CHROMOSPHERIC JET AND GROWING "LOOP" OBSERVED BY HINODE: NEW EVIDENCE OF FAN–SPINE MAGNETIC TOPOLOGY RESULTING FROM FLUX EMERGENCE

Figures 3 and 4 show the evolution of the jet in the Ca ii H line, EUV, and X-rays, and Table 1 summarizes the event time line. Early in the event, from 02:14 to 02:30 UT, there is brief surge-like activity near the limb (Figure 3(a)) and tumbling motions are seen around its base at spicule heights. At the same location, a bundle of material threads of typical width ≲1'' appears at 02:32 UT (see online Animation 1, Figure 3(b)). This bundle, as the direct predecessor of the later jet, extends above the chromosphere at an oblique angle toward the north and gradually swings up clockwise, seemingly unfolding itself and exhibiting oscillatory transverse motions across its axis. At the base of the bundle, a loop becomes visible at 02:44 UT and starts to expand both vertically and laterally (mainly toward the north). From 02:46 to 02:48 UT, blobs of bright emission are seen to swirl up in a helical-like trajectory from the chromosphere into the bundle (Figures 3(c) and (d)). Most of these can also be seen in TRACE 195 Å images (green panels in Figure 4) at lower resolution (1'' versus 02 of SOT) as dark features, because the Ca ii H line emission originates from partially ionized plasma at chromospheric temperatures (≲2 × 10⁴ K; Gouttebroze et al. 1997) that absorbs the background 195 Å emission from hot plasma at coronal temperatures (∼1.5 MK). Some of these absorption features can be seen in X-rays as well (e.g., Figure 4(f)).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Around 02:48–02:49 UT, the swing and rise of the material bundle and the expansion of the loop start to accelerate. At the same time (t₁= 02:49:02 UT), the A4.9 flare sets in near the lower southern leg of the loop. It appears as bright emission in the Ca ii H line (Figure 3), EUV, and X-rays (Figure 4). At t₂= 02:50:32 UT, the loop legs stop their lateral expansion, and soon after (t₃= 02:51:12 UT), the loop appears to rupture and its apex becomes undetectable (Figures 3(f)–(h)). At t₅= 02:52:40 UT, the angle between the axis of the material bundle and the limb reaches its maximum. The bundle then starts to sweep back toward the north in a whip-like manner and rapidly develops into a collimated jet, which eventually extends to a length of 203 Mm or 0.29 R_☉ (Figure 2(e)). The loop continues to "collapse" and material is seen to drain down the northern leg toward a flare kernel only ∼15'' away from the southern leg (Figures 3(i) and (j)). This indicates that the lateral expansion of the loop occurs only in the corona, while the footpoints of the loop remain anchored on the surface.

The jet material is seen as emission in the cool (∼10⁴ K) Ca ii H (Figure 3) and Hα lines and at 304 Å (∼ 10⁵ K) and 171 Å (∼ 1 MK), as absorption in the hotter (∼ 2 MK) 284 Å line (Figure 2), as multiple strands of absorption and emission at intermediate (∼1.5 MK) 195 Å (Figure 4), and as weak emission in XRT SXR (∼ 1–30 MK; Figure 4(g)). This indicates a wide range of temperatures, from 10⁴ to 10⁶ K, in the jet plasma.

The jet exhibits oscillatory transverse motions across its axis as seen by SOT, which we interpreted in Paper I as spins resulting from unwinding twists. Further evidence of this interpretation is found at 195 Å here, where the emission and absorption strands alternate and shadow each other in sequence, as expected for a spinning cylinder in a side view (see online Animation 2).

Later in the event, a flare loop appears in SXR and then EUV (Figure 4). Part of the jet material that fails to escape the gravitational bound returns to the chromosphere along smooth streamlines in the original direction of ascent (Figure 2(f)). The oscillatory transverse motions seen earlier in the jet are no longer present in the falling material (see Figures 1 and 3 in Paper I). At lower altitudes, most of the streamlines bypass a dome-shaped region and outline an inverted-Y geometry (Figure 3(k)).

This event occurred near the limb and thus there was no reliable direct magnetic field measurement available. We then resorted to the potential field source surface (PFSS) model (Zhao & Hoeksema 1994; Schrijver & DeRosa 2003) available in the Solar SoftWare package to obtain indirect information of the magnetic topology in the vicinity of the jet. The modeled field is the extrapolation of a photospheric magnetogram that combines SOHO Michelson Doppler Imager observations within 60° of the disk center and the evolved fields elsewhere according to the flux disperse model.

Figure 5 shows the PFSS model field at 00:04 UT near the time of the event, as seen from different view angles. We noticed a north–south oriented narrow channel (range: 6° in longitude, 18° in latitude) of an open field or a coronal hole located between two NOAA ARs 10940 and 10941. The models up to 7 days prior to the event show that this coronal hole persistently existed, evolved slowly in size and shape, and matched the dark regions seen in SOHO EIT 284 Å images as expected.⁴ We then rotated the 00:04 UT model field to the time of the jet according to the Carrington rotation rate and found that the event was located inside the coronal hole. An example is shown in Figure 3(h), where the coronal hole boundary (dotted line) encompasses the kernels of the flare. This is consistent with the finding of Nitta et al. (2008) for an AR jet and with the expectation that coronal jets tend to occur in open-field regions, as manifested by the ubiquitous polar jets observed by Hinode (Cirtain et al. 2007; Shibata et al. 2007). Close comparison with the PFSS model further indicates that the northwest orientation of the jet is close to those of the open-field lines, again as expected from classical jet models. This can be clearly seen in both SOT Ca ii H and STEREO 304 Å images (Figure 6).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The accompanying A4.9 flare occurred at 02:49 UT and peaked at 02:53 UT in the GOES low channel (1–8 Å) flux. As can be seen in Figures 7(a)–(c), the light curves in the Ca ii H line (SOT) and X-rays (GOES and RHESSI) are very similar, showing a single hump, except for slightly different onset and peak times, presumably due to different instrument response to the varying plasma temperature. The TRACE EUV light curve shows a different trend except during the rise of the flare. The initial decrease from 02:37 to 02:47 UT is due to the unfolding material bundle and growing loop at chromospheric temperatures, which lead to increased absorption of the background EUV emission. The second flux increase (02:56–03:01 UT) is likely the result of the hot X-ray emitting plasma cooling to 195 Å passband temperatures and/or continuous evaporation of chromospheric plasma due to thermal conduction (Zarro & Lemen 1988; Battaglia et al. 2009; Liu et al. 2009d).

Zoom In Zoom Out Reset image size

The flare emission appears near the lower southern leg of the loop and cospatial at various wavelengths, from Ca ii H to EUV and X-rays (Figures 3 and 4). It is seen by RHESSI up to the 6–12 keV band. Fits to RHESSI spectra at the flare peak indicate that the emission is primarily thermal bremsstrahlung from a plasma of temperature T = 12.2 ± 0.6 MK and emission measure EM = (5.3 ± 0.5) × 10⁴⁵ cm⁻³ (see Appendix B). As shown in Figures 7(d) and (e), T and EM have simple temporal profiles. Their differences between GOES and RHESSI likely result from the well-known fact that GOES is more sensitive to cooler plasmas. Note that, because of their sharp occultation at the chromospheric limb, the three bright loops seen earlier by XRT (Figure 4(a)) are most likely located behind the limb. The flare loop happens to be in front of the background loop in the middle (y ∼ −40'') and is located on the visible side of the disk, since its footpoints are inside the limb (Figure 4(l)) and the flare kernels are clearly observed (Figure 3(j)).

As briefly mentioned in Section 2.1, the predecessor of the jet is a bundle of material threads that exhibits transverse motions across its axis. To quantify these motions, we have placed six 3'' wide cuts along the bundle, oriented roughly perpendicular to the local threads as shown in Figure 3(b). Spacetime diagrams (Figure 8) of these cuts clearly show oscillatory transverse motions between 02:36 and 02:48 UT, particularly in the upper portion of the bundle (Cuts 2–5). We selected Cut 3 as an example and fitted those unambiguously identified piecewise tracks with a sine function of time t

$$\begin{equation} s_\perp (t) = s_{\perp 0} + A \sin \left[ {2 \pi (t-t_0) \over P } \right] \,, \end{equation} \tag{ 1 }$$

where s_⊥ is the distance along the cut (⊥ to threads), A is the amplitude, and P is the period. The resulting fits are shown in Figure 9. Compared with those in the late stage of the jet as shown in Figure 4 of Paper I, the oscillation velocities (v_⊥ ∼ 50 km s⁻¹) at the equilibrium position here are about 1.5–2 times larger, while the amplitudes and periods are smaller by similar fractions.⁵

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To track the propagation of these oscillations, we identified the crests of an oscillation in Cuts 2–5 as marked by the plus signs on the left in Figure 8, which exhibit a delay from one cut to the next. We then fitted the separations of the cuts versus the occurrence times of the crests with a linear function. (Here the cut separations are defined as the distance from the center of the cut to the previous neighboring cut.) This yields a speed of v_ph = 786 ± 30 km s⁻¹ for the upward propagation, which is formally identified as a phase speed here. This speed is similar to its later counterpart at the leading edge of the jet (see Figure 3 of Paper I). If these oscillations are MHD waves, the wavelength would be λ = v_phP = v_ph(192 ± 12 s) = 151 ± 11 Mm. We also located the onset (marked by crosses) of the rapid rise of the bundle in each cut. A parabolic fit indicates an acceleration of a = 4.6 ± 0.5 km s⁻² and a final speed of v_ph = 401 ± 39 km s⁻¹ at the highest Cut 5.

Next, we quantify the rise of the material bundle by tracking its apex, lower end, and axis as marked in Figure 3(e). The apex and lower end are identified as the highest and lowest points in altitude of the ensemble of the bundle threads, respectively. The axis is recognized as the bisector of the bundle's angular extent and is characterized by its orientation angle θ_bdl from the limb. Figure 10 shows the history of these geometric parameters, which, together with those shown in Figure 12, are fitted with a piecewise linear and/or parabolic function of time, depending on their temporal evolution, to infer their velocities and/or accelerations.

Zoom In Zoom Out Reset image size

We find a slow-to-fast two-phase evolution⁶ clearly divided by the onset of the accompanying flare at t₁= 02:49:02 UT. (1) Prior to t₁, the apex height h_bdl, axis angle θ_bdl, and displacements of the bundle's lower end from its initial position (Δx_bdl, Δy_bdl) show slow variations (Figures 10(b)–(d)). The clockwise upward swing of the axis is evident in the increase of θ_bdl at a moderate velocity of $\dot{\theta }_{\rm bdl}= (4.9 \pm 0.2) \times 10^{-2} \, {\rm deg}\ {\rm s} ^{-1}$. (2) At the flare onset, all of these quantities but Δy_bdl start to increase rapidly. $\dot{\theta }_{\rm bdl}$ experiences an acceleration and reaches (3.1 ± 0.2) × 10⁻¹ deg s⁻¹ at t₅= 02:52:40 UT when its growth comes to a stop followed by a reverse. Overall, θ_bdl has gained more than 50° in 10 minutes. The bundle's lower end first moves mainly upward (Δx_bdl) with an acceleration of 0.66 ± 0.09 km s⁻² (final velocity: 110 ± 6 km s⁻¹). At t₃= 02:51:12 UT, it turns to a primarily horizontal direction (Δy_bdl, northward) with a comparable acceleration and slight decrease in height by 3'' (Figures 10(d) and 13(c)). At the same time, the top portion of the bundle continues swinging up and southward.⁷ This makes the bundle rotate clockwise, forming an elbow or bend toward the north at t₄= 02:51:44 UT (Figures 3(g) and (h)). Soon after this (02:52:24 UT), the entire material bundle quickly sweeps toward the north like a whip, developing into a collimated jet, and individual threads stretch upward as fast as 176 ± 11 km s⁻¹ (Figure 10(b)).

Another key component of this event is the growing Ca ii H loop that accompanies the rise and eruption of the material bundle. We track the evolution of its apex height h_loop and the lateral span of its two legs (see Figure 3(e)). The latter is derived from the solar-y coordinates of the legs' outer edges since the y-axis here near the equator is almost parallel to the limb.

Similar to the material bundle, the loop apex also experiences two distinct phases divided at the flare onset t₁ (Figure 10(e)): (1) a gradual phase from 02:44 to t₁ with a moderate ascent velocity of 16.2 ± 0.4 km s⁻¹ and (2) an acceleration phase from t₁ to t₃= 02:51:12 UT with an acceleration of 0.73 ± 0.06 km s⁻² and a final velocity of 135 ± 4 km s⁻¹. The second phase ends when the loop appears to rupture and its apex is no longer visible. This coincides with the turning point of the material bundle's lower end mentioned above (Figure 10(d)).

The lateral position y_NLeg of the northern leg of the loop has a similar two-phase evolution with 1–2 times larger velocities and acceleration (Figure 10(f)). In contrast, the southern leg does not show such a distinction and has a uniform low velocity of −5.2 ± 0.1 km s⁻¹, only one-fourth of its northern counterpart (23.3 ± 0.7 km s⁻¹) in the gradual phase. This asymmetry means that the lateral expansion of the loop is primarily on the northern leg moving toward the north. Note that the lateral expansion velocity and acceleration averaged between the two legs roughly equal their vertical counterparts. The lateral expansion, however, ceases earlier than the vertical growth. This occurs in both legs at t₂= 02:50:32 UT <t₃, after which they remain stationary with small fluctuations (standard deviation: 04).

To examine the shape of the loop, we define the aspect ratio of the loop apex height to the half separation of the two legs in the north–south direction, R ≡ h_loop/[(y_NLeg − y_SLeg)/2]. As shown in Figure 10(g), this ratio has a mean of 2.2>1, indicating vertical elongation. The elongation is reduced with R decreasing from ∼ 3.2 to 1.8 during the gradual expansion phase (02:44 UT–t₁). The loop then preserves its shape with a constant aspect ratio ∼1.8 during the common acceleration phase (t₁–t₂), but the elongation slightly increases afterward because of the earlier cessation of the lateral expansion.

Near the end of the visible lifetime of the accompanying loop, the northern leg exhibits interesting behaviors. There are two branches at this leg which initially expand upward during the loop growth. When the elbow of the material bundle appears at t₄ = 02:51:44 UT as noted above, the two branches start to bend and retreat downward. We identified five bright blobs (A–E) as shown in Figure 11, among which blobs A and D represent the visible apexes of the two branches. We then tracked the blob locations with time (Figure 12(a)) and used blob sizes as uncertainties. For each blob, its positions are well represented by a straight line fit, which gives its main direction of motion. We obtained the distance along this direction and inferred the corresponding velocity and/or acceleration (Figures 12(b) and (c)).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We find that both blobs A and D exhibit a downward acceleration followed by a deceleration within ∼1 minute, and their average velocities are nearly −100 km s⁻¹ (see Table 2). Near the onset (02:52:24 UT) of the northward sweep of the material bundle, blob A becomes too vague to be identified, while two new blobs, B and C, appear nearby. They shoot upward with an acceleration a ∼ 11 g_☉ (where g_☉ = 0.274 km s⁻² is the solar gravitational constant) and reach final velocities of 192 ±  18 and 130 ± 18 km s⁻¹. In contrast, as blob D slows down its downfall near t₅= 02:52:40 UT, it turns its horizontal direction from the north to the south (Figure 12(a)) and we assign it a new blob ID named E. This blob resumes a downfall with an acceleration of a = (−4.8 ± 0.7) g_☉ and reaches a velocity of −115 ± 9 km s⁻¹ just before it plunges into spicules in the chromosphere. Since the blobs are tracked by their projected two-dimensional positions in the sky plane, these velocities and accelerations are lower limits of their true values in three-dimensional space. Meanwhile, the solar gravitational acceleration g_☉ is the upper limit of its component along the unknown three-dimensional trajectory. Thus, blob E's downward acceleration is at least fivefold greater than a corresponding free-fall (dotted line in Figures 12(b)). Its final velocity is also 2–3 times larger than those (30–50  km s⁻¹) found at the footpoints of Hα arch filaments (Bruzek 1969; Roberts 1970). Even if a free-fall starts at a higher altitude at the top of trajectory D, it would reach a final velocity of only 42 km s⁻¹, about one-third of trajectory E's value.

As mentioned earlier, jet material bound by gravity falls back to the chromosphere and this continues throughout the duration of this event. The trajectories of the falling material are smooth streamlines, which, at altitudes ≳20'', are almost straight lines in the original direction of ascent. Kinematics of the falling material above this height was investigated in Paper I using spacetime plots (see Figure 2 there). Below this height, the streamlines are curved as if they bypass an unseen dome or null point. This can be clearly seen in the online Animation 1, especially from 03:25 to 03:35 UT.

To highlight these streamlines, we performed running difference between each pair of images 2 minutes apart (every 15 frames at an 8 s cadence) within a selected duration, and superimposed the positive parts of all differenced images. A sample of the results is shown in Figures 13(a)–(c). We then visually traced the streamlines, a collection of which clearly outlines an inverted-Y geometry (Figure 13(d)). By overlaying these streamlines on top of multiwavelength images (Figures 13(e)–(h)), we note the following spatial relationships: (1) some streamlines, particularly those on the right-hand side, are distributed close to the growing Ca loop at its final stage; (2) the inverted-Y goes around the cusp-shaped flare loop seen in SXR; and (3) three branches (labeled b₀, c₁, and a₁) of the streamlines are cospatial with the dark absorption features seen in EUV, which likely represent the same cool material that emits in Ca ii H. We also note that the drift of the material bundle's lower end terminates at 02:54 UT near the bifurcation of the streamlines 1 hr later (Figures 3(l) and 13(c)).

Zoom In Zoom Out Reset image size

Based on the above two inferences, we postulate that the earlier material bundle and later collimated jet represent the outer spine and its neighboring field lines in different stages, while the growing Ca loop is a two-dimensional rendering (projection) of the entire three-dimensional dome or separatrix fan surface (Figure 1). We believe that, among alternative models, this conjecture best explains the observations. The postulated event develops as follows (see Figure 14).

Zoom In Zoom Out Reset image size

When a twisted flux rope emerges into the corona in an open-field environment, the flux rope expands rapidly, driven by its considerable magnetic pressure, and presses onto the ambient field to form a current sheet at the discontinuity between them (Heyvaerts et al. 1977). Magnetic reconnection ensues and results in a fan–spine structure mentioned above. The outer spine originates from the null point and opens to infinity, while the inner spine apparently divides the vault under the dome into two parts seen in two-dimensional projection.

Reconnection outflows continuously pump dense plasma upward along the newly reconnected field lines near the outer spine. Emission from such plasma (hatched region around the spine in Figure 1(a)) could be responsible for the observed material bundle and later collimated jet, whose diffuse upper end and relatively sharp lower end (Figures 3(f)–(h)) are readily explained by the geometry of the outer spine. Reconnection also transfers magnetic helicity from the emerging flux rope into the newly reconnected open field. This and plasma pumping are both manifested in the upward swirling motion of material along the helical trajectory observed around 02:47 UT (Figures 3(c) and (d)). Under the influence of the Lorenz force, the twists in the reconnected open field tend to unwind themselves and drive these field lines to rotate, as found in spinning Hα surges (Shibata & Uchida 1986; Canfield et al. 1996). Such spins propagate up toward the open end in the form of torsional MHD waves, and their two-dimensional projections would appear as traveling sinusoidal oscillations, just as observed here (Section 3.1). The inferred rotational velocities and periods (Figure 9) are comparable to those of transverse oscillations found in X-ray jets (Cirtain et al. 2007), prominences (Okamoto et al. 2007), and coronal loops (Ofman & Wang 2008), which were interpreted as signatures of Alfvén or fast kink mode waves (Vasheghani Farahani et al. 2009). Our inferred phase speed of v_ph = 786 ± 30 km s⁻¹ is also within the range of typical coronal Alfvén or fast-mode speeds. At the same time, the Lorentz force associated with the twists and axial gradient of currents (e.g., see the current density in Figure 9 of Pariat et al. 2009) may have a strong axial component that can drive the upward ejection of material along the axis of the bundle (Bellan et al. 2005).

As the dense emerging flux rope expands and sweeps through the dense lower atmosphere ahead of it, a layer of enhanced local density is expected to form at its leading front. In our case, this layer is the fan surface between the emerging flux and the ambient open field, marked as the hatched region on the dome in Figure 1(a). In two-dimensional projection, it could appear as the growing Ca loop seen here. The expansion rate of the flux rope, presumably depending on the flux emergence rate and the interplay between the strengths of the emerging field and the surrounding field, controls the dynamic evolution. Early in the event (before t₁ = 02:49:02 UT), the flux rope (thus the fan surface) expands at a moderate speed (16.2 ± 0.4 km s⁻¹), giving the ambient field a gentle push. This leads to moderate rates of reconnection and supply of mass and helicity to the material bundle around the spine, which evolves in a quasi-static manner. As sufficient twists have emerged into the corona with time, a kink-like instability can occur and force the flux rope to undergo an accelerating expansion (Fan & Gibson 2004; Pariat et al. 2009; Rachmeler et al. 2010), strongly pushing the ambient field and thus driving rapid reconnection. If we interpret flares to be indicators of rapid energy release by fast reconnection, we may identify time t₁ as the onset of the observed flare heating in this event. In addition, the increased fluxes of mass and helicity transferred to the material bundle through reconnection can no longer relax in a quasi-static manner, eventually resulting in a runaway instability that leads to the collimated jet. This explains the simultaneous transitions from slow to fast evolution for both the material bundle and growing loop (Figure 10). This is also energetically analogous to the two regimes found by Pariat et al. (2010): slow, quasi-static reconnection in the energy-storage phase, and fast, dynamic reconnection in the energy-release phase in a rotating current sheet associated with the kinking flux under the fan surface.

Because of continuous reconnection, the material bundle around the spine is a constantly evolving entity. A given field line in the bundle is illuminated only when reconnection drives dense mass flow along it. It becomes invisible once the temperature leaves the Ca ii H response range (≲2 × 10⁴ K) or the density drops significantly, which could be true at high altitudes or when the reconnection site migrates away. In the former case, the material bundle would correspond to the lower portion of a jet or surge, whose upper portion has upflows at reduced density and brightness. This gives rise to the diffuse appearance of the upper end of the bundle. This also explains why the streamlines of the falling material are distributed in a large volume and in various directions (Figure 13), not just along the path of the final collimated jet. This is because material is continuously ejected upward in different directions when the bundle swings, as what happens when a fireman swings his firehose. In the latter case of reconnection site migration, the visible threads in the material bundle are constantly being replaced and we see different field lines over time as reconnection develops. This could explain why each sinusoidal oscillation (Figure 8) appears ephemerally.

As the system seeks its lowest energy state, the open-ended spine field line can change its orientation, while the null point migrates in space from field line to field line, depending on the dynamic evolution of reconnection (Figure 14). This seems to reflect what we observe here and explain the elbow (Figure 3) and the drift of the material bundle's lower end (Figure 13(c)). The latter is likely located near the null point and its overall upward migration may result from the upward development of reconnection as predicted in the standard flare model (Sturrock 1966; Kopp & Pneuman 1976) or from a cumulative effect of advection motions driven by gas pressure in the neighborhood of the null where the Lorentz force vanishes (Pariat et al. 2009). The low angle of the material bundle in its early stage probably reflects the strong horizontal field at low altitudes. The back-and-forth fast (151 km s⁻¹, see Figure 4 in Paper I) swing of the entire material bundle possibly results from catastrophic release of excessive twists into the open field, similar to the kink or writhe of a flux rope. Also, if we assume the measured twist propagation velocity 786 km s⁻¹ to be the Alfvén velocity v_A, the velocity ratio 151/786 = 0.19 is close to the 0.2 value of the drifting outer spine found in the three-dimensional simulation of Pariat et al. (2009).

Many characteristics of our observations and the above interpretation resemble those in the simulation of Török et al. (2009), particularly their fan–spine topology with a three-dimensional null point resulting from emergence of a bipole into a locally unipolar region and the launch of a torsional MHD wave from the reconnection site. However, in their case, magnetic reconnection develops in two steps and each step leads to the formation of hot loops on each side of the spine, which constitute one-half of a full anemone, consistent with the Hinode XRT observations of a specific jet event. In our case, we found no signature of such two-step reconnection, and we could not identify the expected inner spine emission (hatched with dashed lines in Figure 1(a)) in available data, presumably due to unfavorable temperatures and/or densities under the dome. However, the flare loop in SXR and EUV (Figures 13(g) and (h)) and its right branch in Ca ii H (Figure 4(j)) are located in the expected position to the right-hand side of the null point for the newly reconnected field.

We note that the upward expansion of the emerging flux toward the north has a large component parallel to the ambient field and can readily make its way in a herniation manner, while the expansion toward the south must perpendicularly press the ambient field against strong resistance. This topological asymmetry can therefore preferentially facilitate the northward expansion, which is at least four times faster than its southward counterpart (Figure 10(f)). This is similar to the simulation of the primarily vertical (versus horizontal) expansion of the dome in a vertical ambient field (Pariat et al. 2009). In addition, the null point can be more readily advected in the spine direction, as manifested in its overall upward and northward migration (Figure 13(c)).

The northern leg starts to reverse its expansion and retreat downward at t₄= 02:51:44 UT. In addition, when the material bundle sweeps toward the north, its lower end, assumed to be near the null point, slightly drops in height by 3'' (Figures 10(d)). These occur near the peak of the flare, possibly when a significant amount of energy has been released. This is consistent with the implosion conjecture of Hudson (2000) and its theoretical demonstrations (Zhang & Low 2003; Janse & Low 2007). It predicts that after major energy release, the magnetic pressure (or energy density) in the local volume is significantly reduced and the resulting imbalance with the pressure in the surroundings would push this volume to contract. Observational evidence of this conjecture has gradually emerged (Liu et al. 2009a, 2009c). When this happens, opposite to an expanding twisted flux rope in which twists tend to accumulate in the expanded top portion (Parker 1979, p. 189), more twists will be pushed by the Lorentz force to concentrate in the leg portion when the rope contracts. This explains why twists become more visible in the northern leg later when it contracts. This downward Lorentz force might also provide additional push to the material and help explain the faster than free-fall downflow observed here (Figures 11 and 12).

At the same time, some material is ejected upward at accelerations up to (11.4 ± 3.1)g_☉ (Figures 11 and 12). It is quite puzzling for this to take place simultaneously in the same northern leg as the fast downflow, even though they might occur on different field lines at separate line-of-sight positions. In any case, there are several possible explanations which we cannot distinguish with the available data: (1) if the ejection is on twisted field lines that reconnect with untwisted open-field lines, it could be driven by the Lorentz force associated with the upward relaxation of twists, which seems to happen here (Figures 11(g)–(o)). (2) If the upward ejection and downflow are indeed on the same field lines, they could be the upward and downward branches of the secondary outflows bifurcated from the primary reconnection outflow as predicted in MHD simulations (Yokoyama & Shibata 1996; Moreno-Insertis et al. 2008). (3) Another less likely possibility is that the ejection and downflow are oppositely directed outflows from reconnection occurring at an X-point between them. An X-type null point at this location is, however, not favored by our adoption of the spine–fan topology to interpret this event.

The disappearance of the Ca ii H loop possibly results from the combination of several processes, including mass drainage, temperature variation, and topological change. The first two mechanisms are expected to operate gradually, while the third one could happen catastrophically. First, as a flux rope emerges and plows through the lower atmospheres, dense material is dredged into the hot corona. Part of this material would contribute to the Ca emission seen at the fan surface. Such dense material is expected to be pulled back by gravity and slide down the dome, as seen in Hα arch filaments (e.g., Chou & Zirin 1988). This is because the scale height of the ≲2 × 10⁴ K plasma emitting the Ca ii H line is only ≲1.2 Mm and pressure gradient from gravitational stratification is simply too small to support the weight of plasmas extending to the height of the dome (>10 Mm). In our case, such mass loss is evident in the northern leg of the loop. This would make the apex of the dome to have the lowest density and become less well defined and then invisible first, before the loop legs gradually fade away (Figures 10(e) and (f)). Second, heating of the Ca loop (fan surface) may occur during the course of its rise, possibly as a result of reconnection with the ambient field. Once its temperature rises above the ≲2 × 10⁴ K range, the Ca ii H loop would disappear. This can be seen in Figure 4(h), where the northern leg of the loop is vague in Ca but prominent and bright at 195 Å. Later, because of significant mass drainage that has considerably reduced the density of the loop (fan surface), it would no longer appear as detectable absorption or emission in EUV or SXR, as we see here (Figures 4(j) and (k)). Finally and more importantly, a catastrophic topological change, which may be involved in the launch of the chromospheric jet, can alter field connectivity and contribute to the disappearance of the fan surface in emission.

We note in passing that there is an overarching loop brightening in SXR at 02:44 UT (see Figure 4(b)) that appears simultaneously with the bright Ca ii H loop (dark in EUV) but ∼10'' higher in altitude. The southern leg of the Ca loop appears as dark absorption in SXR (Figure 4(f)), suggesting that the SXR loop brightening is located behind the Ca loop. When the legs of the Ca loop disappear around 02:55 UT (Figure 10(f)), this SXR loop becomes invisible too (Figure 4(j)). These timing coincidences seem to suggest that the SXR loop is of the same emerging flux system as the Ca loop but at higher altitudes and temperatures. However, during its lifetime, the SXR loop hardly changes its size or shape, while the Ca loop has risen more than 20'', well beyond the XRT's 2'' spatial resolution. This contradicts the expectation that the overarching SXR loop would expand together with the low-lying Ca loop if they were of the same emerging flux system. Another possibility is that the SXR loop brightening represents heating during magnetic reconnection in a curved current sheet between the emerging and ambient fields, as suggested by Yoshimura & Kurokawa (1999) who found SXR brightening above emerging Hα arch filaments. However, the location of the event within the inferred coronal hole poses a challenge for the existence of such a hot SXR loop as observed here. The nature of this transient SXR loop brightening thus remains an open question.