TPPWG: Tapered Parallel-Plate Waveguide

• TPPWG is a parallel‐plate waveguide with a variable separation that transforms a free-space THz beam into a tightly confined guided field.
• Its design optimizes impedance matching and field enhancement, achieving peak factors of over 6 with minimal dispersion across a broad THz band.
• TPPWGs enable advanced beam manipulation such as bunch compression and dielectric THz-driven acceleration, serving as both couplers and interaction structures.

A tapered parallel-plate waveguide (TPPWG) is a parallel-plate waveguide formed by two conducting plates whose separation varies along the propagation direction. In the THz literature, the taper has been implemented either as a symmetric horn-like reduction of the plate spacing $h(z)$ or as an exponential profile $y(x)=\pm a e^{b|x|}$ with $a=g/2$, so that a free-space THz beam is transformed into a confined guided field near a narrow interaction gap (Genre et al., 19 May 2026, Othman et al., 2019). In accelerator and ultrafast-beam contexts, the TPPWG serves as a THz coupler, a field transformer, and an interaction structure for compression or acceleration; in related theoretical work, the same geometry is also treated as a platform for transformation-optics tapers, metasurface waveguides, and rigorous scattering analysis (Tichit et al., 2010, Ma et al., 2019, Liang et al., 19 Jul 2025).

1. Definition and modal structure

Electromagnetically, a TPPWG is a metal waveguide formed by two conducting plates with a separation $h(z)$ that is varied along the propagation direction $z$. For THz frequencies, this structure supports primarily the TEM mode, which has no cutoff and very low dispersion (Genre et al., 19 May 2026). In an ideal parallel-plate guide with plate spacing $h$, the higher-order TE$_n$ and TM$_n$ modes have cutoff frequencies

$f_{c,n}=\frac{n c}{2h},\quad n=1,2,\dots$

so tapering alters both the local modal spectrum and the degree of field confinement (Genre et al., 19 May 2026).

Coordinate conventions differ across implementations. In the symmetric accelerator geometry, the taper is described through the plate spacing $h(z)$ and a flare angle $y(x)=\pm a e^{b|x|}$0 (Genre et al., 19 May 2026). In the THz bunch-compression geometry, the plates are tapered in one transverse dimension according to

$y(x)=\pm a e^{b|x|}$1

so that the local plate separation is $y(x)=\pm a e^{b|x|}$2 and the minimum gap $y(x)=\pm a e^{b|x|}$3 defines the interaction region (Othman et al., 2019).

A common misconception is that the existence of a TEM mode makes all TPPWGs effectively single-mode. The accelerator design at $y(x)=\pm a e^{b|x|}$4 THz with an exit gap $y(x)=\pm a e^{b|x|}$5 explicitly notes that higher-order modes can exist in principle, since for $y(x)=\pm a e^{b|x|}$6 the first cutoff is $y(x)=\pm a e^{b|x|}$7 THz; the observed behavior is instead dominated by a mode-matched fundamental TEM-like mode (Genre et al., 19 May 2026). Likewise, the MeV bunch-compression study describes its exponentially tapered PPWG as “dispersion-free” in a practical sense: single-cycle waveforms centered at $y(x)=\pm a e^{b|x|}$8 THz are preserved with minimal dispersion over $y(x)=\pm a e^{b|x|}$9–$a=g/2$0 THz, but some dispersion is observed for the smallest tested gap $a=g/2$1 (Othman et al., 2019).

2. Tapering as impedance transformer and field concentrator

The defining function of the taper is to connect a free-space THz beam to a tightly confined guided mode while suppressing reflection. In the integrated accelerator geometry, the taper serves two main purposes: free-space–waveguide impedance matching and geometric field enhancement (Genre et al., 19 May 2026). The structure was optimized by CST time-domain simulation over a waveguide length range of $a=g/2$2 to $a=g/2$3 mm and a flare-angle range of $a=g/2$4 to $a=g/2$5; the refined optimum was

$a=g/2$6

at a central THz frequency of about $a=g/2$7 THz and an exit gap

$a=g/2$8

for $a=g/2$9 (Genre et al., 19 May 2026). The resulting peak field-enhancement factor was $h(z)$0, and the paper emphasizes that the enhancement does not vary sharply around the optimum, indicating robustness to fabrication and alignment tolerances (Genre et al., 19 May 2026).

Time-domain simulation and electro-optic sampling show the same trend. In the optimized symmetric TPPWG, the field amplitude is enhanced by a factor $h(z)$1 at mid-length and by a factor $h(z)$2 near the exit plane, while the multi-cycle waveform remains centered at $h(z)$3 THz (Genre et al., 19 May 2026). In the exponentially tapered PPWG used for single-cycle THz manipulation, electro-optic sampling in a $h(z)$4 gap measured peak fields of about $h(z)$5 from $h(z)$6 of incident THz energy, corresponding to a measured enhancement of about $h(z)$7 relative to free space (Othman et al., 2019).

The two studies highlight different taper logics. The symmetric $h(z)$8 mm, $h(z)$9 TPPWG concentrates a narrowband, multi-cycle $z$0 THz drive into a dielectric accelerator (Genre et al., 19 May 2026). The exponential PPWG acts as a one-dimensional horn in reverse: it transforms a Gaussian single-cycle THz beam into a stronger quasi-TEM field at the minimum gap while preserving the temporal waveform (Othman et al., 2019). In both cases, the taper is not merely a mechanical transition; it is the element that sets coupling efficiency, local field strength, and usable bandwidth.

3. THz beam manipulation and bunch compression

A TPPWG can operate as an active beam-manipulation structure rather than only as a coupler. In the MeV bunch-compression work, the exponentially tapered PPWG is used to impose a longitudinal energy chirp on a $z$1 MeV electron beam with a single-cycle THz pulse centered at about $z$2 THz (Othman et al., 2019). The beam tunnel has radius $z$3, corresponding to a tunnel cutoff of about $z$4 THz; frequencies below cutoff remain concentrated near the gap, whereas frequencies above cutoff can leak into the tunnel (Othman et al., 2019).

The unshorted single-feed TPPWG provides strong longitudinal field $z$5, but because the THz wave propagates transverse to the beam it also produces a significant $z$6 field and transverse temporal dispersion. To mitigate this, the authors introduce a shorted PPWG, in which the guide is electrically shorted at the beam tunnel location. The reflected wave forms a standing-wave-like field, increases electric-field uniformity across the beam, reduces $z$7, and yields a 50% increase in energy modulation for the same input THz energy (Othman et al., 2019).

The reported performance differences are substantial. With $z$8 of THz energy, the unshorted TPPWG gives an energy chirp of about $z$9 and compresses an initial $h$0 fs rms bunch to about $h$1 fs. Under the same conditions, the shorted TPPWG gives about $h$2, reduces transverse deflection, and compresses the bunch to about $h$3 fs after a $h$4 m drift, a compression factor of about $h$5 (Othman et al., 2019). These results place TPPWGs within the broader class of THz streaking, chirping, and compression devices rather than limiting them to passive guiding.

4. Integrated dielectric THz-driven acceleration

The most explicit accelerator realization of a TPPWG is the dielectric terahertz-driven accelerator that integrates a dual-pillar grating within a symmetric tapered parallel-plate waveguide (Genre et al., 19 May 2026). The TPPWG simultaneously couples two free-space THz beams into the device and enhances the field at the location of the dielectric accelerator. The plates are metallic and treated as perfect conductors in CST; the region between plates is vacuum or air (Genre et al., 19 May 2026).

The integrated dielectric structure is a silicon dual-pillar grating with refractive index $h$6 at THz frequencies. Its key dimensions are tied to the THz wavelength: $h$7

$h$8

For relativistic electrons, the synchronism condition is

$h$9

which reduces to $_n$0 when $_n$1 (Genre et al., 19 May 2026).

Beam-dynamics simulations use a $_n$2 MeV beam with $_n$3 normalized energy spread, $_n$4 fs FWHM bunch length, $_n$5 transverse size, $_n$6 emittance, and bunch charge from $_n$7 pC to $_n$8 pC (Genre et al., 19 May 2026). With a waveguide entrance field of $_n$9 and $_n$0, the local field at the dielectric accelerator is about $_n$1. Over a simulated acceleration length of $_n$2 mm, the structure supports net acceleration, and for $_n$3 input field strength the paper reports

$_n$4

over $_n$5 cm, corresponding to gradients up to $_n$6 (Genre et al., 19 May 2026).

The charge limit is also notable. The study finds negligible beam loading from $_n$7 pC to $_n$8 pC, net acceleration with minimal degradation up to about $_n$9 pC, and strong space-charge degradation at $f_{c,n}=\frac{n c}{2h},\quad n=1,2,\dots$0 pC (Genre et al., 19 May 2026). Phase slippage over the simulated interaction length is about $f_{c,n}=\frac{n c}{2h},\quad n=1,2,\dots$1, so the dominant source of energy-spread growth is the finite bunch length rather than loss of synchronism (Genre et al., 19 May 2026). Experimentally, the waveguide model was validated by electro-optic sampling of an asymmetric fabricated taper, with simulated and measured outgoing waveforms in good agreement (Genre et al., 19 May 2026).

5. Spectral shaping and inverse-design extensions

A distinct line of work uses tapering for spectral synthesis rather than for field concentration. The relevant caveat is explicit: “the paper you provided does not explicitly treat a ‘tapered parallel-plate waveguide’ geometry; all concrete calculations and simulations are for cylindrical dielectric-lined waveguides (DLWs)” (Peetermans et al., 2024). The direct results are therefore not TPPWG results. However, the same source states that “almost all of the physical ideas, design logic, and even several of the key formulas carry over directly to a tapered parallel-plate geometry with only modest modification” (Peetermans et al., 2024).

In the cylindrical work, the design variable is the local resonant frequency $f_{c,n}=\frac{n c}{2h},\quad n=1,2,\dots$2 of the dominant Cherenkov mode, and the spectrum generated by a single electron is written as

$f_{c,n}=\frac{n c}{2h},\quad n=1,2,\dots$3

with $f_{c,n}=\frac{n c}{2h},\quad n=1,2,\dots$4 the density of frequencies produced by the spatially varying structure and $f_{c,n}=\frac{n c}{2h},\quad n=1,2,\dots$5 the coupling amplitude (Peetermans et al., 2024). The inverse-design rule is then to choose the geometry so that

$f_{c,n}=\frac{n c}{2h},\quad n=1,2,\dots$6

where $f_{c,n}=\frac{n c}{2h},\quad n=1,2,\dots$7 is the target spectral envelope, and to sample the taper according to

$f_{c,n}=\frac{n c}{2h},\quad n=1,2,\dots$8

This produced Gaussian and flattop spectra up to about $f_{c,n}=\frac{n c}{2h},\quad n=1,2,\dots$9 THz in cylindrical dielectric-lined waveguides (Peetermans et al., 2024).

A plausible implication for TPPWGs is a planar inverse-design rule based on the local plate spacing $h(z)$0 and, if present, a dielectric thickness profile $h(z)$1. The same source states that for a TPPWG one would define a local plate-spacing profile $h(z)$2, compute the Cherenkov-mode dispersion $h(z)$3 and local coupling amplitude $h(z)$4, then design $h(z)$5 so that the local resonant frequency follows a chosen law such as

$h(z)$6

It further states that in a thin-dielectric-layer limit one expects a resonance condition schematically similar to

$h(z)$7

so that a monotonic, spectrally programmed TPPWG taper should be feasible in principle (Peetermans et al., 2024). Because this is an analogy rather than a demonstrated TPPWG experiment, it should be read as a transfer of design logic, not as a completed planar implementation.

6. Related architectures, theory, and limiting factors

Several adjacent PPWG research programs clarify how broad the TPPWG design space is and where its limits arise.

Theme	Key point	Paper
Transformation-optics taper	Linear, parabolic, and exponential mappings connect widths $h(z)$8 cm and $h(z)$9 cm over $y(x)=\pm a e^{b|x|}$00 cm; the exponential mapping gives the most achievable material parameters	(Tichit et al., 2010)
Metasurface PPWG	Inductive sheets support TM modes, capacitive sheets support TE modes, and reducing separation $y(x)=\pm a e^{b|x|}$01 produces strong coupling and a mixed resonance $y(x)=\pm a e^{b|x|}$02	(Ma et al., 2019)
Near-cutoff slotted PPWG	A localized TE resonance exists slightly below cutoff; 2D FEM gives $y(x)=\pm a e^{b|x|}$03 at $y(x)=\pm a e^{b|x|}$04, and $y(x)=\pm a e^{b|x|}$05 tilt reduces $y(x)=\pm a e^{b|x|}$06 by a factor of $y(x)=\pm a e^{b|x|}$07	(Henstridge et al., 2016)
Rigorous scattering theory	An exact transparent boundary condition based on an electric-to-magnetic Calderón operator yields direct well-posedness and uniqueness for an inverse obstacle problem in a uniform PPWG	(Liang et al., 19 Jul 2025)

These related results sharpen several practical points. First, “minimal reflection” and “low dispersion” are conditional statements. The accelerator TPPWG was designed so that the fundamental TEM-like mode dominates, but higher modes can exist in principle at the chosen gap and frequency (Genre et al., 19 May 2026). The single-cycle compressor preserves waveform fidelity over a broad band, yet the smallest tested gap shows measurable dispersion, and enhancement is strongly sensitive to focus placement at $y(x)=\pm a e^{b|x|}$08 (Othman et al., 2019). Near-cutoff, high-$y(x)=\pm a e^{b|x|}$09 PPWG behavior can become extremely sensitive to plate alignment, as the slotted structure demonstrates (Henstridge et al., 2016).

Second, high-field operation remains materials-limited. The accelerator study states that THz-induced damage in metals and dielectrics is not yet fully characterized, cites practical limits of a few MV/cm for metals and about $y(x)=\pm a e^{b|x|}$10–$y(x)=\pm a e^{b|x|}$11 for silica or silicon before strong conductivity or damage occur, and therefore treats input fields around $y(x)=\pm a e^{b|x|}$12 as conservative while noting that $y(x)=\pm a e^{b|x|}$13–$y(x)=\pm a e^{b|x|}$14 would be desirable (Genre et al., 19 May 2026). This constrains how aggressively a TPPWG can be tapered for field enhancement.

Third, tapering is not confined to geometric metal plates. Transformation-optics tapers replace changing plate spacing by an inhomogeneous anisotropic medium between straight plates (Tichit et al., 2010). Metasurface PPWGs replace the metal walls by penetrable impedance sheets whose local reactance determines whether TE or TM guidance exists, suggesting that a “taper” may also be realized through $y(x)=\pm a e^{b|x|}$15 rather than only through $y(x)=\pm a e^{b|x|}$16 or $y(x)=\pm a e^{b|x|}$17 (Ma et al., 2019). From this perspective, the TPPWG is less a single device than a family of guided-wave transformers in which tapering controls coupling, confinement, dispersion, and, in some implementations, the interaction between THz fields and charged-particle beams.