Introduction

Polyethylene glycol (PEG) is a synthetic polymer which has been used since the 1950s in the archeological conservation of waterlogged wooden artifacts1. PEG replaces water within the wood, preventing warping, shrinkage and cellular structure collapse2. Without the use of a consolidating agent such as PEG, previously waterlogged objects would be subject to potentially catastrophic changes to their structure after their excavation and exposure to dry air and oxygen3. As PEG is non-toxic and relatively cheap, it rapidly replaced previously used conservation agents such as linseed oil, creosote, or alum, becoming one of the most commonly used consolidating agents for wooden objects in museum collections by the end of the 1960s.

Because PEG is a synthetic hydrocarbon, its presence in a sample will adversely affect any attempt at 14C dating. PEG is usually produced from petroleum that is millions of years old and should therefore theoretically not contain any 14C4,5, and younger 14C ages following removal of PEG from wood have previously been demonstrated6. In modern excavations, archeologists will often select wood samples for 14C dating prior to treating artifacts with PEG in order to avoid this problem. As radiocarbon dating of archeological artifacts did not become commonplace until after the widespread use of PEG, however, many important artifacts excavated during the first half of the 20th century were treated with PEG without having been first radiocarbon dated. This makes it difficult to date these objects today. Developing and evaluating accurate procedures for the removal of PEG from wooden objects is, therefore, a critical goal for heritage science research.

This paper reports the results of an experimental evaluation of a chemical pretreatment protocol designed to remove PEG from wooden archeological artifacts. We use wood of a known age obtained from the Gribshunden shipwreck (see below), which was conserved using low molecular-weight PEG-400 at the Blekinge museum. We selected two oak (Quercus robur) and one ash (Fraxinus excelsior) samples; all three are elements of wooden casks recovered from the wreck. The selection of both oak and ash samples was made to ensure repeatability across different wood species. We used a two-stage sequence for removing PEG from the samples, which included treatment with water and organic solvents, and acid-base-acid treatment, followed by cellulose extraction. At the start of the processes and following every subsequent stage, a portion of the sample was put aside for 14C dating and MALDI MS determination of the presence of PEG. Our study evaluates whether our multi-step pretreatment protocol can effectively remove PEG-400 from conserved wood, verified through both MALDI-MS analysis and radiocarbon dating.

Methods

Materials selection

The wooden artifacts selected for analysis in this study were collected from the securely-dated shipwreck of the royal Danish-Norwegian flagship, Gribshunden (Fig. 1) (The authors have obtained written consent from B. Foley and A. Hansson to publish this photo). The ship was built in 1484, launched in 1485, and served a ten-year career until sinking in mid-June 1495 in the Baltic Sea near the town of Ronneby in what is now Sweden7. On its final voyage, the ship conveyed King Hans and several noblemen to a planned six-week political summit in Kalmar with the Swedish, Norwegian, and Danish Councils of State8. During archeological activities conducted on the wreck in 2006, 2019, 2020, and 2021, the field teams recovered 135 elements of wooden casks9. The cask components included the heads (covers) and the staves that form the body of the casks. All but five of these (96.3%) were made from oak. The non-oak elements are three staves of beech and two pieces of ash. The casks from Gribshunden and other barrels recovered from medieval and early modern archeological sites indicate that they typically were constructed from 12–16 staves, depending on the size of the barrel and the width of the staves. In the Gribshunden casks, the few beech and ash elements apparently were mixed into barrels with components otherwise fashioned from oak. The two oak samples in this study are staves of wooden casks recovered from the wreck; the ash sample was also interpreted as a cask element by the excavating archeologist10.

Fig. 1: Photos of cask staves from Gribshunden.
figure 1

(L) Excavation on Gribshunden, with cask components in situ evident in foreground. (R top) Authors B. Foley and A. Hansson examine cask staves (photo published with written consent of B. Foley and A. Hansson). (R bottom) Several staves were recovered from Gribshunden. (Image credits: Left, Brett Seymour; Right top and bottom, B. Foley).

Full size image

For the study conducted here, we selected three of the radially-split cask staves that were recovered during the 2006 excavation and PEG-treated (Table 1)10. To test repeatability across different wood species, two samples are of oak (Quercus robur), and the third is ash (Fraxinus excelsior). All three samples were recovered during excavation in 2006. After recovery they were conserved by treatment with PEG-400 at 10% solution for at least 3 months, then frozen down to −30°C, and finally freeze-dried under vacuum10,11,12. As all samples used in this study were treated with PEG-400, subsequent references to PEG should be read as referring to low molecular weight PEG-400 rather than high molecular-weight variants (see Discussion section for overview of experimental limitations). Both oak samples contained sapwood rings on their outer edges: nine and four rings. These samples were dated by dendrochronology. The ash sample did not contain sapwood rings, and it was not dateable by dendrochronology. However, because it was excavated from the same securely-dated context, and because previous studies have shown the lifespan of medieval casks to be only a few years in their primary use, we are confident that the age of the ash cask stave conforms closely to the other staves recovered from Gribshunden13,14. Further, we hypothesize that in light of the king himself being aboard the ship for its final voyage to a major summer-long political summit, the provisions casks in its hold were likely to have been newly coopered to contain victuals for the royal retinue. In pre-modern Northern Europe, wooden casks were used to store and transport all manner of dry, semi-liquid, and liquid goods. The contents of the particular casks represented in this study are not known. Some barrels recovered from this shipwreck held butchered sturgeon and beef. From their physical characteristics and historical documents listing royal purchases, others are presumed to have contained provisions including beer, and supplies such as gunpowder13,14.

Table 1 Dendrochronological data for the Gribshunden cask staves sampled for 14C dating
Full size table

From each of the three artifacts in this study, we selected five consecutive rings for 14C analysis. In order to obtain sufficient material for dating, the samples were cut with an industrial razor blade from the approximate middle of the staves, where the rings were relatively thick. The cut flakes were divided into inner and outer sample parts, deriving two 14C samples for each stave. This allowed testing for potential PEG concentration variations in the different sections of the staves, although we hypothesized that this was unlikely due to the staves’ form factor. Unlike many other larger PEG-treated archeological wooden artifacts, the cask elements are thin. The 135 staves collected for the original study have an average thickness of 11.8 mm. In fact, the three artifacts tested in this study were even thinner than average, with thicknesses of 10.8, 09.9, and 11.2 mm at their 14C sampling points.

Chemical pretreatment and 14C dating

The six wood samples taken from the outer surface or inner portion of the three staves were split into three replicate samples for 14C dating, including all five annual rings present in the original sample, resulting in a total of 18 samples (Fig. 2, Table 2). The replicate samples were thereafter subjected to three different chemical pretreatment protocols before 14C dating: (1) no treatment (N), (2) treatment with water and organic solvents, and acid-base-acid treatment (W + OS + ABA), (3) treatment with water and organic solvents, acid-base-acid treatment, and cellulose extraction (W + OS + ABA + CE). These treatment protocols were selected as they are within the capacities of most radiocarbon laboratories and have been previously used for PEG removal without any systematic experimental analysis of their effectiveness4. A minor portion of each pretreated replicate sample was put aside for MALDI MS determination of the presence of PEG.

Fig. 2
figure 2

Conceptual workflow diagram showing sampling and treatment steps.

Full size image
Table 2 Position, chemical pretreatment and 14C results for the Gribshunden wood samples
Full size table

Samples subjected to protocol 1 did not receive any chemical pretreatment and were only dried before being weighed in for subsequent graphitization. Chemical pretreatment according to protocol 2 started with agitating thin (<1 mm) shavings of wood with Milli-Q water in beakers at room temperature for 2 minutes. This was done three times for each sample. Subsequently, the same procedure was carried out using acetone (40 °C, 30 min) and methanol (40 °C, 30 min), with each step followed by washing with Milli-Q water. Then, a standard acid-base-acid (ABA) pretreatment was performed using 2% HCl and 2% NaOH at 80 °C. Each of the three steps lasted a minimum of 4.5 h and was followed by washing with Milli-Q water. The remaining wood material was then dried. Pretreatment according to protocol 3 started with all the steps of protocol 2 and continued with cellulose extraction. The wood material remaining after ABA treatment was placed in beakers with Milli-Q water and heated in a water bath to c. 85 °C. After the addition of NaClO2 (80%) and acidification with a few droplets of 3% HCl, the temperature was kept constant for one hour. This procedure was repeated three times, yielding a final NaClO2 concentration of c. 1.4%. The white and fibrous material remaining after bleaching was then washed with Milli-Q water and dried.

Material from all 18 pretreated samples was weighed into tin capsules and combusted in an elemental analyser coupled to an AGE-3 automated graphitization system (Wacker et al. 2010) in which the C of the CO2 produced in the combustion was converted into graphite. 14C measurement of pressed graphite targets was then performed by accelerator mass spectrometry (AMS) using a NEC-SSAMS system15. The blanks measured together with the samples, and used for age calculation, were produced from >100,000-year-old wood through pretreatment and graphitization in the same laboratory environment as the samples, allowing adequate quality control. The resulting 14C ages were calibrated using OxCal online16 version 4.4.4 and the IntCal20 calibration dataset17.

Testing for the presence of PEG

A portion of each sample at each stage of the pretreatment process was tested for the presence of PEG using matrix-assisted laser desorption/ionization mass spectrometry (MALDI MS). We also tested the use of Optimal Photothermal Infrared Spectrometry (OPTIR) for the detection of PEG, but found that the method returned too many false negatives, so we proceeded with only using MALDI MS. MALDI MS is a powerful method for the detection of synthetic polymers such as PEG by measuring the mass and charge of the analyte. The level of detection of MALDI MS for PEG 400 is less than 1 µg/ml18. PEG consists of repeating units that are singly charged and easily separated and recorded using MALDI MS. The spacing of PEG signals in a spectrum is 44.02 Da, which corresponds to the repeating C2H4O units of PEG (Fig. 3).

Fig. 3: Photos of cask staves from Gribshunden.
figure 3

Conceptual drawing showing A Molecular formula of PEG. B MALDI-MS spectrum of dibenzocyclooctynyl poly(ethylene glycol) (DIBO-PEG) acquired on a ToF/ToF instrument, and C Close up of spectrum showing characteristic 44 Da distance between individual peaks. (modified from Wesdemoitis et al.24 and Zheng et al.25—reused with permission from the American Chemical Society).

Full size image

For the MALDI-MS analysis, 30 µl of MilliQ water was added to each wood sample in a 1.5 ml MAXYMum Recovery® microtube from Axygen and sonicated for 30 minutes. As a control, to detect background signals, the same procedure was performed but with only MilliQ water and no wood in the tube. Then 1 µl of the sample solution was mixed with 0.5 µl Matrix solution, consisting of 5 mg/ml α-cyano-4-hydroxy cinnamic acid, 80% acetonitrile, 0.1% TFA, on a MALDI stainless steel plate. MS spectra were acquired using an Autoflex Speed MALDI TOF/TOF mass spectrometer (Bruker Daltonics, Bremen, Germany) in positive reflector mode using a laser intensity of 55%, 10,000 shots collected, and a mass range of m/z 100–2500. As a control, to detect background signals, the same procedure was performed but without any wood sample present in the MilliQ water. All spectra were externally calibrated using Peptide calibration standard II (Bruker Daltonics).

Results

14C dating

Comparison of the results for the three replicate samples from each of the six original wood samples shows that 14C ages, and consequently also calibrated ages, for samples that did not receive any chemical pretreatment (N) are consistently older than for pretreated samples (W + OS + ABA and W + OS + ABA + CE) (Table 2, Fig. 4). The age difference between corresponding N and W + OS + ABA + CE samples varies between c. 600 and 5700 years, based on mean values of maximum and minimum calibrated ages at 1σ. Moreover, calibrated ages for the four N samples for which dendrochronological ages are available are between c. 600 and 3200 years too old. Together, this suggests that the original 14C signal of the wood has been diluted through the addition of a substance containing very little or no 14C, which is consistent with PEG conservation.

Fig. 4
figure 4

Calibrated 14C ages for all 18 triplicate samples from the six original wood samples taken from the outer or inner surface of the three staves. Calibrated age intervals at 1σ (upper) and 2σ (lower) are shown below each probability distribution. Calibration was made using OxCal online16 version 4.4.4 and the IntCal20 calibration dataset17. Results are presented in order according to 14C sample ID number (20447–20464) and separated into groups of three representing the six original wood samples. A vertical line runs through the first 12 samples showing the dendrochronologically determined date of the oak cask staves (1428–1432 and 1429–1433 CE). The pretreatment protocol used for each sample is indicated below the ID number. See Table 2 for further sample information.

Full size image

In contrast to the N samples, those pretreated using the W + OS + ABA protocol show very similar calibrated ages, and these are consistent with the available dendrochronological ages (Table 2, Fig. 4). Except for one sample (LuS-20457), the expected age is within the calibrated 1 σ interval. At 2 σ, all dendrochronological ages of W + OS + ABA samples are within the calibrated interval. The 14C results for samples receiving full pretreatment are very similar to those for corresponding samples not subjected to the final cellulose extraction step (Table 2, Fig. 4). Mean calibrated ages at 1 σ are between 1 and 52 years younger for five of the W + OS + ABA + CE samples but 5 years older for one of them (LuS-20463). Corresponding values at 2 σ are 2-56 years younger and 2 years older.

In combination with irregularities in the calibration curve, minor shifts in 14C age may result in significant changes in the width of calibrated age intervals. This is evident when comparing the results for the replicate samples from three of the six staves and explained by a major wiggle in the 14C calibration curve between c. 1300 and 1400 CE (Table 2, Fig. 5). Although wood formed around 1430 CE, within an adjacent and steep part of the calibration curve, was targeted to achieve as narrow calibrated age intervals as possible, nine of the twelve W + OS + ABA and W + OS + ABA + CE samples are influenced by the 1300–1400 CE wiggle and therefore display two separate calibrated age intervals (Table 2, Fig. 5). Only LuS-20447, LuS-20453 and LuS-20456 display a single calibrated age interval, both at 1 and 2 σ. Interestingly, all three samples received full pretreatment (W + OS + ABA + CE), highlighting the importance of including cellulose extraction in the pretreatment of wood samples.

Fig. 5
figure 5

Individual calibration plots for all 18 14C samples. The obtained 14C age with its normal distribution is shown on the Y axis, and the irregular probability distribution of the calibrated age resulting from the projection of the 14C result onto the calibration curve (included in the plot) is shown on the X axis. Calibrated age intervals at 1 σ and 2 σ are provided in text and shown below each probability distribution plot. Calibration was made using OxCal online16 version 4.4.4 and the IntCal20 calibration dataset17. Each row represents one of the six original wood samples, and each column represents one of the three pretreatment protocols (from left to right: W + OS + ABA + CE, W + OS + ABA, and N). See Table 2 for further sample information.

Full size image

The 14C dating results for the W + OS + ABA + CE samples taken from the outer and inner surfaces of the ash stave should provide a reliable age estimate for this stave, which was not dendrochronologically dated. The calibrated 1 σ intervals for these samples (1310-1397 CE and 1314-1398 CE, respectively) are almost identical, and the same is true for the 2 σ intervals (1300-1406 CE and 1301-1408 CE) (Table 2, Fig. 4). Comparing these results with the dendrochronological dates for the two oak staves (1428-1432 CE and 1429-1433 CE) suggests that the ash stave is slightly older. This is supported by the mean calibrated ages for the W + OS + ABA + CE samples from the three staves. For the outer surface, these are 1426 CE and 1438 CE for the oaks, and 1354 CE for the ash, at 1 σ. Corresponding values at 2 σ are 1424 CE, 1437 CE, and 1353 CE. Based on these 14C results, the rings sampled from the ash wood staves are probably around 60–70 years older than the oak wood. Previous findings confirming the short primary use lifespan of casks (often only 6–8 years as transport containers14) in the late medieval and early modern periods suggest that these ash staves were fashioned from the inner portions of their respective logs13,14.

MALDI-MS detection of PEG

The results of our MALDI-MS analysis showed no detectable PEG in any of our samples following both the ABA pretreatment and cellulose extraction steps of our preparation procedure. Prior to pretreatment, all samples clearly showed the presence of PEG, as indicated by repeating signals at increments of 44.02 Da corresponding to the C2H4O groups units of PEG (Fig. 3). This PEG signal is absent in both the W + OS + ABA and W + OS + ABA + CE samples (Fig. 6 and Fig. 7). These results show that our pretreatment regime removed all detectable levels of PEG from our samples, which corroborates the 14C results presented above. Moreover, it indicates the success of the W + OS + ABA treatment for obtaining accurate dates from PEG-treated wood samples.

Fig. 6
figure 6

MALDI-MS spectrum of untreated (N) samples with regularly spaced peaks clearly indicating the presence of PEG. Insert at bottom highlights the 44.02 Da distance between peaks, which corresponds to the C2H4O units of PEG.

Full size image
Fig. 7
figure 7

MALDI-MS spectrum of samples following W + OS + ABA (left) and W + OS + ABA + CE pretreatment (right). Peaks corresponding to PEG are absent in both. Notice how the spectra for the post-cellulose extraction samples perfectly match the control.

Full size image

Discussion

Contamination with PEG has been a major problem facing scholars seeking to date conserved wooden artifacts for decades2,4,6. This paper has presented two different pretreatment protocols for the removal of low molecular-weight PEG-400 from wood samples and tested their efficiency at PEG removal using both 14C dating as well as MALDI-MS. As discussed above, our results show that both pretreatment protocols were effective at removing PEG for the purposes of obtaining accurate 14C dates on the Gribshunden materials. All of our samples that went through our pretreatment protocols returned radiocarbon dates that are within the error range of the known age of the wood. In addition to these results, some other interesting patterns in the data are taken up in the discussion below.

Comparing the dating results for untreated (N) samples taken from the outer and inner stave surfaces reveals an interesting pattern. For each of the three staves, the outer surface is older than the inner one, and the age difference is of a similar magnitude (c. 1800 and 1500 years for the oak staves, and c. 1600 years for the ash stave) (Table 2, Fig. 4). This indicates a consistent increase in PEG content towards the outer stave surface, which may be an effect of how the staves were handled in the conservation process. Alternatively, the outer wood may have been more susceptible to PEG uptake due to, for example, a larger presence of cracks or a higher degree of decomposition. Another possibility is that the outer layers of wood have been disproportionately affected by erosion bacteria, damaging their cellular structure and increasing their susceptibility to take up PEG19,20. Due to the small sample size of only three staves, this interesting result should be treated as a preliminary finding in need of future verification. The dating results reveal that PEG content was higher in ash wood compared to oak wood, but no firm conclusions regarding species differences in the ability to take up PEG can be made based on this limited dataset.

Both oak and ash are ring-porous species, meaning that the pores at the beginning of the growth season are many times larger than the pores produced at the end of the growth season. Therefore, the amount of pores is much larger in a slow-growing tree than a fast-growing one, which in turn will determine how much PEG the wood can take up. Since the number of rings, and consequently pores, is dependent on the growth conditions during the lifetime of the individual tree, the porosity can differ greatly within the species, complicating any comparisons of average porosity between species. Differences in PEG uptake between samples based on growth conditions would be in addition to PEG uptake differences based on the type of wood and the degree of decomposition that occurred after deposition in an archeological context.

As noted above, one sample (LuS-20463) returned a slightly older date after the W + OS + ABA + CE treatment compared to most samples, which returned younger dates compared to the W + OS + ABA protocol. Using 95% confidence intervals, five samples are in the order of 2-56 years younger, and LuS-20463 is around 2 years older. Therefore, it is possible that this observed pattern is not real and instead due to chance. Nonetheless, it is an intriguing pattern that deserves some discussion. The older age for LuS-20463 could be interpreted to reflect tiny amounts of PEG remaining in some samples after W + OS + ABA treatment and subsequent elimination of these during cellulose extraction. Alternatively, this may reflect the removal of C-containing compounds other than PEG that were either naturally occurring in the wood, introduced during the manufacturing and use of the barrels, formed during decomposition at the seabed, or introduced after excavation of the shipwreck. However, because the observed shift in 14C ages after cellulose extraction is so small and not entirely systematic, it should be interpreted with caution. Both the W + OS + ABA and W + OS + ABA + CE treatments removed all levels of PEG detectable by MALDI-MS, suggesting that in many cases, W + OS + ABA may be a sufficient treatment protocol for the removal of PEG-400 under conditions similar to those tested in this study.

There are several caveats with our study. We selected samples from the Gribshunden shipwreck for our experiment as it is exceptionally well-dated through both historical documentation and dendrochronological analysis. By comparing our 14C results to the known dates of the Gribshunden materials, we could be certain of the effectiveness of our pretreatment protocols. The use of Gribshunden, however, means that our study has certain limitations based on the age and character of our sample material. While our results show promising results for Late Medieval material, further experiments are needed to verify if the same procedures work for materials dating to other time periods. As the efficiency of chemical conservation can depend on the state of cellulose degradation21, it is possible that the process used in this paper would be less effective on older and more degraded woods. A recent study by Manning and colleagues, which involved a warm water bath followed by cellulose extraction, showed similar results for Roman material from the 3rd century BCE, confirmed through comparison to dendrochronological dates22,23. It would be productive for future studies of older materials to experimentally control for the presence of PEG in treated samples using the MALDI-MS analysis described above.

Another potential limitation of our study concerns the types of wood analysed and the type of PEG treatment used by the conservators. Our experiment used wood from both oak (Quercus robur) and ash (Fraxinus excelsior), showing no difference between the results for these two species. Other wood types, however, should also be tested. The exact protocol used for PEG impregnation could also potentially impact the results. The samples tested in this experiment were treated using PEG 400 at a 10% solution for at least 3 months. PEG with higher molecular weights (e.g., PEG 1500, 2000, 4000), or for longer durations, may be harder to remove using the protocols described here. Further testing on high molecular-weight PEGs is needed to determine the best methods for dating wooden artifacts conserved under different conditions. Finally, while our results show positive results for the removal of PEG, we have not tested protocols for the removal of other types of conservation agents, such as linseed oil, sugars, sodium benzoate, or alum. Further protocols need to be developed to assist with the dating of materials treated with these other conservation agents.

Our experimental evaluation demonstrated that a sequential pretreatment protocol combining a Milli-Q water and organic solvent wash with acid-base-acid treatment and cellulose extraction can effectively be used to remove PEG-400 from conserved wood. This was confirmed using both MALDI-MS analysis for the presence of PEG in treated samples as well as through 14C dating and comparison to historically and dendrochronologically determined dates. In all cases, water and organic solvent wash plus ABA treatment were sufficient to remove detectable amounts of PEG-400 for the purposes of 14C dating. This suggests that these relatively simple steps can be followed to accurately date wooden objects that have been conserved using low molecular-weight PEG. While this study demonstrates the effectiveness of pretreatment protocols for the removal of PEG-400 from medieval wood, further studies are needed to determine if the same protocols can be successfully applied to older samples and different PEG types. Nonetheless, the success of the methods described here for the material from the Gribshunden shipwreck suggests that similar approaches could be applied to other excavation contexts. We hope that this experimental demonstration of the effectiveness of this pretreatment protocol can serve as a first step towards the dating of important cultural heritage objects in museum collections throughout the world.