1. We will test here the accuracy of this method for a realistic model of the magnetic field. The regular component is modeled with a bisymmetric spiral configuration, symmetric with respect to the galactic plane (BSS-S) [1]. To model the turbulent component, we use a Gaussian random field with zero mean and root mean square B rms =2  G [3]. For each source considered we simulate 10 CRs with energies randomly chosen between 30 and 300 EeV with an E -2 spectrum at the source and taking into account the magnification of the flux as a function of the energy for that direction. We propagate them through the magnetic field, keeping track of the arrival direction to the Earth. In order to study the effect of the energy and angular resolution we add 1º error in position and 10% error in energy. In Fig. 2 we show the source and CRs arrival directions for 100 randomly selected directions in the sky and in Fig. 3 we show F  1 for these sources. The magnitude of the deflection depends on the region of the sky where the source is located. As an example we show the results for a source located in a region with large deflection (F  1 =3.3º100 EeV): (b,l)=(-60º,220º) and another located in a region with small deflection (F  1 =0.7º100 EeV): (b,l)=(20º,55º). Reconstructing source position and magnetic field with UHECRs multiplets Galactic magnetic field and magnetic lensing The galactic magnetic field has a large scale regular component and a turbulent component, although both of them are poorly known. The regular component follows the spiral arms and has a local value of B reg ≈ 2  G. According to some models it has reversals in direction between neighboring arms. The random component has a root mean square amplitude of B rms ≈ 1-2 B reg and a typical coherence length of 100 pc. The regular component produces the dominant effect on the deflection of high energy charged particles. The deflections caused by the magnetic field can lead to lensing phenomena, with (de)magnification of the flux that modifies the energy spectrum of the source, and to the appearance of secondary images [1-3]. The magnitude of the effect depends on the arrival direction, the ratio between the energy and charge E/Z of the cosmic ray and the magnetic field model. Method Charged particles of different energies coming from the same source suffer different deflections in their way through the Galaxy and are thus observed with different arrival directions. If deflections are small, the arrival direction  of a particle with energy E is related to the source direction by: Considering coordinates  1 along the direction of deflection and  2 orthogonal to it, for high energies we have where we have kept track of the next to leading order term in 1/E. When several CRs from one source are detected, by performing a linear (neglecting the last term in Eq. 2) or a quadratic fit of the position  1 vs. 1/E of the events, the position of the source  1 and the integral of the magnetic field along the line of sight F  1 (  s ) can be reconstructed. *G. Golup, D. Harari, S. Mollerach and E. Roulet CONICET and Centro Atómico Bariloche (CNEA) - Instituto Balseiro, Argentina. References * e-mail: golupg@ib.cnea.gov.ar 1. D. Harari, S. Mollerach, E. Roulet, J. High Energy Phys. 08 (1999) 022 [astro-ph/9906309]. 2. D. Harari, S. Mollerach, E. Roulet, J. High Energy Phys. 02 (2000) 035 [astro-ph/0001084]. 3. D. Harari, S. Mollerach, E. Roulet and F. Sanchez, J. High Energy Phys. 03 (2002) 045 [astro-ph/0202362]. CONCLUSIONS o The median errors of the reconstruction of F  1 applying a linear fit are 0.25 º 100 EeV when no experimental uncertainty is introduced and 0.32 º 100 EeV when a 1 º uncertainty in the position and a 10% uncertainty in the energy are considered (i.e. in 50% of the cases the errors are smaller than these values). Furthermore, the direction of is obtained with a median error of 3.9 º and 5.8 º while the position of the source is obtained with a median error of 0.2 º and 0.5 º without and with experimental resolution and applying a linear fit. o The quadratic fit gives more accurate results than the linear fit when no measurement errors are introduced. However, for the magnitude of the experimental uncertainties considered, the linear fit is more accurate than the quadratic one when the experimental errors are taken into account. o The turbulent component of the galactic magnetic field does not have a significant effect in the reconstruction accuracy, with the exception of some sources near the galactic plane that have multiple images at higher energies than when considering only the regular component. At these energies the images appear near the principal one when comparing to the experimental uncertainties considered here and the effect on the reconstruction accuracy is not large. ABSTRACT We study the possibility to reconstruct the position of UHECR sources and some properties of the magnetic field along the line of sight towards them in the case that several events from the same source are detected. By considering a realistic model for the galactic magnetic field, including both a regular and a turbulent component, we estimate the accuracy that can be achieved in the reconstruction. We analyse the effect of the experimental energy and angular resolution on these results. Results We present the results for the 100 sources with location randomly selected: Centro Atómico Bariloche Figure 1: “Sky sheet”: directions of incoming cosmic rays in the halo that correspond to a regular grid of arrival directions at Earth, for the BSS-S magnetic field configuration with E/Z = 30 EeV. A source located in a fold will have multiple images. Figure 4: Deflection vs. 1/E for (b,l)=(-60º,220º) for (a) only a regular magnetic field and (b) regular and turbulent component plus measurement uncertainties in energy and position. Table 1: Uncertainties in F  1, in the direction of (  ) and in the position of the source (  1,  2 ) for the two examples. Figure 5: Histograms of the uncertainty in F  1,  F  1 (a-b), in the direction of F,  (c-d), and the position of the source,  1,2 (e-f), considering only a regular magnetic field. In the right (left) panels (no) measurement uncertainties have been taken into account. In (a-b) solid lines correspond to the linear fit and dashed lines to the quadratic fit. In (e-f) solid lines correspond to  1 applying a linear fit, dashed to  1 applying a quadratic fit and dotted lines to  2. (1) (2) Figure 2: 100 randomly selected sources (asterisks) and 10 events coming from these sources (circles) in an Aitoff projection of the celestial sphere in galactic coordinates. Figure 3: Distribution of F  1 for the sources considered. The inclusion of a turbulent component does not have a substantial effect on the reconstruction, the main difference comes from adding measurement uncertainties. (a)(b) (a)(b) (c)(d) (e)(f)