Preliminary study of a Georgia O’Keeffe pastel drawing using XRF
and 
XRD
Lynn B. Brostoff
Preservation Research and Testing Division, Library of Congress, 101 Independence Ave. SE,
Washington, DC 20540, formerly Museum Conservation Institute, Smithsonian Institution
Catherine I. Maynor
Smithsonian American Art Museum, Lunder Conservation Center, 750 9th St., N.W., Washington,
DC 20013-7012
Robert J. Speakman
Museum Conservation Institute, Smithsonian Institution, 4210 Silver Hill Rd., Suitland, Maryland 20746

Received 30 September 2008; accepted 20 January 2009
X-ray fluorescence spectrometry 
XRF and micro-X-ray diffraction 

XRD were used to analyze
the composition of pigments on a pastel drawing, Special No. 32, by Georgia O’Keeffe. XRF
analyses showed that, among other pigments present in the drawing, the red, orange, and yellow
pigments may possibly be identified with lead- and chromium-based pigments: lead chromates, red
and yellow lead oxides, and/or lead carbonates, plus calcium-based pastel fillers, such as whiting or
gypsum. XRD examination of a sample removed from a dark mottled area of coral red pastel
confirmed that this pigment layer, which is associated with a darkened appearance and high Pb:Cr
ratios, matches the red lead oxide, minium 
2PbO·PbO2. © 2009 International Centre for
Diffraction Data. DOI: 10.1154/1.3133137
Key words: XRF, XRD, pastel, pigment, red lead, O’KeeffeI. INTRODUCTION
As part of a conservation assessment of Special No. 32,
a Georgia O’Keeffe 1915 pastel drawing on black paper

mounted to paperboard in the collection of the Smithsonian
American Art Museum 
SAAM 
Figure 1, a pigment sur-
vey analysis using non-invasive X-ray fluorescence 
XRF
spectrometry was coupled with minimally invasive micro-X-
ray diffraction 

XRD analysis of a single sample. The
analyses were undertaken by the Smithsonian’s Museum
Conservation Institute 
MCI in order to assist in developing
an approach for the conservation and treatment of the work,
as well as to lay the groundwork for a planned technical
study of O’Keeffe pastels. Non-invasive XRF was chosen as
a survey tool that enabled in situ analysis of the delicate
drawing. XRF allowed rapid preliminary investigation of the
range of possible pigments present and provided a better un-
derstanding of the condition of the piece with regard to select
pigments in red and orange areas that appear to have a dark
brownish surface mottling. The uneven quality of this mot-
tling, and the observation that it appears to be isolated to the
uppermost surfaces of some areas of coral red pastel, sug-
gested that these areas may consist of an altered red pigment,
rather than an intentional application of a darker pastel. XRF
data guided decisions with regard to microsampling for fur-
ther analysis of the dark mottling by 
XRD. Overall, the
goals of the XRF and 
XRD analyses were to identify pos-
sible materials present in the piece and to explore whether
the pastel’s current appearance is a significant departure from
the artist’s original intent.
Natural and synthetic red lead pigments have been in use
since antiquity and their susceptibility to darkening has been
recognized at least since the Renaissance 
Gettens and Stout,
1942. It also has been observed that red lead pigments are
relatively stable when protected from air by a rich organic
116 Powder Diffraction 24 
2, June 2009 0885-7156/2medium such as glue or linseed oil 
Gettens and Stout,
1942. However, the mechanisms and products of red lead
pigment alteration are incompletely understood. Fitzhugh

1986 remarks that the darkening of lead tetroxide 
Pb3O4 or
2PbO·PbO2 may be more complex than simple oxidation to
the lead dioxide 
PbO2. While environmental factors such as
light, humidity, pollution, hydrogen sulfide, and also micro-
bial contamination clearly have potential mechanistic roles,
recent investigations also show that composition and micro-
structural features of the red lead grains, which result from
the manufacturing process, may be determining factors in the
pigment alteration process 
Gettens and Stout, 1942;
Fitzhugh, 1986; Kuchitsu, 1997; Saunders et al., 2002; Aze
et al., 2002. Therefore, the incorporation of red lead into
pastels, which are by nature large amounts of dry pigment
bound by a minimal amount of organic medium, would be
expected to yield an artist’s material that is vulnerable to
alteration and color conversion 
Gettens and Stout, 1942;
Fitzhugh, 1986.
In O’Keeffe’s writings, the artist conveyed that meaning
in her art resides in color and shape itself. This was true
particularly after the pivotal year of 1915, when, in her own
words, “I first had the idea that what I had been taught was of
little value to me except for the use of my materials as a
language—charcoal, pencil, pen and ink, watercolor, pastel,
and oil” 
O’Keeffe, 1976. O’Keeffe used a variety of com-
mercial pastel sticks and appears to have also made pastels
herself, as indicated by materials left in her studio 
Walsh,
2000. Several studies and the record of her working prac-
tices by Caroline Keck 
Walsh, 2000; Barilleaux and Whi-
taker Peters, 2006 provide information about O’Keeffe’s
materials and techniques. However, further investigation, es-
pecially technical study, is warranted. As stressed by Keck
and Walsh 
Walsh, 2000, it was imperative to O’Keeffe that
116009/24
2/116/8/$25.00 © 2009 JCPDS-ICDD
her works remain in pristine condition, especially in terms of
color. Therefore, information about the artist’s materials and
possible changes in color resulting from interaction with the
environment are of great significance to the understanding of
her oeuvre.
II. EXPERIMENTAL
A. XRF analysis
The instrumentation used was an Innov-X Systems Al-
pha Series XT-440 handheld energy dispersive X-ray fluores-
cence spectrometer 
ED-XRF that incorporates a miniature
X-ray tube with a silver anode, a Si-PIN diode detector, and
a primary beam aluminum filter. The XRF will operate be-
tween 10 to 40 kV and 10 to 100 
A; detector resolution is
less than 200 eV. The beam diameter of analysis is approxi-
mately 14 mm. For this research, the instrument was oper-
ated in the “soil mode,” at 35 kV and 13 
A. The purpose of
utilizing these beam settings was to optimize sensitivity to a
broad range of elements, i.e., those found from potassium

K to bismuth 
Bi. The depth of penetration in the area of
examination is dependent on the beam settings, as well as the
elements present, the density of the material, and the loss of
photons from scattering in the air and through the material
matrix itself. Without available purging or vacuum capability
in this analysis, elements lighter than sulfur 
S were not
detectable and elements lighter than calcium 
Ca were ex-
pected to show decreased sensitivity.
The portable XRF unit was used in a stationary mode by
attachment to a specially designed tripod and arm that ex-
tended out over the object, which lay flat on a polymer-
composite laboratory benchtop. The benchtop did not con-
tribute an elemental signature to the analysis, nor did it
contribute significantly to the spectral backgrounds. Further-
more, this horizontal support system, including the backmat,
allowed for the safest handling of the object during analysis.
In this configuration, the XRF instrument was held flush with
the front mat, about 2.0 mm above the surface of the object,
in order not to damage any of the extremely friable media

Figure 2. This precise working distance was maintained
throughout each exposure by keeping the XRF instrument
Figure 1. 
Color online Special No. 32, 1915, pastel on paper, by Georgia
O’Keeffe 
credit: the Smithsonian American Art Museum, Acc. No.
1995.3.2, 1419 1/2 in..stationary and manually repositioning the drawing under-
117 Powder Diffr., Vol. 24, No. 2, June 2009neath the nose of the instrument to the area of interest. Sev-
eral areas also were examined using a faceplate that limited
the analysis to a maximum of 5.0 mm in diameter, and, as
expected, greatly decreased the intensity of the X-ray beam.
Due to the thickness of the faceplate, these spectra were
obtained at an added distance above the object of about 1.0
mm, further limiting detection of X-ray intensity. The win-
dow did not significantly contribute to detection of elements
of interest 
see Results section. Exposures were 180 s for
good peak to background ratio.
The Innov-X software incorporates built-in peak decon-
volution and factory calibration for the beam settings. Based
on an assumption of infinite thickness, empirically-derived,
linear calibration factors are used to calculate the concentra-
tions of detected elements based on the measured intensities
and normalization to the Compton backscattering peak be-
tween 20 to 24 keV. These calculation methods make use of
several assumptions that were not met in practice, and there-
fore do not accurately express elemental quantities. Devia-
tions from ideal sampling geometries that are expected to
cause quantification errors include sample thickness much
less than an infinite thickness, inhomogeneity of materials,
and scattering from separation between the instrument win-
dow and the object. Another problem in interpretation of
software-generated results is spectral interference of peaks
that are poorly resolved and depend on software deconvolu-
tion. For example, there is poor resolution between titanium

Ti K and barium 
Ba L lines, and spectral overlap of lead

Pb L and arsenic 
As K peaks, as well as Pb M lines
and sulfur 
S K lines. Other potential sources of error arise
from matrix effects, including overestimation or underesti-
mation of concentrations due to absorption-enhancement ef-
fects from the particular mix of elements in the sample ma-
trix, especially as these phenomena relate to the presence of
heavy metals like Pb. However, due to the nature of pastels,
which are essentially a thin layer of pigment
s ground with
a minimum of gum binder and differing quantities of chalk
or gypsum, it is reasonable to assume that matrix effects
from the pigment layer or paper are negligible.
B. XRD analysis
XRD analyses were performed on a Rigaku D/Max
Figure 2. 
Color online XRF positioned 2 mm above artwork.Rapid diffractometer with an image plate detector using cop-
117Preliminary study of a Georgia O’Keeffe pastel drawing …
Proposed identification based on color and
elemental composition
Red lead 
Pb 
+Pb white/litharge or talc +ZnO?
Red lead 
+Pb white or talc ?
Chrome orange 
Pb,Cr+gypsum/whiting 
Ca
Chrome orange+gypsum /whiting+red lead
Red lead 
+Pb white or talc?
Red lead 
+Pb white or talc?
Red lead 
+Pb white or talc ? +nearby orange
Red lead+gypsum /whiting+barite 
Ba+ ? 
+TiO2?
Emerald green 
Cu,As+chrome green 
Pb, Cr, Fe

and/or viridian? +gypsum/whiting
Organic blue or synthetic ultramarine 
?
Chrome yellow+gypsum/whiting
Chrome yellow+Sr yellow 
Sr,Cr+barite or
Ba yellow+blue
Chrome orange
Chrome orange+red lead 
+Pb white or talc?
Chrome orange+red lead 
+Pb white or talc?
Red lead 
+Pb white or talc 
? +stray whiting/
gypsum?

near LOD.
118
118
PowderD
iffr
.
,V
ol.24,N
o.2,June
2009
Brostoff
,M
aynor
,
a
nd
Speakm
anTABLE I. XRF results.
Reading Description of area of analysis Cr Fe Cu Zn Asa Pb Otherb
Pb:Cr
rangec
Large window
4 Benchtop LOD LOD LOD LOD LOD LOD
7 12 ply rag matboard LOD LOD LOD LOD LOD LOD Ca, Ti, Fe
10, 11, 29 Average black paper+other support materials 530 1200 51 21 290 Ca, Ti, Sr 0.55
RSD black paper+other support materials 21% 20% 18% 13% 29% 12%
17 Coral-red area w/black paper showing through 600 1400 120 83 6400 9.4 to 12
18 Nearby in coral-red 720 1600 69 40 5400 6.6 to 8.4
19 Orange area with no visible mottling 6200 1000 51 34 7900 Ca 1.1 to 1.4
20 Dark mottling in coral-red stripe over orange 5400 1100 120 35 13000 Ca 2.1 to 2.7
21 Coral-red, little or no dark mottling 600 1200 58 42 12000 18 to 22
22 Coral-red with more dark mottling 560 1300 70 39 11000 17 to 22
23 Dark mottling in coral-red stripe over orange 930 1200 76 33 14000 Ca 13 to 17
28 Brighter red, bottom right 750 1400 83 38 2400 Ca, Ba, Ti 2.8 to 3.6
24 Dark green at top 2900 2700 420 20 800 1900 Ca 0.58 to 0.73
25 Light blue 710 1500 110 36 460 Ca 0.57 to 0.73
26 Yellow 1700 1300 66 24 2000 Ca, Ba 1.0 to 1.3
27 Yellow-green 7500 1600 71 46 8300 Sr, Ba 1.0 to 1.2
Small window
5 Benchtop 
faceplate background LOD 100 40 LOD LOD
13, 30 Average black paper+other support materials 200 500 80 5 100 0.5
RSD black paper+other support materials 4% 1% 4% 141% 27% 24%
31 Orange 
same as 19 2000 400 60 10 2000 Ca 0.8 to 1.2
32 Dark mottling/coral-red stripe over orange 
20 1000 400 90 LOD 3000 2.3 to 3.7
33 Dark mottling/coral-red stripe over orange 900 400 80 LOD 3000 Ca 2.5 to 4.1
34 Dark mottling in coral-red stripe over orange 
23 LOD 400 70 LOD 4000 Ca Pb only
KEY: LOD=limit of defection, RSD=relative standard deviation, bold=greater than LOD 
3max RSD+ave. background, and italics=questionable
aArsenic not confirmed in spectra other than yellow-green.
bNo values available for Ca; other elements detected in raw spectra as relatively trace peaks.
cPb:Cr ratio ranges reported based on 12% variation 
large window or 24% variation 
small window RSD for background paper.
per K radiation, 50 kV acceleration voltage, and 40 mA
current. For these analyses, one tiny sample was mounted
onto a glass fiber using an amorphous cyanoacrylate-based
adhesive and exposed to a microbeam for 15 min using a 0.3
mm collimator, and 5 h using a 0.1 mm collimator. The go-
niometer parameters were chi axis fixed at 45°, omega axis
fixed at 0°, and phi axis spun 360° at 1°/sec. Experimental
patterns were integrated over a large 2 theta region to mini-
mize preferred orientation effects. After background subtrac-
tion, the patterns were qualitatively matched using JADE 7.5
software to known materials in the International Center for
Diffraction Data 
ICDD database.
III. RESULTS AND DISCUSSION
XRF results are provided in Table I according to the
analyses and areas of examination, which are marked as
reading numbers and shown in Figure 3. The table includes
software-determined values only for elements that were veri-
fied by the analyst to be present in the raw spectra; therefore,
only one value for arsenic 
As is reported. Additional peak
identifications are shown in the column marked “other;” here
elements are reported only if they appear above levels de-
tected in the paper and support materials. Note that the
Innov-X system was not set up in this study to determine
values for elements below titanium 
Ti or for strontium 
Sr;
additionally, lack of vacuum in the instrument 
or a helium
purge precluded qualitative detection of light elements, be-
low about sulfur 
S. It must be stressed that the numeric
values listed in Table I for each analyte cannot be interpreted
in a quantitative manner because of non-ideal examination
conditions, as discussed above. However, the software-
generated values, which are normalized to the Compton scat-
tering and scaled according to predetermined sensitivity fac-
tors, may be used for purposes of internal comparison, and
are therefore presented without units, although the software
reports these values as parts per million 
ppm.
The relative standard deviations 
RSDs of elements de-
tected in the paper 
Table I provide a baseline reference with
which to compare levels of elements detected in the pastel
layers. Values determined from a relatively small available
area of unpigmented paper show quite high relative standardFigure 3. 
Color online Areas of analysis.
119 Powder Diffr., Vol. 24, No. 2, June 2009deviations 
RSD; this is most likely due to the typical inho-
mogeneous distribution of trace elements in the paper, as
well as the presence of stray pigment in this area. Although
this raises the limits of detection, elements may still be de-
tected at levels significantly above the background, here
shown in bold when greater than three times the maximum
RSD 
29%, plus the average value detected in the composite
of supporting materials, i.e., the black paper, its paperboard
mount, rag matboard and benchtop; questionable elements
detected at approximately two times the background levels
or less are shown in italics. The RSDs of trace elements
detected in the background are a product of the analytical
method and the paper itself, which, as mentioned, is con-
taminated with stray pigment. On the other hand, the RSD of
Pb:Cr ratios in the paper is more indicative of Pb:Cr varia-
tion in the pastel layer, as measured with this method, and is
more consistent. However this range remains an estimate.
Reflecting this uncertainty, the elemental values are rounded
to two significant figures for the large window and one sig-
nificant figure for the small window. Based on elemental
composition and color, Table I proposes possible pigments
for the pastel colorants in different areas of examination with
varying certainty.
Results indicate the presence of one or more of the fol-
lowing elements in most areas of interest: calcium 
Ca, ti-
tanium 
Ti, barium 
Ba, chromium 
Cr, iron 
Fe, copper

Cu, strontium 
Sr and lead 
Pb; zinc 
Zn detection ap-
peared variable and inconsistent. By selecting slightly differ-
ent spots in the one sufficiently large area of black paper
without visible pastel 
readings 10, 11, 13, 29, Figure 3, it
was determined that elements associated with the various
support materials are as follows: 
1 no significant quantity
of elements is detected in the benchtop; 
2 the matboard has
a measurable trace quantity of Ti 
presumably from TiO2, as
well as variable trace Ca and Fe; 
3 trace Ca, Cu, Sr and
zinc 
Zn are found in the paper/paperboard; 
4 Cr and Pb,
detected in the black paper areas, are possibly associated
with the nearby yellow pastel; and 
5 somewhat higher,
variable levels of Fe are detected in the black paper/
paperboard.
Measurements obtained from the drawing’s red, orange

Figure 4, and yellow 
not shown areas indicate that the
pastel colorants are primarily associated with the elements
Pb, Cr, and Ca. As shown in Tables I and II, this evidence
suggests that O’Keeffe’s pastels may contain pigments made
from lead chromates 
chrome orange, chrome yellow, and
chrome red, lead oxides 
red lead with or without litharge or
massicot, lead carbonates and/or sulfates 
lead whites, and
white calcium-containing pigments such as whiting 
CaCO3
and/or gypsum 
CaSO4·2H2O. This assessment is based on
the color, known formulas of pigments available to O’Keeffe

Table II, and the lack of evidence for significant levels of
potassium 
K, Fe, molybdenum 
Mo, As, cadmium 
Cd, or
mercury 
Hg, elements that would indicate other possible
pigments. Along with clay, both whiting and gypsum are
common ingredients and fillers in pastels. The presence of
Ba may indicate either a pigment or barite, which is a com-
mon extender; the less common extender, calcium chromate,
also could be present, as well as trace amounts of zinc white

ZnO in some cases. It is notable that neither Ca nor Ba is
detected in the coral-red areas. This implies that these pastels119Preliminary study of a Georgia O’Keeffe pastel drawing …
may consist of a clay base plus pigments and fillers, includ-
ing red lead, lead monoxide, lead whites and/or talc 
hy-
drated magnesium silicate.
Analysis of the darker blue-green pastel 
Table I clearly
demonstrates the presence of Ca, Cr, Cu, As, and Pb, as well
as somewhat higher levels of Fe. These results tentatively
suggest a mixture of chrome green 
Pb, Cr, Fe and emerald
green 
Cu, As, plus a calcium-based filler such as whiting or
gypsum. It also is possible that viridian 
hydrated chromium
oxide is present in this pastel color. Results for the yellow-
green area are interesting. In addition to Cr and Pb, trace
amounts of Sr and Ba are indicated. These results suggest
Figure 4. 
Color online Details of XRF spectra of various orange and red
mostly unpigmented black paper and support 
black, reading 10.
TABLE II. Possible lead-, chromium-, and Ca-based
Color Common name
White or grey Lead white 
hydrocerussite
White Lead carbonate 
lead white
White Basic lead sulfate
White Lead sulfates
Yellow Massicot
Yellow-orange Litharge
Red Red lead 
minium
Orange Chrome orange
Yellow Chrome yellow
Yellow Barium yellow
Yellow Strontium yellow
Green Chromium oxide
Blue-green Viridian
Green Chrome green
White Chalk or whiting
White Gypsum120 Powder Diffr., Vol. 24, No. 2, June 2009that the yellow-green pastel contains a mixture of chrome
yellow, strontium yellow 
SrCrO4, barium sulfate 
barite
and/or barium yellow 
BaCrO4, and an undetected blue pig-
ment.
Analysis of the light blue pastel 
Table I does not reveal
notable levels of any element that can be ascribed to the
pastel layer except for Ca. Because the presence of both Fe
and Cu are variable in the background materials, they have a
fairly high limit of detection; the detection of Fe and Cu at
comparable levels in the blue pigment may therefore be as-
signed to the paper itself. Detection of Pb and Cr is also

colored readings 18, 19, 20, and 23 overlaid with spectrum of an area of
ents 
Gettens and Stout, 1942.
Composition
Pb3
CO32
OH2
PbCO3
PbSO4·PbO, typically used with ZnO
PbSO4 
and variants
PbO
PbO
Pb3O4
PbCrO4·Pb
OH2 or PbCrO4·PbO
PbCrO4 and PbCrO4·PbSO4, usually with extenders
BaCrO4
SrCrO4
Cr2O3
Cr2O3·2H2O
Chrome yellow+Prussian blue 
Fe4Fe
CN63
CaCO3
CaSO4·2H2Oareaspigm120Brostoff, Maynor, and Speakman
comparable to that detected in the bare paper. Certain pig-
ments can be eliminated from consideration, including co-
balt, manganese and cerulean blues, so that the blue pastel
colorants may possibly be synthetic ultramarine

Na8–10Al6Si6O24S2–4 or an organic 
carbon-based blue,
plus whiting or gypsum. As mentioned above, with the de-
scribed analytical setup light elements such as those found in
a carbon-based pigment or ultramarine either would not be
detected or would have poor sensitivity, such as in the case
of sulfur. Prussian blue 
ferric ferrocyanide cannot be ruled
out as a component of the blue pigment or as an additive in
the yellow-green pastel, although it is expected that the pas-
tel layers are thick enough in most areas of the drawing that
iron- or copper-based pigments should show clear differen-
tiation from trace levels of these elements in the backing
materials.
Results in Table I further indicate that inter-elemental
ratios, e.g., those taken from the simple mathematical quo-
tient of XRF values for Pb and Cr 
marked Pb:Cr, may be
used to locate areas that possibly contain red lead, as distinct
from lead chromate pigments. The Pb:Cr ratios in Table I are
shown as a range, the latter of which is determined from the
experimental RSD of Pb:Cr ratios in the black paper and
support materials. Note that the RSD of this ratio is less than
the RSD of values for individual elements. Validity of nu-
merical elemental ratios is based on several assumptions, in-
cluding 
1 linearity of calibration factors for each element;

2 homogeneity of the area of examination; and 
3 fair
uniformity in the matrix across the surface of the drawing.
While homogeneity may be compromised by layering and/or
blending of pigments in the pastels, the measured ratios of
Pb:Cr, which are particular to the instrumentation and setup,
appear to differentiate red or orange lead-based pastels from
the lead chromate-based pastels. This is significant due to the
increased stability, in the absence of sulfur in the environ-
ment, of basic orange and red lead chromates 
Gettens and
Stout, 1942; Kühn and Curran, 1986. As discussed above,
the orange area 
reading 19 contains Ca as well as Cr and
Pb, suggesting identification with whiting or gypsum and
chrome orange 
lead chromate. Here the measured Pb:Cr
ratio is 1.1 to 1.4. Because this technique is not quantitative,
the elemental ratio cannot be correlated with the known sto-
ichiometry of pure chrome orange, which is
PbCrO4·Pb
OH2. In other words, the ratio more likely re-
flects a characteristic orange pastel composition in this
analysis. Results for the yellow and yellow-green areas

readings 26, 27 suggest that the pastels contain chrome
yellow 
PbCrO4; here the Pb:Cr ratio is similar to that in the
orange area. While the analysis of reference pigments could
clarify this relationship somewhat, elemental ratios found in
the artwork may also reflect the presence of common con-
taminants or additives, such other lead-rich pigments or a
PbSO4 component, since different shades of chrome pig-
ments are achieved by varying this additive 
Kühn and Cur-
ran, 1986.
Despite limitations in interpretation of the XRF-derived
Pb:Cr ratios, there appears to be consistent separation of the
measured Pb:Cr ratios in differently colored areas. For ex-
ample, compared to the orange areas 
reading 19, spots with
uneven, dark mottling 
readings 17 and 18, 21 and 22 show
dramatically increased Pb:Cr ratios, between 6.6 and 22. The
121 Powder Diffr., Vol. 24, No. 2, June 2009breadth of this range is likely to arise from differences in
application thickness and detection of Cr at background lev-
els in thinly applied coral-red pastel. The Pb:Cr ratios in this
case support the appearance that the same lead-based pastel
is used in coral-red areas of the drawing, all of which exhibit
the uneven, dark mottling. Pb:Cr ratios also appear revealing
in blended or layered coral-red and orange areas. For ex-
ample, in dark, mottled, coral red stripes over orange areas

readings 20, 23, the Pb:Cr ratios are somewhat lower than
coral-red areas 
readings 17, 18, 21,22, which appear to be a
single layer of color. Perhaps more significantly, the Pb:Cr
ratios in dark, mottled areas, particularly the stripe in reading
23, suggest the presence of red lead, possibly mixed with
other lead oxides, carbonates and/or sulfates.
Pb:Cr ratios obtained using the small-spot window plate
on the XRF instrument further suggest red lead in the dark,
mottled stripes and areas of the drawing. As shown in Table
I, peak intensities are about 1/3 of those measured without
the faceplate. Comparison of Pb:Cr ratios in orange and
coral-red areas with and without the small aperture confirms
that the ratios are similar 
readings 19 vs. 31, and readings
32 to 33 vs. 20. This indicates that use of the small window,
and the additional distance to the detector, does not compro-
mise detection of the pigment elements. In reading 34, i.e.,
an area limited to a heavily mottled stripe, only Pb is de-
tected in significant quantities. Although not definitive, the
Pb:Cr ratios thus suggest a link between the dark mottling
and high amounts of Pb, and thus converted red lead oxide
pigment.
One red area appears different both in color and in Pb:Cr
characterization. This is the brighter, more intense red area
examined in reading 28, which does not exhibit dark mot-
tling. Here the ratio is 2.8 to 3.6, with overall lower intensity
of Pb, Cr at background levels, plus a small amount of Ba

and possibly Ti. While this evidence might suggest a thin
application of red lead, the appearance of this red and its
different elemental signature suggest rather that it is an ad-
mixture of red lead with an undetected red colorant and/or
other additive. This points out that Pb:Cr ratios by them-
selves are not indicative of composition.
Following XRF analysis, one tiny sample was removed
from the dark surface in a coral-red area for 
XRD analysis.
Sampling was considered warranted based on the interpreta-
tion of the XRF results and necessary, at this point, to further
understand the condition of the artwork. The sample, about
30 microns in diameter, is shown in Figure 5 mounted on a
glass fiber 
0.1 mm diameter in the instrument. The re-
sulting XRD pattern, obtained with a 0.1 mm collimator, is
shown in Figure 6. The large hump in the raw pattern

above is produced mainly by amorphous scattering from
the glass fiber and adhesive and may be responsible for ob-
scuring small peaks in this region. The background-corrected
pattern 
below provides an excellent match with minium,
the natural version of red lead 
Pb3O4, PDF #01-071-0561;
in fact, the experimental pattern accounts for all the peaks in
the reference pattern for this mineral, shown by the vertical
bars. Thus, 
XRD results confirm XRF results, indicating
that the dark pigment particles in the non-uniformly mottled
areas are associated with red lead 
Pb3O4, properly written as
2Pb2PbO4.
121Preliminary study of a Georgia O’Keeffe pastel drawing …
Red lead is well known to be subject to darkening as
mentioned above. The brown or black products that may
form from the pigment depend on environmental conditions,
including exposure to nitric acid, acetic acid, hydrogen sul-
fide, or simply to air, humidity and light 
Gettens and Stout,
1942; Fitzhugh, 1986. Except for the sulfide product, dark-
ening is thought to arise primarily from the formation of
brown-black lead dioxide 
Gettens and Stout, 1942;
Figure 5. 
Color online Particle from mottled area mounted onto a glass
fiber.Figure 6. 
Color online 
XRD pattern, with and without background subtraction
122 Powder Diffr., Vol. 24, No. 2, June 2009Fitzhugh, 1986. Darkening has also been ascribed to forma-
tion of a mixture of lead monoxide and metallic lead, and the
tendency of a red lead pigment to darken has been related to
the amount of free PbO 
lead monoxide that is naturally
present in the pigment, as well as the composition and mi-
crostructural features of the red lead grains 
Kuchitsu, 1997;
Saunders et al., 2002; Aze et al., 2002. In this context, it is
worth noting that the experimental XRD pattern also is con-
sistent with identification of the minor or poorly crystallized
phase scrutinyite 
PbO2, PDF #01-072-2440 or 00-045-
1416, which is closely associated with plattnerite, a more
commonly referenced lead dioxide alteration product of red
lead 
Fitzhugh, 1986. However, because the most intense
peak in the XRD pattern of scrutinyite is coincident with a
peak in the pattern of minium, and because plattnerite is not
identified, it is difficult to ascertain the presence of the lead
dioxide. In addition, it is noted that the presence of some
PbO in the form of litharge 
PDF #00-005-0561 is consis-
tent as a trace phase in the experimental pattern 
Gettens and
Stout, 1942; Fitzhugh, 1986; no lead sulfide, lead carbonate
or lead sulfates are identified in the pattern.
Despite the lack of evidence for the presence of lead
dioxide, which is a well-known analytical problem, the vir-
tually non-invasive, rapid XRF and 
XRD analysis of the
Georgia O’Keeffe pastel supports visual observations that the
uneven, dark brown mottling on the surface of coral-red ar-
eas is likely to be converted red lead. Most importantly,
analysis supports suspicions gathered from visual observa-
tions that the drawing has changed from its original appear-
ance. The detection of red lead in unmottled areas therefore
has implications for the future care of the drawing, including
protection from environments in which red lead is known to
be unstable, including light, high humidity, and atmospheres
containing sulfur and other pollutants. The preliminary XRF, obtained with 0.1 mm collimator from brownish particle in dark mottling.
122Brostoff, Maynor, and Speakman
survey of representative areas of other pastel pigments in the
drawing suggests numerous pastel compositions and invites
further study of O’Keeffe’s materials.
IV. CONCLUSIONS
Preliminary XRF and 
XRD analyses of the Georgia
O’Keeffe pastel, Special No. 32, provide evidence for the
existence of red lead, an unstable pigment, in coral-red areas
of the drawing. This supports conservators’ suspicions that
dark brown, uneven mottling in these areas may be associ-
ated with red lead pigment, which is well known to be prone
to alteration and darkening. Conversion of red lead pigments
may occur over time in a nonuniform manner from exposure
to air, humidity and light, as is consistent with the appear-
ance of this pastel drawing. The evidence consists firstly of
XRF analysis of relative Pb:Cr ratios, which allows lead-
based pigment composition in red areas to be differentiated
from more stable lead chromate-based pigments in orange
and yellow areas of the drawing. Secondly, XRD analysis of
one microsample removed from a dark mottled area provides
definitive evidence for the presence of red lead oxide, which
is identical to the mineral minium. Neither the expected al-
teration product, lead dioxide, nor lead monoxide is posi-
tively identified, although they could be present as trace
phases in the XRD experimental results. On the other hand,
lead sulfide, lead carbonate, and lead sulphates do not appear
consistent with the XRD patterns. Therefore, results of the
preliminary analysis suggest that the brown mottling may be
converted red lead, although discoloration from an unknown
mold or undetected pollutant products cannot be ruled out.
The detection of red lead in perceived unchanged areas has
further implications for the future care and protection of the
drawing from harmful environments.
Additional pigments and fillers indicated by XRF analy-
sis based on color and elemental composition include
calcium-based pigments such as whiting or gypsum, stron-
123 Powder Diffr., Vol. 24, No. 2, June 2009tium yellow, and chromium-based greens such as chrome
green and probably emerald green. Blue pigments cannot be
proposed with confidence due to limitations of this instru-
ment in detecting light elements, but some blue pigments can
be eliminated, including cobalt, manganese, and cerulean
blues. Due to fairly high levels of Fe and varying trace levels
of Cu in the paper support materials, iron- or copper-
containing pigments cannot be ruled out in some cases; how-
ever, results suggest that the blue pigments, used either alone
or as admixtures in other pigments, are likely to be synthetic
ultramarine or carbon-based. The findings of this preliminary
study of O’Keeffe’s early pastel drawing warrant and invite
further investigation of materials used by this important art-
ist.
Aze, S., Vallet, J.-M., and Grauby, O. 
2002. “Chromatic degradation pro-
cesses of red lead pigment,” ICOM Committee for Conservation pre-
prints, 13th Triennial Meeting, Rio de Janeiro 
ICOM, London, pp.
455–463.
Barilleaux, R. P. and Whitaker Peters, S. 
2006.Georgia O’Keeffe, Color
and Conservation 
Mississippi Museum of Art/Penn State University
Press, Jackson, Mississippi.
Fitzhugh, E. W. 
1986. Artists’ Pigments: A Handbook of their History and
Characteristics, edited by Feller, R. L. 
National Gallery of Art, Wash-
ington, D.C., pp. 109–140.
Gettens, R. J. and Stout, G. L. 
1942. Painting Materials: A Short Ency-
clopedia 
Dover Publications, New York, NY.
Kuchitsu, N. 
1997. “Mineralogical consideration on discoloration of red
lead,” Hozon-kagaku 36, 58–66.
Kühn, H. and Curran, M. 
1986. Artists’ Pigments. A Handbook of their
History and Characteristics, edited by Feller, R. L. 
National Gallery of
Art, Washington, D.C., pp. 187–200.
O’Keeffe, G. 
1976. A Studio Book 
Viking Press, New York.
Saunders, D., Spring, M., and Higgitt, C. 
2002. “Colour change in red
lead-containing paint films,” ICOM Committee for Conservation pre-
prints, 13th Triennial Meeting, Rio de Janeiro 
ICOM, London, pp.
455–463.
Walsh, J. C. 
2000. O’Keeffe on Paper, edited by Fine, R. E., Lynes, B. B.,
Glassman, E., and Walsh, J. C. 
National Gallery of Art, Washington,
D.C., pp. 57–80.
123Preliminary study of a Georgia O’Keeffe pastel drawing …