The La Negra Formation (Figs. 1 A and B) comprises a thick pile of lava flows, largely of high-K basaltic andesites, with thin intercalated sediments. It seems that they erupted in an extensional environment [34] and that, together with their chemical similarity to the Puente Piedra volcanics in Peru [3] and the lavas at Bustamante Hill in Central Chile [22], suggests they were erupted in an ensialic back-arc basin rather than in an actual volcanic arc.

Paleomagnetic sampling was done near of town of Tocopilla (Fig. 1). Rogers et al. [34] proposed a mean age of 187 Ma as the best estimate of the time of emplacement of the lower La Negra volcanic sequence. It is only regrettable that these authors do not report the error on their K–Ar age. We note, however, that the sample dated by Rogers and Hawksworth [34] most probably belongs to site NE06 of our collection (lower part of La Negra sequence [1]). In total, about 280 oriented samples belonging to 32 consecutive lava flows were collected. Commonly, the outcrops extend laterally over a few tens of meters and, in these cases, we drilled typically 6–11 standard paleomagnetic cores per flow. The samples were distributed throughout each flow both horizontally and vertically in order to minimize effects of block tilting and lightning. Cores were obtained with a gasoline-powered portable drill, and then oriented with both magnetic and sun compasses.

Magnetic characteristics of typical samples selected for Thellier paleointensity measurements are summarized in Fig. 2 (upper part) and could be described as follows.

1. Selected samples carry essentially a single and stable component of magnetization, observed upon alternating field (Fig. 2A, sample NE2–011C) treatment. A generally minor secondary component, probably of viscous origin, was present but was easily removed. The median destructive fields (MDF) range mostly in the 40–60-mT interval, suggesting the existence of small pseudo-single to single magnetic domain grains as remanence carriers, which are suitable material for the Thellier paleointensity study [8].

2. Low-field continuous susceptibility measurements with temperature show the presence of a single ferrimagnetic phase with Curie point compatible with Ti-poor titanomagnetite. However, the cooling and heating curves are sometimes not perfectly reversible (Fig. 2B, sample NE2–011D) probably due to the heating in air. In addition, hysteresis measurements were carried out on all samples (not shown). However, we believe that the magnetic domain state estimation using room temperature hysteresis parameters in terms of the plot of magnetization ratio vs. coercivity ratio has no resolution for most of natural rocks [12]. Thus, these data were not used to select the most promising samples for Thellier paleointensity experiments.

In all, we selected 26 samples for the paleointensity experiments having the above-described magnetic characteristics. Remaining samples may be divided in two groups (Fig. 2, middle and lower parts) after their magnetic properties: (i) the samples yield reasonably reversible continuous susceptibility curves (Fig. 2B, sample NE24–168A) associated with the remanence dominated by strong secondary magnetization (Fig. 2A, sample NE24–171A); (ii) the k−T curves of these samples yield the evidence of two ferrimagnetic phases during heating (Fig. 2B, sample NE9–055). The lower Curie point ranges between 350–400 °C, and the highest one is about 580 °C. The cooling curve shows only a single phase, with a Curie temperature close to that of magnetite. Such irreversible k−T curves can be explained by titanomaghemite, which probably transformed into magnetite [24,30] during heating. It is possible that these samples carry chemical remanent magnetization. In addition, this behavior is related by very unstable multivectorial magnetization (new version of Fig. 2A, sample NE09–055B). It is obvious that these samples cannot be used to determine absolute geomagnetic paleointensity.

As part of their efforts to study the details of the volcanic stratigraphy and paleotectonics on Antofagasta region (northern Chile), Alva-Valdivia et al. [1] recently carried out a detailed paleomagnetic study of La Negra Formation. The reliable paleodirections were obtained for 25 sites among 32 sampled (Fig. 3). These directions are considered to be of primary origin because of the occurrence of almost antipodal normal and reversed polarities (Fig. 3). In addition, unblocking temperature spectra and relatively high coercivities point to ‘small’ pseudo-single domain magnetic structure grains as responsible for remanent magnetization. The secondary magnetizations were successfully removed in most cases using the alternating-field demagnetization technique.

The mean paleodirection obtained in this study is $\mathrm{I}=-30.7,\mathrm{D}=11.3°,\mathrm{k}=18,{\alpha }_{95}=6.8°,\mathrm{N}=25$ (Fig. 3). The previous paleomagnetic study provide basically similar directions with $\mathrm{I}=-39.2°,\mathrm{D}=9.1°,\mathrm{k}=11,{\alpha }_{95}=12°,\mathrm{N}=15$ [2]. Combining both data, we obtained a well-defined mean paleomagnetic direction with $\mathrm{I}=-33.9°,\mathrm{D}=10.5°,\mathrm{k}=15,{\alpha }_{95}=6.1°,\mathrm{N}=40$. These directions are in good agreement with the expected paleodirections at about 180 Ma [35] (Fig. 3B). This indicates that studied units do not show evidence for significant tectonic rotation. Fifteen sites yield reverse polarity magnetization and 11 flows are normally magnetized. The tentative direct magnetostratigraphic correlation suggests that La Negra volcanics have been emplaced during a relatively large time span of about 5 Myr [1].

Paleointensity experiments were performed using the Thellier method in its classic form [41]. Heating was made in a vacuum better than 10⁻² mbar. All remanences were measured using both JR5A and JR6 spinner magnetometers. Paleointensity data are reported on the Arai-Nagata [23] plot in Fig. 4 and results are given in Table 1. We accepted only determinations: (1) obtained from at least 5 NRM-TRM points corresponding to a NRM fraction larger than about 1/3 (Table 1), (2) yielding quality factor [6] generally above 5, (3) with positive ‘pTRM’ checks i.e. the deviation of ‘pTRM’ checks were less than 15% and (4) with reasonably linear Zijderveld diagrams obtained from the paleointensity experiments. No significant deviation of NRM remaining directions towards the direction of applied laboratory field was observed. To better illustrate this point, we calculated the ratio of potential CRM(T) to the magnitude of NRM(T) for each double heating step in the direction of the laboratory field during heating at T [9]. The values of angle γ – the angle between the direction on characteristic remanent magnetization (ChRM) obtained during the demagnetization in zero field and that of composite magnetization (equal to NRM(T) if CRM(T) is zero) obtained from the orthogonal plots derived from the Thellier paleointensity experiments – are reported in Table 1. For accepted determinations γ values are all <10.6°, which attests that no significant CRM was acquitted during laboratory heating [9]. This approach is probably more restrictive than simple calculation of angle between the characteristic directions determined by Thellier experiments and by thermal demagnetization in zero field.

Only 11 samples, coming from four individual lava flows, yielded acceptable paleointensity estimates. The almost 95% failure rate that we find in our study is not exceptional for a Thellier paleointensity study of old volcanic rocks, if correct pre-selection of suitable samples and strict analysis of the obtained data are made. The Thellier and Thellier [41] method of geomagnetic absolute intensity determination, which is considered the most reliable one, imposes many restrictions on the choice of samples that can be used for a successful determination [4,5,19,21,27,28].

Although our results are not numerous, some credit should be given because of good technical quality determination, attested by the reasonably high Coe et al.'s quality factors [6]. For the accepted determinations, the NRM fraction f used for determination ranges between 0.30 to 0.75 and the quality factor q varies from 4.4 to 10.7 (generally more than 5). The mean virtual dipole moments (VDMs) are ranging from 3.7±0.9 to $7.1\pm 0.5({10}^{22}$ A m²). This corresponds to a mean value of (5.0±1.8)×10²² A m², which is in agreement with other comparable quality paleointensity determination for Middle Jurassic [20,36]. The site NE20, which is represented only by single determination, was discarded to calculate mean VDM.

La Negra mean VDM is shown in Fig. 5 together with 16 other selected mean VDMs and VADMs (virtual axial dipole moment) for the period 5–200 Ma. We selected data using quite modest criteria [33] demanding (a) the mean based on more than nine successful determinations from at least three cooling units, (b) no transitional data and (c) paleointensity estimates obtained with Thellier method with pTRM checks. The perhaps most striking observation from inspecting Fig. 5 is that, in spite of the many published paleointensity studies, there exists only 17 paleomagnetic dipole moment estimates in the 5–200-Ma period that fulfill the above mentioned modest criteria. Among them, six are from submarine basaltic glass, one from single plagioclase crystal and remaining VDMs come from whole-rock samples. The excellent technical quality of paleointensity data obtained from basaltic glasses, first underlined by Tauxe and Pick [39,40] and later by Juarez et al. [17,18] is not by itself a proof of the geomagnetic validity of the paleostrength found. Goguitchaichvili et al. [10] showed that the magnetite found in this kind of material may be not of magmatic origin, but crystallizes at rather low temperatures as a result of either glass demixtion or alteration. In such a case, the Thellier paleointensity results would underestimate the actual paleofield strength. More data are needed before the intriguing differences between different materials can be put into a geomagnetic context.

Given the large dispersion and the very poor distribution of reliable absolute intensity data, it is hard to draw any firm conclusions regarding the time evolution of the geomagnetic field. As showed by Guyodo and Valet [15], a very large number of records of absolute intensity, reinforced by numerous and precise ages, are necessary to provide a reasonable picture of the field variations trough time.