Multi-spacecraft Observations of the Coronal and Interplanetary Evolution of a Solar Eruption Associated with Two Active Regions

TL;DR

This study tracks the evolution of a solar eruption from the Sun to interplanetary space using multi-spacecraft observations, revealing how magnetic interactions influence CME trajectory, shape, and arrival times at different locations.

Contribution

It provides a comprehensive analysis of a CME linked to two active regions, combining remote sensing and in situ data with modeling to understand its complex propagation.

Findings

CME was deflected eastward and southward by magnetic pressure from active regions.

CME's arrival time at VEX was not earlier despite proximity to the Sun, highlighting shape and direction importance.

ICME was compressed radially but extended transversely in interplanetary space.

Abstract

We investigate the coronal and interplanetary evolution of a coronal mass ejection (CME) launched on 2010 September 4 from a source region linking two active regions (ARs) 11101 and 11103, using extreme ultraviolet imaging, magnetogram, white-light and in situ observations from SDO, STEREO, SOHO, VEX and Wind. A potential-field source-surface model is employed to examine the configuration of the coronal magnetic field surrounding the source region. The graduated cylindrical shell model and a triangulation method are applied to determine the kinematics of the CME in the corona and interplanetary space. From the remote sensing and in situ observations we obtain some key results: (1) the CME was deflected in both the eastward and southward directions in the low corona by the magnetic pressure from the two ARs and possibly interacted with another ejection, which caused that the CME arrived at VEX that was longitudinally distant from the source region; (2) although VEX was closer to the Sun, the observed and derived CME arrival times at VEX are not earlier than those at Wind, which suggests the importance of determining both the frontal shape and propagation direction of the CME in interplanetary space; (3) the ICME was compressed in the radial direction while the longitudinal transverse size was extended.

Tables

Table 1. Table 1: Times and Fitted Parameters of GCS Model for the CME

		Time	Longitude	Longitude	Latitude		Height		Aspect ratio	Half angle
		(STEREO, UT)	(Carrington, °)	(HEE, °)	(HEE, °)		($R_{☉}$)			(°)
SOHO/LASCO C2 and STEREO/COR1
		15:25	76.4	60.4	3		3\@alignment@align.00	0.3	20
		15:35	74.1	58.1	3		3\@alignment@align.29	0.3	20
		15:45	72.2	56.4	3		3\@alignment@align.59	0.3	20
		16:00	71.0	55.3	3		4\@alignment@align.00	0.3	25
		16:10	69.9	54.3	3		4\@alignment@align.41	0.3	25
		16:25	68.0	52.5	3		4\@alignment@align.86	0.3	27
SOHO/LASCO C2 and STEREO/COR2
		16:39	64.9	49.5	2		5\@alignment@align.57	0.4	30
		16:54	63.3	48.1	2		6\@alignment@align.15	0.4	30
		17:24	62.9	48.0	2		7\@alignment@align.35	0.4	30
SOHO/LASCO C3 and STEREO/COR2
		17:39	62.9	48.1	1		7\@alignment@align.79	0.4	30
		17:54	62.7	48.1	0		8\@alignment@align.40	0.4	30
		18:24	62.5	48.1	0		9\@alignment@align.71	0.4	33
		18:39	62.3	48.1	0		10\@alignment@align.28	0.4	33
		18:54	62.2	48.1	0		10\@alignment@align.79	0.4	33
		19:24	60.9	47.1	0		11\@alignment@align.81	0.4	35
		19:39	60.7	47.0	0		12\@alignment@align.71	0.4	35
		19:54	60.7	47.1	0		13\@alignment@align.56	0.4	35
		20:24	59.8	46.5	0		14\@alignment@align.85	0.4	35
		20:39	59.3	46.1	0		15\@alignment@align.64	0.4	35
		21:24	58.6	45.8	0		17\@alignment@align.57	0.4	35

Full text

Sun: coronal mass ejections (CMEs) — Sun: magnetic fields — Sun: activity — solar-terrestrial relations — solar wind

\submitted

Accepted by The Astrophysical Journal on April 11, 2017; unproofed.

1 Introduction

Coronal mass ejections (CMEs) are large expulsions of plasma and rapid releases of magnetic field energy from the solar atmosphere. They are called interplanetary CMEs (ICMEs) when propagating in interplanetary space, which are one of the major disturbances in the solar wind. ICMEs are also significant factors inducing geomagnetic activities when they impact the Earth. From the low corona to interplanetary space, a CME may experience non-radial motion, interact with other CMEs and be deformed due to the interaction with the ambient solar wind, involving plenty of physical processes.

The configuration of the magnetic field surrounding the source region as well as the interaction with other structures in the solar wind are important for estimating the propagation direction of a CME and thus crucial for space weather forecasting. Previous coronagraph observations suggested that latitudinal deflections of CMEs were controlled by the background coronal magnetic field and the overall flow near solar minimum (MacQueen et al., 1986), which was confirmed by the result that CMEs near solar minimum were systematically deflected toward lower latitudes while CMEs during solar maximum could deviate toward either the pole or equator (e.g., Cremades & Bothmer, 2004; Kilpua et al., 2009). Some CMEs were found to be deflected toward the weak magnetic field region associated with the streamer belt or the heliospheric current sheet (e.g., Xie et al., 2009; Kay et al., 2013; Liewer et al., 2015), while Xie et al. (2009) reported that fast CMEs could be deflected away from the streamer belt. The coronal hole with strong open magnetic field is one important factor that can divert a CME (e.g., Gopalswamy et al., 2004, 2005, 2009; Xie et al., 2009; Panasenco et al., 2013). Möstl et al. (2015) and Wang et al. (2015) found that the strong magnetic field of a sunspot in an active region (AR) and the configurations of the magnetic field surrounding the source region play a significant role in channeling the CME. Deflection of CMEs in the corona is likely to occur within a few solar radii from the Sun (e.g., Isavnin et al., 2014; Kay et al., 2015; Wang et al., 2015), while CME-CME interaction may change the propagation direction of a CME in interplanetary space (e.g., Gopalswamy et al., 2001; Liu et al., 2012, 2014; Lugaz et al., 2012).

A plethora of efforts have been made to understand the morphology of the CME in the corona and interplanetary space, which is crucial to determine the arrival time of the CME at a planet. Thernisien et al. (2006, 2009) presented the graduated cylindrical shell (GCS) model employing a croissant-like empirical flux-rope structure to describe the morphology of the CME near the Sun. A compact structure was used to derive CME kinematics (e.g., Sheeley et al., 1999). A circle attached to the Sun, called the harmonic mean method, was also applied to determine the kinematics in the inner heliosphere (e.g., Lugaz et al., 2009). A self-similar expansion (SSE) geometry with an expanding circle in the ecliptic plane was adopted to describe the CME front (Lugaz et al., 2010; Davies et al., 2012; Möstl & Davies, 2013). Möstl et al. (2015) and Rollett et al. (2016) introduced an elliptical shape with a constant angular width to forecast CME arrival time and speed based on single-viewpoint imaging observations. The readers are directed to Liu et al. (2010b) for discussions of some of the geometries. However, a CME could be deformed by interacting with the ambient solar wind (e.g., Riley et al., 2003; Odstrcil et al., 2004; Manchester et al., 2004; Liu et al., 2006; Savani et al., 2010), as well as structures in the solar wind (e.g., Lugaz et al., 2009; Liu et al., 2012, 2015; Kataoka et al., 2015). These interactions can alter their plasma and magnetic field characteristics as well as morphologies in interplanetary space.

In this manuscript we investigate the coronal and interplanetary evolution of a CME launched from a source region linking two ARs. The eruption may occur as a result of magnetic coupling of the two ARs, but it is different from the cases of sympathetic eruptions (e.g., Schrijver & Title, 2011; Jin et al., 2016). The remote sensing and in situ observations from SDO, STEREO, SOHO, VEX and Wind are used in this study. The positions of the spacecraft and planets during the event are indicated in Figure 1. STEREO A was ${\sim}$81° ahead of the Earth and was $\sim$20° west of the source region ($\sim$W62°), and STEREO B was ${\sim}$74° behind the Earth. VEX (Venus) was $\sim$0.73 au from the Sun and $\sim$31° east of the Earth. The spacecraft are at positions that are advantageous for investigating the kinematics and the structure of the CME. The study illustrates how the magnetic field of the two ARs on one side of the source region change the motion of the CME near the Sun. The multi-spacecraft in situ measurements and remote sensing observations place constraints on the structure and the kinematics of the CME, and demonstrate how the frontal shape and propagation direction determine the arrival time of the CME. Furthermore, multiple spacecraft at longitudinally separated positions reveal the longitudinal transverse size of the ICME, which is important for understanding the structure and distortion of the ICME in the solar wind. The evolution of the CME in the corona and interplanetary space is presented in Sections 2 and 3, respectively. We conclude and discuss the results in Section 4. The results of this work shed light on the kinematics and morphological evolution of the CME in the inner heliosphere as well as space weather forecasting.

2 Evolution in the Corona

The CME was launched around 14:30 UT on 2010 September 4 in the rising phase of solar cycle 24, associated with two roughly straight flare ribbons on the eastern edges of two ARs 11101 (W67°N11°) and 11103 (W69°N26°), as shown in the right panel of Figure 2. Figure 2 displays the extreme ultraviolet (EUV) images and magnetograms about the source region from SDO and STEREO A. In the magnetograms, the white (dark) regions represent the positive (negative) magnetic field regions, and the locations of ARs 11103, 11101 and 11102 are marked with “A”, “B” and “C”, respectively. According to the EUV observations of SDO/AIA (Lemen et al., 2012) and STEREO A/EUVI (Howard et al., 2008), the ribbons became visible at wavelength 304 Å at about 14:16 UT and separated eastward and westward, respectively. From the SDO/HMI (Schou et al., 2012) magnetogram overlaid with the intensity contour of the AIA 304 Å image, the magnetic field corresponding to the eastern ribbon is positive and the magnetic field for the western ribbon is negative. The neutral line between the two ribbons is along the meridian and at a longitude of $\sim$W62° ($\sim$77° in Carrington longitudes). Viewing from the viewpoint of STEREO A, a potential-field source-surface (PFSS) model reveals an overlying loop structure above the two flare ribbons as shown in the lower right panel of Figure 2. The associated loop brightening during the eruption was also observed at wavelength 195 Å by STEREO A/EUVI (not shown here). A hot channel, which is considered as an erupting flux rope (e.g., Zhang et al., 2012; Cheng et al., 2014), is not observed in the EUV images. However, an eruption with a two-ribbon flare and loop brightening is considered evidence of an erupting flux rope (e.g., Amari et al., 2000; Qiu et al., 2004). In our case, an erupting flux rope could be inferred from the flare ribbons and the loop brightening, which was lying over the neutral line between the two ribbons. The erupting flux rope probably was of low twist because the magnetic field near the neutral line was relatively weak and not strongly sheared. As can be inferred from the combined EUV and HMI observations in the upper right panel of Figure 2, the erupting flux rope likely extended from the positive polarity region of AR 11103 (marked with “A”) in the north to the negative polarity region of AR 11101 (marked with “B”) in the south. The source region of the CME was expected to be between the two ribbons and adjacent to the eastern edges of the two ARs. ARs 11103 and 11101 are separated by $\sim$15° in latitudes but may be connected through the erupting flux rope that is under the PFFS model revealed loop structure. The maximum magnitude of the HMI measured magnetic field in AR 11101 is about 2100 gauss, and the one in AR 11103 is about 1600 gauss. Möstl et al. (2015) and Wang et al. (2015) reported that the strong magnetic field (over 3000 gauss) of a nearby sunspot could change the propagation direction of the CME near the Sun. In our case, although the magnetic field is not so high, the asymmetric magnetic field configuration close to the source region may also interfere the motion of the CME in the corona.

Figure 3 shows the magnetic energy density distribution surrounding the CME source region at heights of 1.75 and 2.25 $R_{\sun}$. The magnetic energy density $w_{\mathrm{B}}$ is calculated from $w_{\mathrm{B}}={\mathbf{B}^{2}}/{2\mu_{0}}$, where the magnetic field $\mathbf{B}$ is from the PFSS extrapolation and $\mu_{0}$ is the vacuum permeability. In Figure 3, the two vertical dashed lines indicate the approximate positions of the two ribbons, and the source region is between them. For the sake of simplicity, we use contours of the magnetic energy density to infer the general direction of the gradient descent. As displayed in the upper panel of Figure 3, the contour reveals that the gradient descent of the magnetic energy density near the source region is generally eastward at the height of 1.75 $R_{\sun}$. As shown in the lower panel of the figure, the gradient descent near the source region at the height of 2.25 $R_{\sun}$ has a southward component. ARs 11103, 11101 and 11102 marked with diamond symbols in Figure 3 are generally in the regions of higher magnetic energy density on the west side of the source region. This implies that ARs 11103, 11101 and possibly 11102 caused an asymmetric magnetic field configuration near the source region. Previous studies suggested that an asymmetric configuration of the coronal magnetic field could deflect a CME toward the weak magnetic field region (e.g., MacQueen et al., 1986; Kilpua et al., 2009; Kay et al., 2015; Wang et al., 2015). We use the GCS model developed by Thernisien et al. (2006, 2009) to investigate the propagation direction near the Sun, and find that the propagation direction of the CME at 6.15 $R_{\sun}$ (marked with a square in Figure 3) is in the southeast of the source region, which is consistent with the general direction of the gradient descent of the magnetic energy density. The GCS modeling will be detailed in the text below.

The GCS model can determine the height, direction and other parameters of the CME in the corona based on the coronagraph observations from STEREO/SECCHI (Howard et al., 2008) and SOHO/LASCO (Domingo et al., 1995). Two coronagraphs are on board each STEREO spacecraft (Kaiser et al., 2008), of which COR1 has a field of view (FOV) of 1.5–4 $R_{\sun}$ and COR2 has a FOV of 2.5–15 $R_{\sun}$ (Howard et al., 2008). Coronagraph C2 on board SOHO has a FOV of 1.5–6 $R_{\sun}$ and C3 has a FOV of 3–30 $R_{\sun}$ (Domingo et al., 1995). Multi-spacecraft coronagraph observations from three viewpoints place strong constraints on the parameters of the GCS model. The CME appeared in the FOV of STEREO B/COR1 at 15:05 UT, SOHO/C2 (from position angle PA $\sim$270°) at 15:12 UT, and STEREO A/COR1 at 15:37 UT. Running difference coronagraph images from SOHO, STEREO A and B as shown in Figure 4 are used to fit the GCS model parameters of the CME. In every two rows of Figure 4, two sets of identical running difference images are presented, where the GCS modeled CME (green grids) are superimposed on the second set of images. In the running difference images of the top two rows of the figure, we see that the CME was about to enter the FOV of STEREO A/COR1 at 15:25 UT when it was clearly observed by STEREO B/COR1 and SOHO/C2. This strongly restricts the CME propagation direction at that moment with the GCS model, because an underestimated angle between the Sun-Earth line and the propagation direction will bring the modeled CME (the green grids) into the FOV of STEREO A/COR1 then. In the middle two rows of Figure 4, the running difference image of SOHO/C2 shows a loop-like structure in the northwest of the CME. This structure may be a substructure of the CME, and it is not fitted during the GCS model. However, the possibility that the loop-like structure is of another CME cannot be excluded. The front of the CME was distorted and the southern part moved faster than the part above the ecliptic plane as displayed in the images of the bottom two rows of Figure 4. The distortion cannot be fitted by changing the tilt-angle parameter of the GCS model. Only the part of the CME front near and above the ecliptic plane is fitted by the GCS model when the distortion is significant. The distortion may be caused by a streamer interacting with the southern part of the CME from the behind, which can be seen in STEREO/COR2 white-light observations (not shown here). At each time, tilt-angle, aspect-ratio, half-angle and latitude parameters of the GCS model are fixed while only the longitude and height parameters are fitted.

The fitted parameters of the GCS model and their corresponding times are given in Table 4. The propagation direction of the CME at height 3.00 $R_{\sun}$ obtained from the GCS model was 60$\fdg$4 west of the Sun-Earth line. It was close to the longitude ($\sim$W62°) of the source region estimated from the EUV and HMI observations. The angle between the propagation direction and the Sun-Earth line decreased remarkably to 48$\fdg$1 when the CME height reached 6.15 $R_{\sun}$, and then gradually changed to 45$\fdg$8 when the height was 17.57 $R_{\sun}$. During the whole GCS process, the tilt-angle parameter is set to 90° that is consistent with the observed meridian-aligned flare ribbons. The constant tilt angle shows that the CME did not apparently rotate near the Sun. The other three parameters are slightly adjusted according to the data times, and the changes of aspect ratio and half angle could be explained by the expansion of the flux rope. Zuccarello et al. (2012) reported that a CME could be deflected latitudinally about 15° toward the equator within $\sim$3.5 $R_{\sun}$. The roughly constant latitude (around 0°) in Table 4 indicates that the CME in our manuscript had been diverted over 10° southward from the location of source region ($\sim$W62°N15° as indicated with the two vertical lines in Figure 3) before reaching the height of 3 $R_{\sun}$. The GCS modeling results also show that the CME was deflected over 10° eastward at the height of 6.15 $R_{\sun}$. The CME propagation direction at 6.15 $R_{\sun}$ is roughly in the southeast of the source region, which is near the regions of lower magnetic energy density as shown in Figure 3. The deflection direction of the CME in the low corona is roughly consistent with the distribution of the magnetic energy density. The direction variation of the CME revealed by the GCS model is minor above 6.15 $R_{\sun}$ as shown in Table 4, which is consistent with that the rate of CME deflection caused by the coronal magnetic field is more significant within a few solar radii (e.g., Isavnin et al., 2014; Kay et al., 2015).

3 Properties in Interplanetary Space

The interplanetary kinematics of the CME is derived through a geometric triangulation technique initially proposed by Liu et al. (2010a) based on stereoscopic wide-angle imaging observations from STEREO. The triangulation method first utilizes a Fixed $\beta$ (F$\beta$) geometry for the CME front, which assumes a relatively compact structure (Sheeley et al., 1999; Kahler & Webb, 2007) simultaneously observed by two separated spacecraft. Later, Lugaz et al. (2010) and Liu et al. (2010b) incorporated a harmonic mean (HM) approximation, which assumes that the CME front is a sphere attached to the Sun and tangent to the lines of sight of the two spacecraft. The triangulation concept has proved to work well in acquiring CME interplanetary kinematics and in connecting remote sensing observations with in situ signatures (e.g., Liu et al., 2010a, b, 2013, 2016; Lugaz et al., 2010; Möstl et al., 2010; Harrison et al., 2012; Davies et al., 2013; Hu et al., 2016). The triangulation technique with F$\beta$ and HM approximations is detailed and discussed in Liu et al. (2010b, 2013, 2016).

The elongations and times as inputs of the triangulation method are from the observations of STEREO COR2, HI1 and HI2. The angular FOV is 0$\fdg$7–4° around the Sun for COR2, $20{\arcdeg}\times 20{\arcdeg}$ centered at 14° from the center of the Sun for HI1, and 70° in diameter centered at 53$\fdg$7 for HI2 (Howard et al., 2008). The left panel of Figure 5 depicts the running difference images of the CME in interplanetary space taken by HI1 and HI2 of STEREO A and B. The feature near the northern edge of the CME in the HI1 images might be the substructure of the CME that is also observed by SOHO/C2. We cannot rule out the possibility that the feature is of another CME. The northern front of the CME was distorted and the southern part of the front was advancing the other part. Through stacking the running difference intensities of COR2, HI1, and HI2 within a slit along the ecliptic plane, we construct two time-elongation maps (J-maps, e.g., Sheeley et al., 1999) for the CME as shown in the right panel of Figure 5. The red curves in the maps represent the elongation angles of the CME front near the ecliptic plane observed from the viewpoints of STEREO A and B, respectively. STEREO A tracked the CME out to $\sim$50° while STEREO B followed the CME out to $\sim$40°, as indicated by the red curves. The elongations and times are extracted and used as inputs of the triangulation method. The CME kinematics along the ecliptic plane derived from the F$\beta$ and HM triangulations will be compared with the in situ observations. The trajectories of the CME derived from the triangulation with the F$\beta$ and HM approximations, as well as the positions of the spacecraft in the ecliptic plane during the event are displayed in Figure 1. The black diamonds and red crosses in Figure 1 represent the trajectory of the CME apex in the ecliptic plane derived from the F$\beta$ and HM triangulations, respectively. The green and blue circles represent the sizes of the HM assumed spherical front when the CME arrives at Wind and VEX, respectively. The arrows indicate the directions of the apex when the CME arrives the two spacecraft.

The propagation direction, height and speed of the CME in the ecliptic plane derived from the triangulation method and the kinematics in the corona obtained from the GCS model are presented in Figure 6. In the top panel of the figure, the derived propagation direction is given in degrees from the Sun-Earth line. The direction angle acquired from the HM triangulation is roughly twice the one from the F$\beta$ triangulation and more variational, which is noticed in previous studies (Lugaz et al., 2010; Liu et al., 2013, 2016). Near the Sun the longitude is overestimated by the HM approximation which is probably a result of the overestimated CME size as discussed in Liu et al. (2013, 2016). Both the GCS model and the F$\beta$ triangulation reveal that the CME was propagating to the east of the source region, and the direction changed dramatically toward the east in the low corona. The direction angles from the GCS model are between those from the F$\beta$ and HM approximations. The direction angles from the GCS model are closer to those from the F$\beta$ approximation at lower heights, and have a trend to approach to those from the HM approximation at larger heights. The CME was deflected from about 60° to about 50° west of the Sun-Earth line below $\sim$6 $R_{\sun}$. The difference between the propagation directions before and after the data gap in the top panel of Figure 6 implies that the CME may continue to turn eastward in interplanetary space. The final propagation direction angle before reaching the Earth is likely around $20\arcdeg$ west of the Sun-Earth line, to which both the F$\beta$ and HM triangulations seem to converge. Given these variations in the propagation direction, the CME possibly arrived at Venus (VEX) that was longitudinally $\sim$90° distant from the source region (see the text below). There were other ejections launched near AR 11103 and 11102 on September 4, which were all on the west side of the CME of interest. One of these ejections may interact with the CME, which is likely to be the reason for the direction change of the CME in interplanetary space. As shown in the middle and bottom panels of Figure 6, the GCS model and the triangulation method give consistent distances and speeds of the CME below $\sim$20 $R_{\sun}$. The distances and speeds from the F$\beta$ approximation slightly deviate from those from the HM approximation beyond $\sim$130 $R_{\sun}$, which is probably due to the limitation of the geometric assumption for the F$\beta$ approximation at larger distances as discussed in Liu et al. (2013, 2016). Below $\sim$30 $R_{\sun}$, the CME was gradually accelerated to about 600 km s*-1*. Due to the data gap we cannot tell if there is a deceleration process, but the speeds beyond $\sim$60 $R_{\sun}$ do not show much variation ($\sim$620 km s*-1* for F$\beta$ approximation and $\sim$530 km s*-1* for HM approximation as marked in Figure 1).

Figure 7 shows the solar wind magnetic field measurements from the VEX magnetometer (Zhang et al., 2006), where the ICME (with an interval of $\sim$7.5 hr from 11:04 UT to 18:33 UT on September 8) is reported by Good & Forsyth (2016). The high fluctuations of the magnetic field at lower altitudes was probably caused by the induced magnetosphere of Venus. VEX was at a distance of $\sim$0.73 au with a longitude of \sim$$-31° that was over 90° away from the source region. The distance data range of the CME derived from the observed elongations using the triangulation method is not large enough to estimate the arrival times of the CME front at VEX and Wind. Assuming a fixed propagation direction ($\sim$17° from the Sun-Earth line for F$\beta$ and $\sim$24° for HM) outside the data range, we use a linear fit of the distances determined by the triangulations to estimate the arrival times of the CME at VEX and Wind. The HM triangulation suggests that the eastern flank of the CME (the eastern part of the blue circle attached to the Sun in Figure 1) arrived at VEX around 15:53 UT on September 8, about 5 hr later than the start time of the ICME observed at VEX, with a radial speed at VEX of $\sim$310 km s*-1*. Assuming that the compact structure for the F$\beta$ approximation is the dominant apex and the CME front is also a circle attached to the Sun like the HM approximation, we estimate the arrival time at VEX, using the F$\beta$ derived directions and distances, to be 20:52 UT on September 7, about 14 hr earlier than the observed ICME start time. The HM triangulation has well predicted the arrival times of ICMEs in comparison to the in situ observed start times of the ICMEs (e.g., Liu et al., 2010b, 2016). In our case, the small difference between the HM triangulation derived arrival times and the ICME start time suggests that the VEX observed ICME shown in Figure 7 is the interplanetary counterpart of the CME of interest.

The associated in situ measurements at Wind during the event are displayed in Figure 8. Two ICME-like structures are indicated with “I1” and “I2” in the figure, one of which may be associated with the CME of interest. Although the alpha-to-proton density ratio is below the threshold for a typical ICME (e.g., Richardson & Cane, 2004; Liu et al., 2005), the duration of structure I1 ($\sim$6 hr from 19:42 UT on September 7 to 1:39 UT on September 8) can be determined by a combination of the increase of the alpha-to-proton density ratio, the low proton temperature (Richardson & Cane, 1995) and the magnetic field profile. Structure I2 has an interval of $\sim$15.5 hr (from 10:15 UT on September 8 to 1:52 UT on September 9), whose trailing edge may be ambiguous. The HM approximation determines the arrival time of the CME at Wind to around 03:23 UT on September 8 and the estimated radial speed is $\sim$490 km s*-1*. The F$\beta$ approximation gives an arrival time around 18:17 UT on September 7. The HM estimated arrival time is $\sim$8 hr later than the start time of structure I1, and is $\sim$7 hr earlier than the start time of I2. The F$\beta$ estimated arrival time is $\sim$1 hr earlier than the start time of I1, and is $\sim$16 hr earlier than the start time of I2. Given the small difference between the estimated and observed arrival times, it is difficult to determine which of the two ICME-like structures (I1 and I2) is associated with the CME of interest. There was a brief structure (from 12:18 UT to 16:26 UT on September 9) after structure I2, which is likely irrelevant to the CME of interest because of its late start time.

Figure 9 shows the in situ measurements at STEREO A during the event. On September 7 STEREO A observed a shock leading an ICME with the arrival time of 8:29 UT. The ICME interval is from around 13:00 UT on September 7 to around 9:45 UT on September 8, which is determined from the combination of the low-temperature region and magnetic field profile. Both the shock arrival time and the ICME start time are over 80 (60) hr earlier than the F$\beta$ (HM) estimated arrival time. Although the triangulation assumed shape may deviate from the real geometry of the CME, the large difference between the estimated and observed arrival times can hardly associate the STEREO A observed ICME with the CME of interest. This is also consistent with the statement in the ICME catalog of Good & Forsyth (2016) that no in situ ICME signatures observed at STEREO A were likely to be associated with the ICME at VEX in Figure 7. A couple of ejections were launched from source regions longitudinally close to STEREO A on September 4. The STEREO A observed ICME might be associated with one of these ejections, which might interact with the CME of interest and caused the latter to continue turning eastward in interplanetary space.

The HM approximation estimated arrival times of the CME at VEX and Wind are generally consistent with the observed, even though the ICME at Wind cannot be uniquely associated with the CME of interest. This seems to suggest that the assumed spherical geometry is applicable to the eastern part of the CME front in this case. The in situ measurements and the triangulation method both reveal that the arrival time of the CME at Venus (VEX) is not earlier than that at the Earth (Wind), although the Sun-Venus distance ($\sim$0.73 au) was less than the Sun-Earth distance. This is exactly what is shown in Figure 1: the expanding eastern flank of the assumed CME circle touches VEX after the front reaches Wind. Therefore the arrival times of the CME at the two spacecraft with different heliocentric distances can be determined by the joint effect of the propagation direction and the assumed expanding circular shape. However, it is unclear whether the assumed shape could describe the geometry of the western part of the CME because of the lack of consistent in situ measurement of the western part. Assuming that the CME front is symmetric with respect to the final propagation direction, we estimate the longitudinal angular width of the CME front to be probably no less than 100° given that the angular distance between VEX and the propagation direction is $\sim$50°. Let the semi-perimeter of the circle (the leading edge of the flux rope) in the HM approximation represent the longitudinal transverse size, and we estimate the size to be about 1.7 au when the CME arrived at Wind. The radial widths of structures I1 and I2 at Wind (as indicated in Figure 8), obtained from the HM estimated radial speeds multiplied by the durations, are $\sim$0.07 au and $\sim$0.18 au, respectively. Both the radial widths of structures I1 and I2 are far less than the transverse size $\sim$1.7 au. Note that the flux rope is generally perpendicular to the ecliptic plane as can be seen from Section 2, so the transverse size is approximately equal to the transverse size of the cross section of the flux rope in the ecliptic plane. This result suggests that the ICME structure was extremely compressed in the radial direction and extended in the longitudinal direction in interplanetary space.

4 Conclusions and Discussions

We have investigated a CME launched on 2010 September 4 from a source region connecting AR 11101 and AR 11103, based on EUV, magnetogram, white-light and in situ observations from SDO, STEREO, SOHO, VEX and Wind. The PFSS model is used to examine the magnetic field environment near the source region. The GCS model and the triangulation method are applied to analyze the kinematics and structure in the corona and interplanetary space. Below we summarize the results and discuss their crucial implications for understanding:

1. how the propagation direction of the CME in the low corona is changed by the asymmetric magnetic field configuration caused by the ARs on one side of the source region;
2. the determination of both the CME frontal shape and the propagation direction in space weather forecasting; and
3. the deformation of the ICME structure.

The CME was deflected in the low corona by the magnetic pressure from the two nearby ARs in the west, and possibly interacted with another CME in interplanetary space. The erupting flux rope originated from the positive and negative polarity regions belonging to the two different ARs. With all the sunspots on the west side, the source region was adjacent to the eastern edges of the two ARs, which caused an asymmetric magnetic field configuration on the two sides of the eruption and resulted in a deflection of the CME in the low corona. The deflection direction revealed by the GCS model is consistent with the general direction of the gradient descent of the magnetic energy density derived from the PFSS model. This suggests that a CME originated from the edge of an AR is likely to be deflected toward a non-radial motion due to the asymmetric background magnetic field configuration. This is similar to the case of Möstl et al. (2015) and Wang et al. (2015) where the strong magnetic field on one side of the CME source region changed the propagation direction of the CME. However, a CME-CME interaction is possibly involved in the change of propagation direction. Given that there were other ejections with close launch times and source regions to those of the CME of interest, the CME possibly interacted with one of these ejections during the propagation. The deflection and interaction may be the reason that caused the CME to reach VEX that was longitudinally distant from the source region. CME-CME interaction is not unusual and may change the propagation direction and 1-au property of a CME (e.g., Gopalswamy et al., 2001; Lugaz et al., 2009, 2012; Liu et al., 2012, 2014, 2015). This illustrates the necessity of considering both the background magnetic field condition and successive eruptions near the source region in determining the propagation direction of a CME. 2. 2.

It is important to determine both the frontal shape and the propagation direction of the CME in interplanetary space for space weather forecasting. Although the associated ICME at Wind cannot be uniquely identified from the two plausible structures, the HM triangulation with a circular geometry in the ecliptic plane can give arrival times that are less than 8 hr different from observed at VEX and Wind. Both the observed and the triangulation estimated arrival times of the CME at VEX ($\sim$0.73 au) are not earlier than those at Wind ($\sim$1 au), although VEX was closer to the Sun than Wind. These results suggest that the simple expanding sphere attached to the Sun may describe the shape of the eastern CME front in the inner heliosphere. However, the spherical shape is only consistent with in situ measurements at two points, and there is no consistent observations of the western part of the CME. The real geometry of the CME front is not strongly constrained. Furthermore, the CME could be distorted to a complex geometry so that it arrived at VEX later. Also note that the HM triangulation gives a misleading propagation direction near the Sun and large variations in the propagation direction. The CME touched VEX with its eastern flank in this case, and the flank of a CME could also trigger a geomagnetic storm as reported by Möstl et al. (2010). This indicates the importance of determining both the CME frontal shape and propagation direction in forecasting geomagnetic activity. 3. 3.

The structure of the CME in interplanetary space was compressed radially and extended longitudinally, leading to a large transverse size of the flux-rope cross section in the ecliptic plane. Estimate of the transverse size and deformation of the ICME structure in the inner heliosphere could be helpful to improve our knowledge of the interaction between the ICME and the solar wind as well as space weather forecasting. For example, the transverse size can increase the contact area of the interaction between an ICME and the solar wind and the probability of arriving at the Earth, while the deformation could alter the interplanetary plasma and magnetic field properties of an ICME near the Earth. Observations from the Solar Orbiter (Müller et al., 2013) out of the ecliptic plane can be of great advantage to study the kinematics and geometric evolution of CMEs, as well as their interactions with other structures in the solar wind.

The research was supported by the Recruitment Program of Global Experts of China, NSFC under grant 41374173 and the Specialized Research Fund for State Key Laboratories of China. We acknowledge the use of data from STEREO, SDO, SOHO, VEX, Wind, ACE and GOES.