Stimulated electromagnetic emissions spectrum observed during an X‐mode heating experiment at the European Incoherent Scatter Scientific Association

An extraordinary (X‐mode) electromagnetic wave, injected into the ionosphere by the ground‐based heating facility at Tromsø, Norway, was utilized to modify the ionosphere on November 6, 2017. The high‐power high‐frequency transmitter facility located at Tromsø belongs to the European Incoherent Scatter Scientific Association. In the experiment, stimulated electromagnetic emission (SEE) spectra were observed. A narrow continuum occurred under cold‐start conditions and showed an overshoot effect lasting several seconds. Cascading peaks occurred on both sides of the heating frequency only in the preconditioned ionosphere and also showed an overshoot effect. These SEE features are probably related to the ponderomotive process in the X‐mode heating experiment and are helpful for understanding the physical mechanism that generated them during the X‐mode heating experiment. The features observed in the X‐mode heating experiments are novel and require further investigation.

The SEE spectra generated by ordinary (O-mode) polarized EM waves have been studied extensively in the past several decades at different heating facilities, such as those in Tromsø, Norway (Stubbe et al., 1984(Stubbe et al., , 1994Leyser et al., 1990;Fu et al., 2015), in Russia (Leyser et al., 1993;Frolov et al., 1999Frolov et al., , 2001Sergeev et al., 2006), at the High Power Auroral Stimulation Observatory (HIPAS) in Fairbanks, Alaska (Armstrong et al., 1990;Cheung et al., 1997), at the High-Frequency Active Auroral Research Program (HAARP) in Gakona, Alaska (Bernhardt et al., 2011), and in Arecibo, Puerto Rico (Thidé et al., 1989(Thidé et al., , 1995. In past decades, researchers have reported many features of the SEE spectra and their generation mechanisms. Stubbe et al. (1984) introduced several features of SEE spectra, including the "downshifted maximum" (DM) and "upshifted maximum" (UM), which are downshifted and upshifted from the heating frequency f 0 by approximately 9-15 kHz, respectively; the 2DM or 3DM, which are 2 or 3 times the frequency shifts of the DM; and the "downshifted peak" (DP) and "upshifted peak" (UP), which are peaks at 1-3 kHz below or above f 0 . Leyser et al. (1993) separated the continuum spectra into a broad con-tinuum (BC), which extends up to 60-120 kHz on the downshifted side of the heating frequency f 0 , and a narrow continuum (NC), which exists in the frequency range between the DM and f 0 and presents a swift decrease in intensity with an increasing frequency offset. According to investigations of the temporal evolution of SEE features, the NC can be further separated into a ponderomotive NC (NC p by Frolov et al., 2004, or FNC by Leyser, 2001, which is produced by a ponderomotive nonlinearity, and a thermal NC (NC th by Frolov et al., 1997, or SNC by Leyser, 2001, which is related to a thermal parametric instability.
Most of the reported SEE spectra have been observed during Omode heating experiments, and only a few observations have been made during X-mode heating experiments. Thidé et al. (1983) observed narrow peaks that formed a banded structure in the upshifted sideband in an X-mode heating experiment with a heating wave frequency f 0 of 2.759 MHz. Sharma et al. (1993) suggested the SEE spectra observed during the X-mode heating period reported by Thidé et al. (1983) were the result of a parametric decay instability (PDI). They showed that the X-mode EM wave could parametrically decay to an upper hybrid wave and a lowfrequency electrostatic ion-Bernstein wave at different gyroharmonics. In the X-mode heating experiments carried out on October 21, 2012, and October 27, 2013, at Tromsø, Norway, Blagoveshchenskaya et al. (2015 reported that narrow SEE spectra had been observed at Saint Peterburg, Russia (1,200 km away from the EISCAT heating facility), accompanied by SEE spectra within 200 kHz observed at Tromsø. The frequency of narrowband SEE spectral components was within 1 kHz of the HF heating frequency, and these components were considered stimulated Brillouin scattering (Bernhardt et al., 2011;Fu et al., 2015). Blagoveshchenskaya et al. (2015) also reported that a narrow SEE spectral component was observed during the period when HF-enhanced ion lines were observed by the EISCAT ultra high frequency (UHF) radar in the X-mode heating experiment. These results suggested that the narrow SEE spectra may be associated with the ion acoustic wave, electrostatic ion cyclotron wave, and electrostatic ion cyclotron harmonic wave. Blagoveshchenskaya et al. (2017a) demonstrated experimental results on the behavior and intensities of various spectral lines in the narrowband SEE spectra within 1 kHz of the heater frequency, depending on the pump frequency stepping across the fifth electron gyroharmonic frequency observed far away from the heater frequency. Blagoveshchenskaya et al. (2018) analyzed in detail the narrowband SEE spectral features in the course of the X-mode experiment on February 25, 2013, at a pump frequency of 5.423 MHz at Saint Peterburg, which is 1,200 km away from Tromsø. However, Blagoveshchenskaya et al. (2015Blagoveshchenskaya et al. ( , 2017aBlagoveshchenskaya et al. ( , 2018 did not observe any spectral features within 200 kHz near Tromsø that were associated with X-mode heating near the fourth gyroharmonic frequency.
This article presents unique experimental observations of the SEE spectral components during an X-mode heating experiment performed on November 6, 2017, at the EISCAT heating facility at Tromsø, Norway. In Section 2, the experimental facilities and observations are briefly introduced. During the X-mode heating experiment, NC spectra, peaks at ±5.127 kHz, and cascading peaks on both sides of the heating wave frequency were observed. The observed SEE spectra are discussed further in Section 3, and the principal conclusions are summarized in Section 4.

Experimental Observations
The experiment reported here was performed on November 6, 2017, at the heating facility near Tromsø, Norway (69.59°N, 19.21°E), which is run by the EISCAT Scientific Association (Rietveld et al., 1993(Rietveld et al., , 2016. The experiment began at 11:30 UT with a 15-min transmission cycle of 10 min on and 5 min off. The heating frequency was 5.423 MHz, just below 4f ce , using high-gain array 1 with a beam inclination of 12° and 6°, where f ce is the electron gyrofrequency. The difference between the heating waves f 0 and 4f ce was approximately 45 kHz. The period for each incident angle was 30 min, or two transmission cycles. The polarization of the heating wave was set as the X-mode EM wave. All 12 transmitters were used at a nominal 80 kW each, resulting in a gain of 27.4 dB and an effective radiated power of approximately 450 MW. The SEE receiver was an Ettus Research USRP N210 instrument fitted with a global positioning system (GPS) disciplined oscillator to provide the precise time and frequency. It was installed in the Tromsø suburb of Kroken, Norway (69. 68°N, 19.07°E), about 12 km north-northwest of the heating array. A mountain separates the receiver from the heating transmitter. The antenna of the SEE receiver was a Wellbrook active loop antenna of 1 m diameter. The sampled data were processed with the fast Fourier transform algorithm to yield SEE spectra with a frequency resolution of approximately 244.14 Hz.
The EISCAT UHF radar was used to measure the plasma parameters in the ionosphere, such as the electron density and electron temperature, from 50 km to approximately 700 km, with a 5 s time resolution. The plasma parameters were calculated from the EIS-CAT UHF radar spectra with the Grand Unified Incoherent Scatter Design and Analysis Package (Lehtinen and Huuskonen, 1996). A Dynasonde, co-located with the Tromsø heating facility, measured the ionogram at intervals of 2 min (Vicker, 2011). The ionogram provided the critical frequency of the ionospheric layers, the maximum plasma frequencies of the different ionospheric layers, and the maximum heights of the different ionospheric layers. Figure 1 illustrates the ionospheric parameters measured by the UHF radar and the critical frequency of the F 2 layer as measured by the Dynasonde HF sounder at Tromsø from 11:25 UT to 12:00 UT. These included the electron density (Figure 1a), the electron temperature (Figure 1b), the raw electron density (radar echo power profile; Figure 1c), and the critical frequency of the F 2 layer ( Figure 1d). Figure 1b shows that the electron temperature was enhanced by approximately 300-500 K in the range of 150-300 km during the heating periods. In Figure 1a and 1c, the electron density and the raw electron density did not show a considerable change during the heating cycles. As shown in Figure 1d, the heating wave frequency was higher than f O F 2 , which indicates that an O-mode wave penetrated the ionosphere during the two heating periods. The frequency of the X-mode heating wave was lower than f X F 2 in most of the heating cycles, except from 11:34 to 11:38 UT. Although the EISCAT heating facility was unable to produce a perfect O-mode or X-mode wave, any O-mode heating effect could be excluded because of the lower f O F 2 , which was lower than both the heating wave frequency and the upper hybrid frequency. Figure 2 illustrates the observed spectrogram of the heater signal within 200 kHz at 11:30-12:28 UT, in which the pump wave frequency was higher than f O F 2 . As shown in Figure 2a, the incident angle of the heating wave was −12° (pointed to the geomagnetic field) at 11:30-11:55 UT. Figure 1b presents the SEE spectra observed at 12:00-12:25 UT with an incident angle of −6°. SEE spectra around f 0 within ±5 kHz was clearly observed in the four heat-ing cycles under two different incident angles. As shown in Figure 2, the intensity of spectral components in the SEE spectra reached approximately 10 dB. The SEE spectra disappeared at around 12:34 UT as the X-mode critical frequency dropped below the heater frequency.
Figure 3a-f presents observations of the SEEs in the first 5 s of the first heating cycle, i.e., from 11:30:00 to 11:30:05 UT. Figure 3a shows that two peaks occurred at +5.127 kHz and −5.127 kHz from the heating frequency f 0 . Figure 3b indicates that the intens- Universal Time EISCAT Scientific Association EISCAT UHF RADAR The NC is a continuous spectrum that occurs only on the downshifted side of the pump frequency f 0 , usually in the frequency range between the DM and the pump frequency. It showed a swift decrease in intensity with an increasing frequency offset (Leyser et al., 1993;Frolov et al., 1997).   Figure 5a illustrates the variation in intensity of the spectral peaks at ±5.127 kHz with respect to time during the heating period, in which the red and blue solid lines represent the spectra at −5.127 and +5.127 kHz, respectively. The spectral peaks at ±5.127 kHz were sharply enhanced as the heater was turned on and were maintained throughout the 10-min heating cycle. The intensity of the spectral peaks was highly symmetrical in the heating period, except for the first several seconds. Both peaks illustrated an overshoot in intensity within the first 5 s of the heating period. In addi-   tion, the intensity of the DP remained higher than that of the UP throughout the heating period. Figure 5b illustrates the temporal variation of the spectral intensity at five different frequencies picked up from the NC spectra, with the different colors representing different frequencies. The horizontal axis marks the time after heater turn-on. As shown in Figure 5, the intensity of the NC spectrum decreased with the frequency offset. The spectra at all frequencies increased with time in the first 3 s and then decreased to the noise level at 5 s after heater turn-on. The NC spectrum showed an overshoot effect, which is a prominent emission intensity maximum at a finite time after heater turn-on. A similar observation of NC development was reported by Leyser (2001) during O-mode heating experiments. Figure 5c illustrates the temporal variation in the intensity of the cascading peaks, with the different colors representing different frequencies. As shown in Figure 5c, the intensity of the cascading peaks reached a maximum 3 s after heating and then decayed to the noise level within 5 s.
In comparing Figure 3 with Figure 4, it can be deduced that the NC spectra were generated only with the cold start (i.e., without a previous heating period), whereas the cascading spectra were observed in the preconditioned ionosphere (i.e., in the second heating cycle). Both the NC spectra and the cascading spectral peaks show an overshoot effect, as illustrated in Figure 5b and 5c. The two spectral peaks at f 0 of ±5.127 kHz remained throughout the heating period, as shown in Figure 5a.

Discussion
SEEs are a crucial tool for investigating the nonlinear plasma response to an EM HF heating wave because the SEE spectra are sensitive to various wave-plasma interactions and they cover both short-and long-timescale nonlinear processes (Leyser, 2001;Frolov et al., 2004). Features of SEEs have been studied under different pump-plasma interaction conditions, such as spectral behavior in response to different heating wave frequencies (Stubbe et al., 1994), the heating power (Frolov et al., 2004), and pumping by additional waves (Stubbe et al., 1985). In addition, different SEE features have been utilized for studies of short-timescale ponderomotive nonlinear processes, such as Langmuir turbulence (Frolov et al., 2004), and long-timescale thermal nonlinear pro- cesses, such as the excitation of FAIs (Stubbe and Hagfors, 1997;Leyser, 2001). Provided the physical mechanism generating the SEE component is identified, information on the background ionospheric plasma and on nonlinear processes can be deduced by observing the variation in SEE features and the evolution of SEEs under different experimental conditions.
In Figure 3, the NC feature shows the overshoot effect for a cold start, i.e., without a previous heating wave. Figures 3 and 5b show that the NC feature occurred 2 s after pump-on and was quenched 2 s later. The intensity of the NC spectrum decreased with the frequency offset. The intensity of the NC spectrum at each frequency gradually increased to the maximum and decreased to the noise level within 5 s. In previous O-mode heating experiments, the ponderomotive NC was suggested to be the only spectral feature generated for a cold start (Leyser, 2001). It was found to be excited in intensity within a few milliseconds after heater turn-on for a cold start, and it exhibited an overshoot within 10 ms after the heater was switched on (Boiko et al., 1985). The ponderomotive NC was attributed to the development of the PDI below the heating wave reflection height (Frolov et al., 1997(Frolov et al., , 2004. In the O-mode heating experiments, the typical timescale of PDI excitation was several milliseconds (Robinson, 1989). Frolov et al. (2004) reported that in an O-mode heating experiment, the intensity of the ponderomotive NC increased rapidly, reached a maximum in the time interval of 2-10 ms after heater turn-on, and decayed from approximately 0.7 to 4-5 ms. Compared with the ponderomotive NC feature in O-mode heating experiments, the timescale of NC development was much longer in the X-mode heating experiment reported here. In X-mode heating experiments, the PDI was excited for several seconds and even minutes (Blagoveshchenskaya et al., 2017b;Wang X et al., 2018). This conclusion was reached by observing the temporal evolution of the HF-enhanced plasma lines and ion lines, both of which are signatures of PDI excitation in ionospheric heating experiments. As shown in Figures 3 and 5b, the NC was still generated for a cold start and exhibited an overshoot effect. Thus, the NC observed in the X-mode heating experiment may possibly be related to the PDI. Figure 4a illustrates that the heating wave intensity at 11:45:00 UT was lower by 7 dB than its intensity level in the first cycle shown in    Figure 3a. Furthermore, the cascading peaks in Figure 4 occurred only in the preconditioned ionosphere. The cascading peaks in Figures 4 and 5c also exhibited an overshoot effect. It is possible that the cascading peaks may be related to the FAIs induced by the heating wave. Field-aligned irregularities have been observed during X-mode heating experiments at Tromsø, Norway (Blagoveshchenskaya et al., 2011(Blagoveshchenskaya et al., , 2014(Blagoveshchenskaya et al., , 2015. Borisov et al. (2018) suggested using a two-step process to interpret the formation of 10m FAIs by the X-mode heating wave at high latitudes. The first step is the thermal self-focusing instability, and the second step is the formation of electron inhomogeneities resulting from the thermal self-focusing instability or from the gradient drift and current convective instabilities. Otherwise, it should be noted that small-scale FAIs were not excited when powerful X-mode waves were used for pumping at the Sura facility, which is in the middle latitudes (Frolov et al., 2014).
The peaks at f 0 of ±5.127 kHz occurred symmetrically on both sides of the heating wave f 0 and were present during the entire 10-min heating cycle, as shown in Figures 3, 4, and 5a. However, in O-mode heating experiments, the DP/UP usually occurred at about −1 to 3 kHz, and the DM/UM often appeared at about 9 to 15 kHz (Leyser, 2001;Frolov et al., 2004). The DP and UP are sym-metrical features around f 0 (Stubbe et al., 1994;Leyser, 2001). The broad symmetrical structure has been observed to be symmetrically excited at f 0 of approximately ±15 to 30 kHz, which has been recorded only slightly above 3f ce (Stubbe et al., 1994). Lobachevsky et al. (1992) observed an anticorrelation of the SEE spectra at f 0 of ±5 kHz for approximately f 0 to 3f ce on a time scale of several seconds in an O-mode heating experiment performed at Tromsø, Norway. Compared with the SEE spectra observed in the O-mode heating experiment, the persistent feature of the peaks at f 0 of ±5.127 kHz was novel in the X-mode heating experiment. Fejer and Leer (1972) showed that the electron Bernstein mode associated with FAIs may be excited parametrically by an X-mode heating wave at a multiple gyrofrequency near the X-mode heating wave reflection height. This is a ponderomotive process that should occur rapidly after the heater is turned on. The peak at f 0 of ±5.127 kHz may be the excited electron Bernstein wave scattering off the FAIs or low hybrid waves.
The presence of SEE spectra during the X-mode heating cycle is a large unexplored research area (Leyser, 2001). The physical mechanisms that generated the observed SEE spectra, including the NC spectra, the peaks at ±5.127 kHz, and the cascading peaks, are not clear. This implies that more experiments need to be performed to investigate the SEE features in X-mode heating experiments in the future, as well as further theoretical development to explain them.

Summary
Stimulated electromagnetic emissions spectra for O-mode heating experiments have been extensively investigated since the 1980s, whereas SEE spectra in X-mode heating experiments have rarely been studied because of the limited observations of SEE during X-mode heating. This report presents unique SEE observations made during an X-mode heating experiment on November 6, 2017, at the EISCAT heating facility in Tromsø, Norway. The heating wave frequency in the experiment was close to the fourth electron gyroharmonic frequency. The experimental observations illustrated NC spectra, peaks at ±5.127 kHz, and cascading peaks on both sides of the heating wave frequency. Novel SEE features were observed in the X-mode heating experiment. The NC spectra and the cascading peaks presented an overshoot effect, and the peaks at ±5.127 kHz remained during the entire heating cycle. The intensity of the heating wave in Figures 2a and 5a indicated that the NC was observed only for the cold start and that the cascading peaks were observed only for the preconditioned ionosphere. The CUTLASS HF coherent radar, which can detect decameter-scale FAIs above Tromsø, was operating in a nonoptimal mode and frequency (12-12.5 MHz) during this experiment but nevertheless detected weak heat-induced backscatter (T. Yeoman, personal communication), although this observation was too poor for a detailed correlation with the SEE spectra. This result implies that more experiments are needed to investigate the SEE components generated by the X-mode heating wave in the future.

Acknowledgments
EISCAT is an international scientific association supported by research organizations in China (China Research Institute of Radiowave Propagation), Finland (Suomen Akatemia, SA), Japan (Na-