A Palaeomagnetic Study of Minoan Age Tephra from Thera

They concluded that there was a time gap of up to 50 years between the plinian phase and the later phases of the eruption. They also identified remanence directions and intensities obtained from the plinian phase with those from the (LM IA?) destruction levels in central Crete; and related the later (?) Thera eruption phase to (LM IB?) destruction in eastern Crete. It is not clear, from archaeological evidence, if destructions in central and eastern Crete were synchronous, and this magnetic study is the only quantitative evidence for a time gap between the two. However, there is convincing volcanological evidence for a short, single eruption lasting only a few days (Sparks 1985), and there is also no evidence of a time gap in the ash layer preserved in deep-sea cores. This study will attempt to reconcile the conflicting magnetic and volcanological date, by detailed study of the remanence acquisition mechanisms in each of the phases of the Minoan eruption.

INTRODUCTION 

This paper is a preliminary report on an investigation of the volcanology and palaeomagnetism of tephra from the Minoan Age eruption on Thera. This is a relatively new application of palaeomagnetism and we have obtained estimates of the emplacement temperatures of these deposits. Temperature estimates are extremely useful for elucidating volcanological sedimentation processes and distinguishing hot-flow or fall deposits from colder flash-flood or water-saturated material. The volcanological aspects of the study will be combined with these temperature determinations to help reconstruct the eruption in detail, and shed light on the origins of the more controversial Phase 3 and Phase 4 tephra.

The eruption can be separated into five successive phases (Sparks and Wilson, this volume). The first, plinian phase (Phase 1) is unambiguous in origin, but once sea water had gained access to the vent in subsequent phases, the eruption mechanisms became more complicated. Phase 2 contains base surges and contemporaneous plinian deposits. Phase 3 tephra are chaotic, poorly sorted, massive units containing abundant lithic blocks and appear to be low temperature primary pyroclastic flows. Phase 4 deposits are incontrovertibly of hot origin, although the deposition mechanism of intercalated lithic breccias is unknown. The final stage (Phase 5) of the eruption was deposition of alluvial lithic-rich breccias. Tephra from all phases from the eruption contain both pumice and ash and lithic blocks of material erupted in previous events on Thera; the relative proportions are variable within and between deposits.

PRINCIPLES OF EMPLACEMENT TEMPERATURE ESTIMATION BY PALAEOMAGNETIC METHODS

In pyroclastic deposits which have come to rest considerably above ambient temperatures, such as Phases 1 and 4, the mean emplacement temperature of the deposit can be determined by analysis of magnetization in lava clasts incorporated within the deposit. These clasts will have originally been magnetized prior to the eruption, will have been heated during their incorporation into the deposit and will have cooled to ambient temperature in their present position. The heating will have demagnetized a proportion of the original magnetization with blocking temperatures less than and up to the peak temperature reached in the deposit, and this will have been replaced by a new partial thermo-remanence (TRM) on subsequent cooling. The orientation of the original, high blocking temperature remanence will be random between clasts, reflecting the random position of the clasts themselves. The low blocking temperature remanence will, in contrast, have the same orientation in each clast, parallel to the earth's magnetic field during cooling. The temperature required to separate these two components on laboratory thermal demagnetization is a good estimate of the peak temperature of the deposit. McClelland and Druitt (1989) have described this method in detail, using earlier deposits from Thera. Pumice fragments and blocks from the same deposit will have a single direction of remanence, but previous studies on pyroclastic tephra (Hoblitt and Kellogg 1979; Kent et al. 1981; McClelland and Druitt 1989) have shown that the blocking temperatures of this remanence usually extend considerably above the emplacement temperature of the deposit, often up to the Curie temperature of the magnetic mineral carrying the magnetization.

SAMPLING AND EXPERIMENTAL PROCEDURE

At this preliminary stage of the study, we have collected oriented block samples for palaeomagnetic study from equal numbers of pumice and lithic clasts at different localities in Phases 1 to 4. The sampling sites are marked in Fig. 1; single sites in each of Phases 1, 2 and 3 have been sampled in the Thera quarry, three sites of Phase 3 and one of Phase 1 in the Oia quarry and one site in Phase 4 at Monolithos. A total of 61 lithics and 56 pumices have been studied so far. Lithic clasts were oriented by gluing rigid plastic plates (5 x 8 cm approximately) onto the uneven surface of the block, and marking the strike and dip of this plate. Pumice blocks were oriented either by the same method as the lithics or by carving a flat surface with a sharp knife and marking strike and dip directly onto the pumice surface. The size of lithic and pumice blocks collected ranged from 3.5 to 20 cm diameter. Standard (one inch diameter) cores were cut from the blocks back at the laboratory. Magnetization of the samples (NRM) was measured using either a CCL cryogenic magnetometer or a Molspin spinner magnetometer. Samples were thermally demagnetized using a furnace with a residual field of less than 5nT. Vector structure of the remanence was analysed using the line-fitting algorithm of Kent et al. 1983 (the LINEFIND program).

PALAEOMAGNETIC RESULTS

     Presentation of palaeomagnetic results:      The vector component structure of magnetization in a rock can be revealed by thermal demagnetization, where the samples are heated and cooled in a field-free space to increasing peak temperatures and the remaining remanence is measured after each temperature treatment. Heating and cooling in zero-field demagnetizes the grains which have blocking temperatures up to the peak heating field, so layers of remanence are progressively stripped away by this technique. The change of direction and intensity of remanence during thermal demagnetization is represented on a vector plot, where the vector after each treatment is projected onto the horizontal and onto a vertical plane; both projections are shown on a single diagram. Fig. 2 (a) (BMA 1-15) shows a vector plot of a single-component remanence from a pumice sample; here the demagnetization simply decreases the intensity of remanence without significantly changing its direction. One line in each projection is, therefore, traced out by all the demagnetization points joining the NRM point to the origin. Plotting the demagnetization of a two-component remanence on a vector plot results in two lines being raced out by the vector end-points in each projection (Fig. 2 (b), BMD 1-2).

     Lithics:      Temperatures of emplacement can only be determined when two components of remanence are identified on thermal demagnetization of a lithic clast. At temperatures up to the emplacement temperature (T_empl) a low blocking temperature component is demagnetized and is in the direction of the ambient field at the time of emplacement (approximately declination 360°, inclination +60°). Above T_empl the original, now randomly oriented, high blocking temperature component is removed. 67 lithic samples have been thermally demagnetized and can be classified into three types of vector structure. The first class has a well-defined low Tb overprint (360/+60) which contributes 25-50% of the total NRM and has blocking temperatures from room temperature up to 200-400^o C. This is termed a 'large overprint' in Table 1. The second class has a less well-defined overprint which contributes up to a maximum of 10% of total NRM, and has blocking temperatures only up to 200^o C. This is termed a 'small overprint' in Table 1. The third class carries a single-component remanence. Some of these lie in the same direction as the overprint and have been classified as complete remagnetizations, but five samples from Phase 3 have a single-component remanence with different directions, and these have been termed 'no overprint' in Table 1.

Over 50% of lithics from the plinian airfall (Phase 1) have a large overprint magnetization from which the emplacement temperature can be easily estimated. In Fig. 3 (a) the overprint is removed by 350^o C and a westerly, downward magnetization is isolated above this temperature. Fig. 3 (b) shows an example where the overprint is directed to the north-east rather than the north, indicating some movement of the clast after cooling. Again the overprint is removed by 350^o C. Fig. 4 shows histograms of emplacement temperature estimates for each phase, separated into estimates from samples with 'large overprints' in black, and 'small overprints' unshaded. There is no significant difference in temperature estimates from Oia quarry which is further from the vent than Thera quarry. Temperature estimates range from 150-200^o C to 300-350^o C and the distribution peak is at 250-300 ^oC. Phase 2 lithic samples have a similar distribution of 'large' and 'small' overprints to those in Phase 1. Fig. 3 (c) shows a typical 'small' overprint which is removed by 250^o C, and Fig. 3 (d) a 'large' overprint which is removed by 200^o C. Mean temperatures are somewhat lower in Phase 2 than in Phase 1 (Fig. 4).

The distribution of remanence class and temperature estimates is significantly different in lithics from Phase 3. Over 50% of the lithics only have a small overprint and 15% have no overprint. Typically, the overprint is removed below 200^o C, often in the first step (Fig. 3 (f)). 25% of samples have a 'large' overprint and the mean temperature estimated from these is significantly higher than from the samples with 'small' overprints, with a peak at 200-250° C.

Finally, 77% of the lithics from Phase 4 have large overprints (Fig. 3 (g)) and temperature estimates are the highest in the eruption sequence with a mean of 300-350^o C (Fig. 4).

Satisfactory estimates of the direction of the low temperature overprint can be made from all samples with large overprints but only from some samples with small overprints. Table 2 shows the mean overprint directions for lithics from each phase, and for all phases together. The directions from each phase are statistically indistinguishable from each other, for our present data set. The high temperature components are highly scattered, and a Fisher analysis of the directions shows that the vector sum R = 11.3 does not exceed the significance point R_o = 11.4 for N = 50 (Watson 1956) and that there is a 95% chance that the distribution is indeed random.

     Pumices:      Thermal demagnetization of pumice samples does not produce trajectories that are as well-defined as those obtained from lithic clasts; pumices from Phase 4 give the best defined trajectories. The magnetization in pumices is weaker than in lithics by an order of magnitude but experimental noise does not contribute to the scatter since remanence was measured using a cryogenic magnetometer with a sensitivity 100 times greater than the weakest sample. Pumices from Phases 1, 2 and 4 have single-component remanence; typical examples are shown in Fig. 5 (a) and (b). Maximum blocking temperatures of this remanence range from 350^o C to 580^o C. Remanence from pumices of Phase 3 is much less well behaved on thermal demagnetization. Typical vector plots are shown in Fig. 5 (c) and (d) where the trajectory looks like a 'cat's cradle' above 300^o C. Table 2 shows the mean remanence directions from pumices of each phase and for all phases. The directions from each phase are statistically indistinguishable from each other for our present data set. Remanence directions from Phase 4 are much better grouped than from the other phases, as a consequence of the better demagnetization trajectories obtained from these samples.

INTERPRETATION

Lithics with a large overprint and a well-defined, randomly oriented high blocking temperature component have undoubtedly been emplaced above ambient temperature, and the laboratory temperature required to remove the overprint is a good estimate of the maximum temperature reached by the block in situ in the deposit. This maximum temperature may vary considerably between blocks within a deposit, depending on the previous history of each clast. Some blocks may have been considerably hotter than the mean temperature of the matrix of the flow (i.e. the pumices) and will register a high temperature. Others may have been cold prior to incorporation into the deposit and will have heated up to the average pumice temperature before acquiring an overprint on cooling. Therefore, the lowest temperature estimate from a given locality is the best estimate of the mean emplacement temperature of the whole deposit (see McClelland and Druitt 1989 for further discussion).

Interpretation of small overprints with blocking temperatures below 200^o C is more uncertain. Blocking temperature is simply a measure of the difficulty of surmounting an energy barrier which holds a magnetization in a fixed orientation (i.e. the difficulty of changing the direction of a remanence), and therefore blocking temperatures are time-dependent since this is a kinetic process. Magnetic grains with blocking temperatures of 150° C over the laboratory time scale have blocking temperatures of 20° C over a time scale of a few thousand years (Pulliah et al. 1975). Considering material from this recent eruption, the direction of the earth's magnetic field has remained almost constant from the time of eruption until the present. If we consider a block with a single-component remanence oriented in a direction very different from the ambient field, then the magnetization in grains with low energy barriers to change of direction (low blocking temperatures) will gradually change to become parallel to the ambient field. This process is known as magnetic viscosity and magnetization acquired by this process is termed viscous remanent magnetization (VRM). The 'small' overprints may, therefore, be recently acquired VRM rather than TRM overprints due to heating at the time of eruption.

     Phase 1:      The presence of well-defined two-component remanences in lithics from Phase 1 is consistent with the known hot origin of the deposit, and constrains the average emplacement temperature to be 250-300° C. This temperature is not significantly different in Oia and Thera.

     Phase 2:      Again, well-defined two-component remanence demonstrates the hot nature of the deposit, although the introduction of water into the system has decreased the mean emplacement temperature to 150-250° C.

     Phase 3:      Maximum temperature estimates for lithics from Phase 3 have a wide range from zero to 250-300° C. The pumices are also much less efficiently magnetized than in the other phases. There are three possible reasons for this. (1) The matrix was cold and all temperature variation is due to the source temperature variation between blocks. (2) The temperature of the matrix was spatially variable whilst all the lithics were relatively cold; thus some blocks were partially reset whilst others surrounded by cold matrix gained a VRM component or retained a single-component magnetization now randomly oriented between blocks.(3) The temperature of the matrix and lithics is spatially variable.

     Phase 4:      Well-defined two-component remanence is again consistent with the known hot origin of this phase, and we can provide a mean emplacement temperature estimate of 300-350° C for the site at Monolithos. This is the hottest phase of the eruption, and the remanence directions from both lithics and pumices are better grouped than in the other phases.

In all phases, the maximum blocking temperature of the remanence in the pumices is considerably above the estimated emplacement temperature of the deposit. If the remanence in the pumices were purely of thermal origin, then the remanence would not have blocking temperatures which considerably exceed the emplacement temperature. There must be a component ot chemical origin (CRM) in the pumice remanence acquired during cooling in the deposit, with the same direction as the TRM component (see McClelland and Druitt 1989 for further discussion). The mechanism of the CRM acquisition is as yet unclear.

CONCLUSIONS

The TRM technique is a valuable tool for determining emplacement temperature of pyroclastic deposits. Two-component remanences in lithic clasts must be used for these estimates since pumice remanence contains a significant CRM component which has blocking temperatures considerably above the emplacement temperature of the deposit. Our emplacement temperature estimates are systematically lower than those of Downey and Tarling (1984) by several hundred degrees, and we believe that the ash in the Phase 3 deposits could have been cold during emplacement. The difference in these two studies could be explained if Downey and Tarling used maximum blocking temperature estimates from pumices; details of their methodology are not given in their paper. We have not been able to see any difference in direction between remanence in Phase 1 and the overlying phases, which was reported by Downey and Tarling. The pumices in Phases 1, 2 and 3 that we have studied do not have the capacity to record the ambient field direction at the time of deposition with the precision that Downey and Tarling have claimed. We have sampled pumice blocks with diameters greater than 3.5 cm.

This study will now be extended to cover the whole of the island to look at the geographical distribution of temperature in each phase, and to investigate the origin of the intercalated breccias in Phase 4.

Addendum

Professor Tarling commented after the presentation of this paper that he had evidence for hot emplacement of Phase 3. We would like to discuss the interpretation of his evidence.

Previous palaeomagnetic work on the Minoan tephra by Downey and Tarling did not use the same technique as we did for temperature estimation. Their estimates were based on demagnetizations of pumices; however, the technique is only valid for lithic fragments. Work on lithics was limited and included a study of two large 0.5 m blocks from Thera quarry Phase 3. They studied the variation in the direction of the NRM from the surface to the centre of the block. The first 3 cm of the section had NRM directions consistent with the local field, and had been completely overprinted in some way. Deeper inside, the NRM orientations were different. Unfortunately the samples were not thermally demagnetized to split the components which made up the NRM. As a result there is not enough information to construct a thermal profile across the blocks, and Tarling's argument that the NRM change indicates a hot matrix heating a cold block is not persuasive. Moreover, a complete resetting of the NRM would require a very high matrix temperature (500-600° C), which would reset the small blocks we measured. This high matrix temperature would probably also reset the entire block as the cooling of a 20 m thick deposit at that temperature would be slow compared to the heating of a 0.5 m block.

An alternative explanation is a chemical weathering effect producing magnetite or maghaemite in the outer rim of the block. Further work must be undertaken before any clear conclusions may be drawn.

     Phase 4:      H. Pichler and W.L. Friedrich proposed an entirely cold origin for Phase 4, the material at Monolithos being reworked Phase 3. Our temperature estimates from Phase 4 at Monolithos show that these deposits were emplaced hot and are in situ, not reworked. If the deposits had been reworked the low temperature remanence components would have been randomized. This is not the case - see Table 2.

     Phase 1:      Questions were raised about the emplacement temperatures of Phase 1. No charring of wood was observed at Akrotiri (C. Doumas) nor was significant heating of the soil (S. Limbrey). In his discussion of Asama volcano, S. Aramaki (this volume) mentioned that charring only occurred locally around the hottest fragments which came to rest on thatched roofs during the 1783 eruption. It is thus relevant to emphasize that the best estimate of emplacement temperature is thought to be the minimum temperature recorded by any lithics in that phase. Our measurements range from 200-250° C to 300-500° C. The explanation is that a range of initial temperatures is possible for clasts, but during equilibration any clasts with temperatures lower than the equilibration temperature will be heated up. Thus the minimum temperature recorded by the lithic fragments will be the best estimate of the deposit's ambient temperature. The range of temperatures is consistent with the Aramaki observation, where only a few hot fragments produced charring whilst the bulk of the material was too cold.

------------------------------------------------

-------------------------------------------------------