Corresponding author: R. J. Miceli, Earth and Atmospheric Science, Cornell University, 3154 Snee Hall, Ithaca, NY 14853, USA. (email@example.com)
 Artificial field-aligned plasma density irregularities (FAIs) were generated in the E region of the ionosphere above the High Frequency Active Auroral Research Program facility during campaigns in May and August of 2012 and observed using a 30 MHz coherent scatter radar imager in Homer, Alaska. The purpose of this ionospheric modification experiment was to measure the threshold pump power required to excite thermal parametric instabilities by O-mode heating and to investigate the suppression of the FAIs by simultaneous X-mode heating. We find that the threshold pump power for irregularity excitation was consistent with theoretical predictions and increased by approximately a factor of 2 when X-mode heating was present. A modified version of the Another Model of the Ionosphere (SAMI2) ionospheric model was used to simulate the threshold experiments and suggested that the increase was entirely due to enhanced D region absorption associated with X-mode heating. Additionally, a remarkable degree of fine structure possibly caused by natural gradient drift instability in the heater-modified volume was observed in experiments performed during geomagnetically active conditions.
 A hallmark of ionospheric modification experiments is the generation of field-aligned plasma density irregularities, which are sometimes termed “artificial” irregularities or AFAIs (see reviews by Robinson , Frolov et al. , and Gurevich ). The irregularities are believed to be generated by thermal parametric instabilities [Grach et al., 1978; Das and Fejer, 1979; Fejer, 1979; Kuo and Lee, 1982; Dysthe et al., 1983; Mjølhus, 1990] and, having entered nonlinear stages of development, by resonance instability [Vas'kov and Gurevich, 1977; Inhester et al., 1981; Grach et al., 1981; Dysthe et al., 1982; Lee and Kuo, 1983; Mjølhus, 1993]. Most research has concentrated on F region AFAIs, although irregularities can be generated in the E region by pump waves with sufficiently low frequency. Examples of E region AFAI generation have been described by Djuth et al. , Coster et al. , and Hysell et al. , among others.
 Thermal parametric instability involves the linear mode conversion of an electromagnetic O-mode pump wave into an electrostatic mode (an upper hybrid wave, primarily) in the presence of zero-frequency field-aligned plasma density irregularities (the purely growing mode) that grow in amplitude under the action of wave heating, leading to instability. Wave trapping can ultimately occur, giving rise to resonance instability. Whereas the latter can be sustained with very low pump mode amplitudes, the former only occurs when the pump mode amplitude exceeds some threshold. Dysthe et al.  presented a detailed derivation of the threshold for thermal parametric instability in the F region, and Hysell et al.  generalized their work slightly for application in the E region, where the relatively high electron-neutral collision frequency and electron cooling due to inelastic collisions must be considered. Experiments with E region AFAIs showed reasonable agreement with the theory.
 Since the threshold electric field increases with electron temperature, it would seem to be possible to increase the threshold and inhibit thermal parametric instability using X-mode heating at an offset frequency where the X-mode interaction height matches the upper hybrid interaction height of the O-mode emission. The situation is complicated by at least two related factors, however. The first is Dregion absorption, which will also be increased by X-mode heating and inhibit AFAI generation in a manner unrelated to threshold considerations. The second is that the threshold for thermal parametric instability is influenced by the pump standing wave ratio at the upper hybrid interaction height, itself related to local absorption. For E region experiments, all of these factors may be important.
Frolov et al.  tested AFAI suppression in the F region by simultaneous X-mode heating. They concluded that both parametric decay instabilities and thermal parametric instabilities could be suppressed by X-mode heating at the optimal choice of interaction height. They identified three different timescales for the observed X-mode effects. Gustavsson et al.  conducted Fregion experiments at High Frequency Active Auroral Research Program (HAARP), using optical airglow as a diagnostic. In their experiments, X-mode heating was found to reduce 6300 Å emissions. This was interpreted in terms of two of the three aforementioned effects: increased absorption due to X-mode heating and the temperature dependence of the threshold electric field for thermal parametric instability.
 Afterward, Hysell et al.  performed related experiments at HAARP, examining whether X-mode heating altered the threshold for O-mode-induced E region AFAIs observed by coherent scatter radar. They found a drastic increase (more than a factor of 4) in the O-mode pump power needed for AFAI generation when half the HAARP array was emitting X-mode at a selected offset frequency. Simple calculations suggested that all three of the aforementioned mechanisms working together could account for the increase. However, subsequent numerical modeling (described below) showed that such a drastic effect was unlikely. A subsequent review of the experimental procedure raised doubt about whether the effective radiated power (ERP) of the O-mode radiation in the 2010 experiments was consistent during periods with and without X-mode heating. The experiments have consequently been repeated and accompanied by a more comprehensive modeling approach.
 We report on a new series of experiments performed at HAARP in the summer of 2012. For these experiments, O-mode pumping at fixed ERP was accompanied by X-mode heating at an offset frequency. E region AFAIs were detected using the 30 MHz coherent scatter radar. O-mode power levels were varied, and the effect of X-mode heating on the threshold for AFAI generation was reassessed. This time, we find that the power required for irregularity generation increases by a factor of 2 or less when half the HAARP array emitted X-mode at a suitable offset frequency. These findings are more consistent with numerical modeling results.
 In the course of the HAARP experiments, we observed an unusually high degree of spatiotemporal fine structure in the coherent scatter from the heater-modified volume on one occasion. While some structuring is often present in coherent scatter from AFAIs, this was atypical. We present the observations in question and briefly discuss their significance.
 The Ionospheric Research Instrument (IRI) at HAARP (High Frequency Active Auroral Research Program—62.39°N, 145.15°W) was used to generate E region AFAIs. Experiments were performed using O- and X-mode emissions, vertical pointing, and changing, finely graduated O-mode power levels. The O-mode emission frequency was 2.75 MHz for all of our experiments. When X-mode heating was used, the frequency was 3.169 MHz. This is the frequency at which the X-mode reflection height matches the upper hybrid interaction height.
 For the experiments described immediately below, the HAARP antenna array was divided into two subarrays, each an array of 7 × 12 elements. O- and X-mode emissions were generated using the two different subarrays. Using 100% of available subarray power, the 2.75 MHz O-mode and 3.169 MHz X-mode emissions would have had 79.9 and 81.1 dBW ERP, respectively. While the X- and O-mode rays deviate spatially as the waves propagate upward, the deviation is very slight below the upper hybrid interaction height in the E region. Consequently, the main beams of the two subarrays could both be directed toward zenith for these experiments.
 The ionosphere over HAARP was probed with a coherent scatter radar interferometer located on Diamond Ridge overlooking Homer, Alaska. The locus of perpendicularity is at precisely 100 km altitude over HAARP, as required for observing artificial E region FAIs monostatically. The experiments utilized a 13-baud Barker coded pulse with a 9 μs baud length. The interpulse period for the radar experiments was 2.46 ms or 370 km. Doppler velocities as large as ∼1000 m/s can be measured without frequency aliasing, which is necessary for observing natural auroral irregularities, although the Doppler shifts encountered during ionospheric modification experiments are typically an order of magnitude smaller. Specifications for the radar were given by Hysell et al. . The radar was recently relocated to Diamond Ridge and its array of receiving antennas expanded such that the longest interferometry baseline is presently 15 wavelengths long.
 Experiments were performed during the week of 6 May 2012. A geomagnetic storm caused significant absorption during the experiments starting on 8 May. Furthermore, the ionosphere was sufficiently dense during this time for sky waves and attendant radio interference to be present around midday when experiments were occurring. Data discussed in this section were selected from time intervals when absorption and interference were minimal. Additional experiments were subsequently performed the week of 5 August 2012, during geomagnetically quiet conditions.
 In the experiments to test X-mode suppression of AFAIs, O-mode signals were generated at power levels (relative to the maximum available power) that varied in steps according to the schedule shown in Table 1. Power levels were sustained for 10 s intervals. Over time, the power was incremented or decremented in discrete steps according to a quadratic formula so that the electric field incident on the ionosphere varied approximately linearly. A 1 min gap introduced at the end of the cycle provided a total cycle time of 5 min. X-mode heating, meanwhile, was performed at full subarray power throughout every other O-mode heating interval. Consequently, the overall cycle time for the combined experiment was 10 min.
Table 1. O-Mode Heating Power Levels for Onset Threshold Experimentsa
Start Time (s)
O-Mode Power (%)
The second column corresponds to the observations in Figures 1 and 2, while the third column corresponds to the observations in Figure 3. Power step levels were maintained for 10 s intervals. The power percentages are with respect to the power available from an IRI subarray. X-mode emission was at full subarray power.
 Figure 1 shows results from the experiments on 07 May 2012. The figure shows coherent scatter received by the 30 MHz radar versus slant range and time. The range extent of the AFAIs is mainly indicative of the horizontal width of the modified E region along the radar line of sight, which is northeastward. The spatial structure of the AFAI backscatter will be examined in more detail later in the paper. The Doppler shifts are relatively small throughout the modified volume, and only the lowest frequency Doppler bins resulting from spectral analysis are utilized in constructing this figure.
 Close inspection shows that backscatter from AFAIs was detectable within the first 20 s of the heating cycle beginning at 20:55 UT when X-mode heating was absent. This means that AFAIs could be generated using 11% of available O-mode subarray power or less, since 11% was the lowest power level in the schedule. In contrast, echoes could only be detected within the first 40 s of the cycle beginning at 21:00 UT when X-mode heating was present, meaning that 20% of subarray power was required. These results typified the 07 May experiments, which did not include sufficiently low power levels to determine the threshold for AFAI generation when X-mode heating was absent but which did demonstrate that the power threshold level at least doubled when X-mode heating took place. At the end of all the heating cycles, irregularities could be sustained over broad spatial regions at the lowest power level, with and without X-mode heating. In no case did the echo power saturate, i.e., echo power always increased with increasing pump power.
 Figure 2 shows comparable results obtained on 08 May 2012. Geomagnetic activity was already beginning to increase by this time. Here as is often the case during geomagnetically active times, the Doppler shifts of the echoes were significant and also structured. Backscatter power levels and heating efficiency were also somewhat reduced due to increased background absorption. The associated loss in heating efficiency made it possible to determine the threshold for AFAI onset both with and without X-mode heating. In this example, AFAIs were detectable within the first 30 s of the heating cycle (beginning at 21:40 UT) when X-mode heating was absent but just within the first 50 s of the cycle (beginning at 21:35 UT) when X-mode heating was present. The corresponding subarray power fractions required for AFAI generation were 15% and 25%, respectively, the ratio being somewhat less than a doubling.
 Similar experiments except with a heating power schedule that included lower and more finely graduated O-mode power levels were performed during the week of 05 August 2012. The O-mode power levels still followed a quadratic progression, as shown in Table 1. The heater power increased for 2 min from 0% to about 72% of the maximum available subarray power and then decreased for 2 min back to 0%. Figure 3 shows the echoes observed by the coherent scatter radar during these experiments. Irregularities were observed within 40 s for the heating cycle starting at 22:20 UT, in the absence of X-mode emissions. This corresponds to about 14% of the maximum subarray power. For the cycle starting at 22:15 UT, irregularities were first observed at 22:26 UT, 60 s into the experiment. This corresponds to about 24.5% of the maximum subarray power. Like the May experiments, the heating power required to excite irregularities with X-mode emissions was slightly less than twice what was required without them.
 In all cases we observed, the threshold power level for E-region AFAI generation approximately doubled under X-mode heating with a HAARP subarray excited at full power. In the remainder of this paper, we examine the degree to which this finding is consistent with contemporary thermal parametric instability theory.
 According to Dysthe et al.  (with adaptations from Hysell et al. ), the threshold peak electric field for thermal parametric instability in the E region at high latitudes can be estimated from:
Here n∘ is the background electron density, Te and Ti are the electron and ion temperatures, νen is the electron-neutral collision frequency, lcis the electron mean free path, L is the vertical plasma density scale length, kpis the pump mode wave number at the upper hybrid interaction height, and α is the zenith angle of the geomagnetic field. Also, Y≡Ωe/ω, and Z≡νen/ω. The δeexpression is the electron cooling rate due to inelastic collisions with neutrals, adapted from figures given by Gurevich  and accurate for electron temperatures below about 400 K.
 The term involving the factors 2kpδL represents the effects of the finite vertical extent of the interaction region. The significance of this term and the inelastic cooling rate term for E region ionospheric modifications was highlighted by Hysell et al. . Finally, r is the effective reflection coefficient, the ratio of the incident and reflected pump mode wave amplitudes at the interaction height. A pump mode standing wave is crucial for breaking the symmetry that would otherwise prevent thermal parametric instability for a circularly polarized pump wave [Das and Fejer, 1979; Dysthe et al., 1983]. The formulas above assume that the O mode is purely circularly polarized; a different formulation is required for near-vertical incidence at middle and low latitudes.
 Estimating the parameters for E region field-aligned irregularities at 99 km over HAARP and taking Te=Ti= 209 K, L= 6 km, and νen = 4 × 104 s−1predicts (according to the aforementioned theory) a threshold electric field of approximately 190 mV/m in the absence of X-mode heating. This is consistent with our experimental results, taking into consideration magnetoionic effects and nominal D region absorption, which we show how to estimate below. However, this estimate depends strongly on a number of parameters that are difficult to measure or estimate precisely, the vertical plasma gradient density scale length at the interaction height chief among them. A more accurate test of threshold theory can be performed by modifying ionospheric conditions in a limited, tractable way and then measuring the corresponding effect on the threshold for irregularity onset. This is the rationale for the present experiments.
 The threshold theory exemplifies three possible mechanisms for X-mode affects on the efficiency of AFAI generation by O-mode heating. The first is the explicit dependence of the threshold electric field on temperature at the upper hybrid interaction height. The second is X-mode enhanced D region absorption caused by X-mode heating. The third is the change in the effective reflection coefficient that also occurs when absorption near the interaction height increases. Some of these mechanisms can be measured experimentally. For example, Langston and Moore  measure D region absorption caused by X-mode heating. Quantifying the combined affects of these mechanisms, however, requires numerical modeling.
3.1 SAMI2 Model
 In order to evaluate and rank the possible influences on and temperature dependencies of AFAI generation efficiency, we have modified the SAMI2 model described by Huba et al. . SAMI2 solves for ionospheric plasma state variables (ion and electron number density, velocity, and temperature) in a dipole magnetic coordinate system using a semi-implicit parallel transport scheme followed by an explicit scheme for perpendicular transport. Photoproduction, chemical production and loss, and heating and cooling (including inelastic processes) are incorporated in the parallel transport scheme. Neutral state variables and neutral and electric field forcing are incorporated from empirical models. Time stepping is controlled through evaluation of the Courant condition. SAMI2 has been extensively validated [e.g., Huba et al., 2002, 2003].
 By default, SAMI2 incorporates seven ion species (H+, He+, N+, O+, N, NO+, and O), coupled through a system of 21 chemical reactions plus recombination reactions. This system is sufficient for E and F region simulations but lacks negative ions important in the D region. In particular, Enell et al.  have shown how increasing the D region electron temperature during ionospheric modification experiments increases the electron attachment rate and the ratio of negative ions to electrons, a process with implications for the overall absorption rate [see also Rodriguez and Inan, 1994]. We have consequently augmented SAMI2 with the introduction of some simplified negative ion chemistry.
 Following the prescription of Rodriguez and Inan , we have added provisions for a negative ions and positive cluster ions. The production and loss of these species in this prescription are adapted from the simplified set of rate equations given in the reference. Most importantly, the electron attachment rate leading to the production of negative ions is temperature dependent. We neglect transport for the new species and regard them as being locally produced and consumed.
 Two more modifications were made to SAMI2 for this analysis. One was to add a nested fine grid at D and E region altitudes. The grid places 3000 points between 50 and 110 km altitude. This is sufficient for resolving the Airy pattern of the pump mode. The other modification is the addition of a pump mode solver. The solver is based on the work of Thidé and Lundborg  and permits the calculation of the X- and O-mode pump waves based on SAMI2 model parameters (electron density, temperature, gyrofrequency, electron-neutral collision frequency, and magnetic field) at every time step. Electron-neutral collision frequencies were calculated using formulas given by Schunk and Nagy  and were adjusted to account for a velocity-dependent collision frequency using the formulation given by Sen and Wyller . The heating rate, estimated from the magnetoionic solver and reintroduced to the SAMI2 heat transport model, is given by
where n, ω, and I refer to the X-mode index of refraction, frequency, and intensity (Poynting flux), respectively.
3.2 Model Results
 The profiles in Figure 4 show diagnostic parameters for the case of negligible X- and O-mode power emissions. The left-hand panel shows background electron temperature and density profiles as calculated by the modified SAMI2 model. The center and right panels show the amplitude envelopes (solid lines) (relative to their amplitudes at 50 km altitude) and the index of refraction (dashed lines) of the X- and O-mode waves, respectively. Additionally, the right-hand panel shows a third curve (dash-dotted line) representing the value of the index of refraction for the O-mode wave corresponding to upper hybrid resonance. The interaction height is where this curve intercepts the actual O-mode index of refraction.
 Figure 5 shows similar information, this time for the case of X-mode emissions at full subarray power. (Since we are considering the threshold condition, the O-mode power is regarded as being too small itself to modify the ionosphere.) Comparing with the profiles in Figure 4 shows that the O-mode pump power reaching the upper hybrid interaction height has decreased significantly. This is due to the heater-induced absorption between about 70 and 88 km altitude. The O-mode wave amplitude reaching the interaction height with X-mode heating is about 75% what it is in the absence of heating. The power lost to absorption is consequently a little under half.
 The fraction of X-mode pump power reaching the interaction height is also greatly reduced. Consequently, the heating taking place at this altitude is modest. This can be seen more clearly in Figure 5b, which provides more detailed state parameter profiles near the interaction height. The temperature increase due to direct heating here is only about 10 K. While small, this increase was enough to shift the X-mode reflection height downward by about 60 m, making it fall slightly below the upper hybrid interaction height. This is an unanticipated complication, likely rendering the direct heating effect experimentally undetectable. Finally, direct heating near and above the upper hybrid interaction height can be seen to be insufficient to alter the effective O-mode reflection coefficient. The modeling consequently indicates that the only important effect of X-mode emission is to increase the absorption of the O-mode signal below the interaction height.
4 Fine Structure
 Although somewhat outside the main focus of this paper, the fine structure evident in Figure 2 is remarkable and warrants comment. Irregular variations in power and Doppler frequency with range across the heater-modified volume have been observed before with the Homer radar, although seldom as distinctly as in the 08 May 2012 data. It is often the case that the anterior and posterior sides of the modified region exhibit Doppler shifts with different directions, respectively. Banded striations with positive or negative changes in range over time are not uncommon. The fine structure generally but not always follows moderate geomagnetic activity. (We generally cannot run experiments during strong geomagnetic activity due to precipitation-induced absorption, which defeats the heating experiment and our ability to probe the modified region with our radar).
 In the example in Figure 2, the anterior (posterior) side of the heater-modified region was blue (red) shifted. Red- and blue-shifted striations with negative range rates are present during both heating cycles and appear to span the gap between them continuously, seemingly unaffected by the absence of heating. In other experiments, we have observed fine structure immediately upon heater turn-on after long periods of heating inactivity. It does not appear as though the fine structure is caused by the heating itself but appears to be natural.
 In order to inspect the fine structure in more detail, we have computed two-dimensional images of the heater-modified volume using the aperture synthesis imaging technique described by Hysell and Chau . The Homer radar has six antenna groups for reception, affording 15 nonredundant antenna baselines for interferometric imaging. The algorithm used is not diffraction limited and has an azimuth angle resolution as fine as about 0.5° for the highest signal-to-noise ratios we observe in these experiments. This translates to a transverse spatial resolution of about 5 km at HAARP ranges. This is coarser than the radar range resolution but still small compared to the dimensions of the heater-modified volume.
 Figure 6 shows six radar images representative of the two heating cycles in Figure 2. Curves outlining the HAARP subarray radiation pattern are superimposed on the images for reference. Homer lies to the southwest of the region shown in the figure.
 The images show that the echoes are red-shifted (blue shifted) on the northern (southern) side of the heater-modified volume. The demarcation line between red- and blue-shifted echoes is irregular and evolves with time. During the first heating cycle, two intense, compact scattering regions with opposite Doppler shifts appear on in the northwest quadrant of the volume. Over time, these move southward, remaining visible and in tact throughout both heating cycles. The various image features are not spurious but evolve gradually and systematically from image frame to frame.
 Overall, the imagery suggests clockwise circulation (vorticity) concentric with the heater-modified E region volume. Such circulation is not consistent with polarization of the thermally depleted volume by a background transverse electric field. Such polarization would divert the flow within the volume but not cause a concentric vortex [Hysell and Drexler, 2006]. The circulation is consistent with electron diamagnetic drift around the thermally depleted volume. However, diamagnetic drifts should be present routinely, whereas this circulation is not consistently present. Another possibility is polarization by upward field-aligned current. The direction of the attendant polarization electric field in a depleted E region volume would be inward, giving rise to counterclockwise electron convection. This scenario is further consistent with precipitation-induced absorption.
 The irregular flow boundary and the concentrated scattering regions within the larger heater-modified region suggest plasma irregularities created by natural plasma instabilities and simply illuminated by HAARP. The most likely candidate is gradient drift instability operating on precipitation-induced plasma density gradients in the E region in the presence of background convection electric fields. The instability would have been insufficiently strong to produce meter-scale irregularities and VHF backscatter but capable of producing kilometer-scale irregularities in the flow including drifting kilometer-scale vortices. The features in Figure 6 bear at least superficial resemblance to gradient drift waves and instabilities in the equatorial electrojet, only operating at longer scales [Hysell and Chau, 2006].
5 Summary and Conclusions
 The HAARP experiments were originally designed to test the theory of the threshold pump power required for thermal parametric instability and its dependence on electron temperature at the upper hybrid interaction height in particular. One long-term goal of the investigation is the ability to measure electron temperatures and temperature profiles using heating in an altitude regime that is difficult to investigate with conventional remote sensing approaches. However, the modified SAMI2 model indicates that the main effect of X-mode heating in our experiments was to increase D region absorption, with very little X-mode power propagating to or being deposited at the upper hybrid interaction height. This means that the aspect of the theory regarding the threshold dependence on temperature remains unvalidated. In order to compensate for the unanticipated experimental effects we encountered, the experiment could be modified to be run at higher X- and O-mode frequencies to reduce the amount of absorption. Relatively low E region densities restrict the range of frequencies available severely, however. Additionally, the experiment could be improved by varying the X- and O-mode frequency offset so as to make the heating near the X-mode reflection height coincide more closely with the upper hybrid interaction height.
 The experimental methodology described here could also be modified and used as a powerful diagnostic of D region absorption. Our model predicts D region O-mode absorption as a function of X-mode pump power level, and our threshold power measurement can validate the prediction. The accuracy of the prediction rests upon the accuracy of the underlying D region photochemistry model and on the heater-induced variations in the electron density profile it predicts. Precise absorption measurements for a variety of X-mode pump power levels would give incisive diagnostic information and permit precise photochemical model tuning and refinement.
 The experiments also produced evidence of fine structure in the E region convection during geomagnetically active periods which we tentatively attribute to natural gradient drift instability. In order to verify this hypothesis, further experiments should be conducted in conjunction with background convection electric field measurements, since the Doppler shifts we measured and corresponding, small-scale convection perturbations should be directly relatable to the background convection velocity in the case of gradient drift. The imminent launch of the Enhanced Polar Outflow Probe satellite should provide such an experimental opportunity for such an investigation.
 This project was supported by DARPA through contract HR0011-09-C-0099. Additional support came from the High Frequency Active Auroral Research Program (HAARP) and from the Office of Naval Research and the Air Force Research Laboratory under grant N00014-07-1-1079 to Cornell.