---
title: 'EOT-X: A Polysemous Technical Label'
url: https://www.emergentmind.com/topics/eot-x
type: topic
---

# EOT-X: A Polysemous Technical Label

Searching arXiv for “EOT-X” and closely related usages to ground the article.
arxiv_search.query({"search_query":"all:\"EOT-X\" OR ti:\"EOT-X\" OR abs:\"EOT-X\"","start":0,"max_results":10,"sort_by":"submittedDate","sort_order":"descending"})
Cross-checking the cited arXiv records by identifier.
arxiv_search.query({"search_query":"id:1011.3713 OR id:1705.10171 OR id:1212.4227 OR id:1302.5660 OR id:2507.04708 OR id:1604.00970 OR id:2604.12940 OR id:2110.05232 OR id:2112.09283 OR id:2511.00286","start":0,"max_results":20,"sort_by":"relevance","sort_order":"descending"})
EOT-X is a context-dependent research label that appears in several technically unrelated literatures. In optics, it denotes mechanisms and extensions of extraordinary optical transmission through periodic metallic structures, including standing-wave and cavity-induced formulations [1011.3713; 2112.09283]. In semiconductor device physics, it denotes equivalent oxide-thickness engineering in InGaAs gate-all-around nanowire MOSFETs [1212.4227; 1302.5660]. In natural language processing, it names a human-annotated dataset for joint emotion detection and opinion-trigger extraction in e-commerce reviews [2507.04708]. Closely related abbreviations also designate extended object tracking, entropic optimal transport, the Eocene--Oligocene transition, and an Equation-of-Time dynamical framework [1604.00970; 2604.12940; 2110.05232; 2511.00286]. This suggests that EOT-X is not a single standardized term, but a domain-specific shorthand whose interpretation is fixed by disciplinary context.

## 1. Nomenclature and scope

The principal source of ambiguity is that the same letter sequence is reused across fields for distinct technical objects, tasks, and theories.

| Domain | Usage of EOT-X / related form |
|---|---|
| Optics | Extraordinary-transmission-by-standing-waves mechanism; cavity-induced extraordinary optical transmission |
| Semiconductor devices | Equivalent oxide-thickness engineering in InGaAs GAA nanowire MOSFETs |
| NLP | Dataset for Emotion detection and Opinion Trigger extraction |
| Tracking | Extended Object Tracking |
| Statistics / OT | Entropic Optimal Transport |
| Paleobiology | Eocene--Oligocene transition |
| Earth--Sun physics | Equation-of-Time dynamical framework |

A recurring misconception is to treat EOT-X as a single cross-domain acronym. The literature instead shows several unrelated usages. In optics, the label is tied to resonant transmission in periodic metallic structures; in transistor research it concerns gate-stack capacitance scaling; in NLP it denotes a benchmark resource; and in adjacent fields the same base acronym, EOT or EoT, refers to entirely different mathematical or historical objects [1011.3713; 1212.4227; 2507.04708; 1604.00970].

## 2. Extraordinary optical transmission as a standing-wave mechanism

In the unified analytical model of Schwarz et al., extraordinary optical transmission through a corrugated metallic film is explained by standing-wave resonances of higher diffraction orders in a periodic grating rather than by specific polarization-dependent surface-plasmon properties [1011.3713]. The geometry is a one-dimensional metal grating of period $d$, slit width $a$, metal thickness $h$, and surrounding refractive indices $n_1$ and $n_3$, with slit filling index $n_s$. A plane wave of vacuum wavenumber $k_0=2\pi/\lambda$ and incidence angle $\theta$ excites Bloch orders with transverse wavenumbers
$$
\beta_m = k_0 \sin\theta + m\frac{2\pi}{d}.
$$

The key condition is that one of the higher orders, typically $m=\pm1$, is evanescent outside the grating but propagating inside the slit region. It then forms a leaky cavity, and the standing-wave condition along $z$ produces pronounced transmission maxima in the zero-order outgoing beam. In this formulation, extraordinary transmission is a Fabry--Pérot resonance of a Bragg mode. For ideal metal walls, the slit propagation constants are polarization dependent: for TE polarization the fundamental mode has cutoff with $\gamma=\pi/a$ and
$$
k_z^{(\mathrm{TE})} = \sqrt{(n_s k_0)^2 - \gamma^2},
$$
whereas for TM polarization there is no cutoff and
$$
k_z^{(\mathrm{TM})} = n_s k_0.
$$

The effective index of the $m$-th Bragg mode is written as
$$
n_{\mathrm{eff},m}=\frac{\sqrt{(k_z^{\mathrm{prop}})^2+\beta_m^2}}{k_0},
$$
and in practice the first Bragg order dominates. The resonance condition for a slab of thickness $h$ and index $n_{\mathrm{eff}}$ between media $n_1$ and $n_3$ is
$$
2k_z^{\mathrm{prop}}h - 2\phi_1 - 2\phi_3 = 2\pi \ell,\qquad \ell\in\mathbb{N},
$$
with reflection phase shifts $\phi_1$ and $\phi_3$ defined through $\arctan(\Gamma_j/k_z^{\mathrm{prop}})$ for TE or TM cases. The same underlying transcendental equation governs TE and TM polarizations and both subwavelength and non-subwavelength regimes.

The zero-order transmittance is then expressed by a standard Airy formula for a lossy slab,
$$
T(\lambda)=\frac{(1-R_1)(1-R_3)}{1-2\sqrt{R_1R_3}\cos(2k_z^{\mathrm{prop}}h-\phi_1-\phi_3)+R_1R_3},
$$
with $R_j=|r_j|^2$ and Fresnel-type coefficients $r_j$ defined through the effective index and propagation angles. In the subwavelength regime $d<\lambda$, only $m=0$ propagates outside, while $m=\pm1$ remain trapped and generate sharp EOT peaks when the slab-resonance condition is satisfied. In the non-subwavelength regime $d\ge \lambda$, first Bragg orders may propagate externally, and the peaks broaden and merge into full-transmission Wood anomalies, but the same resonance equation continues to predict enhanced transmission lobes.

For the numerical example $d=0.9\,\mu\mathrm{m}$, $a=0.35\,\mu\mathrm{m}$, $h=0.25\,\mu\mathrm{m}$, $n_1=n_3=n_s=1$, and normal incidence, the $m=1$ standing-wave equation predicts TM resonances at approximately $\lambda_0\approx0.82\,\mu\mathrm{m}$ for $\ell=1$ and $\lambda_0\approx1.65\,\mu\mathrm{m}$ for $\ell=2$, in agreement within $\pm10\,\mathrm{nm}$ with rigorous-coupled-wave analysis. For TE polarization, no EOT appears for $\lambda<0.70\,\mu\mathrm{m}$ because $\lambda/n_s>2a\approx0.70\,\mu\mathrm{m}$ marks the cutoff condition, while above cutoff the same framework predicts a peak near $\lambda_0\approx0.85\,\mu\mathrm{m}$, matching RCWA within $20\,\mathrm{nm}$ [1011.3713].

## 3. Metasurface and cavity extensions in the optical literature

Subsequent optical work broadens this resonance picture in two directions: angle-routed decay channels in asymmetric slit metasurfaces and cavity-induced EOT below the usual surface-plasmon boundary.

In the metasurface of Deng et al., a periodic array of narrow metallic slits is embedded asymmetrically with $n_1>n_3$, so that changing the in-plane wavevector $k_x=k_0\sin\theta$ opens distinct phase-space windows for enhanced $0$th-order transmission, total internal reflection, or enhanced $-1$st-order diffraction [1705.10171]. The field in the superstrate and substrate is expanded in diffraction orders, while inside the slit only the fundamental cavity mode is retained. Projecting the boundary conditions yields a $2\times2$ linear system whose nontrivial solution defines the slit-cavity dispersion relation. In the $(k_x,k_0)$ diffraction-order chart, small incident angles place the system in a regime where resonant decay routes mainly into the $0$th transmitted order, producing EOT; larger angles suppress transmission and eventually route decay into the $-1$st reflected order, producing extraordinary optical diffraction. The overlap coefficient into order $m$ is proportional to a sinc factor,
$$
C_m(\theta)\propto \mathrm{sinc}\!\left[\frac{(k_x+2\pi m/p)w}{2}\right],
$$
and the relative magnitudes of $|C_0|^2$ and $|C_{-1}|^2$ determine whether EOT or EOD dominates.

For $p=1\,\mu\mathrm{m}$, $w=0.1\,\mu\mathrm{m}$, $h=1\,\mu\mathrm{m}$, and $(n_1,n_2,n_3)=(2,1.4,1)$, the cavity resonance is located at $\lambda_0\approx3.3\,\mu\mathrm{m}$ and is nearly independent of $\theta$. For $\theta<30^\circ$, both FEM and the analytical model give $T_0>0.9$, $R_0<0.1$, and $\mathrm{FWHM}\sim0.05\,\mu\mathrm{m}$; for $30^\circ<\theta<40^\circ$ the structure behaves as a mirror with $R_0\approx1$; for $\theta>40^\circ$, $D_{-1}>0.9$ while $R_0$ and $T_0$ are near zero. This enables a single planar device to operate as a transmission filter, mirror, or off-axis lens [1705.10171].

A different extension is the cavity-induced EOT reported by Zhang et al., where a holey gold film on a silicon slab forms a weak Fabry--Pérot cavity supporting a bound surface state below the classical first-order surface-plasmon cutoff $f_{\mathrm{pl0}}$ [2112.09283]. The resonance condition is
$$
2k_{\mathrm{Si}}(\omega)h+\phi(\omega)=2\pi m,
$$
with silicon thickness $h$ and phase correction $\phi(\omega)$ from reflection at the perforated metal. The surface-plasmon-like mode is modeled by an effective plasma relation with cutoff
$$
f_{\mathrm{pl0}}=\frac{c}{d\sqrt{\varepsilon_{\mathrm{Si}}}}.
$$
For $d=150\,\mu\mathrm{m}$ and $\varepsilon_{\mathrm{Si}}=11.7$, this gives $f_{\mathrm{pl0}}\simeq583\,\mathrm{GHz}$. Measured transmittance for $h=207$, $196$, and $185\,\mu\mathrm{m}$ shows cavity-EOT peaks below $f_{\mathrm{pl0}}$ at approximately $503$, $525$, and $550\,\mathrm{GHz}$, respectively. As the resonance approaches $f_{\mathrm{pl0}}$, the cavity-induced EOT merges with the first-order surface-plasmon EOT without a transmission zero between them, broadening the bandwidth by roughly one order of magnitude in the THz regime. The same platform supports active modulation with graphene: at $500\,\mathrm{GHz}$ the peak transmittance decreases from about $0.50$ to $0.25$ under a bias sweep from $-0.3$ to $0.5\,\mathrm{V}$ [2112.09283].

Taken together, these optical papers shift the interpretation of EOT-X away from a single plasmonic narrative. One strand emphasizes higher-order Bragg-mode standing waves independent of surface plasmons; another shows angle-controlled routing of slit-cavity decay into transmission or diffraction channels; and a third demonstrates cavity-induced bound states below the classical plasmonic cutoff [1011.3713; 1705.10171; 2112.09283].

## 4. Equivalent oxide-thickness engineering in InGaAs nanowire MOSFETs

In semiconductor device physics, EOT-X denotes scaling of the equivalent oxide thickness in InGaAs gate-all-around nanowire MOSFETs. Gu et al. demonstrated $20\,\mathrm{nm}$--$80\,\mathrm{nm}$ channel-length devices with EOT reduced to $1.2\,\mathrm{nm}$ by combining an ultrathin interfacial $\mathrm{Al_2O_3}$ layer with a higher-$\kappa$ $\mathrm{LaAlO_3}$ dielectric [1212.4227]. The gate stack uses $0.5\,\mathrm{nm}$ $\mathrm{Al_2O_3}$ followed by $4\,\mathrm{nm}$ $\mathrm{LaAlO_3}$, with a $40\,\mathrm{nm}$ tungsten nitride gate deposited without air break. The capacitance-equivalent thickness is
$$
\mathrm{EOT}=t_{\mathrm{Al_2O_3}}\left(\frac{\varepsilon_{\mathrm{SiO_2}}}{\varepsilon_{\mathrm{Al_2O_3}}}\right)
+t_{\mathrm{LaAlO_3}}\left(\frac{\varepsilon_{\mathrm{SiO_2}}}{\varepsilon_{\mathrm{LaAlO_3}}}\right),
$$
so that with $\varepsilon_{\mathrm{SiO_2}}\approx3.9\varepsilon_0$, $\varepsilon_{\mathrm{Al_2O_3}}\approx9\varepsilon_0$, and $\varepsilon_{\mathrm{LaAlO_3}}\approx16\varepsilon_0$, Sample A gives
$$
\mathrm{EOT}\approx0.5\,\mathrm{nm}\cdot\frac{3.9}{9}+4\,\mathrm{nm}\cdot\frac{3.9}{16}\approx1.20\,\mathrm{nm}.
$$

The reported fabrication sequence includes MBE growth on an InP substrate, source/drain implantation, electron-beam lithography defining fins with $L_{\mathrm{ch}}=20$--$80\,\mathrm{nm}$ and $W_{\mathrm{nw}}=20$--$35\,\mathrm{nm}$, dry and wet etching to release the nanowires, $(\mathrm{NH}_4)_2\mathrm{S}$ passivation, ALD growth of the dielectric stack, WN deposition, and subsequent patterning and metallization [1212.4227]. This process yields a conformal wrap-around stack on $20$--$35\,\mathrm{nm}$ wide nanowires.

The electrical consequences are substantial. For the best devices, the authors report subthreshold swing $\mathrm{SS}=63\,\mathrm{mV/dec}$, drain-induced barrier lowering as low as $7\,\mathrm{mV/V}$, $I_{\mathrm{ON}}\approx0.63\,\mathrm{mA/\mu m}$, and $g_m\approx1.74\,\mathrm{mS/\mu m}$ at $V_{DD}=0.5\,\mathrm{V}$; gate leakage remains below $10^{-4}\,\mathrm{A/cm^2}$ for EOT $=1.2\,\mathrm{nm}$ at $V_{gs}$ up to $0.8\,\mathrm{V}$ [1212.4227]. The interface-trap density is inferred from
$$
\mathrm{SS}\simeq60\,\mathrm{mV/dec}\left[1+\frac{qD_{it}}{C_{ox}}\right],
$$
yielding a mean $D_{it}\simeq4\times10^{12}\,\mathrm{eV^{-1}cm^{-2}}$ and a best-device value near $9\times10^{11}\,\mathrm{eV^{-1}cm^{-2}}$.

The companion study on variability shows that placing $\mathrm{Al_2O_3}$ directly on InGaAs is critical [1302.5660]. For $L_{\mathrm{ch}}=20\,\mathrm{nm}$ and $W_{\mathrm{NW}}=20\,\mathrm{nm}$ at $V_{DD}=0.5\,\mathrm{V}$, the Al$_2$O$_3$-first flow gives $57\,\mu\mathrm{A}$ on-current per wire, $165\,\mu\mathrm{S}$ peak transconductance, $\mathrm{SS}=75\,\mathrm{mV/dec}$, and $\mathrm{DIBL}=40\,\mathrm{mV/V}$, compared with $48\,\mu\mathrm{A}$, $155\,\mu\mathrm{S}$, $80\,\mathrm{mV/dec}$, and $73\,\mathrm{mV/V}$ for the LaAlO$_3$-first flow. Across $50$ measured devices, the Al$_2$O$_3$-first process reduces the standard deviation of $I_{\mathrm{ON}}$ by $54\%$, the standard deviation of $g_{m,\max}$ by $64\%$, the standard deviation of SS by $25\%$, and the interquartile range by $46\%$ [1302.5660]. In this literature, EOT-X therefore refers not to transmission phenomena but to aggressive electrostatic scaling via dielectric-stack engineering.

## 5. EOT-X as a dataset for joint emotion and trigger extraction

In natural language processing, EOT-X is a human-annotated dataset introduced for the joint task of Emotion detection and Opinion Trigger extraction in e-commerce text [2507.04708]. The underlying task unifies emotion classification with extraction of the exact contiguous text span that explains why the emotion occurred. The annotation framework uses Plutchik’s eight primary emotions plus a Neutral label when no emotion is present.

The dataset contains $2{,}400$ customer reviews sampled equally from Amazon categories including Beauty, Home, Electronics, and Clothing, as well as TripAdvisor and Yelp, with roughly $6{,}000$--$6{,}500$ labeled emotion occurrences after aggregation and an average of $2$--$3$ emotions per review [2507.04708]. The approximate class distribution is Joy $\approx1{,}839$ instances $(\approx30\%)$, Trust $\approx857$ $(\approx14\%)$, Disgust $\approx760$ $(\approx12.5\%)$, Anticipation $\approx752$ $(\approx12\%)$, Surprise $\approx700$ $(\approx11\%)$, Sadness $\approx589$ $(\approx10\%)$, Anger $\approx406$ $(\approx7\%)$, and Fear $\approx98$ $(\approx2\%)$.

Annotation is performed by three expert raters with graduate-level training in emotion analysis. The instructions require annotators to read the entire review, identify all clearly expressed emotions, extract one or more exact substrings for each emotion, avoid inferring unstated emotions, and avoid discontinuous spans. Edge-case rules specify that nested spans should be resolved by choosing the most informative continuous span, sarcasm should be interpreted for the intended emotion, and the entire review may serve as the trigger for very short texts. The gold standard uses majority vote for emotions and a union of unique trigger spans, with the longest span chosen when overlaps occur. Inter-annotator agreement is reported as Fleiss’ $\kappa\approx0.89$ for emotion labels and $\kappa\approx0.84$ for token-level trigger matching, both described as almost perfect agreement [2507.04708].

The per-review JSON structure includes the review identifier, full review text, and a list of emotion objects with associated triggers. Product- and service-level stratification is used to avoid leakage, producing splits of $1{,}919$ training reviews $(80\%)$, $239$ validation reviews $(10\%)$, and $241$ test reviews $(10\%)$. Emotion detection is evaluated as multi-label classification with precision, recall, and F1; trigger extraction is scored with Exact Match, Partial Match, and ROUGE-1 and ROUGE-L [2507.04708].

This NLP usage is conceptually unrelated to optics or transistor scaling. Here EOT-X is a benchmark resource intended for explainable opinion mining, customer-service modeling, and study of joint emotion--cause extraction, and the associated paper evaluates $23$ large language models and proposes the EOT-DETECT prompting framework [2507.04708].

## 6. Other technical uses of EOT and EoT

Several adjacent literatures use the base acronym EOT or the variant EoT for concepts that are again unrelated to the preceding meanings.

In tracking and sensor fusion, EOT denotes extended object tracking: an object is modeled by a kinematic state $x_k$, an extent state $E_k$, and a measurement set $Z_k=\{z_k^{(j)}\}_{j=1}^{n_k}$, with Bayesian prediction and update driven by an extended-object measurement likelihood [1604.00970]. Two major single-object paradigms are reviewed. The random-matrix approach represents extent by a symmetric positive-definite matrix $X_k$ and approximates the posterior with a Gaussian--Inverse--Wishart density. The random hypersurface model represents star-convex contours through a radial function and applies nonlinear Kalman filtering to a non-additive measurement model. The same review covers multi-object EOT through both Random Finite Set and non-RFS methods, including PHD, CPHD, CBMeMBer, GLMB, LMB, PMBM, JPDA, and MHT, and discusses camera, X-band radar, lidar, and RGB-D measurement models [1604.00970].

In statistics and optimal transport, EOT denotes entropic optimal transport [2604.12940]. For probability measures $\mu,\nu$ on a Polish locally compact space and regularization parameter $\epsilon>0$, the EOT value is
$$
\mathrm{EOT}_\epsilon(\mu,\nu)=\min_{\pi\in\Pi(\mu,\nu)}\int c(x,y)\,d\pi(x,y)+\epsilon\,\mathrm{KL}(\pi\|\mu\otimes\nu).
$$
The corresponding optimal plan has density
$$
\frac{d\pi^\epsilon}{d(\mu\otimes\nu)}(x,y)=\exp\!\left(\frac{f_\epsilon(x)+g_\epsilon(y)-c(x,y)}{\epsilon}\right),
$$
where $(f_\epsilon,g_\epsilon)$ solve the Schrödinger system. The cited paper establishes asymptotic weak convergence for a large class of functionals of the empirical EOT plan, derives uniform confidence bands for colocalization curves, proves bootstrap consistency, and illustrates the theory with Monte Carlo scenarios and STED microscopy data on mitochondrial proteins [2604.12940].

In paleobiology, EOT denotes the Eocene--Oligocene transition, a period of global environmental change coinciding with a major biotic turnover [2110.05232]. The cainotherioid study uses Bayesian birth--death models with $1\,\mathrm{Myr}$ bins, time-continuous environmental regressions, and six cranio-dental traits to infer diversification dynamics across the transition. Speciation declines toward the transition, then rises sharply just after it, reaching about $0.9\,\mathrm{Myr}^{-1}$ at $33$--$31\,\mathrm{Ma}$; extinction peaks near $34\pm0.5\,\mathrm{Ma}$ and again near $27$--$26\,\mathrm{Ma}$; within-clade diversity dependence is negative with $G_{\mathrm{div}}=-2.65$ [2110.05232]. This is a historical-geological usage rather than an acronymic technical framework.

In Earth--Sun physics, the variant EoT is the Equation of Time, and one recent paper develops a unified dynamical framework around it [2511.00286]. The paper writes solar meridian declination and the Equation of Time as Fourier series in the fractional-year angle $x=(2\pi/365)(t-1)$, decomposes the EoT into eccentricity and obliquity terms, defines subsolar-point acceleration $NBI_\alpha$, and argues that the locus $NBI_\alpha(\delta)$ traces a lemniscate whose symmetry mirrors the analemma. It further defines the EoT velocity $\omega^*=d\delta^*/dt$, reports extrema at the solstices and zero crossings at mid-season, constructs derivative chains up to snap for declination, EoT, and Earth’s rotational speed, and identifies eight sharp kinematic intervals over the annual cycle [2511.00286].

Across these literatures, the unifying fact is not conceptual continuity but acronymic collision. EOT-X therefore functions as a polysemous label whose meaning must be resolved from the surrounding technical discourse, equations, and application domain.

Source: https://www.emergentmind.com/topics/eot-x