in a correlated manner as displayed in the plot in Fig 4(b). This re-
—

lationship is also indicated by the clustering in the loading plot

(Fig. 6(b)). Certain minerals of titanium, such as ilmenite (FeTiO3)
Elements detected

Fe (0.6),

found at low concentration in the red ochre, could be the reason.

Mg (1.6)

Pb (0.4)
Mg (1.0)
Hg (8.2)

Provenance of ochre can in principle be determined from such

—

elemental analyses, as well as from comparison of particle size,

crystal quality, and mineralogical composition.[6,18] The contri-
bution to titanium line from the escape peak of iron could be
Al (0.6)

Cu (3.2)
P(2.4)

Si (1.1)

Si (1.8)

problematic in the interpretation of the result. However, this

—

contribution to the titanium counts is small, taking into account

iron's relative concentration and therefore the magnitude of
As (1.2)
As (3.5)

Si (4.4)
Fe (0.2)
Si (1.1)
Si (2.4)

iron's escape peak.

Lead has been detected in various parts of the figural paintings
(e.g. Jesus Christ washing the apostles' feet and the four saints
S (14.6 )

S (10.3 )

mounted on horses), suggesting that lead white has been

S (6.2 )

S (6.9 )

S (7.4 )

S (4.0 )

employed as underpainting. From antiquity, lead white was the

most common white pigment until the mid-19th century, when
it was replaced by less toxic pigments.[19] The naturally occurring
(8.0)
Ca (2.4)

Ca (4.3)

(7.9)
(9.6)

(6.6)

form, cerussite, PbCO3, or hydrocerussite, 2Pb(CO3Pb(OH)2), was

not commonly used. Synthetic lead white was produced by
Ca
Ca
Ca
Ca

ancient technologies [20] and could have been used in the figura-
tive wall paintings. In spot 18, lead is detected at elevated level
A geometric pattern in an arch located in different

and associated with darkened coloration, which could be attrib-

A geometric pattern in arch located in different

uted to the possible oxidation of lead white into plattnerite

(PbO2) or to the formation of galena (PbS) upon interaction with
sulfides.[21–24]
Orange in the capitals of the columns
Yellow in the capitals of the columns
site than spot 9 and 10 in Table 1
site than that of spot 8 in Table 1

The iron XRF signal indicated the presence of iron in yellow-

colored parts. The goethite-based yellow ochre appears to be
White in the geometric figures
Black in the geometric figures
Origin

used. Yellowish-white toned unpainted parts are found to be

due to gypsum altered by surface deposition rich in iron
containing aluminosilicates (spots 2 and 6). In other parts (e.g.
spot 20), yellowish-white appearance is associated with a high
level of lead, suggesting possible use of lead white to impart
opaque white coloration. The current appearance indicates dark-
ening via partial conversion of white lead into other compounds
as well as the accumulation of dust and soot.
The black coloration is found to be related to high levels of
calcium and sulfur, implying the darkening attributed to carbon
Orange

rich particles from lamp black. The black coloration could also
Yellow

Yellow
White
Color

Black
Red

be linked to compounds of lead as indicated earlier for the

466

case of spot 18. Organic materials at the surface could have

wileyonlinelibrary.com/journal/xrs Copyright © 2013 John Wiley & Sons, Ltd. X-Ray Spectrom. 2013, 42, 462–469
Portable XRF for the study of murals of Yemrehanna Krestos Church

Figure 4. Scatter plots showing correlation between elemental concentrations of (a) calcium and sulfur and (b) iron and titanium. The figures reported
by the analyzer as weight percentage are used in the plots.

exacerbated the disfiguration in different parts of the paintings
by promoting accumulation of dust and soot. In the case of the
gray color, spot 19, lead white appears to have been employed
along with a small amount of carbon black.
Geometric figures on the capitals, columns, and arches
The geometric figures and patterns are mostly still brilliant in ap-
pearance (Fig. 5), considering their age, except where dust and
soot are deposited. As opposed to the figural paintings, vermil-
lion, orpiment, and minium were clearly detected (Table 2).
Gypsum is identified as the ground layer preparation material in
the geometrical and figurative paintings alike. It is also used for
white colorations in the painting layers.
The XRF analysis detected lead in details of paintings colored
red orange. Like lead white, the red–orange minium has been
used since antiquity.[25] Minium is obtained by heating lead
white, 2PbCO3Pb(OH)2, or litharge, PbO, to high temperatures
and has been used in manuscript illuminations, polychrome
sculptures, panels, and wall paintings.[20] These tend to darken
upon contact with soluble salt solutions under humid conditions
due to conversion to plattnerite (PbO2), cerussite (PbCO3), and
PbMg(CO3)2, as well as to galena upon contract with sulfide
containing compounds.[20,21,25–28]
X-ray fluorescence spectrometry examination revealed the
presence of high concentration of mercury and sulfur in the red
parts, indicating the presence of vermillion, a pigment well
known to have been employed in paintings in the medieval pe-
riod. The pigment seems to be stable with no exaggerated dark-
ening observed except for a purplish tone in parts of the
geometric interlacing figures. The purplish tint might be due to
partial conversion of vermillion into metacinnabar [29] or applica-
tion of a dye for protection or for esthetic reasons. Few elements
were detected associated with mercury or sulfur; thus, a synthetic
vermillion appears to have been used rather than the natural
mineral cinnabar. The earliest description of the synthesis of this
pigment dates back to the 8th century when it was considered a
novelty.[30]
High levels of arsenic and sulfur were detected in the yellow-
colored parts. The brilliant yellow pigment appears to be the nat-
urally occurring mineral orpiment, although pararealgar is also
possible based on its color and chemical composition. However,
the yellow parts lack the orangish or reddish appearance because
of the minerals associated with pararealgar. The orpiment fading
via conversion to the most probable arsenic oxide supports the
use of orpiment, as this tendency is observed in parts of the
paintings. Various sulfides of arsenic have been employed as pig-
ments since prehistoric times.[31]
Possible interactions of pigments are illustrated in Fig. 5. The
yellow and orange colorations in the geometric patterns are of-
ten surrounded by a black boundary sketch. This minimizes the
direct contact between the components of the pigments such
as orpiment and minium (Fig. 5(a)). No darkening in the colors
is observed in these areas. Where there is no black boundary
sketch, direct contact results in blackening of the minium pig-
ment (Fig. 5(b)). This is in contrast to no visible darkening ob-
served in painting details where there is direct contact between
vermillion and orpiment-based pigments (Fig. 5(c)). No chemical
reactions appear to take place between the two sulfides.
Except for the darkening linked to certain parts of lead-based
pigments and the accumulation of dust and carbonaceous mate-
rials such as soot, carbon black seems to have been used for

Figure 5. Interaction of pigments. (a) In the star figure, where minium and orpiment-based yellow is separated by a black sketch, no enhanced dark-
ening is observed. (b) Darkening is visible in areas of direct contact between orpiment and minium in the close-up view of a carved and painted capital
of a column. (c) Painting details on an arch where there is direct contact between vermillion and orpiment pigments. Note also the indication of fading
467

of orpiment pigment in the yellow coloration in all the three close-up pictures.

X-Ray Spectrom. 2013, 42, 462–469 Copyright © 2013 John Wiley & Sons, Ltd. wileyonlinelibrary.com/journal/xrs

Figure 6. Application of PCA to identify association of elements in different pigments and comparison of the elemental composition pattern of the
figural and geometrical figures in Yemrehanna Krestos Church. The selected elements from those detected are displayed in the loading plot (Fig. 6(b)). In
Fig. 6(a), c stands for geometrical figures and w for figurative ones. The two principal component results obtained from PCA (PC1 and PC2) are presented
in the form of scatter plot—the score plot (Fig. 6(a)).
black coloring. The suggestion is based on the absence of high
concentrations of calcium, phosphorus, or manganese, which
eliminates the use of bone black and manganese-containing–
dark-colored pigments. Carbon black obtained from burning of
wood and other organic materials also contains mineral traces
such as clay and sand as well as hydrocarbons,[22] which can only
be confirmed by complementary analytical methods. The clay
and sand components are detected in the XRF analysis as indi-
cated in Tables 1 and 2.
A comparison of figurative and geometric paintings
Principal component analysis (PCA) has been used to distinguish
between the figurative and the geometric paintings. Weight percents
of Pb, Ti, Cu, Hg Fe, Si, Al, Cl, Zr, P, Sr, and As from 54 XRF shots on
wall paintings and on arch and column figures were selected as
input variables. In the score plot (Fig. 6(a)), the diagonal line clearly
separates the painting programs into two distinct groups.
Location of an independent variable in the loading plot (Fig. 6(b))
indicates the influence of the variable on the principal components,
whereas the clustering in different directions indicates relation-
ships between the elements in the pigments. For example, Al, Si,
Fe, and Ti appear to be related (drawn in the same direction to
more or less the same extent), and they are associated with clay-
based ochre compounds of the figurative paintings. Similarly, Hg
and As have contributed to the separation of the geometric paint-
ing grouping in the score plot. Note mixing of the two types in the
upper right quadrant of the score plot mainly due to the influence
of the common lead element in the most probable lead white and
minium pigments. In some parts (spot 11 and 14, Table 1), minium
appears to have been used in combination with red ochre.
Rapid survey with portable X-ray fluorescence spectrometry
The results displayed on the screen of the XRF analyzer within a
few seconds helped to select areas for analysis with prolonged
time and to optimize parameter settings. In identically colored
areas where similar compositions are observed in real-time anal-
yses, duplicate microsamples or prolonged XRF measurements
are unnecessary. However, similarly colored areas in close prox-
468
imity could show drastic differences in chemical compositions
wileyonlinelibrary.com/journal/xrs Copyright © 2013 John Wiley & Sons, Ltd.
K. F. Gebremariam, L. Kvittingen, and F. G. Banica
and therefore deserved closer examinations. One example is the
exposed white areas belonging to underlying layers of the figurative
paintings (spots 12 and 22), which showed large differences in
chemical compositions (dominantly calcium vs lead-based compounds).
Another example is the dark areas represented by spots 18 and 23.
On the other hand, the portions painted yellow, red, orange, and
white in the geometric figures on the columns, arches, and ceilings
represent similarly colored as well as chemically composed parts.
Such quick survey of the chemical components of similar colored
paintings is essential to ensure and justify the representativeness
of the areas of analysis selected or a few microsamples to be taken
when needed for further analyses. The geometric paintings on the
ceilings and upper portions of some arches close to the ceiling are
not covered in this study.
Conclusions
In this investigation, pigments in the paintings of Yemrehanna Krestos
Church were rapidly and effectively identified, in situ and nondestruc-
tively, using a handheld XRF. Identification and interpretation were
supported by the color of the painted parts and correlation between
elements of the pigments using the semiquantitative data. The
relationships between the concentrations of different elements
detected in various parts of the paintings were found to be helpful
to decipher the nature of the compounds in use, as exemplified by
gypsum. Deterioration products have been tentatively suggested
in some cases on the basis of the elemental composition, the
current color and properties of pigments used in the medieval
period. PCA was applied to identify groupings related to different
sets of palettes constituting two painting programs. Nevertheless,
some pigments were difficult to identify with certainty, as there
are a number of possibilities resulting in similar elemental compo-
sition and coloration. The existence of several painting layers and
thus heterogeneity in composition coupled with the painting
surface condition has complicated the interpretation of the XRF
results. Systematic technical studies on painting programs can
provide valuable information on the nature and origin of pigment
materials to supplement stylistic and iconographic investigations
besides supporting informed conservation of the paintings. Moreover,
X-Ray Spectrom. 2013, 42, 462–469
Portable XRF for the study of murals of Yemrehanna Krestos Church
extensive documentation of painting materials and techniques
could lead to comparative studies with other medieval paintings
in the country and abroad.
Acknowledgements
The authors acknowledge the Ethiopian Orthodox Tewahedo
Church and the Ethiopian Authority for Research and Conserva-
tion of Cultural Heritage (ARCCH) for the facilitation during the
field work, the Department of Chemistry, Faculty of Science,
NTNU, and the NUFU program, financed by NORAD, for financial
support during the field work and analysis as well as Richard
Verley for the linguistic input.
References
[1] Z. Thiessen, Yemrehanna Krestos church – documenting cultural
heritage in Ethiopia, CHWB, Stockholm, 2010. http://www.chwb.
org/dokument/pdf/Yemrehanna%20Krestos.pdf
[2] T. Tamirat, Church and State in Ethiopia, 1270–1527, The Clarendon
Press, Oxford, 1972.
[3] G. Connah, Forgotten Africa: An Introduction to its Archaeology, 1st
edn, Routledge, New York, 2004.
[4] D. W. Phillipson, Ancient Churches of Ethiopia, Yale University Press,
New Haven, 2009.
[5] D. A. Scott, in Comprehensive Analytical Chemistry, Vol. XLII: Non-
Destructive Microanalysis of Cultural Heritage Materials, (Eds: K. Janssens,
R. Van Grieken), Elsevier, Amsterdam, 2004, pp. 474.
[6] C. Herm, Chimia 2008, 62, 887.
[7] A. Kriznar, V. Munoz, F. de la Paz, M. A. Respaldiza, M. Vega. X-ray
Spectrom. 2011, 40, 96.
[8] K. Uhlir, M. Griesser, G. Buzanich, P. Wobrauschek, C. Streli, D. Wegrzynek,
A. Markowicz, E. Chinea-Cano. X-ray Spectrom. 2008, 37, 450.
[9] A. Deneckere, W. Schudel, M. Van Bos, H. Wouters, A. Bergmans, P.
Vandenabeele, L. Moens. Spectrochim. Acta, Part A 2010, 75, 511.
[10] M. Eveno, B. Moignard, J. Castaing. Microsc. Microanal. 2011, 17, 667.
X-Ray Spectrom. 2013, 42, 462–469 Copyright © 2013 John Wiley & Sons, Ltd.
[11] A. Gianoncelli, J. Castaing, L. Ortega, E. Dooryhée, J. Salomon, P. Walter,
J. L. Hodeau, P. Bordet. X-ray Spectrom. 2008, 37, 418.
[12] L. K. Herrera, S. Montalbani, G. Chiavari, M. Cotte, V. A. Sole, J. Bueno,
A. Duran, A. Justo, J. L. Perez-Rodriguez. Talanta 2009, 80, 71.
[13] S. Liu, Q. Li, F. Gan, P. Zhang. X-ray Spectrom. 2011, 40, 364.
[14] F. Rosi, A. Burnstock, K. J. Van den Berg, C. Miliani, B. G. Brunetti, A.
Sgamellotti. Spectrochim. Acta, Part A 2009, 71, 1655.
[15] P. Vandenabeele, R. Garcia-Moreno, F. Mathis, K. Leterme, E. Van
Elslande, F. P. Hocquet, S. Rakkaa, D. Laboury, L. Moens, D. Strivay,
M. Hartwig. Spectrochim. Acta, Part A 2009, 73, 546.
[16] F. P. Hocquet, H. P. Garnir, A. Marchal, M. Clar, C. Oger, D. Strivay. X-ray
Spectrom. 2008, 37, 304.
[17] S. Valadas, A. Candeias, J. Mirao, D. Tavares, J. Coroado, R. Simon, A.
S. Silva, M. Gil, A. Guilherme, M. L. Carvalho. Microsc. Microanal. 2011,
17, 702.
[18] D. Hradil, T. Grygar, J. Hradilova, P. Bezdicka. Appl. Clay Sci. 2003, 22,
223.
[19] G. Van der Snickt, K. Janssens, O. Schalm, C. Aibeo, H. Kloust, M.
Alfeld. X-ray Spectrom. 2010, 39, 103.
[20] R. J. Gettens, G. L. Stout, Painting Materials: A Short Encyclopedia,
Dover Publications, New York, 1966.
[21] C. Miguel, A. Claro, A. P. Goncalves, V. S. F. Muralha, M. J. Melo.
J. Raman Spectrosc. 2009, 40, 1966.
[22] J. P. Lyalikova, N. N. Lyalikova. Stud. Conserv. 1986, 31, 65.
[23] J. L. Perez-Rodriguez, C. Maqueda, M. C. J. De Haro, P. Rodriguez-
Rubio. Atmos. Environ. 1998, 32, 993.
[24] E. Kotulanova, P. Bezdicka, D. Hradil, J. Hradilova, S. Svarcova, T. Grygar.
J. Cult. Herit. 2009, 10, 367.
[25] S. X. Daniilia, E. Minopoulou. Appl. Phys. A 2009, 96, 701.
[26] S. Aze, P. Delaporte, J. M. Vallet, V. Detalle, O. Grauby, A. Baronnet, in
Lasers in the Conservation of Artworks, (Eds: J. Ruiz, R. Radvan, M. Oujja,
M. Castillejo, P. Moreno), CRC Press, London, 2008, pp. 11–13.
[27] G. D. Smith, R. J. H. Clark. J. Cult. Herit. 2002, 3, 101.
[28] S. Sotiropoulou, S. Daniilia, C. Miliani, F. Rosi, L. Cartechini, D.
Papanikola-Bakirtzis. Appl. Phys. A: Mater. Sci. Process. 2008, 92, 143.
[29] P. Pouli, D. C. Emmony, C. E. Madden, I. Sutherland. Appl. Surf. Sci.
2001, 173, 252.
[30] R. J. Gettens, R. L. Feller, W. T. Chase. Stud. Conserv. 1972, 17, 45.
[31] V. Daniels, B. Leach. Stud. Conserv. 2004, 49, 73.
469
wileyonlinelibrary.com/journal/xrs