High-Redshift Quasars at z ≥ 3—III: Parsec-Scale Jet Properties from Very Long Baseline Interferometry Observations

Abstract

High-redshift active galactic nuclei (AGN) provide key insights into early supermassive black hole growth and cosmic evolution. This study investigates the parsec-scale properties of 86 radio-loud quasars at z ≥ 3 using very long baseline interferometry (VLBI) observations. Our results show predominantly compact core and core-jet morphologies, with 35% having unresolved cores, 59% with core–jet structures, and only 6% with core–double jet morphology. Brightness temperatures are generally lower than expected for highly radiative sources. The jets’ proper motions are surprisingly slow compared to those of lower-redshift samples. We observe a high fraction of young and/or confined peak-spectrum sources, providing insights into early AGN evolution in dense environments during early cosmic epochs. The observed trends may reflect genuine evolutionary changes in AGN structure over cosmic time, or selection effects favoring more compact sources at higher redshifts. These results stress the complexity of high-redshift radio-loud AGN populations and emphasize the need for multi-wavelength, high-resolution observations to fully characterize their properties and evolution through cosmic history.

1. Introduction

The study of high-redshift active galactic nuclei (AGN) provides unique insights into not only the early evolution of supermassive black holes (SMBHs) and their host galaxies but also fundamental physics under extreme conditions during a critical period in cosmic history [1,2,3,4]. While multi-wavelength observations have revealed much about black hole accretion processes, very long baseline interferometry (VLBI) observations of relativistic jets offer direct probes of the innermost regions around these early black holes, providing crucial measurements of their energetic output and feedback mechanisms [5,6,7,8].

At lower redshifts, radio-loud AGN constitute approximately 10% of the entire AGN population [9,10]. However, the evolution of this fraction with cosmic time and its dependence on physical parameters remains heavily debated. Recent studies have revealed complex trends, with some suggesting an increase in the radio-loud fraction with optical luminosity [11], while others find evidence for a roughly constant fraction across a wide range of redshifts and luminosities [9,10]. The apparent contradictions in these findings highlight the challenges in disentangling genuine evolutionary effects from observational biases.

Theoretical work suggests that radio jets may play a particularly important role in galaxy evolution at high redshifts through mechanical feedback [6]. The intense energy output from these jets can significantly impact the surrounding interstellar and intergalactic medium, potentially regulating both star formation and black hole growth in the early universe. Understanding the physical properties of these jets is therefore essential for constraining models of galaxy formation and evolution.

Blazars [12], a subset of AGN characterized by relativistic jets oriented close to our line of sight [13,14], are preferentially detected in flux-limited surveys due to Doppler-boosting effects, particularly at high redshifts [15]. This selection bias must be carefully considered when interpreting population statistics. While blazars typically exhibit flat radio spectra (e.g., [16]), a surprising finding from recent surveys is that a significant fraction of high-redshift radio-emitting AGN have been found to display steep or peaked spectra [17].

These spectral characteristics are more commonly associated with either young radio sources or those confined by dense environments [18,19], such as compact steep-spectrum (CSS) and GHz-peaked spectrum (GPS) sources [19,20,21,22]. The study of GPS and CSS sources at high redshifts provides crucial insights into the early stages of jet development and its interaction with the host galaxy (e.g., [23,24]).

The high prevalence of GPS and CSS sources at high redshifts raises fundamental questions about the physical conditions that may favor the formation of these compact radio sources in the early universe. Multiple factors could contribute to this phenomenon, including higher ambient gas densities affecting jet propagation, more frequent galaxy mergers triggering AGN activity, different accretion modes in early SMBHs, and selection effects in current surveys [25,26]. Understanding the relative abundance of these factors is crucial for constraining models of AGN evolution across cosmic time.

In the first paper in this series, Sotnikova et al. [27] (Paper I) presented a comprehensive study of 102 high-redshift ($z\ge 3$) quasars with $S_{1.4\mathrm{GHz}}>100$ mJy observed using the RATAN-600 telescope. They found that nearly 50% of the sample exhibited peaked spectra, while only 24% showed flat spectra typical of blazars, suggesting a significant population of young, evolving radio AGN at high redshifts. These sources offer a unique opportunity for investigate SMBHs during their formative stages. The powerful jet feedback mechanisms in young radio AGN (e.g., [26,28]) provide insights into the co-evolution of black holes and their host galaxies in the early universe.

The second paper in our series [29] (Paper II) focused on the variability properties of these high-redshift quasars, revealing complex patterns of flux density variations across multiple frequencies. This variability analysis provided important constraints on the physical processes occurring in these distant AGN, complementing the spectral classification from the first paper.

In this third paper, we present the first systematic VLBI study of parsec-scale jet properties in a well-defined, flux-limited sample of 86 radio-loud quasars at $z\ge 3$. Using multi-frequency VLBI observations spanning multiple epochs, we aim to address several key questions: what are the predominant morphological characteristics of jets in high-redshift quasars? How do the core properties (size, flux density, brightness temperature) of these distant sources relate to their overall spectral characteristics and variability? For sources with multi-epoch data, what can we infer about jet kinematics and evolution at high redshifts? How do the physical properties of GPS and CSS sources at high redshifts compare to their lower-redshift counterparts?

At the redshifts of our targets ($z=3\text{–}5.3$), these observations probe physical scales of just a few parsec in projection1, providing unprecedented resolution of the jet collimation regions. This allows us to directly study the physics of relativistic jets during an epoch when the Universe was less than 2 billion years old.

By combining our VLBI analysis with previous spectral and variability studies, we aim to present a comprehensive picture of the properties of the radio-brightest high-redshift quasars, from their integrated spectra and variability characteristics to their parsec-scale jet structures. This multi-faceted approach allows us to probe the physical processes governing jet formation and propagation in the early universe, providing valuable constraints on models of SMBH growth and AGN evolution during the crucial transition from cosmic dawn to cosmic noon.

2. Sample

2.1. Sources with VLBI Data

Our study builds upon a well-defined parent sample of 102 high-redshift quasars initially observed by Sotnikova et al. [27]. This parent sample was selected based on three key criteria: redshift $z\ge 3$; 1.4 GHz total flux densities exceeding 100 mJy; declination range from $-{35}^{\circ }$ to $+{49}^{\circ }$. From this parent sample, we identified 86 sources with available VLBI data suitable for detailed structural analysis (Appendix A). While this selection ensures sufficient signal-to-noise ratios for reliable VLBI imaging, we note that the resulting sample cannot be considered strictly complete due to the VLBI data availability criterion. Therefore, the statistical results should be interpreted with appropriate caution.

The redshift range of our targets (z = 3–5.3) encompasses a crucial epoch in cosmic history, spanning from the end of the Epoch of Reionization (which ended around z∼5.3) to the onset of the cosmic noon (around z∼3). This period, which is marked by rapid evolution in both galaxy properties and black hole growth rates, provides a unique window into AGN physics.

Our sample selection process involves several important selection effects that must be carefully considered when interpreting the results. While the radio selection criterion ($S_{1.4\mathrm{GHz}}>100$ mJy) is well defined, the requirement for spectroscopic redshifts introduces an additional selection bias, potentially favoring optically brighter sources. Of the original 102 sources in the parent sample, 16 lack VLBI detections in the Astrogeo VLBI database2. Analysis of their single-dish measurements suggests these sources may have a more extended structure that is resolved out at VLBI scales, implying our VLBI-detected sample could be biased toward more compact sources.

The geodetic/astrometric nature of our VLBI observations introduces further selection effects. These observations typically target compact, bright sources optimal for geometric measurements—a fundamentally different selection than imaging-focused surveys like the MOJAVE (Monitoring Of Jets in Active galactic nuclei with VLBA Experiments) [30,31,32], which often preferentially observed sources with complex, evolving jet structures. This distinction is particularly important when comparing our results with lower-redshift samples (Section 4.3).

Table A1. The observational parameters of 1213 observation epochs ¹. The columns include the source name (1), redshifts (2), observation date (3), bandwidth (4), observation frequency (5), the synthesis beam FWHM and position angle (6, 7, 8), thermal noise level rms (9), integrated flux density, core flux density and core size from Gaussian model fitting (10, 11, 12), the ratio of core flux density versus integrated flux density (13), and the VLBI integrated flux density versus RATAN flux density observed within 3 months (14).

Name	z	Epoch	BW	$\mathit{\nu }$	${\mathit{\theta }}_{\mathit{maj}}$	${\mathit{\theta }}_{\mathit{min}}$	P.A.	rms	${\mathit{S}}_{\mathit{int}}$	${\mathit{S}}_{\mathit{core}}$	Core Size	$\frac{{\mathit{S}}_{\mathit{core}}}{{\mathit{S}}_{\mathit{int}}}$	$\frac{{\mathit{S}}_{\mathit{VLBI}}}{{\mathit{S}}_{\mathit{RATAN}}}$
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)
			(MHz)	(MHz)	(mas)	(mas)	(°)	(Jy beam−1)	(Jy)	(Jy)	(mas)
J0001+1914	3.10	2 January 1996	16.0	8336.89	2.93	0.94	104.1	1.15	0.38	0.38	0.16	1.00	0.99
J0048+0640	3.58	30 April 2004	32.0	8646.22	2.47	1.08	82.90	0.82	0.07	0.05	0.24	0.74	1.32
J0121-2806	3.11	12 May 2005	32.0	8646.22	3.51	1.19	101.8	1.15	0.14	0.14	0.80	1.00	0.87
J0148+4215	3.24	8 February 2012	32.0	8642.24	1.55	1.28	126.25	0.35	0.12	0.12	0.13	1.00	1.11
J0151+2517	3.10	9 July 2005	32.0	8646.22	4.20	1.01	105.7	1.02	0.11	0.11	1.66	1.00	1.07
J0203+1134	3.63	1 October 1998	16.0	8644.23	1.62	0.73	97.94	0.55	0.53	0.38	0.64	0.72	0.95
J0257+4338	4.06	30 June 2005	32.0	8646.22	1.96	0.98	83.80	0.50	0.12	0.10	0.30	0.88	0.88
J0324-2918	4.63	12 May 2005	32.0	8646.22	2.51	0.95	96.40	1.24	0.09	0.09	0.26	1.00	1.07
J0337-1204	3.44	7 May 1997	32.0	8339.47	2.80	1.12	88.75	1.89	0.10	0.10	0.18	1.00	0.45
J0339-0133	3.19	27 March 2007	32.0	8642.24	3.69	2.57	78.5	1.25	0.13	0.13	2.24	1.00	1.00
J0354+0441	3.26	15 July 1995	16.0	8336.89	2.34	1.05	92.44	1.24	0.20	0.2	0.83	1.00	0.93
J0424+0805	3.09	20 July 2005	32.0	8646.22	2.17	0.91	90.50	0.72	0.33	0.33	0.15	1.00	0.87
J0428+1732	3.32	2 January 1996	16.0	8336.89	2.67	0.99	100.75	1.26	0.16	0.16	0.31	1.00	1.03
J0539-2839	3.10	27 August 1997	32.0	8339.47	2.78	0.95	92.4	1.11	0.87	0.87	0.39	1.00	0.73

¹ The whole table can be found at https://doi.org/10.5281/zenodo.14258041.

presents the fundamental properties of our sample. The redshifts (column 2) are adopted from the NASA/IPAC Extragalactic Database (NED), while precise coordinates are derived from the Astrogeo archive. Radio spectral classifications are adopted from Sotnikova et al. [27], and morphological classifications are based on our VLBI image analysis. The sample spans a range of radio spectral types, from flat-spectrum sources typical of blazars to peaked-spectrum objects characteristic of young or evolving radio sources.

Among our sample, several sources are particularly noteworthy, warranting detailed study: (1) the highest-redshift source, 102,623 + 254,259 (z = 5.3), represents one of the most distant known radio-loud AGN. This source exhibits a one-sided core-jet morphology on milliarcsecond scales [33], with its jet oriented close to the line of sight. Its characteristics, including a large bulk Lorentz factor, high-brightness temperature, and fast jet proper motion, make it an exceptional laboratory for studying relativistic jets in the early Universe [34]. (2) 160,608 + 312,504 (z = 4.56, corresponding to J1606+3124) shows a rare compact symmetric object (CSO) morphology at such a high redshift. GPS with a rest-frame turnover frequency of 10.6–17.8 GHz provides important constraints on radio source evolution or confinement models [35]. Notably, this source is probably classified as a galaxy rather than a quasar, adding diversity to our high-redshift radio source population.

2.2. VLBI Data

Our study utilizes VLBI data obtained primarily from the Astrogeo VLBI archive, a comprehensive database containing calibrated VLBI observational data. The Astrogeo archive consists primarily of astrometric and geodetic VLBI observations, which typically have shorter integration times per source compared to dedicated imaging observations. This affects the sensitivity and uv-coverage of the data, resulting in typical image noise levels ranging from 0.1 to 1 mJy beam⁻¹ and a practical flux density sensitivity threshold of approximately 0.05–0.1 Jy at X-band and 0.1–0.2 Jy at S-band. This sensitivity limitation is reflected in the lower flux density cutoff observed in our sample distribution (Figure 1). Weaker sources were less frequently targeted due to correlator constraints and detectable signal-to-noise ratios.

For our sample, we collected all available VLBI images and calibrated visibility data. The observations span multiple frequencies, predominantly including 2.3 GHz (S band), 4.7/5.0 GHz (C band), 7.6/8.4 GHz (X band), 15 GHz (U band) and 24 GHz (K band) for selected sources. This multi-frequency coverage enables detailed spectral properties and investigation of frequency-dependent effects on source structure. The X-band (8.4 GHz) observations form the core of our analysis, providing an optimal balance between resolution, sensitivity, and consistent coverage across the sample.

Figure 1 presents the VLBI flux density distribution of our sample at both X-band (upper panel) and S-band (lower panel). The X-band (∼8.4 GHz) flux density distribution shows an asymmetric profile with a pronounced peak at 0.1∼0.25 Jy and a long tail extending to higher flux densities up to ∼3.0 Jy. The S-band (∼2.3 GHz) distribution exhibits a similar asymmetric pattern but peaks at slightly higher flux densities (0.2∼0.4 Jy).

Several key features are noteworthy. First, both bands show a sharp decline in source counts with increasing flux density, reflecting both the intrinsic rarity of bright radio sources and following the expected luminosity function of radio-loud AGN. Second, the lower flux density cutoff around 0.05–0.1 Jy in X-band and 0.1–0.2 Jy in S-band represents the sensitivity threshold of current VLBI observations rather than an intrinsic feature of the source population. Third, the different distributions between X and S bands provide insights into the spectral properties of these sources, with the systematically higher S-band fluxes indicating predominantly steep or peaked spectra.

The observational data in our study span from the early 1990s to 2022, providing a substantial temporal baseline for many sources. Notably, 62 out of our 86 sources have ≥2 epochs of observations, with some having up to 163 epochs for J0646+4451 (OH 471) [36], 137 epochs for J2129-1538. Our VLBI dataset complements and extends the RATAN-600 observations. While the RATAN-600 data covered a more recent and concentrated period from 2006 to 2019 [27] and from 2006 to 2022 [29], our VLBI dataset provides a longer historical perspective, reaching back to the early 1990s. This extensive temporal coverage is particularly valuable given cosmic time dilation effects. For our highest redshift sources, a 30-year observing baseline in the observer’s frame corresponds to only about 5–7 years in the source rest frame, necessitating long monitoring periods to study structural evolution.

The observations were conducted using various VLBI arrays, predominantly the very long baseline array (VLBA), but also including the European VLBI Network (EVN) and global VLBI arrays. This diversity of arrays provides complementary coverage of different angular scales and sensitivities, enabling robust characterization of source structures across a wide range of spatial frequencies.

This VLBI dataset, combined with our previous single-dish RATAN-600 monitoring and spectral analysis, provides comprehensive view of high-redshift radio-loud AGN properties and evolution. The combination of high angular resolution, multi-frequency coverage, and extensive temporal sampling makes this dataset uniquely suited for investigating the physics of relativistic jets in the early universe.

3. Methods

Our analysis of this high-redshift quasar sample employs a comprehensive methodological approach designed to extract maximum physical insight from the VLBI data while carefully accounting for observational limitations and potential biases. The key analysis steps are summarized below, with detailed procedures provided in Appendix B.

The VLBI data were obtained from the Astrogeo archive, which provides initial calibration for phase and complex gains. For about 12% of observations requiring additional calibration (primarily due to complex source structures or technical issues), we implemented enhanced calibration using the DIFMAP software package [37]. Quality control criteria included verification of amplitude and phase calibration, assessment of uv-coverage and sensitivity limits, and cross-validation with independent observations where available.

Source structures were characterized through model fitting using circular Gaussian components, a choice that provides good approximation for compact radio structures while minimizing free parameters. For multi-epoch sources, kinematic analysis required careful component identification and tracking, considering multiple parameters: component flux densities, sizes, positions relative to the core, and spectral properties where available.

Rest-frame brightness temperatures were calculated for fitted components using standard formulae, and careful attention was paid to resolution limits. For unresolved components, we derived conservative lower limits using the maximum resolution of each observation. Jet parameters were estimated using equipartition brightness temperature as a reference, though we note that this assumption may not hold in all cases, particularly in the most compact regions of AGN jets and in their maximum brightness state (see Appendix B.4 for discussion).

However, we note that while this method is straightforward, it relies on the assumption of equipartition between particle and magnetic field energy densities, which may not hold in all cases, particularly in the most compact regions of AGN jets and in their maximum brightness state. Therefore, these Doppler factor estimates should be considered indicative rather than definitive, and are used in conjunction with other observables to constrain the intrinsic jet parameters.

For sources with multi-frequency data, spectral index maps were constructed using matched-resolution images. This process required careful image alignment and consideration of both thermal noise and systematic calibration uncertainties. Through this systematic approach, we extracted reliable physical parameters while paying careful attention to uncertainties and potential systematic effects.

4. Results

4.1. Source Compactness

Our VLBI imaging analysis reveals distinctive patterns in the morphological properties of high-redshift radio-loud AGN. To quantify source compactness, we calculated the ratio $C=S_{VLBI}/S_{RATAN}$, where $S_{VLBI}$ represents the integrated flux density measured from VLBI observations and $S_{RATAN}$ denotes the flux density measured by RATAN-600. We focused on S band (2.3 GHz) and X band (7.6 and 8.4 GHz) for this analysis, as these frequencies had the highest number of source detections. This parameter C serves as a key diagnostic tool, with values approaching unity indicating highly compact sources where most emission originates from parsec scales. Lower C values suggest the presence of extended emission on scales larger than those probed by VLBI.

To compute the C parameter, we used several strategies to deal with the sparse VLBI and RATAN-600 data. We prioritized contemporaneous observations, interpolated flux densities when frequencies did not match, and applied variability corrections for non-contemporaneous data within 3 months. VLBI-only data provided the lower limits for C. These methods allowed us to calculate consistent C values across our sample, despite the data limitations.

Analysis of the compactness distribution (Figure 2) reveals a pronounced broad peak in our sample. The vast majority of sources cluster tightly in the range of C = 0.8–1.1, with a peak around $C\approx 0.9$, indicating that most of the emission originates from parsec-scale structures which can be detected by VLBI. For the compactness distribution analysis shown in Figure 2, we used contemporaneous single-epoch measurements directly from the Astrogeo catalogue, where C typically ranges from 0.6 to 1.4. The values of $C>1$ may occasionally occur due to source variability between VLBI and RATAN-600 observation epochs, even after applying our 3-month contemporaneity criterion.

Of the 102 parent sample, 16 sources lacking VLBI detections which were excluded in this study would likely show lower C values if they could be measured. Including these sources would extend the distribution toward lower compactness values, though the predominance of compact sources in our sample would likely remain significant.

The overwhelming majority of sources in our sample display remarkably compact structures, with 35% appearing as unresolved cores and 59% showing core-jet morphologies. The compactness parameter serves as an independent metric that corroborates our morphological classification scheme (see the discussion in Section 4.2). Sources classified as core-dominated are expected to exhibit C values approaching unity, while those with significant extended emission should display lower C values. Indeed, the majority of our sources (30%) demonstrate high C values (>0.8), consistent with their classification as core or core-jet morphologies. This high degree of compactness aligns with expectations for a sample of high-redshift quasars, where extended emission is often resolved out in VLBI observations [38,39].

To investigate evolutionary trends in our sample, we examined the relationship between compactness parameter C and redshift (Figure 3). Each source in the plot shows a mean value of C values with standard deviation due to variability across multiple epochs of observation. Across the redshift range (z = 3.0–4.6), the mean compactness values consistently cluster around $C\approx 1.0$, showing no significant systematic evolution with redshift. This stability suggests that the basic structural properties of radio-loud AGN cores remain relatively constant throughout this epoch of cosmic history. Moreover, we observe a remarkable decrease in the dispersion of compactness value C as redshift increases. While sources at $z\approx 3$ show a wide range of C values (roughly 0.3 to 3.0), this spread progressively narrows at higher redshifts. This reduction in scatter could reflect either observational effects or physical processes. From an observational perspective, surface brightness dimming ($\propto {(1+z)}^{-4}$) might preferentially eliminate detection of extended structures at higher redshifts, effectively reducing the range of observable C values. Additionally, our flux-limited sample may preferentially select more beamed, compact sources at higher redshifts, further reducing the observable range of structural variations. The combination of stable mean compactness with decreasing variance provides important constraints for models of AGN evolution, though interpreting these trends requires careful consideration of selection effects and observational biases. Future observations with increased sensitivity to extended emission will be particularly valuable for understanding the reduced scatter at higher redshifts.

Only four sources (J0214+0157, J1457+0519, J1715+2145, J2314+0201, approximately 5% of the sample) show notably lower C values (<0.7), suggesting these are rare cases in which a substantial emission comes from larger-scale structures resolved by VLBI. This strong predominance of compact sources likely reflects both the selection effects of our flux-limited sample and the intrinsic nature of high-redshift radio-loud AGN, where extended emission may be suppressed by increased inverse Compton losses against the cosmic microwave background.

Figure 4 illustrates the relationship between the maximum source compactness (C) and variability index ($V_{S_{8.2}}$) for our sample. The variability index $V_{S_{8.2}}$ is calculated using the Formula (1) in Sotnikova et al. [29] at X band. Our correlation analysis yields a weak positive correlation coefficient of 0.22 between $V_{S_{8.2}}$ and C. While this correlation is not strong, it suggests a subtle tendency for more compact sources to display slightly higher-variability amplitudes. However, the scatter in the distribution is considerable, with variability indices ranging from nearly 0 to about 0.95 across different C values. Some of the most compact sources ($C>1.5$) show moderate-to-high variability ($V_{S_{8.2}}>0.4$), but there are also several compact sources with low variability levels. In contrast with the display in Figure 2, we plot the maximum value of C observed across all available epochs for each source to capture the full range of structural variations. This multi-epoch approach results in some sources showing larger C values (up to ∼3.0) due to significant flux variability over the long monitoring period. These higher values typically correspond to epochs of enhanced source activity where the VLBI flux density temporarily exceeded the reference RATAN-600 measurements. This complex pattern suggests that while source compactness may influence variability behavior, it is likely not the dominant factor determining variability amplitude in high-redshift AGN. Other physical parameters, such as jet orientation, Doppler boosting, and intrinsic source properties, may play equally important or even more significant roles in determining the observed variability characteristics.

4.2. Morphology

The morphological classification of high-redshift AGN presents unique challenges due to their extreme compactness, with many sources appearing unresolved even at VLBI resolutions. This limitation necessitates a comprehensive approach that integrates multiple lines of evidence beyond spatial structure alone.

Our classification methodology combines four key diagnostic indicators: spatial morphology, spectral properties, variability characteristics, and brightness temperature measurements. This integrated approach helps overcome the limitations of traditional classification schemes that rely primarily on a resolved structure. For example, while jet components typically exhibit steep-spectrum optically thin emission above 2.3 GHz (corresponding to ${\nu }_{\mathrm{int}}\ge 10$ GHz in the observer’s system of $z\ge 3$ sources), we find that even core components in our sample can display steep spectra between 5 and 8.4 GHz, likely due to the prevalence of peaked spectral distributions at high redshifts [27], which may indicate either youthful AGN activity or dense ambient medium confinement; in these cases, spectral indices measured between 5 and 8.4 GHz often fall in the steep-spectrum part of these peaked distributions.

To better understand these spectral characteristics, we constructed detailed spectral index maps for each source (Appendix B.6; see the example in Figure A2), providing crucial additional constraints for morphological classification. Core components in our sample typically display high brightness temperatures (${10}^{10}$ K), flat or inverted spectra, significant variability, and compact structure. In contrast, jet components generally exhibit lower brightness temperatures, steeper spectra, and more extended morphologies when resolved.

This multi-parameter classification framework allows us to confidently categorize sources even when traditional structural analysis proves insufficient, providing a robust foundation for investigating the physical properties of high-redshift radio-loud AGN.

Based on these criteria, we classified sources into the following categories: compact core-only sources (C), core-jet structures (CJ), and compact double jets (CD). Figure 5 shows the example images of these types. Class C sources retain point-like morphology at all resolutions, which is identified by its flat or inverted spectrum, high-level variability and high brightness temperature. Class CJ sources have a distinct core and one-sided weaker jet structure, typically showing a steep spectrum and brightness gradient along the jet axis. Class CD sources show two steep spectrum features with comparable brightness, similar to CSOs. For sources with multi-epoch observations, we ensure the classification is consistent across epochs, noting any significant structural changes.

The core-only sources (Class C, accounting for 35%) are characterized by a single, compact component that remains unresolved even at the highest available angular resolution (∼0.5 mas at 8.4 GHz). These sources exhibit median brightness temperatures of $T_b=(8.97\pm 2.15)\times {10}^{10}$ K, approaching but generally not exceeding the inverse Compton limit. This suggests that while relativistic beaming is present, extreme Doppler boosting factors ($\delta >10$) are rare in our high-redshift sample. Some of the class C sources display flat spectra in RATAN-600 observations, suggesting they are likely blazars [40]. A subset (55%) show peaked spectra [27], potentially indicating young or recently reactivated AGN [20].

Core-jet (CJ) sources comprise 59% of our sample. Their total flux density spectra range from flat to peaked, reflecting the complex interplay between core and jet emission at different frequencies. This spectral diversity highlights the importance of high-resolution VLBI observations in disentangling emission components in high-redshift quasars [41].

Core-double jet (CD) sources, while rare in our sample (5%), present intriguing cases for study. These sources typically show GPS-type peaked spectra in their total flux density measurements, suggesting they may represent young radio sources in the early stages of evolution [20,24]. The small fraction of complex or two-sided sources in our sample presents an interesting puzzle. Their rarity could reflect either genuine evolution in radio source properties with cosmic time or observational biases against detecting extended emission at high redshifts. Extended emission could be potentially missed in our observations, suggesting that the true fraction of complex sources might be higher.

Our classification is unavoidably subject to resolution limitations, especially for the most distant sources. Some core-jet sources might be misclassified as compact cores due to these constraints. Future observations with enhanced sensitivity and resolution may reveal additional structural details, potentially leading to reclassification of some sources [42,43,44].

A particularly intriguing finding is the presence of large jet bending (>90°) in five sources in our sample (J0203+1134, J1230-1139, J1405+0415, J1445+0958, J2003-3251), accounting for approximately 6% of our sample. This fraction appears slightly higher than that seen in lower-redshift samples [32]. This comparison warrants careful consideration. The large jet bending may indicate stronger interaction with a denser circumnuclear medium, more frequent merger-induced perturbations or enhanced pressure gradients in the early universe. Our flux-limited sample is inherently biased toward the most powerful radio sources at high redshift, which may be less susceptible to environmental effects due to their higher jet powers. A more complete understanding of jet bending mechanisms across cosmic time would require expanding studies to include weaker radio AGN at high redshifts, where lower-power jets would be more sensitive to interaction with the host galaxy environment and ambient medium. Additionally, the denser environments expected in the early Universe could play a more significant role in jet deflection for these lower-power sources.

This morphological analysis provides insights into jet physics and source evolution in the early universe, though several observational biases must be carefully considered. First, VLBI observations inherently tend to miss low surface brightness emission, particularly at high redshifts where surface brightness sensitivity is further reduced by cosmological dimming. Second, our flux-limited sample preferentially selects brighter sources, which are more likely to have jets aligned closer to our line of sight, resulting in compact nuclei that can outshine extended jet emission due to Doppler boosting. The high median brightness temperature ($T_b=(8.97\pm 2.15)\times {10}^{10}$ K) observed in our Class C sources exceeds equipartition values, suggesting Doppler boosting and thus orientation bias, which is a notable contrast to studies of sources at $z>6$ [45]. Therefore, while the prevalence of compact structures in our sample could indicate either young sources or strong environmental confinement of jets at high redshifts, these interpretations must be weighed against the aforementioned selection and observational biases. These findings, when properly accounting for observational effects, provide constraints for models of AGN evolution and jet formation in the early Universe, though the true distribution of source properties may be broader than what our flux-limited, VLBI-selected sample reveals.

4.3. Jet Proper Motion

Our multi-epoch VLBI observations, spanning 5–15 years depending on the source, provide unprecedented insights into jet kinematics at high redshifts. Our kinematic analysis focuses on 57 sources with multiple epochs of X-band VLBI observations, providing optimal resolution and sensitivity for tracking jet component evolution. Within this subset, 34 sources have three or more epochs, enabling robust proper motion measurements through least-squares fitting of component trajectories. Examples of jet proper motions are shown in Figure 6. Other results can be accessed at Zenodo3. The distribution of jet proper motion is shown in Figure A1.

For sources with only two epochs, we calculated jet proper motions directly from the positional changes of identifiable components. The remaining sources with 3–9 epochs were analyzed using standard component tracking and linear fitting techniques.

Nine sources in our sample have more than 20 epochs of observations, allowing for detailed kinematic analysis. Some of these have been the subject of previous studies (J0339-0133 [46], J0001+1914, J1658-0739 [47], J1405+0415 [48], J0203+1134, J0539-2839, J1354-0206, J2129-1538, and J0646+4451 [36]). These well-sampled cases reveal complex kinematic behavior, with some components showing evidence for acceleration and/or non-linear trajectories.

The measured proper motions range from 0.002 to 0.09 mas yr⁻¹, corresponding to apparent superluminal speeds between 0.2 c and 10.0 c when converted to projected linear velocities. The analysis reveals an intriguing distribution of jet speeds that challenge conventional expectations for high-redshift blazars (Figure A1). This distribution of high-z jets shows a marked difference from lower-redshift samples, where maximum apparent speeds often exceed $10c$. The lower apparent speeds in our high-redshift sample suggest either intrinsically lower jet bulk Lorentz factors or stronger environmental effects on jet propagation.

Figure 7 presents a comparison between jet proper motions in our high-redshift sample and those from the MOJAVE sample (redshift range 0.05 to 3.40, median $z=0.79$). While MOJAVE observations preferentially target sources with complex jet structures, potentially biasing their sample toward more dynamically active AGN, our comparison focuses on maximum jet speeds, as these represent the highest achievable velocities in each population. This approach provides a conservative benchmark for comparing jet kinematics between epochs, as any selection bias toward more active sources in MOJAVE would only strengthen our finding that high-redshift jets show systematically lower speeds. The observed difference in velocity distributions persists even when accounting for these selection effects, suggesting a genuine evolution in jet kinematical properties with cosmic time.

The unexpected prevalence of slow or stationary components could be attributed to several possibilities: faster-moving components might exist on scales smaller than our resolution limit; sources viewed at larger angles to the line of sight will exhibit slower apparent motions; there may be a population of high-redshift sources with genuinely lower jet velocities.

The kinematic data also reveal interesting correlations with other source properties. Sources with lower apparent speeds tend to show higher fractional polarization and more stable flux densities, suggesting these may be viewed at larger angles to the line of sight despite their blazar-like core dominance. Conversely, the sources showing faster apparent motions typically display greater variability and lower polarization fractions, consistent with standard beaming models.

These findings have important implications for our understanding of jet physics in the early universe. The scarcity of very high apparent speeds, combined with the trends in polarization and variability, suggests that either the intrinsic jet properties or their interaction with the environment differs significantly from what we observe in the local universe.

5. Discussion

Our investigation of high-redshift radio-loud AGN using VLBI imaging reveals several notable characteristics that emphasize the profound influence of the early universe environment on AGN jet evolution. Our VLBI imaging reveals a striking dominance of compact structures, with 94% of sources showing either core-only or core-jet morphologies, significantly exceeding the fraction observed in lower-redshift samples, where more extended and complex structures typically dominate [49]. While surface brightness dimming at high redshifts might create bias against the detection of large-scale structures [50]), our analysis indicates that observational effects alone cannot fully account for the pronounced morphological differences between high- and low-redshift AGN populations.

Proper motion measurements further stress fundamental distinctions in jet dynamics. We find that the apparent jet speeds in our high-redshift sample, ranging from 0–10 c, are systematically lower than those in well-studied low-redshift projects such as MOJAVE, where superluminal speeds frequently exceed 10 c [32]). This discrepancy persists even after accounting for flux-limited selection biases [51]), and the absence of very fast jets (${\beta }_{\mathrm{app}}>10c$) [39] strongly suggests intrinsically lower bulk Lorentz factors or more pronounced environmental confinement at early epochs [52]). These findings align with theoretical expectations that the dense protocluster environments prevalent at high redshifts impose significant constraints on jet propagation [53,54]).

In addition to kinematic differences, the relationship between source compactness and variability offers valuable insights. Although only a weak positive correlation emerges between the compactness parameter C and the maximum variability index $V_{s,max}$, this trend implies that more compact emission regions are associated with marginally greater variability amplitudes [55]). Yet the substantial scatter in this correlation indicates that factors beyond emission region size—such as jet composition, magnetic field configuration, or inhomogeneous external media—also influence variability characteristics [56]).

The prevalence of GPS and CSS sources within our sample further enriches our understanding of early AGN evolution. These spectral types may indicate either young AGN or “frustrated” sources confined by dense ambient medium [22,24]). The observed properties could be explained by either evolutionary effects or environmental conditions unique to the early universe. At high redshifts, both scenarios are plausible: rapid supermassive black hole growth and abundant ambient gas could jointly govern jet evolution, hindering large-scale expansion and favoring early-stage or absorbed spectra [5,57].

Taken together, our findings strongly suggest that the early universe exerts a more pronounced influence on AGN jet formation and evolution than previously recognized. Relative to lower-redshift systems, high-redshift jets display systematically lower speeds, more compact morphologies, and a higher fraction of peaked-spectrum sources, consistent with stronger environmental confinement and enhanced absorption processes [26,58]. Dense early environments may inhibit the development of large-scale radio structures and, through free-free absorption or synchrotron self-absorption, support a high occurrence of peaked-spectrum AGN [59,60]). Despite these constraints, the observed diversity in AGN properties suggests that intrinsic processes and line-of-sight effects continue to generate significant observed variability, even under challenging early-universe conditions [61].

Comparisons with established low-redshift samples (e.g., MOJAVE) [30,32] confirm that the absence of ultra-fast jets and extremely high brightness temperatures in our high-redshift sources is not merely a selection artifact, but likely represents a genuinely different population [39]). These results prompt a reevaluation of existing AGN evolutionary models, encouraging new theoretical frameworks that incorporate environmental density and cosmic epoch as key parameters [34]. Furthermore, the systematically lower brightness temperatures in our sample suggest diminished Doppler boosting or altered microphysical conditions in the early Universe [54,62]).

In addition, the high fraction of compact and peaked-spectrum AGN may indicate that we are observing many objects at nascent stages, which have either yet to evolve into large-scale radio sources or have been “stalled” by dense media [24,63]. This provides a unique window into the early growth and feedback of AGN in the formative stages of the universe [5]. Our observations imply that the duty cycles, lifetimes, and evolutionary pathways of early AGN may differ significantly from their low-redshift counterparts, shedding new light on how supermassive black holes accumulated substantial mass so rapidly in the young cosmos [64].

Future VLBI observations with higher sensitivity and longer time baselines will be crucial to capturing subtle kinematic signatures and faint extended structures. Multi-wavelength observations, including millimeter/submillimeter and X-ray data, will help characterize the host galaxy environments, linking jet properties to their cosmic habitats [65,66]).

6. Conclusions

Our systematic VLBI study of 86 high-redshift ($z\ge 3$) radio-loud AGN has provided new insights into relativistic jets in the early universe. Through detailed analysis of their morphology, kinematics, and spectral properties, we find evidence that jet physics at high redshifts may differ significantly from the local universe.

The key findings from our study reveal three notable characteristics. First, the sources show a remarkably high degree of compactness, with VLBI-to-single-dish flux density ratios peaking sharply around 0.9. This compactness likely reflects both selection effects and enhanced inverse Compton losses against the cosmic microwave background at high redshifts. Second, the jet kinematics show predominantly low apparent speeds (${\beta }_{\mathrm{app}}<15c$), in contrast to lower-redshift samples in which faster speeds are common for radio-brightest quasars. Third, the high fraction (∼50%) of peaked and steep-spectrum sources suggests a significant population of young or recently activated AGN and/or sources interacting with dense environments at high redshifts.

These results challenge our current understanding of AGN jet evolution. The combination of slower jet speeds, lower brightness temperatures, and high peaked-spectrum fraction points toward either stronger environmental effects or fundamentally different jet properties in the early universe. The prevalence of young radio sources in our sample may provide crucial insights into the triggering mechanisms of AGN activity during this early cosmic epoch.

Future observations with next-generation facilities like the SKA and FAST Core Array and their VLBI components [42,67], combined with large-sample statistical studies, will be essential for expanding these studies to larger samples and higher sensitivity. These observations, combined with theoretical advances, will help distinguish between evolutionary and orientation effects, ultimately providing a clearer picture of how radio-loud AGN develop and evolve in the early universe.