Md Arifuzzaman Faisal Research Articles

Cold- to Hot-Tube Dispersion Derivation: A Tutorial on Smith-Purcell Radiation Growth Rate

2026-09-26T00:00:00-04:00

Why starting current is really a growth problem

A coherent Smith–Purcell oscillator does not begin at full strength as soon as electrons pass over a grating. The electron beam first interacts with the evanescent slow wave supported by the periodic metal structure. That interaction creates a small velocity modulation. After some distance, the velocity modulation becomes density modulation, or electron bunching. When the bunching becomes strong enough, the beam can transfer coherent energy to the electromagnetic mode and the oscillation grows [1,3,4].

The minimum dc beam current required for this process to build into coherent radiation is the starting current.

A direct way to find it is to repeat particle-in-cell (PIC) simulations at many beam currents and identify the threshold at which a sustained coherent field appears. That method is powerful, but it is expensive when many grating geometries must be tested.

Our published work showed that the same threshold trend can be predicted from the spatial growth rate calculated from the hot-tube dispersion relation [1].

This article develops that connection from the beginning: fields, boundary matching, beam perturbation, hot-tube dispersion, complex roots, spatial growth, and finally the comparison with PIC.

Researcher roadmap: first solve the electromagnetic fields in each region, then match those fields at the grating boundary, then add the electron-beam response, and finally solve the resulting dispersion equation for a complex wavenumber.

What you will get from this article?

By the end, you should be able to identify what is being solved at each stage:

Field problem: write Maxwell-consistent fields in the groove and above the grating.
Boundary problem: match tangential fields to obtain the passive, or cold-tube, dispersion.
Beam problem: linearize the electron continuity and momentum equations.
Coupled problem: insert the beam response into the field matching to obtain the hot-tube dispersion.
Root problem: solve the hot-tube equation for $k=k_r+jk_i$ at a real operating frequency.
Design problem: use $-k_i$ to estimate which geometry should have the lowest starting current.

Step 1: begin with Maxwell’s equations

For the two-dimensional $x$-$y$ model, the relevant TM-like field set can be written using $H_z$, $E_x$, and $E_y$. With harmonic time dependence $e^{-i\omega t}$, Maxwell’s equations are

[\nabla\times\mathbf{E}

i\omega\mu_0\mathbf{H},]

and

[\nabla\times\mathbf{H}

\mathbf{J} - i\omega\epsilon_0\mathbf{E}.]

Before the electron beam is introduced, $\mathbf{J}=0$ in the vacuum regions. Once the beam is added, its ac current becomes part of the second equation.

The main task is to first solve the passive grating fields. We obtain the cold-tube dispersion from the two-region problem, then add the beam and construct the three-region hot-tube problem.

Step 2: write the cold-tube fields in the two regions

For the cold-tube problem, the electron beam is not yet allowed to load the electromagnetic mode. The structure therefore needs only two electromagnetic regions: Region I inside a groove and Region II in the vacuum above the grating. The dispersion relation is obtained by solving Maxwell’s equations in these two regions and matching the tangential fields at the grating surface [1,5,6].

Figure 1. Smith–Purcell grating cold-tube geometry. The groove depth is $h$, groove width is $w$, and period is $L$.

Region I: inside a groove

Let the groove occupy

[-h<y<0.]

Using the groove-mode form employed in the field-matching treatment [5], the fields can be written as

[H_{1,z}(x,y,t)

\bar{H}_{1z,0} \frac{\cosh!\left[k_0(y+h)\right]} {\sinh(k_0 h)} e^{-i\omega t},]

and

[E_{1,x}(x,y,t)

\bar{E}_{1x,0} \frac{\sinh!\left[k_0(y+h)\right]} {\cosh(k_0 h)} e^{-i\omega t}.]

The amplitude factors absorb the normalization used in the full modal solution. The important feature here is the explicit groove-depth dependence and the fact that the tangential electric field satisfies the conducting boundary at the groove bottom.

Region II: vacuum above the grating

For the cold-tube problem,

[0<y<\infty.]

Because the grating is periodic along $x$, the vacuum field is expanded in Floquet harmonics. Since this region extends to $y\rightarrow\infty$, only the decaying solution is physically allowed:

[H_{2,z}(x,y,t)

\sum_{p=-\infty}^{\infty} H_{2,p} e^{-\alpha_p y} e^{i\beta_p x} e^{-i\omega t},]

[E_{2,x}(x,y,t)

\sum_{p=-\infty}^{\infty} \left( -i\frac{\alpha_p}{\omega\epsilon_0} \right) H_{2,p} e^{-\alpha_p y} e^{i\beta_p x} e^{-i\omega t},]

and

[E_{2,y}(x,y,t)

\sum_{p=-\infty}^{\infty} \frac{\beta_p}{\epsilon_0\omega} H_{2,p} e^{-\alpha_p y} e^{i\beta_p x} e^{-i\omega t}.]

Here,

[\beta_p

k+\frac{2\pi p}{L},]

and

[\alpha_p

\sqrt{\beta_p^2-\left(\frac{\omega}{c}\right)^2}.]

For an evanescent surface harmonic we choose the branch with

[\operatorname{Re}(\alpha_p)>0,]

so the field decays away from the grating. There is no growing term $e^{+\alpha_p y}$ in the cold-tube Region II, because such a term would diverge as $y\rightarrow\infty$.

This two-region field picture is the one used to obtain the passive grating dispersion. The third field region appears only after the electron beam is introduced in the hot-tube problem.

Step 3: match the fields at the grating surface

Now the two regional solutions are connected at $y=0$.

Over the groove opening, the tangential fields are continuous:

[H_{1,z}=H_{2,z},]

and

[E_{1,x}=E_{2,x}.]

Over the metallic part of the grating surface, the tangential electric field must vanish:

[E_{2,x}=0.]

Because Region II is written as a Floquet series, these conditions are projected over one grating period. This converts the boundary-matching problem into algebraic relations among the groove amplitude and the harmonic amplitudes $H_{2,p}$ [1,5,6].

After eliminating the amplitudes, the passive or cold-tube dispersion relation is obtained.

Using the normalized quantities

[\bar{\omega}=\frac{\omega L}{c}, \qquad \bar{k}=kL, \qquad \bar{W}=\frac{w}{L}, \qquad \bar{H}=\frac{h}{L},]

define

[P_n=\bar{k}+2n\pi,]

[\theta_n=\frac{P_n\bar{W}}{2},]

and

[\gamma_n= \sqrt{P_n^2-\bar{\omega}^2}.]

The cold-tube relation can then be written [1]

[\frac{\cot(\bar{\omega}\bar{H})} {\bar{\omega}\bar{H}} - \sum_{n=-\infty}^{\infty} \left( \frac{\sin\theta_n}{\theta_n} \right)^2 \frac{\bar{W}} {\gamma_n\bar{H}} = 0.]

This equation answers the first design question:

What electromagnetic surface mode does the grating support before beam loading is included?

Our later two-layer analysis follows the same logic: define the passive field regions first, apply the interface conditions, and then add the beam-loaded regions needed for the hot-tube problem [5].

Step 4: use the beam line to select the operating frequency

The electron beam has axial velocity $v_0$. Its synchronism line is

[\omega=kv_0.]

In normalized form,

[\bar{\omega}

\beta_0\bar{k},]

where

[\beta_0=\frac{v_0}{c}.]

The intersection of this beam line with the cold-tube dispersion curve gives the evanescent-wave operating frequency used for the hot-tube calculation [1].

For the 50-keV beam used in the published study,

[\beta_0=0.4126.]

At this stage, the beam is only selecting the operating point. It has not yet modified the dispersion.

Step 5: introduce the hot-tube field regions

Figure 2. Smith–Purcell grating hot-tube geometry. The groove depth is $h$, groove width is $w$, beam position $a$, and period is $L$.

The cold-tube calculation treated the whole space above the grating as one decaying vacuum region. Once the electron beam is placed at

[y=a,]

that vacuum region must be split at the beam plane.

The hot-tube field problem therefore uses three regions:

Region I: groove, $-h<y<0$;
Region II: vacuum between the grating and beam, $0<y<a$;
Region III: vacuum above the beam, $y>a$.

Region I keeps the same groove field used in the cold-tube problem. The change occurs above the grating.

Hot-tube Region II: between the grating and the beam

Because Region II is now bounded by the grating at $y=0$ and the beam at $y=a$, both exponential solutions are needed:

[H_{2,z}(x,y,t)

\sum_{p=-\infty}^{\infty} \left( H_{2z-,p}e^{-\alpha_p y} + H_{2z+,p}e^{\alpha_p y} \right) e^{i\beta_p x} e^{-i\omega t},]

[E_{2,x}(x,y,t)

\sum_{p=-\infty}^{\infty} \left( -i\frac{\alpha_p}{\omega\epsilon_0} \right) \left( H_{2z-,p}e^{-\alpha_p y} - H_{2z+,p}e^{\alpha_p y} \right) e^{i\beta_p x} e^{-i\omega t},]

and

[E_{2,y}(x,y,t)

\sum_{p=-\infty}^{\infty} \frac{\beta_p}{\epsilon_0\omega} \left( H_{2z-,p}e^{-\alpha_p y} + H_{2z+,p}e^{\alpha_p y} \right) e^{i\beta_p x} e^{-i\omega t}.]

The $e^{+\alpha_p y}$ term is allowed here because Region II is finite: it ends at the beam plane rather than extending to infinity.

Hot-tube Region III: above the beam

For

[y>a,]

the field must again decay as $y\rightarrow\infty$. Therefore,

[H_{3,z}(x,y,t)

\sum_{p=-\infty}^{\infty} H_{3,p} e^{-\alpha_p(y-a)} e^{i\beta_p x} e^{-i\omega t},]

[E_{3,x}(x,y,t)

\sum_{p=-\infty}^{\infty} \left( -i\frac{\alpha_p}{\omega\epsilon_0} \right) H_{3,p} e^{-\alpha_p(y-a)} e^{i\beta_p x} e^{-i\omega t},]

and

[E_{3,y}(x,y,t)

\sum_{p=-\infty}^{\infty} \frac{\beta_p}{\epsilon_0\omega} H_{3,p} e^{-\alpha_p(y-a)} e^{i\beta_p x} e^{-i\omega t}.]

At the beam plane, the tangential electric field is matched across $y=a$, while the ac beam current supplies the corresponding magnetic-field discontinuity. The next steps derive that beam response from the continuity and momentum equations [1,2,5,8].

Step 6: perturb the electron beam

Now include the beam dynamically.

Let the equilibrium electron density and velocity be $n_0$ and $v_0$. Write the beam variables as

[n=n_0+n_1,]

and

[v=v_0+v_1,]

with the small-signal conditions

[\lvert n_1\rvert\ll n_0, \qquad \lvert v_1\rvert\ll v_0.]

Using $\lvert\cdot\rvert$ instead of vertical-bar characters also avoids Markdown interpreting the equation as a table.

The unperturbed beam current density is

[J_0=-e n_0v_0.]

The purpose of the hot-tube model is to find how the small ac quantities $n_1$ and $v_1$ respond to the electromagnetic field and then feed back into Maxwell’s equations. This is the standard small-signal AC space-charge framework used in electron–circuit theory [2,8].

Step 7: linearize the beam continuity equation

Charge conservation gives

[\frac{\partial n}{\partial t} + \frac{\partial(nv)}{\partial x} = 0.]

Substitute

[n=n_0+n_1, \qquad v=v_0+v_1,]

and neglect second-order terms such as $n_1v_1$. The result is

[\frac{\partial n_1}{\partial t} + v_0 \frac{\partial n_1}{\partial x} + n_0 \frac{\partial v_1}{\partial x} = 0.]

Assume the perturbations vary as

[e^{-i\omega t+ikx}.]

Then

[\frac{\partial}{\partial t}\rightarrow-i\omega, \qquad \frac{\partial}{\partial x}\rightarrow ik.]

Therefore,

[-i(\omega-kv_0)n_1 + ikn_0v_1 = 0,]

which gives

[n_1

\frac{k n_0} {\omega-kv_0} v_1.]

This already reveals the importance of synchronism. The beam response becomes strong when

[\omega\approx kv_0.]

Step 8: linearize the electron equation of motion

For the axial ac velocity, the linearized momentum equation is

[m \left( \frac{\partial}{\partial t} + v_0 \frac{\partial}{\partial x} \right) v_1 = -eE_1.]

Using the same harmonic dependence gives

[-im(\omega-kv_0)v_1

-eE_1.]

Hence,

[v_1

-\frac{i e} {m(\omega-kv_0)} E_1.]

Substitute this into the continuity result:

[n_1

-\frac{i ekn_0} {m(\omega-kv_0)^2} E_1.]

The ac current is

[J_1

-e(n_0v_1+v_0n_1).]

The important part is the resonant factor

[\frac{1} {(\omega-kv_0)^2}.]

It is the mathematical signature of strong beam response near synchronism [2,8].

Step 9: convert the beam response into the hot-tube coupling factor

The beam plasma frequency is

[\omega_p

\sqrt{ \frac{e^2n_0} {m\epsilon_0} }.]

For the sheet-beam model used in the paper, beam thickness $\tau$ enters through

[\omega_a^2

\omega_p^2\tau.]

Define

[\bar{\omega}_a

\frac{\omega_a L^{1/2}}{c}.]

The normalized beam–grating distance is

[\bar{A}

\frac{a}{L}.]

For the $n$th spatial harmonic, the beam response enters through [1]

[\Psi

\frac{ \bar{\omega}_a^2\gamma_n }{ (\bar{\omega}-P_n\beta_0)^2 }.]

The beam modifies the corresponding harmonic through

[U_n

\frac{ 1- \Psi\cosh(\gamma_n\bar{A})/e^{\gamma_n\bar{A}} }{ 1- \Psi\sinh(\gamma_n\bar{A})/e^{\gamma_n\bar{A}} }.]

This form makes two pieces of physics visible.

First, the denominator

[(\bar{\omega}-P_n\beta_0)^2]

contains beam synchronism.

Second, the factors involving $\gamma_n\bar{A}$ show why the beam must remain close to the grating: the surface field is evanescent, so coupling weakens with separation [1].

The same region-by-region field construction and beam-loading idea is developed more generally for a two-layer grating in our 2025 IEEE Transactions on Plasma Science paper [5].

Step 10: obtain the final hot-tube dispersion relation

After the beam response is inserted into the boundary-matching system, the final hot-tube relation becomes [1]

[f(\bar{k},\bar{\omega}) \equiv \frac{ \cot(\bar{\omega}\bar{H}) }{ \bar{\omega}\bar{H} } - \sum_{n=-\infty}^{\infty} U_n \left( \frac{\sin\theta_n}{\theta_n} \right)^2 \frac{\bar{W}} {\gamma_n\bar{H}} = 0.]

Compare this with the cold-tube equation. The electromagnetic structure is similar, but now each spatial harmonic carries the beam-loading factor $U_n$.

That is the central mathematical step:

Cold tube: geometry determines the passive mode.
Hot tube: geometry and beam dynamics determine the coupled mode.

How do I solve the hot-tube equation numerically?

A practical workflow is:

Solve the cold-tube dispersion first.
Use the beam-line intersection to choose a real operating frequency $\bar{\omega}$.
Select a harmonic truncation, for example $-N\le n\le N$.
For a trial complex $\bar{k}$, evaluate $P_n$, $\gamma_n$, $\theta_n$, $\Psi$, $U_n$, and then $f(\bar{k},\bar{\omega})$.
Solve the two real equations

\[\operatorname{Re}f=0, \qquad \operatorname{Im}f=0\]

for

\[\bar{k} = \bar{k}_r+i\bar{k}_i.\]

Use several initial guesses because multiple roots can exist.
Retain the physically relevant root and repeat with larger $N$ to verify convergence.
Sweep $h$, $w$, beam current, or beam height only after the root solver is stable.

A useful numerical check is to plot

\[\log_{10}\lvert f(\bar{k},\bar{\omega})\rvert\]

over the complex-$k$ plane. Local minima provide good initial guesses for a Newton, secant, or two-dimensional root solver.

Step 11: interpret the complex wavenumber

At the selected real operating frequency, solve the hot-tube equation for

[k=k_r+ik_i.]

The field varies as

[E(x,t) \propto e^{-i\omega t+ikx}.]

Substituting the complex wavenumber,

[E(x,t) \propto e^{-i\omega t} e^{ik_r x} e^{-k_i x}.]

Therefore,

$k_i>0$ gives spatial decay;
$k_i<0$ gives spatial growth.

The spatial growth rate is

[g_s=-k_i.]

For one example in the paper,

[w=60~\mu\mathrm{m}, \qquad h=40~\mu\mathrm{m},]

the normalized growing root is approximately

[\bar{k}_r=4.3289, \qquad \bar{k}_i=-0.041.]

The negative imaginary part identifies the growing branch [1].

Why the hot-tube real part shifts

The real part of the hot-tube wavenumber is slightly different from the cold-tube beam-line intersection.

This is beam loading. Once the electron beam is allowed to respond dynamically, it perturbs the electromagnetic mode and slightly detunes the real propagation constant [1,8].

The shift in $k_r$ is physically meaningful, but for starting-current optimization the most important quantity is $k_i$, because it controls spatial amplification.

Where the strongest growth occurs

The 2023 paper swept groove geometry while keeping the other parameters fixed [1].

For

[w=60~\mu\mathrm{m},]

the strongest growth occurred near

[h\approx100\text{–}120~\mu\mathrm{m}.]

For

[h=100~\mu\mathrm{m},]

the strongest growth occurred near

[w\approx80~\mu\mathrm{m}.]

These operating points lie near the upper band edge of the cold-tube dispersion curve [1].

Band-edge regions can be especially susceptible to beam-driven instability, which provides a useful physical interpretation for the growth maximum [7].

Step 12: compare the growth rate with PIC starting current

Now comes the device-level test.

The paper compared

[-\bar{k}_i]

from the hot-tube dispersion with the starting current obtained independently from PIC simulations [1,3,4].

The PIC model used a 35-period grating with the same basic beam and geometric parameters.

The result is the key finding:

When the spatial growth rate increases, the starting current decreases.

This agreement is valuable because the two calculations are fundamentally different.

The hot-tube model is a linear frequency-domain calculation.

PIC follows the nonlinear particle and field evolution in time.

Yet both show nearly the same dependence on groove geometry [1].

Why the correlation is exponential

For coherent radiation, the power scales approximately as

[P \propto E^2 \propto N^2 \propto I^2,]

where $N$ is the number of coherently bunched electrons and $I$ is beam current [1,3].

Therefore, at the scaling level,

[E\propto I.]

A spatially growing mode is amplified over interaction length $L_G$ as

[E \propto e^{\lvert k_i\rvert L_G}.]

Combining the two relations gives

[e^{\lvert k_i\rvert L_G} \propto I.]

This explains why the spatial-growth curve and the logarithm of the starting-current curve follow one another so closely [1].

The relation is a scaling argument, not a full nonlinear saturation model. Its value is that it explains why a linear hot-tube calculation can predict a nonlinear threshold trend.

Why this is useful for a new design

Suppose you want to optimize a new Smith–Purcell structure.

A brute-force approach would be:

choose a geometry;
build the PIC model;
sweep beam current;
find the starting current;
repeat for every geometry.

That quickly becomes expensive.

A dispersion-guided approach is more efficient:

derive or compute the cold-tube dispersion;
locate the beam-synchronous operating frequency;
add the beam equations;
solve the hot-tube relation;
extract the growing root;
map $-\operatorname{Im}(k)$ over geometry;
reserve PIC for the best candidates.

Our 2025 two-layer work uses the same basic philosophy for a more complicated structure: the fields are written in multiple regions, matched at all interfaces, and then used to determine how the additional layer changes the cold operating point and hot spatial growth [5].

This is why learning the derivation is useful beyond one grating geometry. The field expansion → boundary matching → beam response → complex root workflow can be reused for many linear free-electron slow-wave devices.

The physical picture in one sentence

The full derivation can be summarized as

geometry defines the fields → boundary matching defines the passive mode → the electron beam loads that mode → the complex hot-tube root gives spatial growth → spatial growth predicts the beam-current threshold.

That is the practical meaning of the imaginary wavenumber. It connects electromagnetic field theory directly to an experimentally relevant device quantity: the current required to start coherent radiation.

References

M. A. Faisal and P. Zhang, “Grating Optimization for Smith–Purcell Radiation: Direct Correlation Between Spatial Growth Rate and Starting Current,” IEEE Transactions on Electron Devices, vol. 70, no. 6, pp. 2860–2863, 2023. DOI: 10.1109/TED.2022.3208846 ↗
Y. Y. Lau and D. Chernin, “A Review of the AC Space-Charge Effect in Electron–Circuit Interactions,” Physics of Fluids B: Plasma Physics, vol. 4, no. 11, pp. 3473–3497, 1992. DOI: 10.1063/1.860356 ↗
P. Zhang, L. K. Ang, and A. Gover, “Enhancement of Coherent Smith–Purcell Radiation at Terahertz Frequency by Optimized Grating, Prebunched Beams, and Open Cavity,” Physical Review Special Topics – Accelerators and Beams, vol. 18, 020702, 2015. DOI: 10.1103/PhysRevSTAB.18.020702 ↗
D. Li et al., “Growth Rate and Start Current in Smith–Purcell Free-Electron Lasers,” Applied Physics Letters, vol. 100, 191101, 2012. DOI: 10.1063/1.4711803 ↗
M. A. Faisal and P. Zhang, “Parametric Analysis on Enhancement of THz Smith–Purcell Radiation by Two-Layer Grating Structure,” IEEE Transactions on Plasma Science, vol. 53, no. 6, pp. 1170–1185, 2025. DOI: 10.1109/TPS.2025.3567163 ↗
H. P. Freund and T. M. Abu-Elfadl, “Linearized Field Theory of a Smith–Purcell Traveling Wave Tube,” IEEE Transactions on Plasma Science, vol. 32, no. 3, pp. 1015–1027, 2004. DOI: 10.1109/TPS.2004.827612 ↗
D. M. H. Hung et al., “Absolute Instability Near the Band Edge of Traveling-Wave Amplifiers,” Physical Review Letters, vol. 115, 124801, 2015. DOI: 10.1103/PhysRevLett.115.124801 ↗
P. Wong, P. Zhang, and J. Luginsland, “Recent Theory of Traveling-Wave Tubes: A Tutorial-Review,” Plasma Research Express, vol. 2, 023001, 2020. DOI: 10.1088/2516-1067/ab9730 ↗

License. Unless otherwise noted, the text and original illustrations in this tutorial are licensed under Creative Commons Attribution 4.0 International (CC BY 4.0) ↗. Third-party material and cited publications remain subject to their respective copyright and license terms.

How Grating Geometry Sets Frequency and Starting Current in Smith–Purcell Radiation

2026-09-24T00:00:00-04:00

Why grating geometry matters

Smith–Purcell radiation (SPR) is produced when an electron beam moves close to a periodic metallic grating. The basic picture is simple: the moving electrons excite electromagnetic fields near the grating, and part of that interaction appears as radiation at a wavelength set by the beam velocity, the grating period, the diffraction order, and the observation angle [1,2].

The familiar Smith–Purcell relation is

[\lambda

\frac{L}{n} \left( \frac{1}{\beta} - \cos\theta \right),]

where $L$ is the grating period, $n$ is the diffraction order, $\beta=v/c$ is the normalized beam velocity, and $\theta$ is the observation angle.

This equation is important, but it does not contain the groove height $h$ or groove width $w$. That can give the impression that these dimensions are secondary. In a coherent Smith–Purcell oscillator, they are not. The groove geometry changes the electromagnetic slow-wave mode supported by the grating, and this changes both the operating frequency and the strength of the beam–wave interaction [1,3,4].

In our published study, the grating period and electron-beam properties were kept fixed while $h$ and $w$ were varied. The starting current changed by orders of magnitude [1]. The central question was therefore not simply “what wavelength is radiated?” but rather:

How does the detailed grating geometry change the mode that the beam interacts with?

The answer begins with the cold-tube dispersion relation.

The cold-tube problem: first find the passive grating mode

“Cold tube” means that the electron beam is not yet included in the electromagnetic dispersion calculation. We first ask what surface mode the periodic metallic grating supports by itself.

The analysis uses normalized quantities

[\bar{\omega}=\frac{\omega L}{c}, \qquad \bar{k}=kL, \qquad \bar{W}=\frac{w}{L}, \qquad \bar{H}=\frac{h}{L}.]

For the rectangular grating model used in the paper, the cold-tube surface-mode relation is [1]

[\frac{\cot!\left(\bar{\omega}\bar{H}\right)} {\bar{\omega}\bar{H}} - \sum_{n=-\infty}^{\infty} \left( \frac{\sin\theta_n}{\theta_n} \right)^2 \frac{\bar{W}} {\gamma_n\bar{H}} = 0,]

with

[\theta_n=\frac{P_n\bar{W}}{2}, \qquad P_n=\bar{k}+2n\pi,]

and

[\gamma_n= \sqrt{P_n^2-\bar{\omega}^2}.]

The physical idea behind this equation is easier than the expression itself. The field inside each groove is matched to the Floquet harmonics above the periodic grating. The groove width, groove height, and period therefore enter directly into the allowed surface-wave dispersion [1,5].

The paper used the dominant TEM-like groove mode for this calculation. A sufficiently large number of spatial harmonics was retained numerically to ensure convergence [1].

The beam line selects the operating point

After the cold dispersion curve is known, the electron beam is introduced only as a synchronism line. For a beam with velocity $v_0$,

[\omega = kv_0.]

In normalized form,

[\bar{\omega}

\beta_0\bar{k},]

where

[\beta_0=\frac{v_0}{c}.]

For the 50-keV beam used in the paper,

[\beta_0=0.4126.]

The intersection of this beam line with the cold-tube dispersion curve gives the evanescent surface-mode operating frequency $f_{\mathrm{ev}}$ [1].

This is the first key design result: even with fixed beam energy and fixed grating period, changing $h$ or $w$ moves the intersection point. Therefore, groove geometry changes the operating frequency.

The coherent SPR of interest in the study occurs at the corresponding harmonic of the evanescent mode [1].

Why deeper or wider grooves lower the operating frequency

The numerical sweeps in the paper show a clear trend. As either the groove height $h$ or groove width $w$ increases, the operating frequency generally decreases, while the corresponding evanescent wavelength increases [1].

The reason is visible directly in the cold-tube dispersion curve. Increasing the groove dimensions changes the upper band edge. The band edge becomes lower and more rounded, so the same electron beam line intersects the curve at a smaller frequency.

That gives a practical tuning mechanism. Two gratings can have the same period $L$ and use the same beam energy, yet operate at different frequencies because the groove dimensions are different.

This is why the basic Smith–Purcell wavelength equation is not a complete device-design rule. It captures the radiation kinematics, while the cold-tube dispersion captures the actual slow-wave mode supported by the structure [1,3].

Geometry can even change the wave direction

The slope of the dispersion curve gives the group velocity,

[v_g

\frac{d\omega}{dk}.]

For many of the geometries in the paper, the slope at the operating point is negative, which is characteristic of backward-wave behavior.

However, for deeper grooves around

[h=120\text{–}200~\mu\mathrm{m},]

the slope becomes positive in the reported sweep [1].

That means a purely geometric modification can move the beam–grating interaction from a backward-wave-like region to a forward-wave-like region, even though the beam energy and grating period remain unchanged.

This is a strong reminder that $h$ and $w$ do not make only small corrections. They can change the qualitative electromagnetic character of the interaction.

The same geometry also changes spatial growth

Cold-tube dispersion tells us where the beam and grating can interact. It does not yet tell us how strongly the coupled wave will grow.

That requires the hot-tube model, where the beam dynamics are included self-consistently. The hot solution gives a complex wavenumber

[k=k_r+jk_i.]

For the field convention used in the paper,

[E(x,t)\propto e^{-j\omega t+jkx},]

[E(x,t) \propto e^{-j\omega t+jk_r x} e^{-k_i x}.]

If $k_i<0$, the field grows with distance. The spatial growth rate is therefore

[g_s=-k_i.]

The important result is that the growth rate also depends strongly on $h$ and $w$ [1].

For the parameter sweeps reported in the paper, strong growth occurs near groove heights around

[h\approx100\text{–}120~\mu\mathrm{m},]

for the fixed-width sweep, and near

[w\approx80~\mu\mathrm{m},]

for the fixed-height sweep [1].

These high-growth points occur when the operating point approaches the upper band edge of the cold dispersion curve. Band-edge regions are known to be particularly sensitive to beam-driven instability [6].

Why this matters for starting current

The starting current is the minimum dc beam current required for coherent oscillation to build. Below this current, the beam may still interact with the grating, but the modulation is not strong enough to produce sustained electron bunching and coherent radiation [1,4].

The paper compared the calculated spatial growth rate with starting currents obtained independently from particle-in-cell simulations using a 35-period grating [1,3].

The result is very clear: the geometry dependence of the starting current closely follows the geometry dependence of the spatial growth rate.

Geometries with larger $-k_i$ require lower starting current.

This means that groove optimization can be organized as a two-step physics problem:

use the cold-tube dispersion to find the operating frequency;
use the hot-tube growth rate to find the geometry that reaches coherent oscillation most easily.

This is much more efficient than using a full PIC starting-current search at every point in a large geometry sweep.

A practical optimization workflow

A useful design workflow is therefore:

1. Choose the beam energy and grating period

These set the basic beam line and periodic scale.

2. Sweep groove height and width

For each pair $(h,w)$, calculate the cold-tube dispersion.

3. Find the beam–mode intersection

Use

[\bar{\omega}=\beta_0\bar{k}]

to determine $f_{\mathrm{ev}}$.

4. Check the desired radiation frequency

Reject geometries whose operating point is outside the target range.

5. Solve the hot-tube dispersion

Obtain

[k=k_r+jk_i.]

6. Rank the remaining geometries by $-k_i$

Larger $-k_i$ means stronger spatial amplification.

7. Use PIC only for final verification

PIC is then applied to a small set of promising designs rather than the entire parameter space.

This strategy keeps the physical meaning visible at every step. Geometry first changes the passive mode, and the changed mode then modifies the beam-driven growth.

The main design lesson

The most important lesson is simple:

A Smith–Purcell grating is not defined only by its period.

The groove width and height reshape the slow-wave dispersion, move the operating frequency, change the group velocity, change the spatial growth rate, and ultimately change the starting current.

For THz device design, the useful geometry is therefore not simply the one that satisfies the Smith–Purcell wavelength relation. Similar slow-wave-structure considerations and three-dimensional beam requirements are also important when the design moves toward a practical device [7,8]. It is the geometry that places the desired frequency at a strong beam–wave operating point with sufficiently large spatial gain.

That is why cold-tube and hot-tube dispersion should be used together [1]. The first tells us where the device operates. The second tells us how strongly it can grow.

References

M. A. Faisal and P. Zhang, “Grating Optimization for Smith–Purcell Radiation: Direct Correlation Between Spatial Growth Rate and Starting Current,” IEEE Transactions on Electron Devices, vol. 70, pp. 2860–2863, 2023. DOI: 10.1109/TED.2022.3208846.
S. J. Smith and E. M. Purcell, “Visible Light From Localized Surface Charges Moving Across a Grating,” Physical Review, vol. 92, p. 1069, 1953. DOI: 10.1103/PhysRev.92.1069.
P. Zhang, L. K. Ang, and A. Gover, “Enhancement of Coherent Smith–Purcell Radiation at Terahertz Frequency by Optimized Grating, Prebunched Beams, and Open Cavity,” Physical Review Special Topics – Accelerators and Beams, vol. 18, 020702, 2015. DOI: 10.1103/PhysRevSTAB.18.020702.
V. Kumar and K.-J. Kim, “Analysis of Smith–Purcell Free-Electron Lasers,” Physical Review E, vol. 73, 026501, 2006. DOI: 10.1103/PhysRevE.73.026501.
H. P. Freund and T. M. Abu-Elfadl, “Linearized Field Theory of a Smith–Purcell Traveling Wave Tube,” IEEE Transactions on Plasma Science, vol. 32, pp. 1015–1027, 2004. DOI: 10.1109/TPS.2004.827612.
D. M. H. Hung et al., “Absolute Instability Near the Band Edge of Traveling-Wave Amplifiers,” Physical Review Letters, vol. 115, 124801, 2015. DOI: 10.1103/PhysRevLett.115.124801.
S. C. Yurt et al., “Similarity of Properties of Metamaterial Slow-Wave Structures and Metallic Periodic Structures,” IEEE Transactions on Plasma Science, vol. 44, pp. 1280–1286, 2016. DOI: 10.1109/TPS.2016.2535305.
K.-J. Kim and V. Kumar, “Electron Beam Requirements for a Three-Dimensional Smith–Purcell Backward-Wave Oscillator for Intense Terahertz Radiation,” Physical Review Special Topics – Accelerators and Beams, vol. 10, 080702, 2007. DOI: 10.1103/PhysRevSTAB.10.080702.

When Does a Second Grating Layer Improve THz Smith–Purcell Radiation?

2026-09-24T00:00:00-04:00

A second layer can help — but only when the physics says it should

A two-layer grating looks like an obvious upgrade to a conventional Smith–Purcell structure: place a second periodic metallic grating on the other side of the electron beam and create a stronger electromagnetic interaction region.

The important result of our 2025 IEEE Transactions on Plasma Science study is that the story is not that simple.

A two-layer grating can strongly enhance coherent Smith–Purcell radiation, increase the spatial growth rate, and reduce the starting current. But the enhancement is not universal. Whether the second layer helps depends on the grating geometry, the electron-beam energy, the operating point on the dispersion curve, and the resulting spatial growth rate.

That leads to a much more useful design question:

For a given beam energy and grating geometry, should the device use one grating layer or two?

The paper develops a physics-based way to answer that question.

Figure 1. Original article graphic comparing the two configurations. In the two-layer structure, the electron beam travels between a lower and an upper periodic grating. The relevant geometric controls include groove width $w$, grating period $L$, layer heights $h_1$ and $h_2$, and beam distances $a_1$ and $a_2$.

Start with the Smith–Purcell interaction

When an electron beam travels close to a periodic grating, it can excite an evanescent surface wave. The familiar Smith–Purcell wavelength relation is

[\lambda

\frac{L}{n} \left( \frac{1}{\beta} - \cos\theta \right),]

where $L$ is the grating period, $n$ is the diffraction order, $\beta=v/c$, and $\theta$ is the radiation angle.

For coherent backward-wave-oscillator-type Smith–Purcell radiation, however, knowing the radiation wavelength is only the beginning. The practical device must also develop enough beam bunching and wave amplification for coherent oscillation to start.

That introduces the starting current: the threshold beam current required to excite sustained coherent radiation.

Reducing this current is valuable because a lower beam current can reduce beam loss, reduce heating of beam-carrying components, and help extend cathode lifetime.

What does the second grating layer actually change?

The second layer changes the electromagnetic boundary conditions seen by the surface mode.

In a single-layer structure, the beam couples to the slow evanescent wave associated with one periodic grating. In the two-layer configuration, the beam is placed between two periodic metallic surfaces. The additional layer changes the allowed electromagnetic modes and therefore changes the grating dispersion.

The consequence is important:

[\text{grating geometry} \rightarrow \text{dispersion} \rightarrow \text{beam–wave coupling} \rightarrow \text{spatial growth} \rightarrow \text{starting current}.]

The second layer is useful only when that chain moves the operating point into a region with stronger beam-driven growth.

Step 1 — cold-tube dispersion finds the operating point

The cold-tube problem is solved without dynamic beam loading. It describes the wave propagation supported by the grating structure itself.

For the two-layer grating, the cold-tube model is obtained by solving Maxwell’s equations in the different field regions and enforcing continuity of the tangential field components at the grating interfaces.

The main geometric variables are

[w,\qquad h_1,\qquad h_2,\qquad a_1,\qquad a_2,\qquad L.]

Changing these parameters shifts the cold-tube dispersion curve.

The electron beam introduces the beam line

[\omega = k v_0.]

Its intersection with the cold-tube dispersion determines the evanescent-wave operating frequency.

For the 50-keV example with

[w=60~\mu\mathrm{m}, \qquad h_1=h_2=40~\mu\mathrm{m},]

the calculated evanescent-wave frequencies were approximately

[f_{\mathrm{ev,1}}=0.717~\mathrm{THz}]

for the single-layer grating and

[f_{\mathrm{ev,2}}=0.68~\mathrm{THz}]

for the two-layer grating.

Because the radiating Smith–Purcell wave occurs at the second harmonic in this configuration, the corresponding theoretical radiation frequencies are about

[1.43~\mathrm{THz} \quad\text{and}\quad 1.36~\mathrm{THz}.]

The PIC simulations gave approximately $1.39$ and $1.32~\mathrm{THz}$, respectively. The small shift is consistent with beam loading, space charge, and nonlinear effects that are present in PIC but absent from the passive cold-tube calculation.

This agreement is useful because it shows that the simpler dispersion calculation can locate the operating frequency before a full PIC study is performed.

Step 2 — hot-tube dispersion tells us whether the wave grows

The hot-tube model includes the electron beam.

The field equations are coupled to the beam continuity equation and equation of motion. At a real operating frequency from the cold-tube solution, the hot-tube equation is solved for a complex wavenumber,

[k = k_r + j k_i.]

With the field convention used in the paper,

[E(x,t) \propto e^{-j\omega t+jkx} = e^{-j\omega t} e^{j k_r x} e^{-k_i x}.]

Therefore,

[k_i<0]

corresponds to spatial growth, and the magnitude

[-k_i]

is the spatial growth rate.

This is the central design quantity.

A larger spatial growth rate means that the beam-driven mode amplifies more strongly along the interaction length.

Step 3 — spatial growth predicts the starting-current trend

The paper compares the hot-tube spatial growth rate with starting currents obtained independently from PIC simulations.

For the two-layer structure, the same correlation previously identified for single-layer gratings remains clear:

The grating parameters that produce a larger spatial growth rate also produce a lower starting current.

For the 50-keV optimization study, the second-layer height was found to be favorable near

[h_2=40~\mu\mathrm{m}.]

With that value fixed, strong growth was obtained when the groove width was approximately

[w=60\text{–}80~\mu\mathrm{m}]

and the first-layer height was close to

[h_1=100~\mu\mathrm{m}.]

For the reported optimized case,

[w=60~\mu\mathrm{m}, \qquad h_1=100~\mu\mathrm{m}, \qquad h_2=40~\mu\mathrm{m},]

the PIC-calculated starting current reached approximately

[I_s=12.5~\mathrm{A/m}.]

The same geometry also corresponded to the strongest spatial-growth region in the theoretical calculation.

A dramatic example — but it should not be generalized blindly

One of the clearest PIC examples used

[w=60~\mu\mathrm{m}, \qquad h_1=h_2=40~\mu\mathrm{m}.]

At a beam current of $2000~\mathrm{A/m}$, introducing the second grating layer produced approximately a fourfold increase in radiation intensity compared with the single-layer structure.

The two-layer structure could also operate at only $500~\mathrm{A/m}$ while producing a radiation level comparable to the single-layer structure operating at $2000~\mathrm{A/m}$.

Figure 2. Article visualization of the approximate comparison reported in the paper for $w=60~\mu\mathrm{m}$ and $h_1=h_2=40~\mu\mathrm{m}$. The values are normalized to the single-layer, $2000~\mathrm{A/m}$ case.

This example shows why the two-layer concept is attractive.

But the more important result of the paper is that the same improvement does not occur everywhere in parameter space.

The most useful design rule: compare the growth rates

Let

[

k_i

_{\mathrm{two}}]

and

[

k_i

_{\mathrm{single}}]

represent the spatial-growth magnitudes for otherwise comparable two-layer and single-layer designs.

The paper shows the following selection rule:

[\frac{|k_i|{\mathrm{two}}} {|k_i|{\mathrm{single}}} >1 \quad \Longrightarrow \quad \text{two-layer structure tends to require lower starting current},]

while

[\frac{|k_i|{\mathrm{two}}} {|k_i|{\mathrm{single}}} <1 \quad \Longrightarrow \quad \text{single-layer structure tends to require lower starting current}.]

The PIC starting-current calculations followed this prediction across the tested comparison points.

This is much more useful than assuming that adding another grating automatically improves performance.

Figure 3. A practical design workflow based on the paper. First calculate the cold-tube operating point, then solve the hot-tube complex wavenumber, compare spatial growth, and choose the structure that gives the stronger growing branch for the intended operating condition.

Why the upper band edge matters

The calculations show that the strongest spatial growth generally occurs when the beam-synchronous operating point approaches the upper band edge,

[\bar{k}\approx\pi.]

This region is particularly susceptible to beam-driven instability.

For some grating dimensions, the two-layer structure shifts the operating point closer to this high-growth region and therefore reduces the starting current.

For other dimensions, the second layer shifts the operating point away from the favorable band-edge region. In those cases, the single-layer configuration can have a higher growth rate and a lower starting current.

That explains why geometry optimization cannot be separated from the choice of one or two layers.

The second layer shifts the useful beam-energy range

A particularly important result appears when the beam energy is varied.

Across the parametric study, the spatial-growth bands of the two-layer structures shift toward lower beam energy compared with the corresponding single-layer structures.

This means that, for many geometries, the two-layer configuration can produce stronger spatial growth at lower beam energies.

At higher beam energy, the situation can reverse and the single-layer structure can have the larger growth rate.

This gives the second layer a practical role:

It provides an additional degree of freedom for moving a strong-growth operating region toward the beam energy available in the device.

The second layer is therefore not simply an enhancement layer. It is a dispersion-engineering tool.

What happens when the upper grating is moved away?

The distance between the beam and the second grating also matters.

As the upper layer is moved far from the electron beam, its electromagnetic influence becomes weak. The calculated two-layer cold-tube dispersion then approaches the single-layer dispersion, and the corresponding spatial-growth bands also begin to overlap.

Physically, this is exactly what should happen:

[a_2\rightarrow\text{large} \quad\Rightarrow\quad \text{second layer becomes electromagnetically irrelevant}.]

This limiting behavior provides another useful consistency check for the two-layer theory.

Material conductivity matters too

The idealized dispersion theory is useful for identifying the underlying physics, but a real THz grating is not a perfect conductor.

The paper therefore also used CST PIC simulations to examine the effect of finite grating conductivity.

For both single-layer and two-layer structures, increasing conductivity increased the output power and reduced the starting current.

The physical reason is straightforward: lower-conductivity materials dissipate more electromagnetic energy as Joule heating, leaving less field energy available for beam–wave interaction.

For the tested geometry, the two-layer structure retained a higher output power and lower starting current because it was operating in a parameter region with the stronger spatial-growth rate.

This shows that electromagnetic loss and dispersion optimization have to be considered together in a practical THz source.

A useful workflow for future grating designs

The study suggests a computationally efficient strategy for designing Smith–Purcell structures:

Choose candidate values of $w$, $h_1$, $h_2$, $a_1$, $a_2$, and $L$.
Solve the cold-tube dispersion.
Find the beam-synchronous operating frequency.
Solve the hot-tube dispersion for the growing complex root.
Map $-k_i$ over the geometry and beam-energy range.
Compare the single-layer and two-layer growth rates.
Use PIC simulations only for the most promising candidates.
Include realistic material conductivity when moving toward a practical device.

This approach is considerably more informative than sweeping starting current by brute-force PIC simulation for every possible geometry.

The broader lesson

The most important idea is not simply that a two-layer grating can produce more radiation.

It is that the number of grating layers changes the dispersion landscape, and the device performance follows from where the beam interacts with that landscape.

For one operating point, the second layer can place the beam near a high-growth branch and dramatically reduce the required current.

For another operating point, the same additional layer can move the system away from the favorable region.

So the design principle is:

Do not choose the number of grating layers first. Choose the operating point first, calculate the spatial growth, and let the beam–wave physics determine the structure.

That principle can be useful not only for Smith–Purcell sources but also for more complicated periodic slow-wave structures used in free-electron radiation devices.

References

M. A. Faisal and P. Zhang, “Parametric Analysis on Enhancement of THz Smith–Purcell Radiation by Two-Layer Grating Structure,” IEEE Transactions on Plasma Science, vol. 53, no. 6, pp. 1170–1185, 2025. DOI: 10.1109/TPS.2025.3567163 ↗
S. J. Smith and E. M. Purcell, “Visible Light From Localized Surface Charges Moving Across a Grating,” Physical Review, vol. 92, no. 4, p. 1069, 1953. DOI: 10.1103/PhysRev.92.1069 ↗
M. A. Faisal and P. Zhang, “Grating Optimization for Smith–Purcell Radiation: Direct Correlation Between Spatial Growth Rate and Starting Current,” IEEE Transactions on Electron Devices, vol. 70, no. 6, pp. 2860–2863, 2023. DOI: 10.1109/TED.2022.3208846 ↗
P. Zhang, L. K. Ang, and A. Gover, “Enhancement of Coherent Smith–Purcell Radiation at Terahertz Frequency by Optimized Grating, Prebunched Beams, and Open Cavity,” Physical Review Special Topics – Accelerators and Beams, vol. 18, 020702, 2015. DOI: 10.1103/PhysRevSTAB.18.020702 ↗
D. Li et al., “Growth Rate and Start Current in Smith–Purcell Free-Electron Lasers,” Applied Physics Letters, vol. 100, 191101, 2012. DOI: 10.1063/1.4711803 ↗
J. P. Verboncoeur, A. B. Langdon, and N. T. Gladd, “An Object-Oriented Electromagnetic PIC Code,” Computer Physics Communications, vol. 87, pp. 199–211, 1995. DOI: 10.1016/0010-4655(94)00173-Y ↗
Y. Y. Lau and D. Chernin, “A Review of the AC Space-Charge Effect in Electron–Circuit Interactions,” Physics of Fluids B, vol. 4, no. 11, pp. 3473–3497, 1992. DOI: 10.1063/1.860356 ↗
D. M. H. Hung et al., “Absolute Instability Near the Band Edge of Traveling-Wave Amplifiers,” Physical Review Letters, vol. 115, 124801, 2015. DOI: 10.1103/PhysRevLett.115.124801 ↗

Why Anisotropic Conductivity Changes Current Crowding in 2D Thin Films

2026-09-24T00:00:00-04:00

Why current refuses to travel straight down

A vertical metal contact seems geometrically simple: current enters through the electrode and should flow downward into the semiconductor. That intuition works reasonably well for isotropic bulk materials. It becomes misleading in layered two-dimensional materials, where electrical conductivity can differ by orders of magnitude between the crystal plane and the out-of-plane direction.

In materials such as graphite, MoS$_2$, and WSe$_2$, carriers move far more easily along the atomic layers than across them. The underlying crystal structure explains why. Strong in-plane bonding supports efficient lateral transport, while the weak van der Waals coupling between layers makes vertical motion comparatively difficult.

When a metal electrode is placed on such a material, the injected current therefore takes a path of lower resistance: it spreads laterally beneath the contact before penetrating downward.

That redistribution is the origin of strong current crowding near contact edges.

Anisotropy changes the field itself

The electrostatic problem is governed by the conductivity tensor,

$ \nabla\cdot\left(\boldsymbol{\sigma}\nabla\Phi\right)=0. $

For a layered medium, the in-plane and out-of-plane conductivities are different. Equivalently, the resistivity tensor contains distinct lateral and vertical components.

The paper defines anisotropy ratios such as

[\alpha = \frac{\rho_{\parallel}}{\rho_{\perp}},]

so a small $\alpha$ corresponds to much easier in-plane conduction.

As $\alpha$ decreases, the potential distribution stretches laterally. Rather than dropping rapidly through the thickness, the potential remains correlated over a longer distance along the plane. The current-density pattern follows that field structure.

The consequence is subtle: stronger in-plane conductivity can reduce the overall spreading resistance while simultaneously increasing the localization of current near the contact perimeter.

Those are not contradictory results. One describes the total voltage drop; the other describes where the current is concentrated.

Why edge crowding matters

Current crowding is not only an electrical-resistance problem.

A highly localized current density produces localized Joule heating. In real contacts, this can accelerate electromigration, interfacial degradation, and thermally activated failure. A contact can therefore have an acceptable terminal resistance while still containing microscopic regions that carry much larger current density than the average value suggests.

For atomically thin and layered materials, this distinction becomes especially important because the geometric dimensions are already comparable to the characteristic current-spreading length scales.

A model that captures only terminal resistance may miss the spatial physics that controls reliability.

The top and bottom layers both matter

The exact framework in the paper treats both members of the vertical contact as anisotropic materials. Each layer has independent in-plane and out-of-plane resistivity, and the interface can have a finite specific resistivity $\rho_i$.

This is more general than treating a metal electrode as perfectly equipotential or replacing the semiconductor by a single effective resistivity.

The calculations show that for a fixed out-of-plane resistivity, lowering the in-plane resistivity increases lateral current spreading and sharpens the edge crowding. At the same time, the dominant lever for reducing total spreading resistance in a vertical contact is often the out-of-plane resistivity of the lower layer rather than pushing the in-plane resistivity lower and lower.

That distinction has an important design implication: a material can be extremely conductive laterally and still remain difficult to contact vertically.

Why isotropic intuition can fail

A common shortcut is to replace an anisotropic material with an effective isotropic resistivity. That may reproduce one limiting resistance value, but it cannot reproduce the actual field distribution in a finite device.

The paper shows that isotropic models fail particularly when lateral boundaries, thin-film thickness and strong anisotropy all matter at once.

This is why the full tensor treatment is useful. It predicts not only the total spreading resistance but also how the potential and current redistribute throughout the layered structure.

From contact optimization to reliability

The contact-design message is therefore broader than “reduce contact resistance.”

A robust vertical contact should minimize the total resistance and avoid unnecessarily severe current localization. Lowering interfacial resistivity remains important, but material anisotropy and geometry determine how the injected current reaches the channel after crossing the interface.

For highly anisotropic 2D materials, the current path is inherently nonlocal. The electrode edge, film thickness, lateral extent, and vertical resistivity all participate in the final electrical behavior.

That is why current crowding in layered contacts is fundamentally a field-distribution problem, not just a lumped-resistance problem.

PASCHEN-1D: An Open 1D Plasma Discharge Solver for Research

2026-09-22T00:00:00-04:00

Project credit. PASCHEN-1D was led by Dr. Asif Iqbal, with Prof. Peng Zhang as corresponding author. The other coauthors are Yves Heri, Bingqing Wang, Lan Jin, and Md Arifuzzaman Faisal. I am happy to have been part of this project and to have contributed as a coauthor.

What is PASCHEN-1D?

PASCHEN-1D is a one-dimensional plasma-discharge simulation code for researchers studying gas breakdown, discharge evolution, electrode emission, and plasma–circuit interaction.

The code combines four important pieces of physics:

[\text{charged-particle transport} + \text{electric field} + \text{surface emission} + \text{external circuit}.]

Instead of building these models separately, PASCHEN-1D solves them together in one framework.

This makes the software useful for researchers who need a physics-based 1D discharge model but do not want to build a complete solver from the beginning.

Official PASCHEN-1D resources

The official project information is provided by the Plasma, Beams, and Interface Science Group at the University of Michigan:

Official PASCHEN-1D page — University of Michigan PBI Group ↗

The source code, example configurations, notebooks, documentation, and releases are available on GitHub:

PASCHEN-1D source code on GitHub ↗

If you are planning to use PASCHEN-1D in research, I recommend starting with these two links.

What can PASCHEN-1D simulate?

PASCHEN-1D is designed for plasma problems where the dominant variation can be represented along one spatial direction, for example between two planar electrodes.

It can be used to study:

gas breakdown;
DC plasma discharges;
pulsed discharges;
Townsend and glow-discharge behavior;
photoemission-driven discharges;
secondary-electron-emission effects;
field-emission and thermionic-emission effects;
plasma–surface interaction;
plasma connected to external electrical circuits;
effects of dielectric layers near electrodes;
parameter scans over voltage, pressure, gap length, emission parameters, transport models, or circuit components.

The code is especially useful when many cases must be simulated and a full 2D or 3D model would be unnecessarily expensive.

What physics does the solver use?

The main plasma model is a one-dimensional drift–diffusion–Poisson system.

The charged-particle densities evolve through continuity equations of the form

[\frac{\partial n_s}{\partial t} + \frac{\partial \Gamma_s}{\partial x} = S_s,]

where $n_s$ is the density of species $s$, $\Gamma_s$ is its flux, and $S_s$ represents sources or losses.

The electric potential is obtained self-consistently from Poisson’s equation,

[\frac{d^2\phi}{dx^2}

-\frac{\rho}{\epsilon_0},]

and the electric field follows from

[E=-\frac{d\phi}{dx}.]

So the plasma changes the electric field, and the electric field changes the motion and production of charged particles.

That feedback is essential in breakdown and discharge simulations.

Surface emission is included

Electrode emission can strongly influence whether a discharge starts and how it evolves.

PASCHEN-1D includes several surface-emission options, including:

secondary electron emission;
field emission;
thermionic / Richardson–Dushman emission;
prescribed current-density emission;
pulsed photoemission models.

This is useful for researchers studying systems where the electrode is an active part of the plasma physics.

For example, a laser-driven discharge can require photoemission, while a high-field microgap may require field emission.

Plasma and external circuits can interact

In a real experiment, the voltage across a plasma is not always identical to the voltage produced by the source.

The plasma current interacts with resistors, capacitors, inductors, dielectric layers, and other circuit elements.

PASCHEN-1D includes configurable circuit coupling so that

[\text{plasma dynamics} \longleftrightarrow \text{circuit dynamics}.]

The GitHub version includes explicit, implicit, and modified-nodal-analysis-based circuit options, together with configurable lumped-circuit topologies.

This makes PASCHEN-1D useful for problems where the electrical network is part of the discharge physics rather than just an imposed boundary condition.

What results can you obtain?

A PASCHEN-1D simulation can provide quantities such as:

electron density $n_e(x,t)$;
ion density $n_i(x,t)$;
electric potential $\phi(x,t)$;
electric field $E(x,t)$;
discharge current;
applied and gap voltages;
electron and ion fluxes;
ionization quantities;
electron and ion transport coefficients;
time histories of plasma and circuit quantities;
spatial profiles at selected times;
time- or cycle-averaged profiles;
derived sheath diagnostics.

The repository provides separate Jupyter notebooks for temporal plots, spatial snapshots, and averaged quantities.

Transport data and local-field models

PASCHEN-1D can use user-defined transport and ionization models or table-based local-field data.

The current code supports local-field electron kinetics using transport and ionization tables, including BOLSIG+-style swarm-output data.

Ion transport can also be configured using tabulated mobility and diffusion data.

This allows researchers to replace simple analytic transport expressions with data appropriate to the gas and discharge conditions being studied.

Example cases are already included

The GitHub repository includes ready-to-run configuration examples for several discharge problems, including cases based on:

argon;
nitrogen;
deuterium;
helium.

This is useful for a new user because you do not need to begin from an empty input file.

Start with a supplied case, run it successfully, and then modify the geometry, gas properties, voltage waveform, emission model, transport data, or circuit for your own research problem.

How to start using PASCHEN-1D

A simple workflow is:

Visit the official University of Michigan PBI tools page ↗.
Download or clone the PASCHEN-1D GitHub repository ↗.
Install the Python dependencies described in the repository.
Open run_paschen_1d.ipynb.
Select one of the supplied configuration modules.
Run the simulation.
Use the diagnostics notebooks to inspect the results.
Modify the configuration for your own research case.

The repository also provides a command-line runner for users who prefer to run cases without Jupyter.

Who may find PASCHEN-1D useful?

PASCHEN-1D can be useful for students and researchers working in:

low-temperature plasma physics;
gas discharge physics;
electrical breakdown;
plasma–surface interaction;
electron emission;
laser-driven photoemission and plasma formation;
pulsed-power systems;
microgap discharges;
plasma electronics;
plasma–circuit coupling;
computational plasma physics.

It is particularly attractive when the important physics is approximately one-dimensional and when researchers want to perform many parameter scans efficiently.

When should you use a different model?

PASCHEN-1D is a fluid, one-dimensional model.

A different simulation approach may be more appropriate when your problem requires:

strongly two- or three-dimensional geometry;
kinetic electron distribution functions that cannot be represented adequately through the available closures;
detailed particle kinetics requiring PIC or Monte Carlo methods;
multidimensional magnetic-field effects.

The advantage of PASCHEN-1D is not that it replaces every plasma code. Its advantage is that it provides a relatively lightweight, configurable framework for a broad class of 1D discharge problems.

Code availability and license

PASCHEN-1D is publicly available for research use through GitHub:

https://github.com/pbis-umich/PASCHEN-1D ↗

The repository currently identifies the software license as CC BY-NC 4.0. This permits reuse under the license conditions but restricts commercial use. Always check the current repository license before using, modifying, or redistributing the software.

The repository also includes a CITATION.cff file so the software citation can be imported directly from GitHub.

Cite PASCHEN-1D

If you use PASCHEN-1D in research, please cite the associated paper:

A. Iqbal, Y. Heri, B. Wang, L. Jin, M. A. Faisal, and P. Zhang,
“PASCHEN-1D: A one-dimensional fluid plasma solver with multi-mechanism surface emission and flexible external circuit coupling,”
Computer Physics Communications, 2026.
DOI: 10.1016/j.cpc.2026.110404 ↗

One software package, many discharge problems

For someone beginning a related research project, the main value of PASCHEN-1D is simple:

You can start from an existing, documented plasma solver and concentrate on your physical problem instead of first writing an entire discharge code.

If your system can be represented reasonably well in one dimension and involves gas breakdown, transport, electrode emission, or external-circuit coupling, PASCHEN-1D is worth exploring.

Start from the official PBI project page ↗ and the GitHub repository ↗.

When Classical Spreading-Resistance Approximations Fail in Ultrathin Films

2026-09-22T00:00:00-04:00

A famous (1/a) law with hidden assumptions

One of the most familiar results in spreading-resistance theory is the inverse-radius scaling

[R_s \propto \frac{1}{a}.]

For a sufficiently large diffusive medium, this behavior is physically intuitive: a larger contact gives current more area through which to enter the material, so the constriction resistance decreases.

In anisotropic layered materials, a commonly used form replaces the bulk resistivity by an effective value and writes

[R_s \approx \frac{\rho_{\mathrm{eff}}}{4a}.]

Our 2026 study shows that this expression remains useful — but only as a limiting case. When films become thin or laterally confined, the same formula can fail even if the experimental data look approximately linear over a narrow range.

When the classical limit works

The exact anisotropic solution was benchmarked against spreading-resistance measurements on highly oriented pyrolytic graphite.

In those measurements, the contact radii ranged from roughly 100 to 500 nm, while the graphite flake had a lateral extent much larger than the contact and a substrate thickness of about 2 mm.

This is close to the wide-and-thick asymptotic regime:

[b\gg a, \qquad h_2 \text{ large}.]

Under those conditions, both the experimental trend and the exact anisotropic model recover the apparent (1/a) behavior.

For graphite, the measured anisotropy ratio used in the comparison was approximately

[\alpha_2 \approx 2.73\times10^{-5},]

which is extremely anisotropic. Yet the classical scaling still emerges because the surrounding geometry is large enough to let the current spread toward its asymptotic pattern.

What changes when the device becomes finite

The agreement does not survive arbitrary scaling.

When the surrounding lateral dimension $b$ approaches the contact dimension $a$, current can no longer spread as if the material were infinite. The device boundary begins to reshape the field.

The paper shows that deviations become pronounced as (b/a) is reduced, with strong finite-boundary effects appearing below the wide-device regime. The conclusion section emphasizes that once (b/a) falls below roughly 20, or the film becomes only a few nanometers thick, the simple diffusive approximation can no longer be trusted.

At large (b/a), the resistance eventually saturates with increasing lateral extent because the remote boundary has become irrelevant.

Thin films introduce another constraint

Thickness matters for the same reason.

In a very thin lower layer, the current cannot penetrate deeply before encountering the bottom boundary. The spreading pattern becomes more two-dimensional and remains sensitive to the lateral geometry.

The MoS$_2$ benchmark illustrates this clearly. For a moderately anisotropic multilayer device with $b/a\approx20$ and $h_2=120~\mathrm{nm}$, the data can still display an apparent (1/a) dependence.

But if the lateral extent is reduced to about $b=5a$, or the thickness is reduced to only a few layers, such as $h_2=7~\mathrm{nm}$ in the model comparison, the exact solution departs substantially from the classical line.

Why an effective isotropic resistivity is not enough

A tempting workaround is to replace the anisotropic material by an effective scalar resistivity, for example using a geometric mean of the in-plane and out-of-plane components.

That may reproduce the classical asymptote but does not reproduce the finite-device field.

The paper demonstrates cases where the isotropic model fails to match the exact anisotropic result even when the same nominal effective resistivity is used.

The problem is structural: anisotropy changes the direction of current flow. A scalar replacement cannot reproduce that directional redistribution once boundaries become important.

The danger for parameter extraction

This matters because spreading-resistance measurements are often used to infer intrinsic material properties.

If a finite ultrathin device is fit with

[R_s=\frac{\rho_{\mathrm{eff}}}{4a},]

the fitted $\rho_{\mathrm{eff}}$ can absorb geometry effects that do not belong to the material itself. Back-calculating in-plane or out-of-plane resistivity from that fit can therefore give a serious error.

The safer approach is to fit the resistance using the actual device dimensions and the full anisotropic field solution.

A useful way to remember the result

The classical (1/a) relation is not “wrong.”

It is the correct asymptotic result for a contact embedded in a sufficiently wide and thick medium.

The mistake is treating that asymptote as a universal contact law.

As devices move toward ultrathin layers, short channels and strongly anisotropic materials, the geometry stops being a small correction and becomes part of the physics that determines the measured resistance.

Exact Spreading Resistance in Vertical Contacts to Anisotropic Thin Films

2026-09-15T00:00:00-04:00

Building an exact model for a finite anisotropic contact

Spreading resistance is the extra voltage drop created when current leaves a finite contact and redistributes into a larger conducting region. Classical expressions are compact and useful, but they usually assume isotropic transport, idealized geometry, or semi-infinite media.

Those assumptions become questionable for modern layered devices. A vertical contact to a thin MoS$_2$, WSe$_2$, graphite, or other van der Waals layer may have strong conductivity anisotropy, a finite channel width, nanometer-scale thickness, and a nonzero interfacial resistance.

Our published framework was developed to keep all of those ingredients in the same field solution.

The governing equation keeps the tensor physics

In each material region, the potential satisfies

$ \nabla\cdot\left(\boldsymbol{\sigma}\nabla\Phi\right)=0. $

The conductivity tensor is diagonal in the principal material directions, so the in-plane and out-of-plane conductivities can be specified independently.

For two stacked layers,

$ \boldsymbol{\rho}1= \operatorname{diag}\left(\rho{1,t},\rho_{1,z}\right), \qquad \boldsymbol{\rho}2= \operatorname{diag}\left(\rho{2,t},\rho_{2,z}\right). $

where (t) denotes the lateral coordinate and (z) the vertical direction.

The two regions are joined through a finite specific interfacial resistivity $\rho_i$, which permits a potential discontinuity proportional to the normal current density.

This boundary condition is important because the interface is not simply added after the field solution. A finite interface resistance can alter the way current spreads in the adjoining materials.

Cartesian and cylindrical contacts

The theory is derived for both planar and disk-shaped contacts.

In Cartesian geometry, a contact of half-width $a$ injects current into a lower region extending to a half-width (b>a). In cylindrical geometry, the same variables become the contact radius and surrounding device radius.

The total contact resistance can be decomposed conceptually as

[R_c = R_s + R_{\mathrm{interface}} + R_1,]

where $R_s$ is the spreading resistance, $R_{\mathrm{interface}}$ is the interface contribution, and $R_1$ is the vertical resistance of the upper contact material.

This decomposition makes it possible to see which physical contribution dominates for a given geometry and material set.

Why exact geometry matters

A semi-infinite approximation cannot know when a real device edge is close to the contact.

If the peripheral boundary is far away and the lower layer is thick, the current has enough space to spread naturally. But when (b/a) becomes small or the film thickness is reduced, the boundaries constrain the current paths.

The exact solution captures those finite-size effects automatically because the device dimensions enter the field expansion and boundary conditions.

This is especially important in nanoscale contacts, where a contact radius, channel width and film thickness may differ by only one or two orders of magnitude rather than by many decades.

Anisotropy becomes a geometric transformation of the field

One useful way to interpret the mathematics is that anisotropy effectively stretches one coordinate relative to another.

When lateral resistivity is much smaller than vertical resistivity, equipotential contours spread much farther along the plane. The current density becomes correspondingly concentrated near the contact edges.

The exact field solution therefore connects a measurable terminal quantity — spreading resistance — to the spatial distribution that causes it.

That is an advantage over a purely lumped model.

Recovering familiar limiting cases

A good general model should reduce to known results when its extra physics is removed.

The framework recovers previously established isotropic finite-boundary solutions and the known single-layer anisotropic limit. In the wide-and-thick regime, it also approaches the classical diffusive (1/a) scaling used for spreading resistance.

These limits are valuable checks because they show that the exact anisotropic formulation extends rather than replaces the classical theory.

Why the model is useful experimentally

The field solution is not only intended for forward prediction.

Because the model explicitly includes $a$, $b$, $h_1$, $h_2$, $\rho_i$, and the anisotropic resistivities, it can be used to interpret experimental resistance data without folding geometry into an uncontrolled effective parameter.

That matters when extracting intrinsic transport properties. If an experiment lies outside the wide, thick, semi-infinite regime, fitting it with a simple (1/a) law can return a misleading in-plane or out-of-plane resistivity.

The exact model provides a physics-based alternative in which material anisotropy and actual device dimensions remain visible throughout the calculation.

Analytical Theory vs FEM for Anisotropic Vertical Contacts

2026-09-13T00:00:00-04:00

Why solve the same contact two different ways?

Analytical theory and finite-element simulation are sometimes presented as competing approaches. For electrical contacts in anisotropic thin films, they are more useful when treated as complementary tools.

An exact field solution exposes scaling and limiting behavior. FEM provides direct spatial visualization and can be extended to geometries that become cumbersome analytically.

In our anisotropic-contact study, we used both to test whether the same physical model produced the same resistance and current-crowding behavior.

The analytical side

The theory begins with the conductivity-tensor form of Laplace’s equation,

$ \nabla\cdot\left(\boldsymbol{\sigma}\nabla\Phi\right)=0. $

Both the top and bottom layers can be anisotropic, each with independent in-plane and out-of-plane resistivities. A finite specific interface resistivity $\rho_i$ is included, and finite lateral and vertical boundaries are retained.

Exact field expansions are derived for Cartesian contacts and cylindrical disk contacts.

From the potential, the current density follows from

[\mathbf{J}=-\boldsymbol{\sigma}\nabla\Phi.]

This provides both the total resistance and the full spatial distribution of current.

The FEM side

The same geometry and material tensors were then implemented numerically in COMSOL.

FEM does not require the field to be expressed as a closed analytical series. Instead, the domain is discretized and the anisotropic conduction equation is solved directly.

That makes it particularly good for plotting equipotential contours and current-density maps. The edge crowding caused by high in-plane conductivity becomes visually obvious: potential spreads laterally beneath the contact, while current injection becomes increasingly concentrated near the perimeter.

Agreement is more valuable than either method alone

Across the parameter sweeps reported in the paper, the exact analytical results and FEM calculations show excellent agreement.

That agreement matters for two reasons.

First, it validates the algebra and boundary treatment of the analytical solution.

Second, it validates the numerical implementation against a solution that does not depend on mesh refinement or solver configuration in the same way.

When two mathematically independent routes converge, confidence in the physical result is much stronger.

Experimental benchmarks add a third layer

The study also compares the theory with published measurements on graphite and multilayer MoS$_2$.

For highly oriented pyrolytic graphite, the experimental contact sizes and thick substrate place the system close to the classical wide-and-thick limit. The exact anisotropic solution reproduces the observed inverse-radius trend using the measured anisotropy, without introducing fitting parameters to force agreement.

For multilayer MoS$_2$, the same framework captures the regime where the classical diffusive model works and predicts where it will fail as the film becomes thinner or the lateral extent becomes smaller.

This three-way comparison — analytical theory, FEM, and experiment — is stronger than validating one numerical model against another.

What each method is best at

The analytical solution is ideal for:

rapid parameter sweeps;
identifying asymptotic scaling;
isolating the role of anisotropy;
fitting experimental data efficiently;
serving as a reference solution for numerical codes.

FEM is ideal for:

visualizing local current density and potential;
adding geometric details;
treating nonuniform interfaces;
exploring device shapes outside the separable analytical geometries.

The two approaches therefore serve different stages of the same design workflow.

No-fitting-parameter validation is important

One particularly useful feature of the experimental comparison is that the model does not need an arbitrary fitting factor to recover the reported graphite and MoS$_2$ trends.

That makes the comparison a test of the physical field formulation rather than merely a curve-fitting exercise.

It also helps reveal why a simpler isotropic effective-medium model may appear adequate in one geometry and fail badly in another. The full theory knows the actual device width, film thickness and conductivity tensor, so it can distinguish a true material effect from a finite-geometry effect.

A broader modeling lesson

For nanoscale electronic devices, numerical simulation is increasingly easy to run. That convenience can make it tempting to treat the numerical output as the final answer.

The contact problem illustrates why an analytical reference remains valuable.

A field solution tells us why a trend occurs, which parameters control it, and what happens in limiting cases. FEM then provides the flexibility to move beyond the idealized geometry without losing that physical intuition.

The strongest model is therefore not necessarily the most computationally elaborate one. It is the one that can be checked from multiple independent directions.