Corresponding author: A. Shindin, Radiophysics Faculty, Lobachevsky State University of Nizhni Novgorod, 23 Gagarina Av., 603950 Nizhni Novgorod, Russia. (email@example.com)
 A technique for studying inhomogeneous structure and motions in the ionosphere by measurements of the amplitude and phase characteristics of wideband radio pulses is developed and demonstrated in investigations of the HF-pumped ionospheric volume above the Sura facility. For the study of the vertical structure (electron density profile and vertical velocities) with high altitude and temporal resolution, a Tikhonov regularization algorithm for the solution of the inverse problem using phase sounding data is implemented. Horizontal velocity data are obtained by means of a correlation analysis of the reflected signals received by the diversity technique. The application of the technique described allows measurement of the HF pump-induced modifications of the electron density profile, particularly the plasma expulsion from the plasma resonance and upper hybrid resonance regions, with high accuracy.
 A powerful O-mode electromagnetic pump wave transmitted vertically into the bottomside ionosphericF region plasma excites a wide range of plasma processes leading to the appearance of artificial ionospheric turbulence, i.e., generation of different HF and LF plasma modes, plasma density inhomogeneities of scales from tens of centimeters to kilometers, enhancement of the electron temperature, electron acceleration and ionization, etc. [Gurevich, 2007].
 The pump-plasma interaction is known to be strongest near the pump reflection heightzr at which fp(zr) equals the pump frequency f0, and near the upper hybrid (UH) resonance height zUH where fUH = fp(zUH) = (f02 − fc2)1/2 [here fp = (e2N/πm)1/2, and fc = eH/2πmc is the electron plasma frequency and the electron cyclotron frequency respectively, e and m is the electron charge and mass, N the electron density, c the speed of light, H the geomagnetic field strength]. This corresponds to existing theoretical concepts [Al'ber et al., 1974; Perkins et al. 1974; Grach et al., 1977; Vas'kov and Gurevich, 1977] and is confirmed by investigations of the HF-pumped ionospheric volume by multifrequency Doppler sounding (MDS) [Vas'kov et al., 1986; Berezin et al., 1991; Lobachevsky et al., 1992], which have revealed plasma expulsion from the resonance regions. Some MDS experiments were performed also for pump frequencies close to the 3rd [Lobachevsky et al., 1992], 4th [Grach et al., 1997] and 5th harmonics of the electron cyclotron frequency nfc. During MDS experiments few low-powerO-mode waves with frequenciesfi, i = 1, 2, . ., k around f0 were used to probe different parts of the ionospheric plasma in or near the interaction regions (k = 8 in Berezin et al. , Lobachevsky et al. , and Grach et al. ). Measurements of time variations of probe wave phases allowed measurement of the density profile modifications. The difference between adjacent probe wave frequencies (fi+1 − fi) in these experiments was ≥20 kHz in order to cover an extended altitude region. However, with such a large separation between the probe wave frequencies, the distance between neighbor “probing altitudes” in the ionosphere was typically 0.5–1 km. To increase the altitude resolution, in order to study in detail the evolution of the electron density profile, we need a smaller separation between the probe wave frequencies, but keeping a wide frequency range, providing therefore probe wave reflection from an extended altitude range. For this purpose a method of sounding the HF-pumped ionosphere by powerful short (broadband) radio pulses was developed [Sergeev et al., 2007] and applied in experiments at the Sura facility in 2008 [Shindin et al., 2010] and 2010. During the experiments the Sura facility transmitters were used both for radiation of the pump wave and the probing waves. An additional advantage was realized due to use of broadband digital receiver with a high dynamic range for data registration.
 In this paper the method of sounding and data processing is described for the results obtained in the experiment performed in September 2010. The pump and probing wave radiation scheme is described in section 1. Section 2 contains a brief description of the regularization algorithm for reconstruction of the electron density profile in the ionosphere from the data obtained; the corresponding experimental results are presented in section 3.
 The use of powerful broadband pulses and broadband radio receiver also allows the improvement of efficiency and altitude resolution of the horizontal velocity measurements by correlation analysis of the reflected pulses received by spatially separated antennas (the diversity technique). In section 4 we present results of reconstruction of the 3D velocity field in the ionosphere during HF pumping, where the vertical motion was obtained by MDS and the horizontal motion by correlation analysis.
2. Experimental Setup
 The experiments were performed at the Sura radio facility near Nizhniy Novgorod, Russia (geographic coordinates 56.15°N, 46.10°E) in September 24, 2010. Due to ionospheric and interference conditions, the pump frequency was chosen to be f0 = 4740 kHz. Three 30 min pumping sessions at 16:49, 17:49 and 18:49 LT were carried out with the time schedule of the radiation illustrated by Figure 1. Two Sura transmitters radiated the quasi-continuous pump wave at the frequencyf0 with an effective radiated power P≈ 60 MW. During the 20 min long quasi-continuous pumping, a high duty cycle pulse radiation (pulse durationτ = 70 ms, interpulse period T= 100 ms, the off-duty ratioQ = T/τ ≈ 1.4) was radiated to produce the ionospheric disturbances. Simultaneously, short (τ ≤ 200 μs) and powerful (P ≈ 20 MW) pulses with the same interpulse period T = 100 ms (Q ≥ 500) at two frequencies fDW = f0 and fDW = f0− 200 kHz were radiated as diagnostic waves for probing the structure of the pumped volume. The low duty cycle was used during the whole 30 min session including 20 min of the quasi-continuous pumping. During the pumping, the short pulses were radiated within 30 ms pauses. The two diagnostic waves were radiated by one of the two ‘pumping transmitters’ atfDW = f0 and by the third Sura transmitter at fDW = f0 − 200 kHz. The power of the sounding transmitters was sufficient to create a wide spectrum of diagnostic waves (up to 300 kHz for each transmitter), with a small average power 〈P〉 = P/Q≤ 40 kW, far below the thresholds of the generation and maintenance of the pump-induced ionospheric plasma instabilities [Gurevich, 2007; Al'ber et al., 1974; Perkins et al., 1974; Grach et al., 1977; Vas'kov and Gurevich, 1977]. Therefore, under the combined radiation mode, the quasi-continuous pump wave created a perturbation in the ionosphere, particularly at the plasma resonance (pump wave reflection level) and at the UH resonance, while the pulse sounding simultaneously provided a diagnostics of the heated volume of the ionosphere [Berezin et al., 1991; Sergeev et al., 2007].
 The size of pump-induced plasma perturbations along the geomagnetic field is determined by electron thermal conductivity and plasma diffusion. It reaches several tens of kilometers and is comparable to the characteristic scale of the ionospheric layer. Changing the frequency mismatch Δf = fDW − f0 makes possible to investigate the properties of the plasma turbulence near the center (where fDW = f0) and at the periphery of the pumped volume by measuring the amplitude and phase characteristics of the probe waves. The radiation scheme of the Sura facility used in the experiments and described above is shown in Figure 1a. The diagnostic pulse radiated in the short 30-ms pause of quasi-continuous radiation mode at a frequencyf0 is shown on the time zoom. An example of real oscillogram for this time interval is shown at Figure 1b. It is well seen the ground wave (GW) signal and the sky wave multiple reflection signal of the 200-μs short pulse. The spectrum of the 1st reflection signal which contains two diagnostic c pulses at frequencies f0 = 4740 kHz and fDW= 4540 kHz at the beginning of quasi-continuous heating is shown inFigure 1c. A background noise spectrum is also shown here. The use of broadband radio receiver and a specially developed signal processing algorithms have allowed the study of the evolution of amplitude and phase of the various spectral components of the diagnostic (probing) reflected signal, which passed the pumped volume twice, in a wide (totally ∼500 kHz) band [Sergeev et al., 2007]. The spectral width of the reflected probing signal for each of Sura transmitters (up to 300 kHz) was provided by the high effective power of the Sura transmitters and the dynamic range of broadband receiver ∼96 dB. The frequency resolution of the analysis was chosen to be 1 kHz, the temporal resolution was determined by the interpulse period T = 100 ms. The results of measurements of the temporal evolution of phase φi(t) for different spectral components (i) of ionospherically reflected signal (Figure 2a) provided data for further reconstruction of electron density profile in the ionosphere and its temporal evolution N(z, t) by using an algorithm described in section 2. The initial (reference) profiles were obtained by processing ionograms registered during the experiment every 10 min by ionosonde. Temporal variations of the reflection heights ΔZ(fi, t) of different spectral components at fi, obtained by solving the inverse problem of MDS (reconstruction of electron density profile N(z, t)) allow easy calculation of vertical velocities of the displacement ∂ΔZ/∂t at different fi.
 The reception of the reflected sounding wideband pulses with three separated antennas allows us to determine the horizontal velocity components in the ionosphere by correlation analysis of signals from different antennas. Three squares of 4 dipoles each were selected at three of the corners of the full diagnostic antenna array (24 dipoles), namely the southern (1), central (2), and eastern (3) squares were used as the receiving antennas (see Figure 1d). The distance between the electrical centers of the squares was d = 84 m.
 The use of two kinds of analysis methods to be described gives the three-dimensional velocity field of the plasma in the perturbed region: horizontal components by the correlation measurements of signals from different antennas and the vertical component by the phase measurements.
3. Algorithm for Inverse Problem Solving for Electron Density Profile Dynamics Based on Phase Sounding Data
 For solving the inverse problem of reconstruction of the temporal evolution of the electron density profile in the HF-pumped volume based on the phase (Doppler) sounding of the ionosphere, we consider the case of vertical pumping assuming the horizontal size of the pumped volume to be determined by the facility antenna pattern and large enough to consider the propagation of the pump and sounding waves in plane-layered medium (plasma) with electron density profileN(z, t), where z is the height, t is the time.
 Let the ionospheric perturbations and, therefore, the electron density profile N(z, t0) changes start at the time t = t0. After some time τN a new profile N(z, t1) will be established. The effect of the ionosphere pumping on the profile can be characterized by the perturbation function
which has to be determined on the basis of the phase measurement data. Let's make some simplifying assumptions with respect to the unknown function ΔN(z, t), which take into account the main properties of real perturbations. First, let's suppose that ΔN(z, t) is nonzero in the altitude range [z1, z2], being ΔN(z, t) = 0 outside of this interval. We also assume that the functions ΔN(z, t) and N(z, t) are smooth functions of z. So, geometric optics can be used for calculating parameters of the sounding radio waves. In this case, propagating from the ground up to their reflection points zref and back, each of the sounding waves at the angular frequency ω = 2πf suffers the phase incursion [Ginzburg, 1964]:
where ωp(z, t) = 2πfp is the angular plasma frequency, n(ω, ωp) is a wave refractive index. The reflection altitude zref is determined by the condition
Formulas (2) and (3) need to be written separately for each magnetoionic component. In general, the reflective index of ordinary (O) and extraordinary (X) waves in a magnetized plasma is given by [Ginzburg, 1964]:
where X = ωp2/ω2, Y2 = ωc2/ω2, ωc = 2πfc is the cyclotron frequency, θ is an angle between the geomagnetic field and wave vector. For the vertical propagation above the Sura facility θ = 18.5°. The minus sign in equation (4) corresponds to the refractive index of an ordinary wave and the plus sign to the extraordinary wave.
Here K (ω, ωp) = dnX,O (ω, ωp)/dωp is a kernel of the integral equation, g(ω) is the angular plasma frequency at the reflection point, which for an O-wave isg(ω) = ωand for an X-wave isg(ω) = (ω(ω − ωc))1/2, ΔZ(ωp, t1) − z(ωp, t0) is the altitude shift, i.e., the difference between the sounding radio wave reflection heights at the current and initial times. In the expression (5) we have changed variables of the integration in comparison with (2): we integrate over the plasma frequency ωp instead of the altitude z. It is taken into account that at the reflection point n(ω, g(ω)) = 0, and at the entrance to the plasma layer (when z = z1 and ωp = ω1) ΔZ(ω1) = 0. The left hand side y(ω) in (5) is to be determined from the experimental data.
 For the exact expression of the refractive index for magnetized plasma (4), the equation (5) cannot be reduced to the Abel integral equation. In this case, we can apply regularization algorithms developed by Tikhonov et al. . The efficiency of solving equation (5) by these algorithms was investigated by Kim  numerically for specified models of ΔZ(ωp). According to Tikhonov et al.  an approximate solution to the integral equation (5) by the regularization method can be obtained by looking for the minimum of the smoothing functional:
where α is a regularization parameter,
is a residual norm in the space L2, the left hand side of (5)yδ being considered to include errors of phase measurements, and
is the norm of the solution in Sobolev space W21. The regularization parameter is chosen in accordance with the generalized discrepancy principle by solving the equation for the generalized discrepancy:
The generalized discrepancy principle allows us to match the accuracy of the solution ΔZ with the accuracy δy of the left hand side in (5). Functional (8) involved in (6) with a weight α restricts the solution ΔZ(ωp) and its first derivative ΔZ′(ωp) preventing strong variations of ΔZ(ωp) due to errors of phase measurements.
 After performing the finite difference approximation of the functional Mα(ΔZ) (6), the problem of finding its minimum is reduced to solving a system of k algebraic equations for ΔZ(ωi), i = 1 … k for each time, where k is the number of sounding frequencies:
where Bα = B + αC, B and C are the coefficient matrixes, D is the vector involving the experimental data; B, C and D are obtained by the procedure of finite difference approximation of M(ΔZ).
 Solving (10) for each time and thus calculating ΔZ(ω, t) we are able to follow the evolution of the electron density N(z, t) over the height from the initial electron density profile N0(z). For this we shall transform ΔZ(ω, t) to ΔZ(N, t) by using the univocal relation between the plasma frequency at the radio wave reflection point and electron density, for the ordinary polarized wave ω = ωp = (4πe2N/m)1/2. If z(N0) is the altitude as a function of electron density for the initial profile, then
is the dependence of the reflection height of radio wave on the density. Then we find the required distribution N(z, t) by calculating the inverse of (11).
4. Electron Density Profile Reconstruction
 For reconstruction of the temporal evolution of the electron density profile in the HF-pumped ionosphere by measurements of the phase changes for different spectral components of the wideband diagnostic signal we used a spectral decomposition of over 500 frequencies with a stepδf = 1 kHz. The temporal resolution was determined by the interpulse period T = 100 ms. The data processing algorithm can be described as follows. For each pumping session, we had taken the start time of the data registration at t0 = −60 s. The start time of pump wave radiation was denoted as t = 0. Then we created an array ΔΦ(ω, t) = φ(ω, t) − Φ0(ωmin, t) for phase incursion of probe waves φ(ω, t), defined by (2), where Φ0(ωmin, t) is the phase of the lowest frequency of the probe waves ωmin (Figure 2a). In the Figure 2a the phase variation ΔΦ of different spectral components versus time t are shown, the bold lines here and below correspond to the frequency components reflected at the plasma resonance [f = fp(z0) = f0] and upper hybrid resonance [f = fUH = fp(zUH)]. In our experiment fmin = ωmin/2π, = f0 − 350 kHz. The probe wave at this frequency was reflected a few kilometers below the upper hybrid resonance of the pump wave. According to Kim and Panchenko , the subtraction of Φ0(ωmin, t) allows us to remove from the influence of processes that change the plasma density in lower layers of the ionosphere. Such processes in the heating experiments are a diffusion of plasma along the geomagnetic field and a disturbance of the ionization-recombination balance in the lower ionosphere due to the temperature dependence of the recombination coefficient, which provided an enhancement of the electron density [Gurevich, 1978].
 For each time sample, the integral equation (5) was solved using the Tikhonov regularization method. An example of the evolution of the reflection heights for different frequencies is shown in Figure 2b. Here the pump-induced variations of the reflection altitudes for different frequencies of sounding wave ΔZ(ω, t) are shown for the time interval t = 0–1200 s after subtraction of the “initial” altitude shift ΔZ(ω, 0) obtained at the time of the pump wave switch on. The data presented were averaged over 21 frequency samples ωi using running average. This was necessary to smooth the effects of multipath propagation of the probe waves in the perturbed region of the ionosphere. In the Figure 2b, altitude variations are shown for individual spectral components with a step size of 20 kHz. For clarity, a constant shift of 600 m at t = 0 is added to each component. From the figure it is well seen that, on the background of the natural trend, an increase of reflection altitudes by up to 1000 m during the pumping is observed for the whole range of analyzed frequencies. After the pumping switch off (t = 1200 s), by contrast, there is a decrease of reflection altitudes. This corresponds to a decrease (expulsion) of plasma density during the pumping and an increase (inflow) after the pump switch off. The characteristic times of development and relaxation of the perturbations are about few tens of seconds. This is illustrated in more detail in Figure 2cwhere the evolution of the reflection heights is zoomed for the time of the quasi-continuous pumping switch on (t = 0) and switch off (t = 1200 s) with a constant shift of 300 m at t = 0 and t = 1200 s. In addition, in Figure 2dthe behavior of reflection heights near the plasma resonances at the switch on of the quasi-continuous pumping is shown with maximal temporal resolution (0.1 s) and frequency resolution of 5 kHz (with 3-kHz running average) and for a constant shift of 75 m att = 0. It is seen that the density disturbances appear already during the 1st second of the pumping near the plasma resonance and after 2–3 s of pumping near the upper hybrid resonance.
 Knowledge of the N(f) allows us to translate height variations to perturbations of the plasma density. Figure 3a shows the altitude profile of the relative variation of the plasma density ΔN/N with respect to the initial value N(t = 0, z), ΔN/N = [N(t, z) − N(t = 0, z)]/N(t = 0, z), at the 110th second of the pumping, t = 110 s. The altitude of the pump wave reflection and upper hybrid resonance at t = 110 s is denoted by z0 and zUH, respectively. From the figure it is seen that during the pumping the most intensive plasma density perturbations occurred near the plasma resonances and are negative (ΔN < 0) with an average of about 1.5% of the initial profile. Characteristic spatial scales of the perturbations are 200–500 m. The same values for the spatial scales and density perturbations were obtained in the experiment of September 2008 [Shindin et al., 2010]. Notice that during the long (20 min.) pumping the quasi-regular structure of the electron density profile is destroyed (seeFigures 2b and 3b), most probably, due to excitation of large scale plasma irregularities.
 On the basis of ΔZ(ω) calculations, it is easy to restore the temporal evolution of the electron density profile N0(z) obtained from the ionogram by replacing data array ΔZ(ω, t) to N(ω, t). It should be noted that the accuracy of determining the profile from ionograms is low (the error in determining the height can be several kilometers), whereas the relative changes in the profile by means of the phase measurements are determined much more accurately. An example of the dynamics of the reconstructed profile with a step of 120 s is shown in Figure 3b. In the figure, the successive profiles N(z) are shifted for clarity along the absolute scale of the heights by 1 km per 60 s downward relative to the pumping switching on time t = 0. The thick profiles in the panel correspond to the pump wave switch on and switch off. The six last profiles were obtained with the reference ionogram at t = 1740.
5. Vertical and Horizontal Plasma Motion in the HF-Pumped Ionosphere
 Knowledge of the temporal evolution of the reflection altitude of the different probe waves (different spectral components of wideband diagnostic pulses) allows us to construct the evolution of the vertical velocity in the ionospheric plasma VV = ∂ΔZ(ω, t)/∂t. Here VV is the velocity of the vertical motion of plasma density at a certain magnitude.
Figure 4 shows the vertical velocities VV for the time −5 s < t< 10 s for three successive pumping sessions (each session contains 20 min of quasi-continuous pumping followed by 10 min of pulse diagnostics). Positive velocity values correspond to upward motion of a certain plasma density level while negative values correspond to downward motion. It is seen that velocities up to ±40 m/sec are observed already during the first second for probe waves with frequencies near the pump wave frequencyf0(i.e., waves that reflect near the pump reflection point), and in a wider frequency range in the second-third seconds. After 2–3 s the plasma expulsion from the upper hybrid resonance region began. It appears that the perturbations at the UH resonance start at that altitude since the onset of the perturbations at the lower and higher altitudes is increasingly delayed with increasing distance from the UH resonance (see alsoFigure 2d). It is also clearly seen that the vertical motion becomes increasingly turbulent in the successive sessions.
 An overview of the vertical velocity during these pumping session is displayed in Figure 5 in the left column (running averaging was used over 151 time points (corresponding to 15 s) and 5 probe wave frequencies (corresponding to 5 kHz interval of averaging).
 For calculating the horizontal speed and direction of motion of the diffraction pattern, we used a correlation analysis described in Briggs et al.  and Kazimirovskiy and Kokourov . For this it's needed to determine the time lag τi−j, for which the correlation coefficient ρi−j (τ) between signals Ri and Rj recorded at the antennas i and j attains its maximum. (i, j = 1,2,3). Here
t is the time, τ is the time lag between the signals. Denoting the drift velocity of the diffraction pattern as VH, and the drift direction clockwise from the north as Φ, for the layout of antennas corresponding to Figure 1d we obtain:
where the indices 1, 2 and 3 correspond to the north, central and west antennas respectively, and d is the distance between the antenna centers. (Figure 1d).
 The values obtained for the magnitude (VH) and direction (Φ) of the velocity of the horizontal motion of the diffraction pattern at the reflection heights of the probe waves at different frequencies in a 500 kHz bandwidth are shown in Figure 5 in the middle and right columns. For this figure, 500 kHz probe frequency range corresponded to ≈8–11 km altitude range. The reflection heights of the pump wave (f0 = 4740 kHz) for the three successive sessions were 207 km, 219 km and 217 km respectively). It is seen that a sharp velocity jump is observed when the pump wave switches on. However, we believe that it's spurious, since the simultaneous development of the anomalous absorption of the pump wave and probing signals from all three antennas. From the middle column of Figure 5 it is seen that VH decreased with the probe wave frequency and its reflection altitude increase. Notice that in earlier observations performed under natural conditions (without pumping), an inverse dependence of the horizontal velocity VH on the altitude was observed in Grishkevich and Mityakov . In our experiments we found weak variation of the velocity magnitude of the diffraction pattern and a more significant change of velocity direction when pumping was turned on (for t > 0) and turned off (t > 1200 s, stronger effect). In addition, at times t∼ 500–800 s after turn on, approximately in the middle of the all three pumping sessions, noticeable changes of the magnitude and direction of the horizontal drift and vertical velocities were observed. This could be a sign of large-scale wave motions in the ionosphere. Also, a correlation between the horizontal drift and vertical random motion can be noticed during and after the pumping (Figure 5). The detected variations of the magnitude and direction of the drift due to the pumping switch on and off do not have an explanation yet.
 In this paper we briefly describe a method using applications of broadband radio signals to diagnose structure and dynamics of the HF-pumped volume of the ionosphere. The method is based on sounding the ionosphere by short (wideband) pulses radiated by the heating facility during pauses in the HF pumping, as well the use of broadband radio receivers and digital data processing. The results of applying this method to data obtained at the Sura facility are also presented. They concern the pump-induced electron density profile dynamics as well the vertical and horizontal plasma motion in the heated volume of the ionosphere.
 The spectral width of the registered signal, which can be easily translated into the range of reflection heights of the signal spectral components, is determined by the duration and intensity of the pulse, as well as the dynamic range of the receiving equipment. The use of a broadband radio receiver and digital processing methods allows us to analyze the phase variation of the different spectral components with high frequency (δf) and time (δt) resolution. In our case the time resolution is determined by the interpulse period. In contrast to novel digital ionosondes with time resolutions ∼10 s and MDS facilities of the previous generation with a small number (up to 9) of carrier frequencies of the sounding waves, specific values δt and δf (0.1 s and 1 kHz in our experiments) were determined from the data analysis of the temporal (>δt) and spatial scales of the processes responsible for the redistribution of the electron density in the ionosphere. The solution of the inverse problem for the electron density profile reconstruction using phase measurements is performed with the Tikhonov regularization method. The horizontal velocity data were obtained by means of the diversity technique and correlation analysis of the ionospherically reflected signals. The obtained experimental results, such as the expulsion of ionospheric plasma from the plasma resonance and the UH resonance of the pump wave, characteristic values of the perturbations of the plasma density and the time of their development and relaxation qualitatively correspond to the results obtained by other diagnostic methods of ionospheric plasma such as incoherent [Duncan et al., 1988] and field aligned scattering of radio waves [Gurevich, 2007; Frolov et al., 1997], in situ measurements [Kelley et al., 1995], stimulated electromagnetic emission measurements [Gurevich, 2007; Leyser, 2001], as well as with the use of the MDS facilities of the previous generation [Vas'kov et al., 1986; Berezin et al., 1991; Lobachevsky et al., 1992; Grach et al., 1997]. However, owing to small number of sounding waves in previous MDS facilities the regions of plasma and UH resonances had been investigated in separate experiments. In addition, the spatial and temporal resolution in our measurements and, hence, the accuracy of determining the electron density profile is much higher.
 The joint application of the phase sounding (MDS technique) and a reception of the reflected sounding wideband pulses with three separated antennas (a diversity technique) allows us to analyze both vertical and horizontal motions in the ionosphere by phase and correlation analysis and reconstruction of the 3D velocity of the plasma in the perturbed region. So far only a few successful experiments to investigate the ionospheric plasma during HF pumping have been performed by the method described. To develop an empirical model of plasma motions in the HF-pumped ionosphere it will be necessary to conduct further experiments under different ionospheric conditions. Work on detailed interpretation of the data obtained is also required.
 We plan to apply the method described in the paper in further ionospheric modification experiments for different ionospheric ambient conditions at different heating facilities.
 The work was supported by RFBR grants 11-02-00125 and 12-02-00513.