Figure 1 is the SXT images from Aug. 16 to Aug. 19 to show evolution of the coronal structure. Sigmoidal structure was shown on Aug. 19.
Figure 2 is a sequence of H_alpha images taken at BBSO. Filaments developed in this period and formed a sigmoidal structure on Aug. 19
The magnetograms taken by SOHO/MDI show the evolution of magnetic field in this period figure 3 . The corresponding active region is AR8668.
The alignments of magnetic field and SXT image (figure 4 ) and magnetic field and H_alpha observation (figure 5 ) are done by referring solar limb.
We use different models to reconstruct the 3D magnetic field above the active region AR8668. Under potential field model, the magnetic force lines are calcuated from the logitudinal magnetic field taken by SOHO/MDI (figure 6a ), and are viewed from different viewsight (figure 6b ). Comparisons with SXT and H_alpha observations are shown in figure 7a ,figure 7b , figure 7c , and figure 7d , respectively. The white lines represent open-up magnetic force lines. The force lines match some structures in SXT image but certainly do not match the sigmoid structure, as, theoretically, sigmoidal structure represents highly twisted magnetic tubes(Hood et al, 1992).
Linear force-free field is computed from the observed magnetogram taken by SOHO/MDI (figure 8a , figure 8b ). The comparisons with SXT and H_alpha observations are shown in figure 9a , figure 9b , figure 9c and figure 9d . It is much better to fit the SXT structure than the potential one, but still does not match well.
Based on a non-force-free-field model proposed by Low (1992), we computed the magnetic lines of force from SOHO/MDI magnetogram, considering weak electric current across magnetic field (figure 10a , figure 10b ). The comparisons with SXT and H_alpha observations are presented in figure 11a , figure 11b , figure 11c and figure 11d ), respectively.
Using the boundary element method, we are able to calculate non-linear force free field lines boundaried by vector magnetogram in the photosphere(Yan and Sakurai, 2000). Because the field of view is small and cannot cover the whole area, we fill out other parts with the LFFF force lines calculated from SOHO/MDI magnetogram (figure 12a , figure 12b ). The red lines are non-linear force free field lines and the green and white lines are LFFF force lines. These non-LFFF force lines match part of the sigmoid structure ( figure 13a ,figure 13b , figure 13c , and figure 13d ).
If we think that this helical magnetic field is imbeded in potential field, we can obtain such combined picture (figure 14 ). The green and yellow lines are potential field and the red lines are NLFFF force lines. The structure from different viewsights can be seen in figure 15a , figure 15b , and figure 15c .
Regarding none of our calculations fitting the sigmoid structure, we investigate this structure with high resolution SXT images. figure 16 is partial SXT image with high resolution, overplotted by CDS mg10 image, quoted from WSM3 homepage ( sigmoid working group ). Obviously, this sigmoid structure consists of at least three different bright loops. The force lines based on LFFF, non-force-free-field, and NLFFF models fit two of them (see figure 17 , figure 18 , and figure 19 ).
The NLFFF force lines can be seen from different viewsights ( figure 20 ). It is worthy to note that part of these force lines that match very well with SXT bright loops have high twist (figure 21 ), which may cause kink instability. In fact, the assembly of the dips of these force lines just outlines the filament (figure 22 ), which disintegrated on August 19 ( disintegration ) and eventually erupted later( eruption ), causing a SF flare at 08:27UT on 20 Aug. The calculated NLFFF force lines after this event became open-up( open-up ). Obvious change of magnetic field in this period are also found ( figure 23 ) in vect magnetograms. The positive Q-component and U-component patches (noted as q1 and u1) gradually faded and eventually disappearred. Decrease of Q and U means decrease of the inclination angle of magnetic field, and this observation consists with our calculation--open-up of magnetic field. We will give a detail analysis in another paper.
