Introduction

The cosmogenic nuclide 14C is produced in the Earth’s atmosphere through neutrons captured by nitrogen nuclei. These neutrons are generated by cosmic rays interacting with the atmosphere. Through the global carbon cycle, some of 14CO2 produced in the atmosphere can be retained in annual tree rings1,2,3. Therefore, 14C concentrations in annual tree rings indicate the intensity of cosmic rays. The flux of cosmic rays reaching the Earth is modulated by time variations of geomagnetic and heliomagnetic fields. The international radiocarbon calibration curve IntCal13 contains tree-ring records of 14C data with a typical five-year resolution extending to 13,900 years before the present4. The IntCal13 curve shows variations due to solar and geomagnetic activities on a decadal to millennial time scale. However, there are only a few annual 14C data measured from tree rings. So some rapid 14C increases caused by cosmic-ray events may be hidden in the IntCal13.

Interestingly, a large amount of cosmic rays can be generated on a short time scale by high-energy phenomena, such as supernovae (SNe)5 and large solar proton events (SPEs)6,7,8. Meanwhile, the energy deposited in gamma-rays of SNe1 and short gamma-ray bursts (GRBs)9 can also cause a rapid 14C increase. Accordingly, rapid increases of 14C content in tree rings are valuable tools to explore high-energy phenomena occurred in ancient times.

Recently, two events of rapidly increasing 14C content occurring in ad 774–775 and ad 993–994 were found using Japanese trees6, 10. More recently, a sharp increase of 14C content about 10‰ within about 3 years was found at bc 66011. The ad 774–775 event was confirmed by other tree samples in different places8, 12, 13, indicating that this event is worldwide. Several possible causes for these events have been proposed. Supernova remnants occurring at ad 774, ad 993, and bc 660 have not been observed and historical record of SNe has not been found. So, the three reported events are unlikely caused by SNe11, 14, 15. The other possible explanations of the three events are short GRBs14, 16 or large SPEs6, 8, 11, 12. The 10Be measurements in ice cores from Antarctica, Greenland, and Arctic also show a spike around ad 77517, 18, which indicates that a large SPE is the most likely explanation. But whether short GRBs could generate substantial increase in 10Be remains unclear14, 16. Meanwhile, the local event rate of short GRBs is much smaller than the rate of 14C increase events. Other attempts have been made to search for rapid 14C increases. For example, the 14C content has been measured in bristlecone pine tree-ring samples in bc 2479–2455, bc 4055–4031, bc 4465–4441, and bc 4689–468119. But no large 14C increases during these periods are found. It is possible that there were other rapid 14C increases in the past, undetected due to the lack of annual 14C measurements. In order to find more rapid increases in 14C data at annual resolution, we select the periods during which the 14C value increases significantly in the IntCal13 data. There are two intervals where the increase rate is >0.6‰ per year between bc 3380 and bc 3370. It is possible that larger annual changes hide in the averaged five-year data.

Here, we report the measurement of 14C content for an ancient buried tree in China during the period bc 3388–3358 to search 14C increase events, and find a rapid increase from bc 3372 to bc 3371. Considering the occurrence rate of the rapid 14C increase events, the 14C event could originate from a large SPE.

Results

Wood sample

We use a Chinese wingnut (Pterocarya stenoptera) tree, which was found in the city of Yichang, Hubei Province (30°31′N, 111°35′E), China. The sample of Chinese wingnut is housed in the Yichang museum. The tree is shown in Supplementary Fig. 1. This type of wood was formed in the following way. Living trees were buried in river bed or low-lying place by an earthquake, floods, or debris flows. Then, after thousands of years of carbonization process, this type of wood would be formed. The carbonization process is the early stage of the coalification process20, which can be taken as the slow hydrothermal carbonization process21, and the annual rings are preserved. The buried wood was first introduced to the West by Ernest Henry Wilson in 191322. It is regarded as priceless raw material for carving. It also has substantial artistic and scientific research values, such as revealing forest information, for studying paleoclimate, and speculating on natural disasters.

Unlike very aging trees, the ancient buried woods are relatively common in nature. The period of buried woods spans a long era back. On the other hand, unlike coal, the buried woods still contain the structure of the trees. Therefore, the ancient buried woods are good samples of the carbon abundance research on past epochs up to tens of thousand years ago, as well as on other plant archeology. For example, the use of the sample may extend the data of IntCal13.

The tree was dated with tree-ring records using standard dendrochronology. We used the master chronology of tree-ring widths from California (https://www.ncdc.noaa.gov/data-access/paleoclimatology-data). The program dpLR is used to perform the dendrochronology23. We found that the correlation value with the master is 0.525. The possible age error of the wood sample is about 2 years. Considering the age error of 14C dating is about 20 years, the age from dendrochronology is consistent with that from 14C dating. We separated annual rings carefully using a knife. The cellulose samples are prepared by standard chemical cleaning methods. The tree rings are measured using the Accelerator Mass Spectrometry (AMS) method at the Beta Analytic radiocarbon dating laboratory (http://www.radiocarbon.com/). In order to cross-check our results, we also measured another sheet of wood from the same tree (as a different sample) at the Institute of Accelerator Analysis laboratory (IAA) (http://www.iaa-ams.co.jp/indexen.html).

Measurement data

In general, AMS measures the fraction of modern carbon F, δ 13C, and Δ14C. The definitions of these values can be found in ref. 24. First, we measure the 14C content between bc 3388 and bc 3358 every 5 years. Then, the yearly measurements of 14C content from bc 3379 to bc 3365 are performed, because the 14C increase rate is >0.6‰ per year in this period. Figure 1 shows the variation of 14C content of the tree rings for the period bc 3388–3358. The filled circles are measured results from the Beta Analytic radiocarbon dating laboratory and the open circles are measured from the Institute of Accelerator Analysis laboratory. The two series of data are consistent with each other (Tables 1 and 2). So the measurement results are reproducible. We found an increase of 14C content of about 9.4‰ from bc 3372 to bc 3371. After the increase, a gradual decrease over several years due to the carbon cycle is observed. The significance of this increase with respect to the measurement errors is 5.2σ. The profile of this 14C event shows a rapid increase within about 1 year followed by a decay due to the carbon cycle, which is similar to the ad 774–775 event. In order to estimate 14C production required for this event, we use the four-box carbon cycle model to fit the Δ14C data. The four reservoirs are troposphere, stratosphere, biosphere, and surface ocean water25. The transfer coefficients of carbon from one reservoir to another used in this work are the same as Miyake et al.6. We assume all the 14C is injected into troposphere instantaneously. The best-fitting result is shown as the solid line in Fig. 1. The best fit by the weighted least-squares method yields a net 14C production of Q = (7.2 ± 1.2) × 107 atoms cm−2. According to the calculation of Usoskin et al.8, the 14C production for the ad 774–775 event is (1.3 ± 0.2) × 108 atoms cm−2. Therefore, the intensity of this event is about 0.6 times as large as the ad 774–775 event. In order to compare our results with IntCal134, we average the annual data to obtain a series with five-year resolution. The result is shown in Fig. 2. Considering the measurement errors, the two sets of data are consistent with each other. We also compare our data with the original tree-ring data26, 27 of IntCal13 in Fig. 3. Interestingly, our measured results well agree with the original data of IntCal13.

Fig. 1
figure 1

Measured 14C content. Measured results of Δ14C for the tree rings using the AMS method at the Beta Analytic radiocarbon dating laboratory (filled circles) and the Institute of Accelerator Analysis laboratory (open circles). The typical error of a single measurement is about 2.5‰ for filled circles and 2.0‰ for open circles. In order to obtain a smaller error for this 14C increase event, several measurements for bc 3371 and bc 3372 are performed. The solid line is the best fit for filled circles using the four-box carbon cycle model with a net 14C production of Q = (7.2 ± 1.2) × 107 atoms cm−2. Uncertainties (s.d.) are based on error propagation including measurement errors of the fraction of modern carbon F

Full size image
Table 1 Measured results in the Beta Analytic laboratory
Full size table
Table 2 Measured results in the Institute of Accelerator Analysis laboratory
Full size table
Fig. 2
figure 2

Comparison with IntCal13 data. Comparison of our five-year average of Δ14C data measured in the Beta Analytic laboratory (open squares) with the IntCal13 data (filled squares)4. They are generally consistent with each other considering the measurement errors. Uncertainties (s.d.) of our data are based on error propagation including measurement errors of the fraction of modern carbon F. The errors of filled squares are adopted from IntCal134

Full size image
Fig. 3
figure 3

Comparison with IntCal13 original data. Comparison of our data measured at the Beta laboratory (filled dots) with the IntCal13 original data (The UW Quaternary Isotope Laboratory (QL, open squares)26, The Queen’s University of Belfast (UB, open dots)27). For the three Δ14C value measured in the same year, they are consistent with each other within measurement errors. The value of Δ14C at bc 3385 from QL is between the nearby two points measured in Beta laboratory. Therefore, our measured results well agree with the IntCal13 original data considering the measurement errors. Uncertainties (s.d.) are based on error propagation including measurement errors of the fraction of modern carbon F

Full size image

Discussion

The rapid 14C increase around bc 3372 must be caused by cosmic high-energy phenomena. The solar cycle cannot produce this large increase. There are several plausible origins for this event.

GRBs are the most powerful electromagnetic explosions in the Universe28, 29. According to their duration T, they can be divided into long (T ≥ 2 s) and short (T < 2 s) GRBs. Because the intensity of this event is less than that of ad 774–775 event, the energy of a typical short GRB located at a few kpc can provide necessary energy14 for this event. The previous three 14C events may not be caused by short GRBs17. So, if a short GRB causes this event, it implies that one short GRB explodes in our galaxy about 5,000 years. But the local rate of short GRBs pointed to the Earth is ~10−5 yr−1 7, 15. So, the short GRB hypothesis is largely ruled out.

SNe are also powerful explosions with high-energy emissions. For a supernova, the 14C increase is attributed to both high-energy photons and cosmic rays, but only the high-energy photons would be abrupt. Previous work has shown that a rapid 14C increase of 6‰ occurred 3 years after the SN 1006 explosion1. However, this result is challenged by a recent study30. The 14C increase event reported in this paper occurred about 5,300 years ago, at a time from which there is no human historical record. Based on the above calculation, the gamma-ray energy required for this event in the atmosphere is about 1024 erg. The typical total energy of a supernova is 1051 erg. If a fraction of its total energy, \(\eta _\gamma \simeq 1\%\), radiates in gamma-rays, then the supernova must be closer than 326 pc. From the Chandra Catalog of Galactic Supernova Remnants (http://hea-www.cfa.harvard.edu/ChandraSNR/snrcat_gal.html), we found five supernova remnants with distances closer than 400 pc. The possible ages for these five supernova remnants are: t = 391 kyr for G006.4 + 04.931, t = 4.4 × 109 yr for G014.7 + 09.132, t = 3.1 × 106 yr for G047.3–03.833, t = 340 kyr for Geminga34, and t = 2,000–13,000 yr for G266.2–1.2 (Vela Jr.)35. Indeed, nearby SNe are rare within the last 300 kyr36. Interestingly, the Vela Jr.37 locates at hundreds of parsecs and its age is 2,000–13,000 years35. From this point of view, a supernova origin of the bc 3372–3371 event appears to be plausible. Unfortunately, η γ is much smaller than 1%38, so a supernova origin for this event becomes highly implausible.

The most probable origin is a large SPE. Usually, SPEs are associated with solar flares and coronal mass ejections39. Due to the uncertainty of the SPE energy spectrum, the estimated fluence of SPE caused the ad 774–775 event varies by as much as two orders of magnitude8, 40. Based on more realistic models, Usoskin et al.8 found that the ad 774–775 event could be explained by a large SPE, which was about 50 times larger than the SPE in ad 19568. So, the SPE associated the bc 3372–3371 event is about 30 times larger than the SPE in ad 1956. According to calculations by Usoskin et al.8, the ad 774–775 event required a, SPE fluence (>30 MeV) of 4.5 × 1010 cm−2. Because this event is about 0.6 times as large as the ad 774–775 event, this event should correspond to an SPE fluence about 2.7 × 1010 cm−2, if the same spectrum is assumed. The SPE fluence above 30 MeV of Carrington event is 1.9 × 1010 cm−2 41, which is about 0.7 times as large as that of bc 3372–3371 event. So, rapid increase of radioisotopes should be detectable around ad 1859. However, neither 14C nor 10Be peaks were found around ad 185910, which may be attributed to the different spectra of SPEs. If the bc 3372–3371 event is of solar origin, the associated SPE must be extremely powerful. In modern society, such extreme events would damage electronic and power systems42, deplete atmospheric ozone43, and possibly affect the weather44. Based on the four events, the probability of large SPEs is about 7.4 × 10−4 yr−1. If we assume the energy of an SPE is comparable to that of an X-ray flare, the occurrence frequency of large SPEs is consistent with the frequency of superflares on solar-type stars45.

In conclusion, we find a rapid increase of about 9‰ of 14C content in buried tree rings from BC 3372 to 3371. Whether this event is worldwide is unknown. Therefore, measuring the 14C content of trees in other places around this period is important. The most likely origin of this event is a large SPE. In the future, the measurements of radiocarbon concentration in tree rings are important for studying the large SPEs and cosmic gamma-ray events.

Methods

AMS measurement at the Beta Analytic laboratory

Cellulose in tree rings is extracted in the following steps: (1) washing with distilled water; (2) soaking in HCl, NaOH, and HCl solutions; (3) bleaching with hot NaClO2 and washing with boiling distilled water. Then, the material was combusted to CO2 and converted to graphite using standard procedures. The graphite powders produced are pressed into AMS targets and measured using an AMS system at Beta Analytic radiocarbon dating lab in Miami. The 14C/13C ratio of the sample is compared to known standards (OxalicI and II, National Institute of Standards and Technology standards SRM 4990B and 4990C, respectively), and the result corrected to the measured value of δ 14C made off line on a stable isotope mass spectrometer, giving a value for fraction of modern carbon.

AMS measurement at the IAA

Pretreatment: (1) Rootlets and granules were removed using tweezers. (2) The acid–alkali–acid (AAA) pretreatment process was used for eliminating carbonates and secondary organic acids. After the treatment, the sample was neutralized with ultrapure water, and dried. In the acid treatments of the AAA, the sample is treated with HCl (1 M). In the standard alkaline treatment, the sample is treated with NaOH, by gradually raising the concentration level from 0.001 to 1 M. If the alkaline concentration reaches 1 M during the treatment, the treatment is described as AAA in the table, while AAA if the concentration does not reach 1 M. (3) The sample was oxidized by heating to produce CO2 gas. (4) The produced CO2 gas was purified in a vacuum line. (5) The purified CO2 gas sample was reduced to graphite by hydrogen using iron as a catalyst. (6) The produced graphite was pressed into a target holder with a hole of 1 mm diameter for the AMS 14C dating, using a hand-press machine.

Measurement: The graphite sample was measured against a standard of Oxalic acid (HOxII) provided by the National Institute of Standards and Technology (USA), using a 14C-AMS system based on the tandem accelerator. A blank for the background check was also measured.

Data availability

The data that support the findings of this study are available from the corresponding author upon request.