Luminescence chronology of Pleistocene loess deposits from Romania: testing methods of age correction for anomalous fading in alkali feldspars

Quaternary Science Reviews () 97 97 Luminescence chronology of Pleistocene loess deposits from Romania: testing methods of age correction for anomalous fading in alkali feldspars S. Balescu a, *, M. Lamothe b, N. Mercier a, S. Huot b, D. Balteanu c, A. Billard d, J. Hus e a Laboratoire des Sciences du Climat et de l Environnement, av. de la Terrasse, Domaine du CNRS, 998Gif sur Yvette, France b Universit!e duqu!ebec "a Montr!eal, D!epartement des Sciences de la Terre, CP888 suc. Centre ville HCP8 Montr!eal, Qu!ebec Canada c Institut de G!eographie, Acad!emie Roumaine, str. D. Racovita, 77 Bucarest, Romania d Laboratoire de G!eographie Physique, UMR 89, CNRS, place A. Briand, 99 Meudon, France e Centre de Physique du Globe, 7 Dourbes, Belgique Abstract The IRSL dating method is applied to silt-sized alkali feldspar grains ( mm) from Upper and Middle Pleistocene loess deposits of southeastern Romania (sites of Tuzla and Giurgiu Malu-Rosu), using both the multiple aliquot g dose technique and the SAR method. All samples show evidence of anomalous fading; the measured IRSL ages are in correct stratigraphic order but systematically underestimate the expected geological age (up to %). Three protocols of age correction for the observed fading have been tested: the correction of Huntley and Lamothe (Can. J. Earth Sci. 8 () 9), the correction of Mejdahl (Quat. Sci. Rev. 7 (988) 7) and the correction model suggested by Wintle (Quat. Sci. Rev. 9 (99) 8). The three protocols yield corrected IRSL ages that are in better agreement with the expected ages. This study demonstrates the potential of the IRSL dating method to provide chronological information on Upper and Middle Pleistocene loess deposits of Romania. r Elsevier Science Ltd. All rights reserved.. Introduction Loess sequences are one of the most extensive continental record of paleoenvironmental and paleoclimatic changes during the Quaternary. Unfortunately, the chronology of Middle Pleistocene loess still remains problematic; the luminescence dating of these old sediments is indeed complicated by the widespread observation of anomalous fading which leads to severe underestimates of age. Correction of the luminescence ages for this specific type of decay remains therefore a critical issue for sediments older than ka. This study is the first attempt to assess luminescence ages for Romanian loess deposits as there is still limited information on the absolute age. In this paper, we present results of an optical dating study carried out on the long loess paleosol sequence of Tuzla located in southeastern Romania along the Black *Corresponding author. UFR de G!eographie, Laboratoire de Pr!ehistoire et Quaternaire, Universit!e de Lille, Villeneuve d Ascq 9, France. E-mail address: sanda.balescu@wanadoo.fr (S. Balescu). Sea shore (Fig. ). This m thick profile consists of 7 paleosol complexes below the surface soil, and 7 interbedded loess horizons (Fig. ). IRSL ages measured on alkali feldspar grains are consistently far too young (by more than %) compared to expected geological ages. In the light of this severe age underestimation, three protocols of age correction for fading will be tested on these loess: the correction for anomalous fading proposed by Huntley and Lamothe (), the correction for long-term fading of Mejdahl (988, 989) and the correction model suggested by Wintle (99) and Musson and Wintle (99).. Materials The present study focuses on the Middle Pleistocene loesses of Tuzla deposited during the penultimate and antepenultimate glacial epochs (i.e. OIS and 8, respectively). The stratigraphical scheme proposed by Conea (99) for the loess complex in Romania is mainly based on geomorphological analysis of the Quaternary terrace, 77-79//$ - see front matter r Elsevier Science Ltd. All rights reserved. doi:./s77-79()-8

98 S. Balescu et al. / Quaternary Science Reviews () 97 97 UKRAINE MOLDAVIA HUNGARY Loess and loesslike materials Mountain region Sampling site YUGOSLAVIA km Bucarest Mostistea Koriten GIURGIU BULGARIA Danube BLACK SEA Constanta TUZLA Fig.. Location of the sampling sites. Extension of loess and loess-like materials in Romania modified after Conea (99). Ka 7 9 788 O.I. Stage 7 8 Depth (m) 7 8 9 9 TUZLA TUZ TUZ TUZ 7 B/M TUZ7 8 and on the lithology and number of loess/paleosol horizons accumulated on these terraces. An absolute chronology for these loess is still missing yet, but a correlation with the marine d 8 O records has been recently suggested, based on magnetic analysis of the loess paleosol profile of Mostistea (Fig., in the southeastern Danube plain, SE of Bucarest; Panaiotu et al., ). The magnetic susceptibility variations in the Mostistea profile are in close correspondence with both the marine d 8 O records and the susceptibility variations observed in the Bulgarian loess section of Koriten (NE Bulgaria, Fig. ) where the S L S L S L S L S L S L S L 7 S 7 Fig.. Loess section at Tuzla. Soil of chernozem type Brown-red soil Loess Luminescence sample Bruhnes Matuyama paleomagnetic boundary has been identified within the loess horizon L7 (equivalent to L7 at Tuzla). These results have been used to support the assignment of paleosol layers S, S and S to OIS, 7 and 9, respectively. Samples TUZ and TUZ were collected within the loess horizons L and L (Fig. ). For the sake of comparison, optical dating has been extended to: () Upper Pleistocene loess deposits including: (a) sample TUZ from the uppermost loess horizon L of Tuzla (Fig. ) and (b) sample MALU collected cm below the Upper Palaeolithic occupation level of Giurgiu-Malu Rosu (Fig. ) which provides an independent age control ( C dating of charcoal, Alexandrescu, 99 998). () Infinitely old feldspars from the lowermost loess horizon L7 of Tuzla (sample TUZ7, Fig. ) supposedly assigned to OIS on paleomagnetic evidence (pers. comm., J. Hus, ). In this study, all IRSL measurements are performed on silt-sized alkali feldspar grains ( mm). These were extracted by dry sieving and subsequently treated with HCl (%) and H O to remove calcium carbonate and organic matter. The heavy liquid isolation of alkali feldspars (density o.8 s.g.) was achieved by using sodium polytungstate (Mejdahl, 98). X-ray fluorescence analysis performed on the separated alkali feldspar grains has shown that these are mainly composed of albite (B%), microcline (B7%), orthoclase (B7%) but also contain a small amount of sanidine (B%). Comparative GLSL measurements have also been made on similar silt-sized quartz grains extracted from the Upper Pleistocene loess.

S. Balescu et al. / Quaternary Science Reviews () 97 97 99. Methods TUZ.. ED determination For the equivalent dose determinations on feldspar grains (Table ) we used both: the multiple aliquot additive g dose (MA) method on multigrain aliquots ( mg, aliquots) (Fig. a and b) and the singlealiquot regenerative-b dose (SAR) method on very small aliquots (about grains per aliquot, aliquots) (Fig. a and b) using the protocol of Wallinga et al. () without however any cut heat, based on the observation that the phosphorescence related to the test dose is negligible compared with the IRSL signal (Tx). For the samples investigated, the preheat treatment does not seem to be a critical factor since when the same C for s preheat treatment is applied for both dose and test dose measurements (as suggested by Lamothe et al., ), we obtain similar EDs for TUZ (77Gy without cut heat; 7 Gy with the C preheat). ED determination has also been performed on quartz grains using the SAR protocol of Murray and Wintle (). IRSL measurements for the MA method were performed on the automated Daybreak TL reader (at Montr!eal); the SAR measurements on feldspars and quartz were performed on a Ris^ DA reader (at Gif sur Yvette) using an infrared laser diode and blue emitting diodes (LEDs), respectively. Luminescence signals were detected using a U filter (quartz) or a blue transmitting Corning 7 9/Schott BG9 filter combination (feldspars). Laboratory irradiations of the multiple and single aliquots were performed with a Co g source (.7 Gy/min, Montr!eal) and a 9 Sr b source (9 Gy/min, Gif sur Yvette), respectively... Dose rate determination The concentrations of U, Th and K were measured by neutron activation analysis and high-resolution gamma Normalised IRSL day days days ED day = ± Gy ED days = ± Gy ED days = ± Gy - (a) Normalised IRSL Gamma dose (Gy) TUZ Gamma dose (Gy) days days days ED days = ± 9 Gy ED days = ± Gy ED days = ± Gy - (b) Fig.. Comparative IRSL additive growth curves obtained for different storage times after laboratory irradiation: (a) sample TUZ and (b) sample TUZ. Samples have been preheated at C for 8 h. s shine normalized to natural. Table Equivalent dose estimates Sample Method Mineral ED (prompt) (Gy) ED (delayed ) (Gy) ED (delayed ) (Gy) MALU IRSL-SAR Alkali-feld 7 (n=) GLSL-SAR Quartz 87 (n=) TUZ IRSL-MA Alkali-feld 7 ( day) 97 ( days) 7 ( days) IRSL-SAR Alkali-feld 77 (n=) GLSL-SAR Quartz 7 (n=) TUZ IRSL-MA Alkali-feld 9877 ( days) 7 ( days) 7 ( days) IRSL-SAR Alkali-feld 97 (n=) TUZ IRSL-MA Alkali-feld 87 ( days) 77 ( days) 7787 ( days) IRSL-SAR Alkali-feld 7 (n=) TUZ7 IRSL-MA Alkali-Feld 78787 ( day) 778 ( days)

97 S. Balescu et al. / Quaternary Science Reviews () 97 97 Corrected IRSL (a) Corrected IRSL (b) LN/TN ED 8 7 LN/TN Regenerative dose (Gy) ED TUZ Regenerative dose (Gy) TUZ Fig.. Equivalent dose measured using the SAR procedure: (a) sample TUZ and (b) sample TUZ. IRSL was measured for s at C, using a preheat of C for s. Sensitivity was monitored using the response to a test dose without cut heat. After measurement of the natural signal, increasing regenerative beta doses were applied; this was followed by a zero-dose and a repeat point corresponding to the second regenerative dose. spectrometry. These are in good agreement and demonstrate that all samples are in radioactive equilibrium. The dose rate estimates are reported in Table.. Results and discussion.. Luminescence age estimates The ED values obtained on feldspar using the SAR protocol show good internal consistency for each studied loess sample (s ranging between % and 8%, Table, n ¼ ). This is a strong indication that the feldspar grains have been efficiently bleached at the time of deposition. The ED obtained with the SAR protocol are in good agreement with those obtained by the MA method (see Table ). As shown in Table, the IRSL ages are stratigraphically consistent but increasingly underestimated with depth, up to %. It is worth pointing out that a similar age underestimation has been found in several other loess sections of Central Europe (Musson and Wintle, 99), and in particular in Hungary (Frechen et al., 997), when using the IRSL signal of polymineral fine grains ( mm). As shown in Table, IRSL ages are systematically lower than the GLSL age from quartz using the SAR method and the radiocarbon age from charcoal. Moreover, all samples from Tuzla show a significant decrease of induced luminescence with time as expected for anomalous fading (Fig. a and b). These results suggest therefore that fading is the most probable cause of our systematic age underestimation. Table Dose rate estimates Sample Mineral Ext. alpha (Gy/ka) a Int. beta (Gy/ka) b Ext. beta (Gy/ka) c Ext. gamma+cosmic (Gy/ka) d Total dose rate (Gy/ka) MALU Alkali-feld.7..7..7..7..97.7 Quartz.7..7..7..7.7 TUZ Alkali-feld.7..7..87..7..7.7 Quartz.7..87..7..7. TUZ Alkali-feld.87..7..77..7..7.7 TUZ Alkali-feld.7..7..7..7..7.7 TUZ7 Alkali-feld.97.9.7..7..887..7. a External alpha dose rate. b Internal beta dose rate from inherent K in alkali feldspar grains. c External beta dose rate corrected for beta attenuation and water attenuation. d Gamma dose rate corrected for water attenuation. Mean internal K content: 7%; mean water content: 7% (total weight% H O). The dose rates were calculated using Gr.un s software.

S. Balescu et al. / Quaternary Science Reviews () 97 97 97 Table Comparison of IRSL ages without and with correction for fading, with independent age estimates Sample Apparent IRSL age estimates (prompt) (ka) Apparent IRSL age estimates (delayed) (ka) Corrected IRSL age (ka) Huntley and Lamothe () Corrected IRSL age (ka) Mejdahl (988) Corrected IRSL age (ka) Wintle (99) GLSL age estimates on quartz (ka) Expected geological age MALU 97 (SAR) 7 7 ( C cal) >.797 >.7 years TUZ 7 (MA day) 87 (MA days) 7 7 78 O I Stage 7 (SAR) 779 TUZ 97 (MA days) 79 (MA days) >7 7 77 O I Stage ( ka) TUZ 7 (MA days) 777 (MA days) >87 7 77 O I Stage 8 ( ka) TUZ7 77 (MA day) 779 (MA days) O I Stage (8 ka).. Tests of age correction for fading Three different approaches have been tentatively tested to correct for the observed fading. (a) The correction for anomalous fading of Huntley and Lamothe The correction of Huntley and Lamothe () relies on assumptions () that the observed decrease of luminescence of feldspars follows the logarithmic decay law proposed by Visocekas (979) (athermal quantum tunnelling process), and () that the logarithmic time dependence can be extrapolated to geological times (i.e. beyond ka or more). However, this correction is restricted to the low-dose linear portion of the dose response curve. In the present study, two different approaches for measuring the fading rate (i.e. the g value of Huntley and Lamothe or the luminescence loss in percent per decade, a decade being a factor of in time since irradiation) have been tested on sample TUZ using the regenerative signal. Method : After completion of the normal SAR (Li/ Ti) measurement routine on the Ris^ reader, the aliquots are given an additional regenerative dose, preheated ( C for s) and are subsequently stored at room temperature. After storage, the samples are measured again using the SAR protocol (Li/Ti, s), fully IR bleached ( s with the Ris^ reader) and then reirrradiated, preheated and stored for another period of time. No cut heat was performed. Samples have been stored successively for, and days after b irradiation. Method : Six natural aliquots, previously fully bleached under a sunlamp, are given a regenerative dose, preheated at C for s and subsequently stored. After storage, the samples are measured again using the SAR protocol (Li/Ti, s) and then reirrradiated, preheated and stored for another period of time. No bleach was performed prior to reirradiation (Auclair et al., ). The same preheat treatment ( C for s) was applied to both dose and test dose. The samples have been stored for different periods of time from. up to h. Method : natural aliquots, previously bleached overnight under a sunlamp, are given a regenerative dose, preheated at C for s, and then measured with short shines repeatedly at different times (,, h and days). Three aliquots of untreated natural feldspar grains are used to allow for variations in the sensitivity of the apparatus and signal erosion. This experiment follows method (b) of Huntley and Lamothe (). When applying method to TUZ, we obtained an average g value of.7.7% per decade (n ¼ ). Method yielded for MALU, an average g value of.7.% per decade. By contrast, Method yielded highly scattered g values for both MALU and TUZ; the estimated average g values were.7% per decade (n ¼ ; MALU) and.7.% per decade (n ¼ ; TUZ). These average g values are thus self-consistent, but the error margins on the g values remain far too high for Method. The similarity of the g values obtained on both Upper Pleistocene loess deposits from two adjacent sites is highly encouraging. Further investigation is now needed to () identify the cause of the observed scatter, () improve the precision on the measured g value, and () test the accuracy of the measured g value. Several approaches for measuring g values have been reported by Auclair et al. (). As shown in Table, the corrected SAR IRSL age of TUZ (g ¼ :7:7% per decade) is now closer to the quartz optical age. The corrected SAR IRSL age of MALU (g ¼ :7:% per decade) is in better agreement with the radiocarbon age, by contrast with the quartz optical age which appears to be too young. For the Middle Pleistocene loess deposits, the correction of Huntley and Lamothe could not be applied since

97 S. Balescu et al. / Quaternary Science Reviews () 97 97 they are not on the linear portion of the dose response curve. However, assuming that samples TUZ, TUZ and TUZ have similar geological provenance, the prompt additive IRSL ages ( days delay) of TUZ and TUZ have been tentatively corrected using the g value of TUZ. The corrected additive IRSL ages of TUZ and TUZ, 7 and 87 ka, respectively, have to be seen as minimum corrected ages since this correction is not expected to be valid for old sediments. Consequently, we further applied the fading correction procedures of Mejdahl and Wintle as follows, both relying on analysis of alkali feldspars from infinitely old deposits of similar provenance and equivalent dose rate, i.e. the lowermost loess horizon L7 of Tuzla (sample TUZ7). (b) The long-term fading correction of Mejdahl The mean lifetime for fading of the IRSL signal of TUZ7, when using the MA method, is estimated at t ¼ 77 ka, following the procedure described in Mejdahl (988, 989). Using this lifetime, we obtained for TUZ and TUZ corrected additive IRSL ages in better agreement with the geological estimates (see Table ). Also, when applied to TUZ, it yielded a corrected additive IRSL age slightly lower than the SAR IRSL corrected age. The protocol we used to derive the correction factors is fully described in Balescu et al. (997). It is applied to the delayed IRSL ages estimates. (c) The correction model of Wintle Assuming that the luminescence signal is controlled by a time-dependent signal loss that has a lifetime of t; Wintle suggested that the luminescence age curve of known geological age could be approximated by the following equation (see Fig. ): T ¼ tð e t=t Þ; Luminescence age estimate (ka) TUZ TUZ TUZ τ = ka Feldspar-prompt Feldspar-delayed Quartz MALU 8 Geological estimated age (ka) TUZ7 Fig.. IRSL age curve obtained for samples of known geological age if the IRSL signal is controlled by a time-dependent signal loss that has a lifetime of t ¼ ka. where T is the measured luminescence age, t is the geological age, t is the mean life of the luminescence signal (Wintle, 99; Musson and Wintle, 99). The lifetime t of the IRSL signal of TUZ7, when using the MA method, is estimated at 7 ka. This is similar to the upper age limit determined for the Hungarian loess by Frechen et al. (997) (i.e. 7 ka estimated on fine grains using the IRSL additive method; weeks delay after irradiation). When using a lifetime of ka, the correction model applied to the delayed IRSL ages, results in corrected additive IRSL ages similar to those obtained using Mejdahl s correction (see Table ).. Conclusions The IRSL ages measured on alkali feldspars from Upper and Middle Pleistocene loess deposits of Romania are systematically younger than expected from stratigraphical and geochronological evidence. IRSL analysis of the lowermost loess horizon of Tuzla (L7, B8 ka) suggests an upper IRSL age limit of about ka. Anomalous fading is believed to be the most probable cause of our systematic age underestimation. The correction for anomalous fading of Huntley and Lamothe () applied to SAR IRSL ages of Upper Pleistocene loess, has been successful in yielding a corrected age estimate in better agreement with independent age. But further work is clearly needed to test the accuracy and improve the precision of the measured g values. By contrast, for Middle Pleistocene loess, this correction being no longer valid, we applied the correction for long-term fading of Mejdahl (988, 989) and the correction model suggested by Wintle (99) with a lifetime of ka; they yield similar corrected additive IRSL ages that are much closer to the expected geological age of the investigated samples. At this stage of our research, the corrected IRSL ages of the Middle Pleistocene loess are to be seen as the best preliminary estimates of their depositional age. Future research will be directed towards investigation of adjacent loessic sections to further test the accuracy of these three fading correction protocols and develop a chronostratigraphic framework for the paleoclimatic history of southeastern Romania and the Black Sea region. Acknowledgements The authors are very grateful to M. Auclair (UQAM, Montr!eal) for her valuable help and for neutron activation analyses. We would like to thank M. Preda (UQAM) for X-ray analyses, J.L. Reyss (LSCE, Gif sur Yvette) for gamma spectrometry and M. Laithier

S. Balescu et al. / Quaternary Science Reviews () 97 97 97 (UQAM) for preparing the illustrations. We wish to thank the anonymous referee for helpful suggestions and comments on the manuscript. This research was carried out as part of Project 78 of the French Romanian Cooperation. The funding was provided by the Direction des Relations Internationales du CNRS (Centre National de Recherche Scientifique), the Romanian Academy of Sciences, the Laboratoire de G!eographie Physique de Meudon (UMR 89 CNRS et Universit!e Paris ) and the Institute of Geography of Bucarest. References Alexandrescu, E., 99 998. Observatii asupra industriei litice de la Giurgiu-Malu Rosu. Buletinul Muzeul Teohari Antonescu, anul II IV, Nr., pp. 7. Auclair, M., Lamothe, M., Huot, S.,. The measurement of anomalous fading for feldspar IRSL using SAR. Radiation Measurements, in press (PII:S-87()8-). Balescu, S., Lamothe, M., Lautridou, J.-P., 997. Luminescence evidence for two middle Pleistocene interglacial events at Tourville, northwestern France. Boreas, 7. Conea, A., 99. Profils de loess en Roumanie. In: Fink, J. (Ed.), La stratigraphie des loess d Europe, Bulletin de l Association fran@aise pour l!etude du Quaternaire, Suppl. INQUA, pp. 7. Frechen, M., Horvath, E., Gabris, G., 997. Geochronology of Middle and Upper Pleistocene loess sections in Hungary. Quaternary Research 8, 9. Huntley, D.J., Lamothe, M.,. Ubiquity of anomalous fading in K-feldspars and the measurement and correction for it in optical dating. Canadian Journal of Earth Sciences 8, 9. Lamothe, M., Duller, G.A.T., Huot, S., Wintle, A.G.,. Measuring a laboratory radiation dose in feldspar using SAR. First North American Luminescence Dating Workshop, Tulsa, Abstracts. Mejdahl, V., 98. Thermoluminescence dating based on feldspars. Nuclear Tracks,. Mejdahl, V., 988. Long-term stability of the TL signal in alkali feldspars. Quaternary Science Reviews 7, 7. Mejdahl, V., 989. How far back: life times estimated from studies of feldspars of infinite ages. In: Synopses from a Workshop on Long and Short Range Limits in Luminescence Dating. Occasional Publication 9, The Research Laboratory for Archaeology and the History of Art, Oxford University, Oxford. Murray, A.S., Wintle, A.G.,. Luminescence dating of quartz using an improved single-aliquot regenerative-dose protocol. Radiation Measurements, 7 7. Musson, F.M., Wintle, A.G., 99. Luminescence dating of the loess profile at Dolni Vestonice, Czech Republic. Quaternary Science Reviews,. Panaiotu, C.G., Panaiotu, E.C., Grama, A., Necula, C.,. Paleoclimatic record from a loess paleosol profile in southeastern Romania. Physics and Chemistry of the Earth (A), 89 898. Visocekas, R., 979. Miscellaneous aspects of artificial TL of calcite: emission spectra, athermal detrapping and anomalous fading. PACT, 8. Wallinga, J., Murray, A.S., Wintle, A.G.,. The single-aliquot regenerative-dose (SAR) protocol applied to coarse-grain feldspar. Radiation Measurements, 9. Wintle, A.G., 99. A review of current research on TL dating of loess. Quaternary Science Reviews 9, 8 97.