In next-generation wireless networks (6G), moving to millimeter-wave (mmWave) and Terahertz (THz) frequencies is essential to unlock massive bandwidths and ultra-high data rates. To overcome the severe path loss at these high frequencies, systems rely on massive MIMO architectures with large antenna arrays to focus signals into highly directional, narrow beams.
However, a fundamental physical mismatch arises when combining wideband signals with traditional analog array hardware. This mismatch results in beam squint (or the beam split effect), where different frequency components of the same transmission steer toward entirely different spatial directions. If left uncompensated, this effect causes catastrophic signal loss at the edges of the frequency band, creating a strict bottleneck on achievable throughput.
The Physics of Beam Squint
To understand why beam squint occurs, we must look at how an antenna array steers a wireless beam. A standard phased array steers a signal by introducing a progressive phase shift between adjacent antenna elements.
For a signal to propagate at a specific steering angle $\theta$, the phase shift $\Delta \phi$ between elements separated by a distance $d$ must satisfy:
$$\Delta \phi = \frac{2\pi d}{\lambda} \sin\theta = \frac{2\pi f d}{c} \sin\theta$$
where $\lambda$ is the wavelength, $f$ is the frequency, and $c$ is the speed of light. Rearranging this for the steering angle yields:
$$\sin\theta = \frac{c \cdot \Delta \phi}{2\pi d \cdot f}$$
In conventional hybrid analog-digital beamforming architectures, the analog phase shifters are frequency-independent; they apply a single, flat phase shift $\Delta \phi$ across the entire signal bandwidth. As the equation demonstrates, if $\Delta \phi$ is fixed, the steering angle $\theta$ becomes a function of the frequency $f$.
As Gonzalez-Coma et al. and He et al. point out, traditional narrowband channel assumptions assume that the spatial steering vector remains constant across the bandwidth. While this assumption holds for narrow bands, it breaks down entirely in wideband regimes.
Because the beamwidth of a massive antenna array is incredibly narrow, even a slight frequency-dependent shift causes the beam to miss the intended receiver at the band edges. As documented by Dai et al., this spatial dispersion can cause upwards of an 80% loss in array gain at the peripheral frequencies of a wideband Terahertz link.
Traditional Phase Shifters (Frequency-Independent):
Low Frequency f_min -------------------------> [ Tilts Left ]
Center Freq f_c -------------------------> [ Aligned to Receiver ]
High Frequency f_max -------------------------> [ Tilts Right ]
Result: Signal energy splits spatially; band edges miss the target.
The geometry of the array also changes how this distortion manifests. While linear arrays experience a straightforward angular shift, Guo et al. demonstrate that in Uniform Circular Arrays (UCAs), beam squint causes severe pattern distortion rather than a simple spatial translation, leading to a total dissipation of beamforming gain away from the center frequency.
Hardware Solutions: True Time Delay and Movable Antennas
To maintain high array gain across an entire wideband channel, the beamforming hardware must introduce frequency-dependent phase shifts that scale linearly with $f$, effectively canceling out the frequency term in the denominator of the steering equation.
True Time Delay (TTD) Networks
The most direct hardware solution is replacing or augmenting phase shifters with True Time Delay (TTD) units. Instead of shifting the phase by a fixed angle, a TTD unit introduces a fixed time delay $\Delta t$. Because phase shift is related to time delay by $\Delta \phi(f) = 2\pi f \Delta t$, the phase shift automatically scales with frequency. Substituting this back into the steering equation yields a frequency-independent angle:
$$\sin\theta = \frac{c \cdot (2\pi f \Delta t)}{2\pi d \cdot f} = \frac{c \cdot \Delta t}{d}$$
Dai et al. proposed a hybrid Delay-Phase Precoding (DPP) architecture that inserts a small network of TTDs between the RF chains and the phase shifters, achieving over 94% of the optimal array gain across massive bandwidths. Similarly, Wang et al. leverage TTDs to resolve the beam split effect in Simultaneously Transmitting and Reflecting Surfaces (STARS).
Because equipping every single antenna element with a TTD unit is prohibitively expensive and power-hungry, Liao et al. introduced a Grouping True-Time Delay Unit (GTTDU) framework, which groups antennas into sub-arrays that share a single delay unit, balancing hardware efficiency with squint compensation.
Movable Antennas (MA)
An alternative paradigm avoids complex time-delay circuitry altogether by using mechanical flexibility. By dynamically adjusting the physical positions of the antenna elements, systems can alter the geometric path distance $d$ between elements.
As explored by Zhu et al. and Zhu et al., this physical displacement creates a spatial delay that acts as a natural mechanism for frequency-dependent phase compensation. By optimizing the physical positions of the elements, the array can align signal phases across a broad frequency spectrum for specific user locations without relying on high-frequency analog TTD components.
Movable Antennas (Mechanical Compensation):
[Antenna 1] <---> [Antenna 2] <---> [Antenna 3]
\ / /
\ / / (Adjusting physical spacing 'd'
\ / / compensates for frequency drift)
v v v
Receiver
Algorithmic and Signal Processing Mitigations
When hardware modifications are constrained, the beam squint effect can be managed through advanced signal processing and algorithmic design at the digital baseband.
Analog Beam-Broadening
Instead of trying to perfectly align narrow beams at every frequency, one approach is to widen the beam's spatial footprint so that it encompasses all frequency-dependent shifts. Gao et al. developed an analog beam-broadening technique via subarray aggregation. By dividing a large array into smaller subarrays and applying coordinated phase shifts, the system creates a broader beamwidth that maintains consistent coverage across the entire signal bandwidth, eliminating the need for TTD hardware at the cost of peak directional gain.
Frequency-Dependent Dictionaries and Interpolation
Beam squint severely complicates channel estimation. Because the apparent angle of arrival shifts across subcarriers, the channel's spatial sparsity pattern is broken, causing energy diffusion in the angular domain.
To resolve this, Zhong et al. developed an architecture utilizing frequency-dependent dictionaries. These dictionaries mathematically warp and align the angular grids across different subcarriers, transforming a smeared signal representation back into a sharp, sparse pattern that standard compressed sensing or deep learning models can easily recover.
In a similar vein, Lin et al. introduced multi-dimensional spatial interpolation. This technique mathematically reconstructs wideband signal components at a single, consistent reference frequency, enabling the coherent integration of spatial information across the entire bandwidth to achieve high-precision parameter estimation.