IMF By effects in the plasma flow at the polar cap boundary

We used the dataset obtained from the EISCAT Svalbard Radar during 2000–2008 to study statistically the ionospheric convection in a vicinity of the polar cap boundary as related to IMFBy conditions separately for northward and southward IMF. The effect of IMF By is manifested in the intensity and direction of the azimuthal component of ionospheric flow. The most significant effect is observed on the day and night sides whereas on dawn and dusk the effect is essentially less prominent. However, there is an asymmetry with respect to the noon-midnight meridian. On the day side the intensity of By-related azimuthal flow is maximal exactly at noon, whereas on the night side the maximum is shifted toward the post-midnight hours (∼03:00 MLT). On the dusk side the relative reduction of the azimuthal flow is much larger than that on the dawn side. Overall, the magnetospheric response to IMF By seems to be stronger in the 00:00–12:00 MLT sector compared to the 12:00–24:00 MLTs. Quantitative characteristics of the IMF By effect are presented and partly explained by the magnetospheric electric fields generated due to the solar wind and also by the position of open-closed boundary for different IMF orientation.


Introduction
The structure of high-latitude ionospheric plasma convection is controlled by the IMF and depends primarily on the IMF strength and orientation in the GSM Y-Z plane (Heppner and Maynard, 1987).If IMF B z is directed southward or equal Correspondence to: R. Lukianova (renata@aari.nw.ru) to zero, the ionospheric projection of the magnetospheric plasma circulation is a two-cell convection pattern with a fairly homogenous central anti-sunward flow.The viscous processes at the magnetospheric boundary form the similar, though less intensive, convection pattern.If B z is directed northward and strong enough, additional flow cells with a sunward cross-polar cap flow are developed.
The IMF B y modifies the twin-cell convection patterns and leads to the dawn-dusk asymmetry which is, in general, antisymmetric in the opposite hemispheres.In a given hemisphere the effect is quasi-mirrored with respect to the noon meridian for opposite signs of B y .Basic physical ideas of the IMF B y effects on magnetospheric electrodynamics were suggested more than 30 yr ago (Nishida, 1971;Jørgensen et al., 1972;Stern, 1973;Leontyev and Lyatsky, 1974).In previous studies the IMF B y -related dawn-dusk asymmetry in polar cap convection and the associated Svalgaard-Mansurov effect (Svalgaard, 1973) have been widely attributed to the magnetic tension forces acting on open field lines.The paradigm is as follows (Jørgensen et al., 1972;Khan and Cowley, 2001).The flows are believed to be due the effect of the magnetic tension force acting on the newly reconnected flux tubes in the presence of IMF B y .For B y > 0, open tubes are pulled westward in the Northern Hemisphere and added to the dawn side of the northern tail lobe, while simultaneously being pulled eastward in the Southern Hemisphere and added to the dusk side of the southern tail lobe and vice versa for B y < 0. Thus, the anti-sunward flow over the polar cap is asymmetric with stronger flows on the dawn side in the Northern Hemisphere and on the dusk side in the Southern Hemisphere for B y > 0, and vice versa for B y < 0. The overall effect on open field lines is an additional westward (eastward) circulation in the Northern Hemisphere if B y > 0 (B y < 0), and vice versa in the Southern Hemisphere.R. Lukianova and A. Kozlovsky: IMF B y effects in the plasma flow at the polar cap boundary The B y -related effect may be also interpreted in such a way that the convection asymmetry is formed by two elements (Nishida, 1971;Leontyev and Lyatsky, 1974;Kozlovsky et al., 2003).Firstly, the IMF B y component irrespective of its sign causes anti-sunward transpolar flow due to reconnection at the flanks of the magnetosphere.In the Northern Hemisphere this flow is stronger on the dawn (dusk) side.It adds a westward (eastward) circulation when IMF B y is positive (negative).Secondly, the IMF B y generates a voltage between the two magnetospheric lobes.The IMF is frozen in the solar wind plasma that moves away from the Sun with a typical speed of ∼300 km s −1 .As a result, the IMF generates an electric field E sw = −V sw × B IMF in the Earth-fixed framework.In particular, the IMF B y (positive direction from dawn to dusk) generates an electric field directed from South to North.From the voltage between the two polar caps, one can expect a radial electric field and, consequently, an azimuthal (around the pole) plasma flow, which is opposite in the northern and Southern Hemispheres.For instance, a negative IMF B y produces an eastward (westward) plasma flow within the northern (southern) polar cap.
In a steady-state case, such oppositely directed flows should not co-exist in the closed magnetic field line regions of the conjugate hemispheres because of high conductivity along the magnetic field lines.Hence it was commonly believed that the asymmetric plasma flows occur within the open polar caps forming lobe convection cells.However, recent observations have shown that the anti-symmetric electric field penetrates to the region of closed magnetic field lines.This phenomenon may be explained by the anomalous resistance associated with the inter-hemispheric field-aligned currents (Kozlovsky et al., 2003).Sandholt and Farrugia (2009) showed that the situation may be even more complicated.Namely, plasma convection on open field lines can be subdivided in two distinct stages corresponding to "newly open" and "old open" (reconnection has occurred some time earlier) field lines.Convection channels in the second stage play an important role in the excitation of an IMF B y -related dawn-dusk convection asymmetry.These flow channels indicate momentum transfer from the solar wind along the "old open" field lines associated with polar rain precipitation.In addition, the inter-hemispheric asymmetry may be due to the specific mapping of the reconnection electric field to the ionosphere.This was shown by Watanabe at al. (2007), who studied a topological model of the IMF-magnetosphere reconnection in many details.
Qualitatively, the basic effects of IMF B y on the high-latitude ionospheric convection have been understood by now.However, we need quantitative characteristics which are important for better understanding of the highlatitude magnetosphere-ionosphere coupling and for calibrating models of ionospheric electrodynamics.An interesting study on the plasma flow related to the IMF B y was performed by Khan and Cowley (2001), who analyzed a database of 300 h of tristatic ionospheric velocity measure-ments obtained overhead at Tromsø (66.3 • magnetic latitude) by the EISCAT UHF radar system.The authors found that significant flow variations with IMF B y occur predominantly in the midnight sector (21:00-03:00 MLT) and also in the pre-dusk (16:00-17:00 MLT) sector.The flows are directed eastward for IMF B y positive and westward for IMF B y negative, and with magnitudes of the order of 20-30 m s −1 nT −1 in the midnight sector and 10-20 m s −1 nT −1 in the pre-dusk sector.At other local times the IMF B y -related perturbation flows are much smaller, less than 5 m s −1 nT −1 .These observations relate to the region of closed magnetic field lines.
In the present paper we study the B y effect at essentially higher latitudes, in a vicinity of the polar cap boundary.We utilize a large (more than 1300 h) database collected during 8 yr at the EISCAT Svalbard Radar (around 75 • MLat).Our aim is to infer quantitative characteristics of the east-west ionospheric plasma flow associated with IMF B y .After description of the collected data (Sects.2 and 3) we present the statistical results of the dependence of the flow speed on IMF B y in different MLT sectors (Sect.4).In Sect. 5 we show the dawn/dusk asymmetry feature in association with the IMF B y .Discussion is given in Sect.6.The final section is a summary of the obtained results.

Data and method
The EISCAT Svalbard Radar (ESR) has a unique location that is convenient for monitoring the ionospheric plasma parameters in a vicinity of the polar cap boundary.For the present study the data were obtained from the Common Program 2 (CP2) ESR measurements.During 2000-2008, the radar was working more than 1300 h in the CP2 mode with the beam being periodically alternated between three positions (one vertical and two elevated at 63 • or 66 • to horizon with azimuth angles 172 • and 144 • , respectively).Under an assumption that the plasma flow is spatially uniform over the region of ESR observations and does not change during the scanning cycle, vector of the large-scale plasma flow in the F-region was calculated and the electric field vector was inferred, assuming that the F-region plasma experience the E ×B drift.These measurements refer to the heights of 200-300 km with spatial resolution of the order of 100 km in the horizontal plane.These data are related to about 74.5 • MLat.The scanning cycle was typically about 6 min.We used all triads of consequent antenna positions to calculate the vector flow.In this case each a measurement contributes to three data points.Thus, although the sampling rate of the ESR data is about 2 min, the data was averaged over the interval of about 6 min.
The ESR data are uniformly distributed along local time.The majority of the data (57 %) were obtained near the autumn equinox (September and October).For summer conditions (June-July) there existed 24 % of the data, while 16 % of the data were obtained during February-March (spring), and only 3 % of the data were obtained during winter (December).
For the present study we have calculated the azimuthal eastward component of plasma flow, V E , (i.e.along L-shell), which is at 20 deg.from geographic East to North.The plasma flow data were combined with the data on IMF Y-and Z-components (GSM) obtained from the OMNI database.The OMNI provides interplanetary parameters referred to the Earth's orbit (1 AU).We utilised the data averaged over 5min interval, which match to the radar data.The analysis is made separately for the northward IMF (B z > 0) and the southward IMF (B z < 0).
The main parameters studied are correlation and linear regression of V E with IMF B y .The correlation coefficient C corr calculated in the conventional way, and confidence intervals for the correlation coefficients are calculated using the following formula: where n is the number of data points and t γ is the inverse standardized normal distribution (for the 95 % confidence interval, t γ = 1.96).The linear regression coefficient K is calculated by the least square method to get a linear approximation where V E0 is a background part of azimuthal flow that does not depend on a sign of IMF B y .This flow basically relates to the two-cell convection pattern generated by the reconnection and viscous interaction on the magnetopause.Note that in the Northern Hemisphere, a duskward (i.e.positive) IMF B y generates a westward (i.e.negative) plasma flow and, vice versa, a negative IMF B y generates an eastward (i.e.positive) flow around the polar cap.Hence, the coefficient K is normally negative.Because of that we use reverse axis direction for plotting K below.Strictly speaking, the correlation coefficients are also negative, however, we use below the absolute value of the coefficients to characterize just a degree of linear dependence.

Response time of ionospheric convection to IMF B y
The first our step is to determine the time lag between the IMF B y conditions at 1 AU and the corresponding ionospheric plasma flow at the ESR location.The ionospheric convection response to a change in the IMF as measured by the spacecraft is a sum of the following time intervals: (1) the propagation time for the IMF to reach the at the front side magnetopause (assumed to be typically located at X GSE = 10 Re), (2) the magnetosphere-ionosphere communication time, and (3) the ionospheric convection reconfiguration time (Zhang et al., 2007).
In the present study the time interval ( 1) is neglected because we use the OMNI data which provides interplanetary parameters referred to the Earth's orbit.Thus we need to estimate the sum of intervals ( 2) and (3).There are two possible approaches to determine the convection response to a change in the IMF.The first is to examine each individual event of the IMF change to perform a search of the associated details of the response.Such an analysis is more appropriate for the study of the response to a sudden change in the IMF orientation (e.g.Hairston and Heelis, 1995;Ridley et al., 1998;Kabin et al., 2003;Lyons et al., 2003).The second approach is to average the data for revelation the most common features.Such approach is used for construction of the statistical models with binning of both IMF and ionospheric data (e.g.Heppner and Maynard, 1987;Papitashvili and Rich, 2002;Zhang et al., 2007).In the frame of this approach the slow variations of the IMF may be accounted for.
In order to determine the lag time which provides the largest acceleration of the azimuthal flow in response to a strengthening of the IMF B y , we compared the measured V E values with time series of the IMF B y .We considered various lags between the two data sets, from −100 to 300 min through 5 min.For each lag time, correlation (C corr ) and regression (K) coefficients between V E and IMF B y were calculated, separately for southward and northward IMF.Due to the Earth rotation, the radar data come from different MLT.With no separation on MLT, both the correlation and regression coefficients show maxima at lag times 35 min and 15 min under the south-and northward IMF conditions, respectively.However, the convection on day side may respond faster than that one on the night side (Lockwood et al., 1989).As one might expect different lag times for the day and night sides, we also calculate the coefficients separately for the 06:00-18:00 and 18:00-06:00 MLT sectors.In Fig. 1 the regression coefficient versus the lag time is presented.The left and right panel represents the southward and northward IMF conditions, respectively, whereas the top and bottom panel corresponds to the day and night side, respectively.Vertical lines indicate the lag times of 35 min (left panels) and 15 min (right panels).
One can see that under the southward IMF conditions the shape of dayside and nightside curves are quite similar with the lag time of about 35 min for both.When the IMF is northward, the situation is less clear and it is rather difficult to distinguish single maxima.The most probable dayside lag time can be determined at ∼15 min.The nightside lag time is rather uncertain within 0-70 min and there is no essential K dependency on the lag time within this interval.The lag time show a tendency to be larger at the night side, however, the maxima are rather flat.Below we assume the 15 min lag time for the northward IMF and 35 min for the southern IMF that is quantitatively consistent with previous estimates.We should emphasize that our aim here is to determine only the lag time that should be accounted for comparison the ionospheric data at ESR location with the OMNI data.These lags do not necessarily correspond to the rearrangement of the entire convection pattern in response to the IMF B y changes since at a given UT interval the radar is able to measure the response at the particular location but not an overall change in large-scale convection.However, the above consideration suggests that the B y -related flow averaged along the latitudinal circle near the polar cap boundary may be considered as established (relative to OMNI data) after a suitable lag time of 35 min and 15 min for IMF B z < 0 and B z > 0, respectively.

IMF B y control of the azimuthal plasma flow in different MLT sectors
The ESR measurements cover the entire MLT range with a fairly uniform local time distribution.We separated the whole set of velocity data into 3 h wide sectors centred at 00:00, 03:00, 06:00, 09:00, 12:00, 15:00, 18:00 and 21:00 MLTs. Figure 2 presents the dependence of the azimuthal plasma flow V E upon the IMF B y , when the IMF B z is southward.Each plot represents the corresponding MLT sector and the plots are organized according to the normal MLT circle so that the flows at different locations can be easily compared.Each dot in the plot represents the average over 100 actual data points.To obtain them, the data were sorted according to IMF B y in ascending order in a given MLT sector.After that an averaged value of V E and B y for each 100-point data sequence was calculated.Distributions of the dots demonstrate clearly a linear dependence of V E upon IMF B y .Namely, the eastward (westward) V E increases with the increase of negative (positive) B y .Straight line in each plot shows the linear approximation.It is easy to see, however, that the slope of the fit line (represented by K), its position about the axis and the scattering of points (represented by C corr ) are not the same at different MLT.In the dawn side the slope is steep, implying a stronger response of V E to the B y change.In the dusk side the slope is smaller.In the 12:00, 15:00 and 18:00 MLT sectors V E values are mostly negative, which indicate the predominantly westward flow that is likely a part of low-latitude return flow.In the noon  sector there are only two data points with positive (eastward) flow.This depicts the overall prevalence of the dusk cell that adds more westward flow near the polar cap boundary.
The prevalence of the westward flow at noon may be associated with the inherent asymmetry of the two-cell convection pattern observed even for B y = 0.The cells are not exactly symmetric, with a larger cell and a higher potential on the dusk side.The effect was schematically described first by Atkinson and Hutchison (1978).These authors pointed out that the day-night conductivity gradient in the polar cap Eregion ionosphere squeezes the antisunward convective flow to the dawnside of the polar cap.This is a likely cause of the lack of the mirror symmetry in the flow pattern that, as theory indicates, should exist in away and toward solar sectors in the Northern Hemisphere in summer.The day-night conductivity gradient in the E-region ionosphere causes such an effect due to the Hall current closure of the Region 1 field-aligned current.Ruohoniemi and Greenwald (2005) mentioned the fact that the statistical convection pattern for B y < 0 is most asymmetrical during winter, whereas that for positive B y > 0 is most asymmetrical during summer.Note that the majority of the ESR data are from summer/equinox.Sandholt and Farrugia (2007) presented the case studies of the dynamical evolution of dayside poleward moving auroral forms (PMAFs) in relation to plasma convection during intervals of southeast and southwest IMF orientations when |B y /B z | > 1.The observations were made during winter season in Ny Alesund, Svalbard at 76 • MLAT, i.e. almost at the same site as the ESR measurements.It was shown that PMAF activity is closely associated with vortical flows in the dawn-and duskcentered convection with clear B y -related dawn-dusk asymmetries.However, the large-scale convection does not show the identically mirrored response to the opposite signs of B y .
Thus a prevalence of the westward flow in Fig. 2 may be caused by the larger widening of the dusk convection cell in comparison with the dawn cell especially at the sunlit ionosphere.
Similar analysis was made for the northward IMF conditions.The results are presented in Fig. 3 which is organized in the same way as the Fig. 2.Even visual inspection of the plots reveals a weaker (relative to the B z < 0 case) relationship between V E and B y , especially in the dusk-postnoon  sector, where V E is westward irrespective of a B y sign, and near dawn where V E is primarily eastward.Such flow is likely a manifestation of the two-cell convection pattern with sunward (anti-sunward) flow on dusk (dawn).More tight dependence between V E and IMF B y is seen at the noonprenoon and midnight-premidnight hours.
The values of regression coefficient K and correlation coefficient C corr for the linear approximation of V E and IMF B y relationship are given in Table .The first column in the table indicates the central hour of 3-h MLT intervals where the data were collected.The second column presents the intensity of the background azimuthal flow V E0 .The third and the fourth column presents the regression and correlation coefficient, respectively.The fifth column indicates the total duration of the observations for each MLT sector.Totally, there were 1343 h of observations.Among them, 725 h were during periods of IMF B z < 0, and 618 h were during periods of IMF B z > 0. Mean IMF B z value for the two subsets of data was −2.6 and +2.8 nT, respectively.shows the position of the polar cap boundary which will be discussed in Sect. 5.For all parameters the eight points actually obtained for the corresponding MLT sectors are connected by the smooth curve using the Fourier interpolation.
From the two upper plots of Fig. 4 one can see the following.Both the regression and correlation coefficients show similar behaviour under the northward and southward IMF conditions, although the B y -related azimuthal flow is stronger when B z < 0. The MLT profile of K indicates that the IMF B y effect is maximal near noon and post-midnight (∼03:00 MLT) where its magnitude is about 60 m s −1 nT −1 for B z < 0. If B z > 0, K is about 40 m s −1 nT −1 for both the day and night sides.The deepest minimum is seen on dusk (∼18:00 MLT) for both signs of IMF B z .The other minimum seen at the morning hours (06:00-09:00 MLT) is less prominent for B z < 0 and essentially more prominent for B z > 0. Similarly, C corr exhibits two maxima and two minima.The first maximum is exactly at noon, while the second is slightly shifted towards the early morning hours.A remarkable asymmetry feature for both parameters, K and C corr , is that the dawnside minimum is less prominent than the duskside one.

Position of the polar cap boundary
The demarcation between the closed field lines at lower latitudes and the first open magnetic field line at higher latitudes defines a boundary enclosing the polar cap (known as the polar cap boundary, PCB).As mentioned in Introduction, the IMF B y generates the round-pole plasma flow within the region of open geomagnetic field lines in the polar cap and also in some part of closed magnetosphere adjacent the polar cap boundary.The flow is directed eastward (westward), if the IMF B y is negative (positive).Obviously, the flow intensity should increase with the magnitude of IMF B y , and that is seen in Figs. 2 and 3.
The IMF B y also affects both the size and the shape of polar cap (e.g. Park et al., 2006).Also, the auroral oval and the polar cap as a whole are shifted towards dawn (dusk) for B y > 0 (B y < 0) (Cowley et al., 1991;Trondsen et al., 1999).Hence, if the PCB moves with IMF B y changing, the location of the ESR appears to change with respect to the PCB.As a result, the radar may leave the polar cap and enter the region of the closed field lines.The PCB motion certainly contributes to the regression/correlation coefficient between V E and B y .The effect is the most expectable in the dusk and dawn sides, where the corresponding eastward and westward low-latitude return flow is a part of the two-cell convection.
Figure 5 shows, in frame of MLT and MLat, an average PCB estimated from the Tsyganenko and Sitnov (2005) (TS-2005) geomagnetic field model with the GEOPACK-2008 software (http: //geo.phys.spbu.ru/∼ tsyganenko/modeling.html).To obtain the PCB line, the ionospheric footprint of the last closed field line for each MLT hour was calculated first and then the points were smoothed by the Fourier interpolation.The mean IMF conditions during the period of observations, i.e.B z = ±2.7 nT and B y = ±3.4nT are taken as the model input parameters.Left and right panels of Fig. 5 show the PCB for B y < 0 and B y > 0, respectively.Red and blue line corresponds to B z > 0 and B z < 0, respectively.In each plot the dashed circle shows the latitude of the ESR location.From this figure one can see that according to the magnetic field model for B z = −2.7 nT, the ESR is preferably located on closed field lines in the dayside region and within the polar cap in the nightside region.For the northward IMF conditions (B z = +2.7 nT) the ESR is mostly located on closed field lines being closer to PCB on dusk (dawn) side for B y < 0 (B y > 0).
Even visual inspection of Figure 5 reveals (at least for B z > 0) that the PCB under the opposite B y conditions is not exactly mirrored with respect to the noon-midnight meridian.For instance, the comparison of the PCB at MLT = 06:00, for B y < 0, and MLT = 18:00, for B y > 0, shows the latitudinal displacement of 1-2  The largest PCB shift is seen at about 06:00 and 18:00 MLT.Comparison of the two lower plots in Fig. 4 shows that the largest shift of PCB corresponds to the minima in K.It is not surprising because essential contribution from the low-latitude return flow just on the dusk and dawn sides allows the PCB displacement.Within the two-cell convection pattern the flow reversal occurs in a vicinity of PCB.In the dusk side the return flow equatorward of PCB is directed to West while the transpolar flow poleward of PCB is directed to East.On the dawn side the directions are opposite.Let us consider the idealized situation when the ESR is situated, for instance, near the dawn meridian.If IMF B y becomes positive, it results in two effects.Firstly, an additional westward around-pole flow appears within the polar cap.Secondly, the whole polar cap is shifted toward dawn along the 06:00-18:00 MLT meridian.The ESR turns out within the region where the B y -induced westward flow is superimposed with the eastward transpolar flow.This leads to reduction in V E and decrease of the regression/correlation coefficient in the dusk side.If IMF B y becomes negative, the polar cap is shifted toward dusk side so that the effect is weak.Similar explanation is valid in the case when the ESR is situated near the dusk meridian.Overall, the IMF B y -V E dependence is the weakest at 06:00 and 18:00 MLT where the PCB displacement is maximal.
However, it is noteworthy that the duskside PCB displacement is larger than the dawnside one.The difference is more clearly seen, if B z > 0 (Fig. 4, lower panel).Figure 4 shows an additional asymmetry with respect to the noonmidnight meridian that can be hardly explained solely by the dawn-dusk displacement of the PCB.Indeed, an averaged value of the regression coefficient K in the dawn side (00:00-12:00 MLT) exceeds that in the dawn side (12:00-24:00 MLT).For both B z < 0 and B z > 0, the magnitude of K on the dawn side is almost twice larger than that on the dawn side (∼60 vs. ∼30 m s −1 nT −1 ).Similarly, the correlation between V E and B y is significantly higher on the dawn side.
6 Discussion Khan and Cowley (2001) have employed a database of ∼300 h of ionospheric velocity data obtained overhead of Tromso (63.3 • MLat) and have analysed them trying to detect the presence of flow effects associated with the IMF B y .It was found that significant flow variations with IMF B y occur in the midnight sector and also pre-dusk.The flow velocity and IMF B y are connected by the proportionality factor of 20-30 (10-20) m s −1 nT −1 at midnight (pre-dusk).At other local times the IMF B y -related perturbation flow are much smaller.
We extended the analysis to higher latitudes, where the IMF B y effect is the most prominent.The ESR (74.5 • MLat) measurements provide an opportunity to quantify the dependence of the azimuthal flow speed V E upon IMF B y just in the vicinity of the polar cap boundary.The velocity data are obtained from all MLT sectors.The fact that at some time the region of radar measurements is within the polar cap and at the other time it is on the closed field lines provides an additional interesting information on the convection flow, the polar cap displacement and the response of the magnetosphereionosphere system to the IMF B y .
The performed analysis revealed a linear dependence between V E and IMF B y at the majority of MLTs.This implies an acceleration of the eastward (westward) flow with an increase of negative (positive) IMF B y .The average proportionality factor between the two parameters is 48 (24) m s −1 nT −1 under the southward (northward) IMF conditions.But the fit line has different slopes at different MLTs implying the tightest relationship between V E and IMF B y near noon/post-midnight and the weakest one on dusk and dawn.
We found that the main factor modifying the response of V E is the displacement of the polar cap boundary along the dusk-dawn meridian controlled by the IMF B y .The PCB displacement can explain the existence of two minima and two maxima in the MLT profile of the V E and B y relationship.On the dawn and dusk sides the PCB shift reaches several degrees in latitude, according to TS-05 model.The radar therefore appears in the region where the B y -related azimuthal flow and the DP2 transpolar flow are oppositely directed, resulting in its mutual suppression.On the other hand, a tight relationship between V E and IMF B y is obtained at noon and near midnight, where the TS-05 shows no PCB shift.These observational results support the idea of the B ygenerated interhemispheric voltage, radial electric field and the round-pole plasma flow in the ionosphere (Lyasky, 1978;Kozlovsky et al., 2003).
In addition to the dawn/dusk minimum and noon/midnight maximum in the MLT profile a remarkable asymmetry feature is revealed in the sensitivity of the azimuthal flow to the change of IMF B y .Particularly, putting a separation line along the noon-midnight meridian and comparing the dusk and dawn sides, one can notice that for a given intensity of IMF B y the V E in the 00:00-12:00 MLT sector is generally larger than the corresponding V E in the 12:00-24:00 MLT sector.Also, the night-time maximum is not exactly at midnight, but shifted towards the post-midnight hours.
The dawn-dusk asymmetry can be associated with the large-scale convection.It is well known that the full convection patterns do not demonstrate a simple mirroring across the noon-midnight meridian if a sign of B y is changed (Ruohoniemi and Greenwald, 2005;Papitashvili and Rich, 2002;Lukianova and Christiansen, 2006;Lukianova et al., 2008).In the Northern Hemisphere, the duskside vortex of convection pattern expands over the noon meridian even without influence of IMF B y .The effect is interpreted as a result of the interaction of the initially mirror-symmetric IMF B y pattern with the ionospheric conductivity distribution or the magnetospheric topology.Large and intense duskside vortex contributes to the B y -related azimuthal flow just on the dawn side, where the response of V E to the IMF B y change is more pronounced.Interestingly, the TS-05 model predicts some dawn-dusk asymmetry of the PCB shift related to IMF B y .The PCB shift is slightly larger (by about 1 • MLat) on the dusk side than on the dawn side, the effect is seen better, if B z > 0.
The IMF B x component may play a role.Using global MHD simulations Peng et al. (2010) investigated the IMF B x effect on SW-M-I coupling.Although the analysis was limited to the case of IMF B y = 0 an interesting result has been obtained.In particular, for low Mach numbers and IMF B z < 0 the transpolar potential decreases with increasing B x .Peng et al. (2010) showed that in the presence of a positive (negative) IMF B x the merging line shifts northward (southward) on the day side and southward (northward) on the night side.This shift increases with increasing magnitude of IMF B x .In the Parker spiral the IMF B x is usually associated with the IMF B y .It is easy to imagine that in the presence of a finite B y < 0 the dayside reconnection points in both hemispheres would shift northward.It would add the eastward circulation on the dusk side at higher latitudes in the Northern Hemisphere while in the Southern Hemisphere the westward circulation would be added on the dawn side at lower latitudes.In the presence of B y > 0 the westward circulation will be added on the dawn side at lower latitudes in the Northern Hemisphere.In this case a non-zero B x component may enhance the interhemispheric asymmetry.In particular, in the Northern Hemisphere the effect may result in larger sensitivity of the ESR to the meridional flow on the dawn side.
However, so far we have not found a reasonable explanation of why the nightside maximum is not exactly at midnight but is shifted to ∼03:00 MLT while the paired dayside maximum is exactly at noon.The effect may be caused by unaccounted PCB motion related to localized processes which manifest itself mostly in the (pre-)midnight sector.The postmidnight sector can be the quietest region where the B yrelated azimuthal flow is not suppressed by other kinds of

Fig. 1 .
Fig. 1.The regression coefficients against the lag time for southward (left) and northward (right) IMF.Top and bottom panels correspond to the day and night side, respectively.Vertical lines indicate the lag times 35 min and 15 min.
E−ward flow (m/s) versus IMF By (nT) at 74.4 CGM Lat.

Fig. 2 .
Fig. 2. Dependence of the eastward plasma flow (V E ) on the IMF B y in the eight 3-h MLT sectors for the IMF B z < 0.
E−ward flow (m/s) versus IMF By (nT) at 74.4 CGM Lat.

Figure 4
Figure 4 summarises the information presented in Figs. 2 and 3 and shows the quantitative characteristics of the sensitivity of the azimuthal flow to the change of IMF B y at different MLTs.The values of C corr (upper panel) and K (middle panel) in different MLT sectors, separately for IMF B z < 0 and B z > 0, are presented.The lower panel of Fig. 4shows the position of the polar cap boundary which will be discussed in Sect. 5.For all parameters the eight points actually obtained for the corresponding MLT sectors are connected by the smooth curve using the Fourier interpolation.

1312R.Fig. 4 .
Fig. 4. The values of correlation C corr (upper panel) and regression K (middle panel) coefficients for the V E and IMF B y relationship for different MLT, separately for IMF B z < 0 and B z > 0. The bottom panel shows the shift of polar cap boundary caused by IMF B y changes for different MLT.Red and blue curve corresponds to B z > 0 and B z < 0, respectively.

Fig. 5 .
Fig. 5. Position of the polar cap boundary (in the MLT-MLat framework) according to the TS-2005 model for IMF B y = −3.4nT and B y = +3.4nT (left and right panel, respectively).Red and blue curve corresponds to B z > 0 and B z < 0, respectively.Dashed circle indicates the position of ESR.

Table 1 .
The background flow, regression (K) and correlation (C corr ) coefficients for the linear approximation of the relationship between V E and IMF B y ; total duration of the observations for each MLT sector.