Introduction

Radiocarbon 14C is produced in the Earth’s atmosphere by nuclear interactions with galactic cosmic rays, most of which are charged particles. The flux of cosmic rays is modulated by the solar magnetic activity and the geomagnetic field. Radiocarbon oxidizes in the atmosphere to form 14CO2 and is taken up by trees as a part of the global carbon cycle. As 14C is a radioisotope with a half-life of 5,730 years, the 14C content in tree rings provides a record of cosmic ray intensity and solar activity over a few tens of millennia.

While the solar magnetic activity and the geomagnetic field modulate the cosmic ray background, high-energy phenomena, such as gamma-ray events and large solar proton events (SPEs), can produce a large number of cosmic rays all at once. Then, if such events have occurred in the past, the 14C content in tree rings is possible to record the rapid increase due to the events.

Recently, we found a rapid increase in 14C content within 1 year from AD 774 to 775 (ref. 1). The possible cause of the event is a supernova or a large SPE. However, neither a local supernova nor a large SPE is likely responsible, because of a lack of record of corresponding supernova or supernova remnant and too high energy for SPE1 (ref.1). There have been some attempts to specify the cause of the AD 775 event. Melott and Thomas2 have reexamined the flare energy of the AD 775 event by assuming a directional flare with opening angle of 24°, and concluded that implied energy of SPE is reduced to ~1033 erg. This is about 1/100 energy of our calculation1, which presumes that flare particles propagated isotropically. Then SPE appears to be the possible cause of the AD 775 event2. Some studies also show the SPE origin3,4. On the other hand, although we considered only the normal supernova origin1, Hambaryan and Neuhauser5 claim that a short gamma-ray burst (GRB) (<2 s) can explain the AD 775 event. Although the cause of this event can be explained by a large SPE2,3,4 or a short GRB5, a more detailed discussion of the cause is difficult because the rate of occurrence of such rapid 14C events remains unknown. We need to know whether more events similar to the AD 775 event exist in the past 14C record. Although the 14C increment around AD 775 is the largest class in the last three millennia from IntCal09 data (which is a data set back to about 25,000 years BP)6, there are a large number of period when the 14C content has not been measured with 1-year resolution. Therefore, it is possible that smaller increases than the AD 775 event are hidden in these unmeasured periods. These small increases will not be detected until after 1-year resolution measurements.

Here, we report the measurement of 14C content during an extended period from AD 822 to 1021 to search other 14C increase events, and found another rapid increase from AD 993 to 994. Considering the occurrence rate of the 14C increase events, the large SPE is a plausible cause of the 14C events.

Results

Measurement data

We collected a new series of biennial measurements of 14C content in the Japanese cedar tree (Cryptomeria japonica), which is the same tree as that used for the AD 775 event1, from AD 822 to 1021. In addition to these measurements, we also measured annual resolution data from AD 991 to 1001 in the same tree. The information about the sample is shown in Supplementary Table S1 and Supplementary Fig. S1. As the 14C data for AD 600 to 820 have been already measured1 (Miyake, F., Masuda, K. and Nakamura, T. manuscript in preparation), we obtained a continuous series of AD 600 to 1021. We have updated our previous series1 by measuring new data (for even years: AD 752–768, 782–788 and 794–820, and for odd years: AD 767 and 769). As data for AD 896 and 898 do not exist, we obtained values of 14C content (which is described as Δ14C (refs. 7) for those years by a linear interpolation method. As the data for overlapping years match within measurement errors, these series of measurements are reproducible and we applied weighted averaging for the same year. Figure 1 shows the consolidated Δ14C data of some series for the period of AD 600 to 1021.

Figure 1: 14C data from AD 600 to 1021.
figure 1

Diamonds show our results, which are obtained by a weighted mean of the 2–5 series. The blue curve represents the IntCal09 data set6. The errors are the resultant of error propagation. The error for a sample is a statistical error from a Poisson distribution, and the error for the standard sample is the greater one of either averaged statistical error from Poisson distribution of Δ14C for the six standard samples or the s.d. of values of 14C/12C for six standard samples.

The first half of Fig. 1 shows a grand solar minimum from AD 650 to 720 (Miyake, F., Masuda, K. and Nakamura, T. manuscript in preparation) and a rapid increase of 14C content from AD 774 to 775 (refs. 1). The second half of Fig. 1 shows a rapid increase of 9.1% in 14C content from AD 993 to 994. This is a clear increase that is outside the measurement error (5.1σ). A comparison of the AD 775 and the AD 994 events is shown in Supplementary Table S2 and Fig. 2. The shapes of the two series are very similar, that is, a rapid increase within 1 year followed by a decay owing to the carbon cycle. The scale of the AD 994 event is 0.6 times as large as the AD 775 event.

Figure 2: Comparison of the AD 775 and the AD 994 peaks.
figure 2

Squares show the AD 775 series from AD 770–800, and diamonds show the AD 994 series from AD 989–1019. The zero level of the vertical axes is shifted to be the weighted mean value of AD 770–774 for the AD 775 series and AD 989–993 for the AD 994 series. The errors are the resultant of error propagation. The error for a sample is a statistical error from a Poisson distribution, and the error for the standard sample is the greater one of either averaged statistical error from Poisson distribution of Δ14C for the six standard samples or the s.d. of values of 14C/12C for six standard samples.

Comparison with IntCal data set

Although the quasi-decadal IntCal09 data set6 shows a small 14C enhancement around AD 994 (3 permil increase from AD 980 to 995), this increase is hardly distinguishable from many other variations. To compare our results (obtained from Japanese trees) with IntCal98 (obtained from North American and European trees)8, we averaged the yearly data to obtain a series with a decadal time resolution. The result from AD 605 to 1015 is shown in Fig. 3. The two series show similar variations to each other; however, our results are 2.1 permil smaller than IntCal98 on average (this is 7.0σ significance level). This is considered as a regional effect. As Yaku-Island, where our sample trees lived, is located in southern Japan and is surrounded by ocean, its atmosphere is affected by marine low 14C level gas9.

Figure 3: Comparison of our result with IntCal98 data.
figure 3

The vertical axis represents the 14C content (in Δ14C), and the horizontal axis represents the calendar year. Open squares show the IntCal98 data and filled diamonds show the decadal average of our data (AD 600–1021).

Discussion

The 14C content measurements for the recorded ages of supernova explosions (SN1006, SN1054, SN1572, SN1604 and SN1885) (refs. 10, 11 12) and the emergence years of large solar flares (the Carrington flare (SPE1859) that occurred in AD 1859, and SPE1460 that was detected by annual 10Be data3,13,14) (refs. 12, 15, 16) have been conducted (as shown in Supplementary Fig. S2). We also measured the 14C content around SN1054 and SPE1859 using another Japanese cedar tree (Tree-C, information of this sample is shown in Supplementary Table S1). However, none of the 14C contents show rapid increases within 1 year. There is no 14C increase within 1 year during other periods, from AD 1374 to 1954 (refs. 12, 15, 16, 17). Only rapid 14C increases within 1 year in AD 775 and 994 have been found during about 1,600 years (when we have 1- or 2-year resolution 14C data).

Another cosmogenic nuclide, 10Be, in the Antarctic Dome Fuji ice core also shows increases in the flux corresponding to around AD 775 and AD 994 (refs. 18). Figure 4 shows 10Be flux data for AD 700–1100. The ages of the 10Be data are determined by matching the production rate pattern of 10Be with the 14C production18. The increasing rates are 7.2 × 103 (atoms per cm2 per year/year) from AD 770 to 785, and 6.2 × 103 (atoms per cm2 per year/year] from AD 985 to 995. The scale of the increase around AD 994 is 0.86 times as large as that around AD 775. This value is consistent with the ratio for 14C events (0.6 times larger) because 10Be data have a lower time resolution (~;10 years resolution) than that of 14C data. If the causes of two events are different, the difference between the ratios of 14C and 10Be production rates is expected. This difference is occurred by energy spectrums or particle species of the origin events. From the consistency of increasing ratio of AD 775 and 994 between 14C and 10Be, the cause of the two events must be same.

Figure 4: 10Be data from ice core of Dome Fuji in Antarctica.
figure 4

The vertical axis represents the 10Be flux, which is calculated from the snow accumuration rate, estimated by the three-point (1.5 m: ~;30 years) averaged δ18O (refs. 5). The horizontal axis represents the calendar year. Each point is corrected by a 10Be–14C correlation age model. The two arrows show the ages of AD 785 and AD 995. The 10Be flux increments of the two data is 1.08 × 105 (atoms per cm2 per year) from AD 770 to 785 and is 0.72 × 105 (atoms per cm2 per year) from AD 985 to 995.

Possible causes of these 14C events are large SPEs or cosmic gamma-ray events1. For gamma-ray events, there are supernova explosions and GRBs5. The supernova remnants corresponding to AD 775 and 994 have not been detected (http://hea-www.harvard.edu/ChandraSNR/snrcat_gal.html (Chandra Supernova Remnant Catalog))19 and historical documentation has not been found20. Therefore, a supernova origin is quite unlikely1,5. Although only the normal supernova origin was considered in Miyake et al.1, Hambaryan and Neuhauser claim that a short GRB (<2 s) can explain the AD 775 event5. In case of a short GRB, its spectral hardness is consistent with the differential production rates of 14C and 10Be, and the absence of historical records of a supernova or a supernova remnant is consistent with a short GRB5. Although they claim that the observed rate of short GRBs (one event in 3.75 × 106 years5) and that of 14C events (one 14C event in 3,000 years) are consistent within 2.6σ (refs. 4), the finding of the second 14C event makes a 14C event rate large, and the consistency between the observed rate of short GRBs and the 14C event rate becomes worse (the probability of a short GRB rate with one 14C event in 1,500 years is 0.04%). Adding to this, it is possible that the 14C event rate is larger because there are many periods without a 1-year resolution measurement of 14C content in the 3,000-year period. The actual 14C event rate should be 1/800 years (detected event/measured period with 1–2-year resolution), and the probability is 0.02%. Melott and Thomas2 also discuss the event rate of short GRB (a probablility of order 10−4 over 1,250 years); however, their rate is lower than Hambaryan and Neuhauser. Then, the short GRB is less likely to be the cause. In addition, Hambaryan and Neuhauser claim that it is possible that more short GRBs exist than those observed, which would explain the inconsistency in the event rate between the observed short GRB rate and the 14C event rate; however, additional studies are necessary to confirm this claim.

Next, we consider the SPE origin. An emergence of SPE is considered to be closely bound to solar activity, such as solar flares and CMEs. To know the solar activity during 8–10th centuries, we have examined the IntCal data set6. The variation of 14C content (which is closely related to the flux of galactic cosmic rays reaching the earth) is mainly modulated by the solar magnetic activity and the geomagnetic field. The 14C variation in the past several hundred years reflects the solar magnetic activity as indicated by the sunspot records. In particular, periods of solar inactivity known as grand solar minima can be identified as large peaks in the past 14C content record. The period from the late 13th century to the early 19th century is known as ‘little ice age’, which includes the Wolf, Spörer, Maunder and Dalton solar minima. On the other hand, the 8–10th centuries have no grand solar minimum. There have been some attempts to reconstruct the solar activity by using 14C data set21,22, and they showed higher solar activity levels during the 8–10th centuries than that during the 13–19th centuries, on average. This fact may explain why the 8–10th centuries have two rapid increases of 14C content and after the 11th century there were no such events.

Based on a reassessment of an energy spectrum of SPE, production calculations of 14C and 10Be, and a deposition model, Usoskin and Kovaltsov3 claimed that the AD 775 event can be explained by an extreme SPE that was about 50 times larger than the largest SPE in AD 1956. From this aspect, the AD 994 event is about 30 times larger than SPE 1956. The energetic level does not have a serious effect on living matter on the earth2. Also, according to Usoskin and Kovaltsov3, the occurrence rate of the SPE775 event is 10−4 per year and that of the SPE994 event is 10−3 per years. It is possible that these events occur within 200 years assuming that SPEs are mutually independent. Although they claimed that there is no apparent relationship between the occurrence of SPE and the solar activity level, or it is proposed that large SPEs are occurred more likely during grand solar minima23,24,25, we doubt these claims because the two events occurred in the non-solar minimum period.

Considering the 14C event rate and higher solar activity in the 8–10th centuries, a solar origin is a plausible cause of 14C increase events. Detection of the second 14C event indicates the possibility that a lot of smaller 14C increases are hidden in the periods when the 14C content has not been measured with a 1-year resolution. In the future, it will be necessary to conduct investigations of 14C records during additional unmeasured periods with a 1-year resolution.

Methods

Sample preparation

To measure 14C content in tree rings, we have to extract graphite from the wood samples. First, we separated each annual ring using a cutter knife. Then, we obtained cellulose, which does not move between rings after the rings are formed, by applying a chemical wash for each wood slice. The chemical wash consists of ultrasonic cleaning, acid–alkali–acid treatments, sodium chlorite treatment and neutralization process. The treated material was combusted to CO2 and purified in vacuum lines. Finally, purified CO2 is graphitized by hydrogen reduction under the catalytic influence of iron powder.

AMS measurement

We measured the 14C content in extracted graphite using an accelerator mass spectrometer (AMS) at the Center for Chronological Research26. As AMS provides a relative measurement, six standard samples (NIST SRM4990C oxalic acid, the new NBS standard) were measured in the same batch. Two blank samples were also measured to determine the background (commercial oxalic acid was obtained from Wako Pure Chemical Industries). The concentration of 14C expressed as Δ14C, which is the age- and isotopic fractionation-corrected value, was calculated according to the method by Stuiver and Polach7. The typical precision of 14C content measurements was 2.8‰.

Additional information

How to cite this article: Miyake, F. et al. Another rapid event in the carbon-14 content of tree rings. Nat. Commun. 4:1748 doi: 10.1038/ncomms2783 (2013).