Ground-based radar unveils natural electron Bernstein waves in the ionosphere of Earth Authors: Yonghui Wang, Xinan Yue, Yulun Wu, Yihui Cai, Feng Ding, Juha Vierinen, Ingemar Häggström, Erhan Kudeki, Ning Zhang, Junyi Wang, Mingyuan Li, Baiqi Ning, Lin Jin, Chi Wang Published: 2026 ABSTRACT AbstractElectron Bernstein waves (EBWs) are electrostatic modes that propagate perpendicular to the magnetic field in magnetized plasmas, serving as a diagnostic tool for understanding plasma properties and the energetic particles that drive energy transport. However, their naturally occurring harmonic structures have long eluded ground-based radar detection due to their faint signals and the requirement for an observational geometry nearly perpendicular to the geomagnetic field. Here, we report high-resolution broadband spectral observations of naturally excited EBWs in Earth’s ionosphere using the Sanya Incoherent Scatter Radar. Our measurements reveal a series of discrete, well-defined spectral lines corresponding to the first seven electron gyroharmonics, a result consistent with theoretical predictions. Our analysis suggests that these EBW modes, enhanced by suprathermal photoelectrons, become observable under near-perpendicular geometry, where Landau damping is reduced. The frequency structure of these harmonics acts as a precise diagnostic ruler that enables the retrieval of electron density and perpendicular electron temperature, parameters that are difficult to retrieve with ion line techniques under such geometric constraints. This work provides ground-based radar observations of naturally excited EBWs and shows their use for diagnosing plasma parameters in magnetized plasmas.Subject terms: Astrophysical plasmas, Astronomical instrumentation Introduction The ionosphere1, a tenuous ionized layer that couples Earth’s atmosphere to the magnetosphere, plays an important role in radio communication, navigation, and satellite operations, and provides a natural environment for studying plasma processes2,3. Among the plasma waves found there, electron Bernstein waves (EBWs), electrostatic oscillations that propagate with the wave vector k nearly perpendicular to the magnetic field B, have drawn decades of theoretical attention4 because of their harmonic structure and their relevance to fusion plasma heating5, energy transfer in the solar wind and magnetosphere6,7, wave-mode conversion8, and plasma parameter diagnostics in planetary magnetospheres9. Seminal studies10–12 predicted that, in a magnetized plasma, the ordinary plasma line can be separated by narrow frequency gaps at integer multiples of the electron gyrofrequency, nfce, where fce is the electron gyrofrequency and n is the harmonic number. However, despite early indications from satellite-based active measurements13 and a report of plasma line splitting near 2fce at Arecibo14, altitude-resolved observations of this predicted fine harmonic structure in the ionosphere have remained elusive. Incoherent scatter radar (ISR), also known as collective Thomson scattering in fusion plasma research, has been used in ionospheric physics since the late 1950s15. ISR transmits a radio wave that is Bragg-scattered by electron density fluctuations in the ionospheric plasma. Through analysis of the returned incoherent scatter spectra, researchers can retrieve plasma parameters16,17. Among these spectral features, the plasma line is particularly important18–27. It corresponds to high-frequency Langmuir waves associated with electron density oscillations. In ionospheric plasma under thermal-equilibrium conditions, where suprathermal electrons are absent, plasma lines are usually too weak to exceed the radar detection threshold in routine observations; thermal-level detections have been reported only with the high-power, large-aperture Arecibo radar28. Furthermore, in near-perpendicular propagation regimes, a departure from the k⊥B geometry can increase Landau damping and suppress the wave signal29–32. Under daytime nonthermal conditions, the detectability of plasma lines can improve33. Suprathermal photoelectrons34, generated by solar extreme-ultraviolet (EUV) radiation, can enhance Langmuir waves through weak Cherenkov emission while reducing Landau damping26,35. Even with this natural enhancement, observing EBW-associated fine structure requires a radar to satisfy several technical requirements simultaneously: (1) sufficient transmitter power and antenna sensitivity to detect the inherently weak plasma line echoes, (2) broad receiver bandwidth and high range resolution to resolve the closely spaced Bernstein harmonics, and (3) the ability to point the radar beam nearly perpendicular to the geomagnetic field, thereby minimizing Landau damping. These combined requirements explain why EBWs have been difficult to observe in the ionosphere. In addition, ground-based radar observations provide the key advantage of altitude-resolved measurements. Spacecraft provide only single-point measurements along their trajectories, whereas ground-based radars can probe ionospheric profiles over a broad altitude range when the required sensitivity, bandwidth, range resolution, and viewing geometry are achieved. Here, we present broadband, altitude-resolved observations from low latitude using the Sanya Incoherent Scatter Radar (SYISR; see Supplementary Fig. 1)36–38, a facility of the Chinese Meridian Project for space physics39, that address these observational limitations. Using high-range-resolution alternating code techniques40–43, we observe a series of resolved frequency gaps between 2fce and 7fce that are consistent with classical EBW theory. These observations are also in agreement with the fine structure predicted for Earth’s ionosphere by Salpeter12, and we use this spectral structure to infer the electron density and perpendicular electron temperature. Moreover, the EBW spectrum is sensitive to background density perturbations associated with atmospheric gravity waves, suggesting its potential use as a tracer of ionosphere-atmosphere coupling dynamics. These results demonstrate the capability of ground-based radars to probe ionospheric plasma wave processes and establish the utility of SYISR for resolving k⊥B incoherent scatter spectral characteristics. Results We observed harmonic spectral structures associated with naturally excited EBWs in the ionosphere. By directing the SYISR radar beam perpendicular to the geomagnetic field (Fig. 1), we observed plasma lines over a continuous slant-range interval from 100 to 1000 km. The measurements show that the plasma lines contain frequency gaps near integer harmonics of the electron gyrofrequency. A general feature of the observations is that the gap width, denoted by Δω, decreases with increasing n. In addition, multiple plasma line peaks, with up to three peaks observed at the same time, occur within specific slant-range intervals. Observational geometry for daytime ground-based radar detection of ionospheric Bernstein-wave signatures. a Conceptual overview. Solar extreme-ultraviolet (EUV) radiation (yellow rays, upper left) ionizes the upper atmosphere, forming the ionospheric plasma (magenta shaded layer). Photoelectrons (pink spheres) can enhance plasma oscillations through Cherenkov excitation; their guiding-center motion along the geomagnetic field lines (red solid arcs) is shown by the yellow sinusoidal curve. The blue sinusoidal curve denotes plasma waves that propagate perpendicular to the geomagnetic field. Incoherent scatter radar detects Thomson backscatter from thermal electron (blue spheres) density fluctuations with wave vectors along the line of sight. b Observing geometry of the Sanya incoherent scatter radar (SYISR; 18.3∘N, 109.6∘E). The beam points at 358∘ azimuth (close to the geomagnetic meridian) and 60∘ elevation. Colors along the beam encode the aspect angle α (colorbar, lower right), the angle between the radar wave vector k and its component k⊥ perpendicular to the geomagnetic field vector B (inset diagram). The beam is very close to perpendicular to B (α close to 0∘) at a slant range of 320 km, and α stays below 6∘ from 100 km to 1000 km. c Plasma line spectrum measured by SYISR (slant range 300–1000 km; frequency 1–5 MHz). The plasma line frequency varies with range because the electron density on which it depends varies with altitude. Spectral gaps occur near the electron gyrofrequency harmonics nfce (n = 2, 3, 4), flanked by resonance peaks. These features are identified as signatures of electron Bernstein waves (EBWs). Source data are provided as a Source Data file. Observations during the sunrise period on 18 September 2023 provide a representative example of these harmonic features, which appeared over slant ranges from 300 to 1000 km (see Supplementary Video 1 for their evolution after sunrise). In a snapshot at 06:32 LT (Fig. 2a), plasma lines are resolved in the frequency bands between adjacent harmonics of the electron gyrofrequency, from the fundamental band to the fifth harmonic band (Fig. 2b). Each harmonic is flanked by a forbidden gap where the plasma lines completely vanish. We measured the gap widths to be ~900, 450, 170, and 70 kHz for n = 2, 3, 4, and 5, respectively, showing a decrease in gap width with increasing n (Fig. 2c). The curvature of the spectral lines also follows theoretical predictions, displaying asymptotic behavior as they approach the corresponding electron gyrofrequency harmonics. Notably, plasma lines on both sides of a given harmonic appear simultaneously within the same slant-range interval. In some slant-range intervals, this coexistence involves more than two spectral peaks. For example, at slant ranges from 540 to 580 km, three distinct plasma line peaks associated with the second and third harmonics of the electron gyrofrequency are observed simultaneously (Fig. 2c). The time-frequency-intensity overview further shows the persistence of these harmonic structures during the interval considered (see Supplementary Fig. 2). Range-dependent plasma line spectral features observed by SYISR in a near-perpendicular geomagnetic-field geometry. a Frequency-range-intensity plot of the plasma line spectrum observed during sunrise at 06:32 local time (LT; UTC+8) on 18 September 2023. The data were obtained using a 640 μs 32-bit strong alternating-code waveform with a range resolution of 3 km and an integration time of 5 min. The spectrum spans 12 MHz, from −6 MHz to 6 MHz, covered by three 4 MHz receiver filter banks, and slant ranges from 250 km to 1000 km. b Enlarged view of the positive frequency plasma line spectrum from 1 MHz to 6 MHz in panel (a). Dashed lines mark the electron gyrofrequency harmonics nfce, where fce is the electron gyrofrequency and n = 2, 3, 4, 5. c Frequency offsets of the plasma line peaks extracted from panel (b). Only the peak frequency offsets are retained to reduce the visual influence of background noise and emphasize the plasma line distribution and associated gap structures. Source data are provided as a Source Data file. The observations also show variability in EBW-associated spectral structures under different ionospheric conditions and at different local times. Daytime measurements on 7 June 2024 show contrasting plasma line characteristics above and below 350 km (Fig. 3a; see Supplementary Video 2 for their evolution over the course of the day). At 08:08 LT, frequency gaps are present for harmonic numbers 3–7, whereas the gap near harmonic number 8 is small or not clearly resolved. Below 350 km, the plasma lines are more intense close to the electron-gyrofrequency harmonics than in the corresponding topside region, but their intensity is weaker near the center of the harmonic bands. Similar behavior is observed during the post-sunrise period at 05:40 LT (Fig. 3b) and during the pre-sunset period at 17:33 LT. At slant ranges from 120 to 160 km, an oblique plasma line trace, which is consistent with modulation by an atmospheric gravity wave44, shows a spectral gap as it crosses the third harmonic of the electron gyrofrequency (Fig. 3c). This feature suggests that neutral density perturbations associated with the gravity wave may modulate the local electron density and shift the plasma line frequency, thereby making the EBW-associated forbidden band apparent. Diurnal evolution of plasma line spectral features observed by SYISR in a near-perpendicular geomagnetic-field geometry. a Frequency-range-intensity plot of the plasma line spectrum observed by the Sanya incoherent scatter radar at 08:08 local time (LT; UTC+8) on 7 June 2024. Clear plasma line features are visible from 100 km to 1000 km in slant range and from 2 MHz to 9 MHz in frequency. The observation used a 320 μs, 32-bit strong alternating-code waveform, with 1.5 km slant-range resolution, 4 min integration, and a 16 MHz receiver bandwidth. b Plasma line spectrum observed during sunrise at 05:40 LT on the same day, shown over slant ranges from 300 km to 600 km. Dashed lines mark the electron gyrofrequency harmonics nfce for n = 4 and 5, where fce is the electron gyrofrequency. c Plasma line spectrum observed pre-sunset at 17:33 LT on the same day, shown over slant ranges from 100 km to 300 km. Dashed lines mark nfce for n = 3, 4, 5, 6. In addition to the frequency gaps near nfce, a wavelike structure is present in the plasma line feature between 120 km and 180 km in slant range, coinciding with the frequency gap around 3fce. Source data are provided as a Source Data file. Discussion Our observations provide evidence for naturally occurring EBWs in the ionosphere and help address a long-standing theoretical expectation that plasma lines in a magnetized plasma can separate into harmonic structures, as proposed by Salpeter12. The harmonic structures captured over a broad altitude range, especially the frequency gaps at integer harmonics of the electron gyrofrequency and their systematic dependence on harmonic number, agree quantitatively with EBW dispersion theory for near-perpendicular wavevector-magnetic-field geometry. The observation of such waves in Earth’s ionosphere using an ISR indicates a practical approach to accurately measure the thermal motions of electrons perpendicular to the geomagnetic field. In turn, this may also provide a basis for remote sensing of plasma instabilities, and wave-particle interactions in Earth’s ionosphere. Our numerical simulations support this interpretation. The simulations reproduce the observed frequency gaps and their varying widths, and show the coexistence of multiple EBW modes (Fig. 4a). This phenomenon arises because the solutions to the dispersion relation concentrate their power near the upper-hybrid resonance frequency, leaving other harmonic echoes with negligible intensity. Although the strictly perpendicular geometry is reached only at a specific altitude, simulations based on the generalized EBW dispersion relation show that these frequency gaps remain the dominant spectral feature for small deviations from perpendicular geometry (Fig. 4c). These findings expand our understanding of ionospheric plasma lines by demonstrating that their spectral fine structure at high harmonics is essentially the manifestation of EBWs, beyond split features near the second harmonic of the electron gyrofrequency reported in previous studies 14. Our analysis also indicates that the gap width is only weakly affected by collisions under the conditions considered. Collisions mainly act on the imaginary part of the dispersion-relation solution, which controls wave damping. Because the F-region ionosphere is nearly collisionless, collisions are unlikely to control the observed spectral fine structure (see Supplementary Fig. 3). Theoretical modeling of plasma line excitation and Bernstein-wave resonance. a Salpeter dispersion relation for propagation perpendicular to the geomagnetic field (α = 0∘). The horizontal axis is a dimensionless parameter proportional to electron density, and the vertical axis is the squared electron gyroharmonic number n. Crosses mark resonant solutions, color-coded by calculated relative echo intensity. The dash-dotted line marks the upper-hybrid frequency. Vertical dashed lines show four representative density-parameter values, 3, 8, 15 and 24, for which paired resonant solutions occur on both sides of the n = 2, 3, 4 and 5 electron gyroharmonics. b Photoelectron fluxes during sunrise from the photoelectron-flux module of the SAMI2-PE ionospheric model. Energy-altitude-intensity maps are shown for 06:00 local time (LT) on 18 September 2023 and 08:15 LT on 7 June 2024. The modeled flux is stronger above 400 km at 06:00 LT and concentrated near 200 km at 08:15 LT. c Complex roots of the generalized electron Bernstein-wave dispersion relation for electron plasma frequencies near 3fce, where fce is the electron gyrofrequency. The roots are shown as the aspect angle α decreases from 90∘ to 1∘. Vertical dotted lines, from left to right, mark nfce for n = 1, 2, 3; the horizontal dash-dotted line marks ωi = 0. As α approaches perpendicular propagation, the root near 3fce separates into two branches and moves closer to the real axis, indicating reduced damping. Source data are provided as a Source Data file. Our analysis suggests that photoelectrons generated by solar EUV radiation are an important energy source for these waves. The observed spatiotemporal variations in plasma line intensity correlate with modeled photoelectron flux45. The modeled flux increases after sunrise, and the altitude of peak production moves downward as solar illumination reaches deeper atmospheric layers, leading to an enhanced population of energetic electrons below 300 km (Fig. 4b; see Supplementary Video 3 for the evolution throughout the day). A related question is how these electrostatic waves avoid strong Landau damping under near-perpendicular geometry. Our analysis reveals that the ambient magnetic field transforms what would be heavily damped Langmuir waves into a series of weakly damped EBW modes as the aspect angle approaches zero (Fig. 4c, see Supplementary Fig. 4 for the aspect angle dependence of this reduced damping near other harmonics of the electron gyrofrequency). Suprathermal electrons with energies up to 100 eV can then excite these modes through Cherenkov or cyclotron resonance46, increasing their intensity above the thermal-equilibrium level to amplitudes detectable by SYISR. This interpretation is supported by observations in the bottomside ionosphere, where enhanced spectral power within a 1 MHz band around nfce and an intensity asymmetry favoring lower frequencies (for n < 7) point to a downward-propagating 30, low-energy photoelectron population with energies less than 20 eV. This is also consistent with the cyclotron resonance condition where v∥ is the electron velocity parallel to the geomagnetic field, ω is the wave angular frequency, ωce is the electron angular gyrofrequency, and k∥ is the component of the wavevector parallel to the geomagnetic field. The comparison between observations and theory shows quantitative agreement. By using model altitude profiles of electron density and electron temperature, together with the aspect angle, as inputs to the general EBW dispersion relation, the model produces a theoretical spectrogram that matches the measured spectral structure and supports the identification of the observed features as Bernstein waves (Fig. 5a). This comparison also provides a diagnostic approach for ionospheric plasma parameters. The harmonic structure of EBWs acts as a natural ruler for plasma parameters, overcoming the limitations of traditional ISR ion line methods in near-perpendicular geometry47,48. Using the observed frequency gaps, we derived the perpendicular electron temperature, with an average value of about 3200 K (Fig. 5b). To further validate this diagnostic approach, we extracted a one-dimensional line spectrum at a slant range of 540 km. The simulated spectrum calculated with these inverted parameters reproduces the observed spectral line and the frequency gaps (Fig. 5c). This capability to estimate the perpendicular electron temperature provides a basis for future multi-beam or multistatic experiments to investigate electron temperature anisotropy in the ionosphere49,50. Near-perpendicular Bernstein-wave modeling and plasma-parameter inversion in the ionosphere. a Range variation of resonant solutions calculated from the generalized electron Bernstein-wave dispersion relation between 120 km and 1000 km, where the aspect angle is below 6∘. The background electron density Ne and electron temperature Te are taken from the International Reference Ionosphere (IRI). Dot-dashed lines mark nfce for n = 1, 2, 3, 4, 5, where fce is the electron gyrofrequency. The dashed line marks the upper-hybrid frequency, and the dotted line marks the plasma line frequency. b Comparison between the IRI input profiles and the profiles inferred from the dispersion-relation inversion. The inverted parameters are the electron density Ne and the perpendicular electron temperature T⊥. c Comparison between the observed plasma line spectrum and the theoretical spectrum at 540 km. The theoretical spectrum is calculated from the generalized electron Bernstein-wave dispersion relation using the inverted Ne and T⊥. Source data are provided as a Source Data file. In summary, our observations using the low-latitude SYISR provide broadband, high-resolution spectral evidence of ionospheric EBWs under near-perpendicular geomagnetic field geometries. These findings not only deepen the understanding of wave-particle interactions and suprathermal electron dynamics but also establish a method for inverting electron density and temperature, addressing the limitations of traditional ion line techniques. Furthermore, once the theoretical framework describing how suprathermal electrons enhance these plasma lines is further developed, this approach may also be useful for photoelectron diagnostics. Beyond Earth’s ionosphere, the diagnostic approach demonstrated here can be adapted to study magnetized plasmas in planetary magnetospheres and other astrophysical environments. Additionally, the observed EBWs behavior under low-temperature ionospheric conditions, together with the sensitivity of Landau damping to the magnetic aspect angle, may also provide a reference for laboratory studies of wave heating and mode conversion. Collectively, these results show how high-resolution ground-based radar observations can complement theoretical and experimental studies of fundamental plasma processes. Methods Experiment setup and observations The experiments were conducted using the Sanya Incoherent Scatter Radar (SYISR; 18. 3∘N, 109. 6∘E), an advanced phased-array system in Hainan, China36,37. The radar operated in the 430–450 MHz frequency band, with a peak transmitted power greater than 4 MW and a low-noise receiver with a system temperature, Ts, of about 120 K. For the electron Bernstein wave (EBW) observations, the radar beam was fixed at an azimuth of 358∘, aligned with the magnetic meridian, and an elevation of 60∘. This pointing direction was selected so that the radar wavevector, k, was perpendicular to the geomagnetic field vector, B, at an altitude of about 320 km, corresponding to an aspect angle, α, of 0∘. Here, the aspect angle is defined as the angle between the radar backscatter wavevector, k, and the direction normal to the geomagnetic field. The aspect angle remained smaller than 6∘ over the 100–1000 km observation range. We present data from two representative observation campaigns. The first campaign was conducted on 18 September 2023 from 05:00 to 08:30 local time (LT; UTC+8). The radar used a 442 MHz carrier frequency and a 640 μs, 32-bit strong alternating-code pulse, giving a 3 km range resolution and a 15 ms interpulse period (IPP). The receiver was configured with three filter banks centered at 438, 442 and 446 MHz. Each filter bank had a bandwidth of 4 MHz and a sampling rate of 4 MHz, giving a combined frequency coverage from 436 to 448 MHz. The second campaign was conducted on 7 June 2024, starting at 00:00 LT and lasting 24 h. In this campaign, the radar used a 439.8 MHz carrier frequency, a 320 μs, 32-bit strong alternating-code pulse and a 10 ms IPP, giving a 1.5 km range resolution. The received signal was recorded with a single 16 MHz wideband receiver centered at 444.8 MHz and sampled at 16 MHz. Signal processing and power spectral density estimation The received analog signals were first downconverted to an intermediate frequency of 80 MHz and digitized. The digitized signals were then further downconverted in hardware using cascaded integrator-comb (CIC) and finite impulse response (FIR) filters. The FIR decimation factor was fixed at 5, whereas the CIC decimation factor was an adjustable integer. This configuration produced complex voltage data consisting of in-phase and quadrature (IQ) components, with output sampling rates and bandwidths of up to 16 MHz. In this study, the sampling rate was 4 MHz for each filter bank in the 18 September 2023 campaign and 16 MHz for the wideband receiver in the 7 June 2024 campaign. The software processing in this study started from the complex voltage IQ data. Frequency-domain alternating-code decoding was applied to estimate power spectra with a range resolution corresponding to the code bit length51. The decoded spectra were then combined to obtain range resolved power spectral density (PSD) estimates. A Hamming window was applied before the fast Fourier transform (FFT) used for spectral estimation. The spectra were accumulated over 4-5 min integration intervals. During each integration interval, interference targets were identified and removed before accumulation to form the incoherent scatter power spectra. The spectra were then calibrated to physical units of Kelvin per MHz using the system noise temperature of about 120 K. For the three-filter-bank data from 18 September 2023, the processing was applied to each channel independently. The resulting spectra were aligned by their timestamps and concatenated to form a continuous 12 MHz spectrum covering 436–448 MHz. Theoretical models and numerical computation To interpret the observations, we used three theoretical descriptions of high-frequency electron oscillations in a magnetized plasma. We first establish the fundamental plasma parameters used throughout these models. The Bragg scattering wavenumber is defined as k = 4π/λr, where λr is the radar wavelength. The wavevector k is resolved relative to the geomagnetic field vector B as Here, k⊥ and k∥ are the components perpendicular and parallel to B, respectively, and α is the magnetic aspect angle. Key electron parameters used in the theoretical models include the electron angular gyrofrequency ωce = qeB/me, the electron gyroradius Re = vth/ωce, the electron plasma frequency ωpe, the electron most probable thermal velocity vth, and the Debye length λD. The latter three quantities are given by Here, qe, me, Te, and Ne are the elementary charge, electron mass, electron temperature, and electron density, respectively. The quantities B, kB, and ϵ0 are the magnetic field strength, Boltzmann constant, and vacuum permittivity, respectively. For reference, The simplified plasma line dispersion relation used in this study is The upper-hybrid angular frequency is defined as For the general incoherent scatter dispersion relation with collisions, following the Kudeki formulation24,52,53, we used a model based on the general framework of incoherent scatter theory that includes Coulomb collisions. The dispersion relation is denoted as K(ω) and is given by Here, ω is the angular frequency, j is the imaginary unit, and J(ω) is the Gordeyev integral. For a magnetized collisional plasma, the Gordeyev integral is expressed as This integral depends on the mean squared displacements parallel and perpendicular to the magnetic field, Here, ν is the electron Coulomb collision frequency, γ=tan−1(ν/ωce), and C=(kBTe/me)1/2 represents the root mean square velocity along the radar line of sight. The Gordeyev integral was evaluated numerically using a chirp-z transform algorithm54,55. This formulation was used to examine how the dispersion-relation solutions vary in the complex plane with magnetic aspect angle or collision frequency. For the general electron Bernstein wave dispersion relation, following the Harris formulation56–58, we considered the collisionless case with small deviations from perpendicular propagation. The general EBW dispersion function is Here, b=k⊥2Re2/2, ζ0 = ω/(k∥vth), and ζn = (ω + nωce)/(k∥vth). The function Z is the plasma dispersion function, and In is the modified Bessel function of the first kind. The quantity 1/∣H(ω)∣ was used to examine the altitude dependence of the modeled wave spectral peaks in the ionosphere. For the perpendicular propagation incoherent scatter dispersion relation, following the Salpeter formulation12, we considered the ideal limit in which the wavevector k is strictly perpendicular to the geomagnetic field vector B, corresponding to k∥ = 0. The dispersion function is denoted as S(ω) and is written as Here, the dimensionless parameters are a = ω/ωce, β = 1/(kλD), and ϵ = 1/(kRe). The relative intensity evaluated at the solution frequency ων0 is This formulation was used to analyze the distribution of dispersion-relation solutions with respect to frequency and electron density, and to compare the relative magnitudes of the corresponding modeled spectral peaks under strictly perpendicular conditions. The dispersion relations were evaluated numerically using custom Python code built on standard scientific-computing libraries. These scripts are publicly available and freely accessible, as detailed in the Code availability section. Plasma parameter inversion method The retrieval of ionospheric parameters from the observed plasma line peak frequency offsets relies on the requirement that, when substituted into the Salpeter dispersion relation, the two harmonic branches extracted on either side of nfce yield the same value of β2/2ϵ2, a quantity determined by the electron density Ne. For example, as can be readily seen in Fig. 4a, the two dispersion branches on either side of the a2 = 4 (i.e. 2fce) resonance share the same value of β2/2ϵ2 = 3. Setting the Salpeter dispersion relation S = 0 and rearranging gives We denote the left-hand side by x = 2ϵ2/β2, which is proportional to B2/Ne, and the right-hand side by M. In M, ϵ2 is proportional to B2/Te, while a2 is determined by the observed plasma line frequency offset. Suppose two harmonic-branch offsets, f1 and f2, are observed on either side of nfce. From these we obtain a12 and a22, which yield M1 and M2. Because the two branches must satisfy M1 = M2, the retrieval reduces to finding the electron temperature Te for which (M1−M2)2 is minimized and approaches zero; the magnetic field B is taken from the International Geomagnetic Reference Field (IGRF) model. In practice, we construct a uniform grid of Te and search for the grid point that minimizes (M1−M2)2. This grid-search minimization yields the retrieved Te. The corresponding x is then obtained from M = x; since x is determined by Ne, substituting back into its definition directly yields the retrieved electron density Ne. Applying this procedure point by point to the two peak frequency offsets of the nfce harmonic branches as they vary with altitude produces the retrieved electron density and electron temperature along the ionospheric profile. This grid-based inversion procedure is implemented in Python using standard scientific-computing libraries and, as stated in the Code availability section, is publicly available and freely accessible. External models To examine the possible role of suprathermal photoelectrons in the observed Bernstein wave power variations, we used the photoelectron flux calculation module in the SAMI2-PE model 45,59. SAMI2-PE denotes SAMI2 is Another Model of the Ionosphere with Photoelectrons. This plasma transport model, specifically optimized for the equatorial ionosphere, provides high-fidelity simulations of the photoelectron flux generated by solar photoionization. The modeled photoelectron flux was then compared with the observed temporal evolution of Bernstein wave power to evaluate their temporal consistency. The local geomagnetic field strength B was calculated using the 13th generation International Geomagnetic Reference Field model, abbreviated as IGRF-1360. The IGRF-derived magnetic field was used to compute the electron gyrofrequency, fce = ωce/2π. The electron gyrofrequency was used as a reference parameter for the harmonic structure of Bernstein waves. For comparison with empirical ionospheric conditions, we used the International Reference Ionosphere 2020 model, abbreviated as IRI-2020 61. Electron density and electron temperature profiles were extracted from IRI-2020 for the time and location of the observations and used as reference background profiles. All numerical simulations and data analyses were implemented in the Python programming language62. The computational workflow relied on the open-source libraries NumPy63 for numerical operations, SciPy64 for scientific algorithms, Matplotlib65 for scientific visualization, and pyFFTW66 for efficient Fast Fourier Transforms. We gratefully acknowledge the contributions of the open-source communities that develop and maintain these tools. Supplementary information Supplementary Information Description of Additional Supplementary Files Time-evolving dual-sided plasma line spectra on 18 September 2023 Time-evolving single-sided plasma line spectra observed on 7 June 2024 Modeled photoelectron flux spectrum throughout 18 September 2023 Transparent Peer Review file Source data Source Data