Abstract
Since its introduction in 2009, application of the rehydroxylation (RHX) technique for dating
fired-clay ceramics has been controversial, with very few satisfactory dating results collected in
the interim. The stability and efficiency of this technique has been called into question by several
investigators in the last few years, who have struggled to reproduce and validate this new dating
method. Based on our new mass gain measurements for ca. 2000-7000 years old ceramic
artifacts, the reproducibility in the RHX process rate is analyzed and discussed. Timespan
analysis was performed, and age uncertainty related to RHX dating technique was evaluated by
considering the error propagation. The results show poor reproducibility of the RHX process in
the samples of the same origins, which give new evidence for a revision of the RHX protocols.
Introduction
Rehydroxylation (RHX) dating was proposed in 2009 as an easy and reliable technique in dating
sherds of unglazed archaeological pottery.
regained by refired ceramics through rehydration (Stage I) and rehydroxylation (Stage II)
1
processes, 2 and then modeling the Stage II process by a power law-correlation: 4 , where
is the change in sample mass and t is the time after firing or reheating of the ceramic.
3, 4, 5
The mass gain term, m, can be defined in terms of measured mass (in grams) or as a fractional
mass gain value relative to the starting mass of the sample (g/g); the latter is used here. The
roots of this power law correlation have been proposed to originate from a single file diffusion
mechanism
for water molecules migrating through the clay and meta-clay structures in
ceramics. Under these assumptions, the time elapsed since the last firing of the ceramic artifact
can be estimated from the accurate measurement of Stage II mass-gain kinetics. 1
Since its introduction, great effort has been invested in this new dating method by several
archeological and materials research groups. Unfortunately, only ten fired clay fragments have
been convincingly shown to be successfully dated,
Attempts by others have failed,
7, 8, 9, 10, 11
continue to include: i) the stabilization of sample mass after the removal of physically bonded
water; ii) the universality of reheating conditions used to remove chemically bonded water; iii)
the stability and self-calibrating ability of the technique;
8
1
12
13
fluctuations was demonstrated, providing strong, but still qualitative, experimental support for ceramic
artifacts dating through examination of the rehydroxylation process.
14
To succeed with the RHX dating method, at a minimum, the following conditions need to be met:
a)
b)
c)
d)
all water, including the hydroxyl water that is chemically bonded to clays, is totally
removed during refiring;
e)
the water gains are obtained for refired ceramic under constant humidity and
temperature (the same as in step d); and
f)
Although these conditions are quite well recognized by the research community working on the
RHX dating method, little was done to critically examine the effects these conditions have on
dating. Most of the research followed blind protocols of set time and temperatures during the
experimentation suggested in original publication by Wilson et al.,
followed from the same research group. 3, 4, 5, 15, 16, 17, 18 Because no other research team was able
to successfully date ceramic artifacts through the protocols proposed by Wilson et al., the RHX
dating method remains controversial. Only poorly reproducible results were reported so far, 8, 9, 10,
3
14, 19
with many requiring variations on the original 4 power law. In our previous contributions,
1
we have used a generalized power law ( ) with a variable exponent (n), to describe the RHX
1
behavior of ceramics without identifying their ages. 10 The vs. curves produced for several
samples demonstrated improved linearity in RHX Stage II. 9, 12
The purpose of this contribution is to examine the reproducibility of the Stage II kinetics in the
RHX dating protocol for a group of small sherds from the same parent fragment after refiring at
500oC. The intragroup repeatability was examined for four such sets of ceramic artifacts, all
treated identically. Understanding the reproducibility of RHX dating raises new questions about
the validity of the dating protocol suggested in the past.
Experimental
In our study, four sets of sherds of fired clay ceramics from Mediterranean locations were
examined (Table 1). Ceramic samples were collected and provided for this research by
Prof.Kostalena Michelaki from Arizona State University. The sherds were cleaned to remove
loose debris and dirt, and washed with deionized water. Larger fragments, typically squareshaped cross sections, were cut into approximately 10 10 5 mm pieces with a water-cooled
diamond saw. Masses of the final samples varied from 0.4 to 1.0 g. One fragment of each set of
samples of the same origins was dried at 110C for 5 to 7 days as a control sample. Four to six
other samples were furnace dried overnight at 300C. This elevated drying temperature was used
here to speed up the drying process. We found previously (unpublished) that some ceramic
artifacts hold some amount of physically bonded water very strongly and drying them at 110120oC is not always sufficient to remove the totality of physically bonded water.
After drying, all samples were placed in a glove box (Coy Laboratory Products, Grass Lake, MI)
equipped with a Citizens CM-11 microbalance (Citizen Scale Inc., Edison, NJ) having 1 g
precision. The glove box was set for 202% RH (relative humidity) for the duration of the whole
experiment and located in a lab with an average ambient temperature of 202C. The mass of
each sample was recorded for several days until it reached its equilibrium mass, previously
defined as 2 by Wilson et al.,
15
no larger than 0.01%. Then all samples were fired in a box furnace at 500C for 1 to 3 days and
placed back into the glove box, again at 20C and 20% RH. Samples were labeled with their
drying temperature and refiring times; for example, 300C-2a would mean that the sample was
heated at 300C, equilibrated to find 2 , and then fired for two days at 500C. The refired
ceramic mass 0 was critical to all other experimentation, as it was the value to which changes
in mass were normalized to yield a fractional value of . Sample weight was monitored
periodically after refiring; every 15 min in the first couple of hours, then every 1-2 hours, and
finally at least once a day for several weeks. X-ray diffraction (XRD) measurement was
performed on a small sample surface by using Scintag XDS2000 / X-ray diffractometer with
Cu K radiation ( = 1.540562 ). The scans were performed continuously from 15 to 70o in 2
at a speed of 0.6o/min with a step size of 0.02o. The XRD instrument was operated at a tube
potential of 45 kV and 35 mA filament current. Peaks were identified using the DMSNT (Scintag
Inc.) software package in concert with the Powder Diffraction Database Search (PDDS) software
(Scintag Inc. and Radicon Ltd., St. Petersburg, Russia). The PDDS software was used to search
the Joint Committee on Powder Diffraction Standards-International Center for Diffraction Data
(JCPDS-ICDD) database.
Statistical Analysis
Linear regression analysis was used in this study to analyze the linear, Stage II portion of the
1
versus 4 curves.
1
Timespan analysis was performed by comparing the RHX rate () against the 4 . The RHX rate
1
() was calculated by the slope of the fractional RHX mass gain with respect to 4 , and was
calculated at intervals that began with the last data points and was gradually expanded to include
data back to t = 0 (using the slope from the end approach).
calculated near the end of the experiment experience large perturbations due to the small number
of data points considered; as the timespan analysis approaches the transition to Stage II mass
gain the value of becomes more statistically robust.
Random error propagation analysis
20
dating outcome. It was assumed that the archaeological age of the sample (ta), is described by
Equation 1:
2 4 4
= (
(1)
2 + ( )2 2
= ( )2 2 + (m )2 m
m2
4
m
4
(2)
The standard error values for the coefficients and 4 were determined by using the "LINEST"
function in Microsoft Excel. Standard error for 2 was estimated as the fractional measurementto-measurement deviation of the equilibrium mass (0.01%). Sources of systematic-type error
are not included in this approach.
Seven fragments from sample RLG 019 show a typical two-stage progression of mass with t1/4 .16
Neither drying at 300C, nor prolonged firing time showed a significant effect on the RHX rate.
The timespan analysis of the RHX curve (shown in solid lines) indicates that a trend is present
for the RHX rate () throughout the experiment. Instead of reaching an ideal plateau near the
Stage II transition, appears to increase systematically as the interval of analysis increases up to
the time where Stage II presumably begins. The observed scattering and fluctuations of over
the entire timespan analysis could be argued to increase uncertainty in the calculated age.
For sample RLG 024, evolution of fractional mass gain does not strictly follow the two-stage
pattern; it displays both concave and convex behavior at different times. As the observed
curvature for all six specimens does not occur at the same time, it is likely that the disturbance
comes from the intrinsic instability of the sample and the RHX technique. It is rather unlikely
that this could be attributed to perturbation of the environment, as all samples within the set were
simultaneously submitted to concurrent temperature and humidity fluctuations. With respect to
the RHX rate over time, samples 300C-1c and 100C-1 show a plateau in from 3.0-4.0 and 3.54.2 hr1/4, respectively, while the others do not appear to stabilize at all in the timespan analysis.
Because external conditions were identical, the different RHX behavior of samples from the
same sherds [300C-1 (a, b, c, d, and e)] indicates the poor reproducibility of the technique.
Overall, the RHX rates of RLG 024 specimens appear to scatter more than those from RLG 019,
and lead to larger average age uncertainties. The vast scattering can also be evidenced from the
wide range of 4 and value listed in Table 3. The higher drying temperature, again, had no
discernable effect on the RHX rate. The change in drying conditions appeared to both increase
the RHX rate from 96 g/h1/4 to 400 g/h1/4, as well as decrease it to 47 g/h1/4. The fact that
values nearly ranged over a full order of magnitude is worrisome; in principle, this value should
be practically constant.
Fractional mass gain for sample RLG 007 resembles sample RLG 019, in that both sets of
samples exhibit a typical two-stage progression of mass with time1/4. The timespan analysis over
the entirety of Stage II (Figure 4) shows RHX rates that are relatively flat with a narrow
spreada better outcome than for other samples. Also like RLG 019, the value of in the
timespan analysis appears to depend strongly on the time at which Stage II is assumed to begin.
The timespan and linear regression analyses indicate that the scattering for a family of specimens
cut from the same larger sherd differs from sherd-to-sherd, which is in agreement with recent
results shown by LeGoff and Gallet.
wide as 25%, but samples 100C-1 and 300C-1e, they exhibit nearly identical RHX behavior.
Taken in whole, results in RLG 007 again indicate poor stability in the RHX process.
For sample RLG 014, a plateau in the RHX rate () was observed in Stage II at 2.5-3.5 h1/4. Even
though relatively constant values were observed, their magnitudes vary widely, by as much as
82%. Again, the poor sample-to-sample consistency cannot be explained by random error or
fluctuations in either temperature or RH conditions. The poor reproducibility instead points to
intrinsic instabilities in the RHX process. It should be mentioned that samples 110C-1 and 300C2a exhibited an unexpected mass decrease near the end of the experiment. It is thought that this
was a result of sample disintegration.
For all four sets of samples, the scattered nature of the RHX process introduce 10-40% of
random uncertainty in the calculated ages and the estimated ages are much younger than their
documented ages (Table 1). (However, in the case of RLG 007, the absolute calculated dates
should be much shorter due to RHX measurements being performed at temperatures substantially
higher than the parent sherds ELT. In this case, the sample ages cannot be compared to the
documented age.) This consistent outcome points to some source of systematic error (>99%) in
the RHX dating process that caused a mismatch between the calculated and archaeological ages.
There was also a noticeable amount of scatter within each family of samples, especially RLG
007 and 014, which had calculated ages that spanned two orders of magnitude. The scatter could
be explained, in part, by the pronounced scatter behavior of in Stage II as well as the
subjectivity involved in selecting a Stage II transition in samples that did not show a RHX rate
plateau. If one desired, dates could be made to appear older or younger by arbitrarily selecting a
cut-off point along the timespan analysis that gave the desired RHX rate () and value of 4 .
The data illustrate the troubling scatter in RHX rates and calculated dates between samples cut
from the same parent sherds, and processed in an identical manner.
Conclusions
By examining the RHX rate for all the four series of samples, few examples of consistent RHX
rates, with respect to both time and other samples within a family, were identified. Replicate
experiments show striking scatter at two drying conditions (110C and 300C) and different
refiring conditions (500C for 1 to 3 days), with no discernable effect of either condition on
RHX kinetics. High intragroup variation of RHX rate within a batch of specimens cut from the
same parent sherd is indicative of poor intrinsic Stage II stability when analyzed in the t1/4 power
law framework. The poor stability and the scattered nature of the RHX process introduce 10-40%
of random uncertainty in the calculated ages. This is in addition to unidentified systematic errors
(in some cases corresponding to >99% error) that result in shorter-than-actual calculated ages for
10
all specimens examined here. All the results are indicative of inherent, unexplained complexities
in applying the RHX dating technique to archaeological fired-clay ceramics.
Acknowledgements
We thank Kostalena Michelaki from Arizona State University for providing the fragments used
in the current study, and our collaborator Carl Lipo and Elizabeth Niespolo from California State
University at Long Beach for helpful discussions over the course of RHX experiment. This
project was supported by an Archaeometry Research Award from the United States National
Science Foundation (No.1219540).
References
1. M. A. Wilson, M. A. Carter, C. Hall, W. D. Hoff, C. Ince, S. D. Savage, B. McKay, and I. M.
Betts, "Dating fired-clay ceramics using long-term power law rehydroxylation kinetics,"
Proc. R. Soc. London, Ser. A, 465[2108] 2407-15 (2009).
2. A. Hamilton and C. Hall, "A Review of Rehydroxylation in Fired-Clay Ceramics," J. Am.
Ceram. Soc., 95[9] 2673-78 (2012).
3. M. A. Wilson, W. D. Hoff, C. Hall, B. McKay, and A. Hiley, "Kinetics of Moisture Expansion
in Fired Clay Ceramics: A (Time)1/4 Law," Phys. Rev. Lett., 90[12] 122503-1-03-4
(2003).
4. S. D. Savage, M. A. Wilson, M. A. Carter, W. D. Hoff, C. Hall, and B. Mckay, "Moisture
expansion and mass gain in fired clay ceramics: a two-stage (time)(1/4) process," J. Phys.
D: Appl. Phys., 41[5] 055402(4pp) (2008).
5. S. D. Savage, M. A. Wilson, M. A. Carter, B. McKay, W. D. Hoff, and C. Hall, "Mass Gain
due to the Chemical Recombination of Water in Fired Clay Brick," J. Am. Ceram. Soc.,
91[10] 3396-98 (2008).
11
12
13
14
Site and
Archaeological
Context
Location
ELT/ oC
RLG007
Penguila,
Alicante,
Spain
10.9
RLG014
Umbro Greek/
T2 N73 E45
Context 237
Bag 2348
Bova Marina,
Calabria, Italy
19.520.5
RLG019
Umbro
Neolithic
T6 N6 E59
Sherd #142158
Bova Marina,
Calabria, Italy
19.520.5
Age:
Achaeometric
or
Calendrical
and RHX
target in
Years before
2014
Neolithic
5,590 +/-40
BP,
TRHX: 5,6145,694
Classical
Greek
425-350 BC
TRHX: 2,4392,364
Main Phases
Quartz, illite2M1,
calcite
Quartz, illite2M1,
anorthite, albite,
Early to
Quartz, illiteMiddle
2M1,
Neolithic
anorthite, albite
5,500-5,000
Cal BC
TRHX: 7,5147,014
RLG024
Penitenzeria;
Bova Marina,
19.5Early/Middle
Quartz, illiteT4; E40 N46
Calabria, Italy
20.5
Neolithic
2M1(2),
Sherd #20385,500-5,000
anorthite, albite
66
Cal BC
TRHX: 2,4392,364
Sample dates include expected dates expressed as calibrated or uncalibrated radiocarbon dates or using
the known date range of cultural phases at the specific site. These dates are then converted into RHX
Target Age (TRHX) understood as years before A.D. 2014. Samples RLG014, RLG019, and RLG024
were recovered during excavations by the Bova Marina Archaeological Project team. Details about the
excavation contexts for the sample sherds have been reported elsewhere 21, 22, 23. Sample RLG007 is from
the site of Mas DIs which has also been described in the literature 24, 25.
15
Nominal
Mass
0.780520
m4105[g/g]
105[(g/g)/h1/4]
Age/year
1053 9
160 2
2.1 0.2
300C-1a
0.693347
932 10
156 3
1.6 0.1
300C-1b
0.762487
956 9
161 2
1.3 0.2
300C-2a
0.520244
936 10
163 3
2.2 0.2
300C-2b
0.814421
949 9
157 2
1.7 0.2
300C-3a
0.895630
929 11
156 3
2.1 0.3
300C-3b
0.504413
873 1
156 3
1.7 0.2
16
Nominal
Mass
m4105[g/g]
105[(g/g)/h1/4]
Age/year
100C-1
0.80787
698 7
10 2
74
300C-1a
0.684012
807 14
54
31
300C-1b
0.624774
664 16
19 4
1.3 0.4
300C-1c
0.92682
849 10
20 3
0.5 0.1
300C-1d
0.886827
610 8
31 2
1.4 0.5
300C-1e
0.811570
834 10
40 3
43
17
Nominal
Mass
m4105[g/g]
105[(g/g)/h1/4]
Age/year
100C-1
0.635143
864 11
130 3
38 1
300C-1a
0.591287
990 13
135 3
0.5 0.1
300C-1b
0.750737
782 10
137 3
24 2
300C-1c
0.622529
739 9
133 2
23 2
300C-1d
0.603122
754 11
119 3
290 30
300C-1e
0.577538
870 11
121 3
33 4
18
Nominal
RLG 014
Mass
100C-1
m4105[g/g]
105[(g/g)/h1/4]
Age/year
252 8
55 3
685 27
300C-1a
0.659297
0.622409
257 5
59 1
34 9
300C-1b
0.656815
336 12
35 2
91
300C-2a
0.458795
237 7
59 2
600 200
300C-2b
0.682371
200 6
62 2
15 3
300C-3a
0.629391
242 6
11 4
157 23
19
20
Figure 2. RHX mass gain and RHX rate of RLG 019. After drying at 110C (100C-1) and 300C,
they were fired at 500C for 1, 2, or 3 days samples denoted as 300C -1a(b), 300C -2a(b),
300C -3a(b), respectively.
21
Figure 3. RHX mass gain and RHX rate of RLG 024. After drying at 110C (100C-1) and 300C,
they were fired at 500C overnight samples denoted as 300C-1 (a, b, c, d, e)).
22
Figure 4. RHX mass gain and RHX rate of RLG 007. After drying at 110C (100C-1) and 300C,
they were fired at 500C overnight, which was denoted as 300C-1 (a, b, c, d, e)).
23
Figure 5. RHX mass gain and RHX rate of RLG 014. After drying at 110C(100C-1) and 300C,
they were fired at 500C for 1, 2, 3 days, which was denoted as 300C -1a(b), 300C -2a(b), 300C
-3a respectively.
24