---
title: Three-Level Encapsulation (TLE) Overview
url: https://www.emergentmind.com/topics/three-level-encapsulation-tle
type: topic
---

# Three-Level Encapsulation (TLE) Overview

Three-Level Encapsulation (TLE) denotes several distinct constructs across contemporary research rather than a single unified method. In the cited literature, the term refers to a formally verified, bitmask-based encoding scheme for three contiguous hierarchy levels in full-stack systems; a liquid-liquid process that yields a single inner core surrounded by three shell layers; a graphene/MoS2/graphene “sandwich” for radiation-resistant electron microscopy; an oil-encapsulated bubble-in-water three-phase compound; and a three-level polarization-state construction in plasmonic photonics [2510.00002] [2408.09026] [1310.4012] [2505.18339] [1806.00619].

## 1. Domain-specific meanings of TLE

The shared acronym masks substantial disciplinary divergence. In software engineering, TLE is a schema-level encoding pattern for hierarchies. In fluid mechanics and encapsulation science, it denotes multilayer shell formation around a core or bubble. In 2D materials microscopy, it names a three-layer heterostructure. In plasmonic photonics, it denotes a qutrit-style construction built from Stokes-parameter channels.

| Domain | Meaning of TLE | Three levels |
|---|---|---|
| Hierarchical data systems | Bitmask-based encoding of bounded hierarchies | Grandparent table, parent columns, child bitmasks |
| Liquid-liquid encapsulation | Triple-layered encapsulation | Core + three shells: $L_{c,o}$, $L_{s,1}$, $L_{s,2}$ |
| STEM/TEM of MoS2 | Full graphene encapsulation | Graphene / MoS2 / graphene |
| Bubble encapsulation | Oil-encapsulated bubble-in-water compound | Water, oil droplet, air bubble |
| Plasmonic photonics | Three-level polarization-state encoding | $|S_1\rangle$, $|S_2\rangle$, $|S_3\rangle$ |

A recurrent misconception is to treat TLE as a cross-disciplinary standard. The cited works do not support that interpretation. They use the same term for unrelated three-level organizations, and the technical content is domain-local rather than transferable by acronym alone [2510.00002] [2408.09026] [1310.4012] [2505.18339] [1806.00619].

## 2. TLE as a formally verified hierarchical encoding pattern

In scalable full-stack software engineering, TLE is defined as “a compact, bitmask-based encoding scheme that turns deep hierarchical relationships into single-row, constant-time operations.” Its canonical structure contains exactly three contiguous levels: Grandparent (Level $N$) as a table name, Parent (Level $N+1$) as column names in that table, and Children (Level $N+2$) as the cell value of each parent column, encoded as a bitmask. A parent at Level $N$ can anchor the next TLE instance at Level $N+1$, yielding recursive extensions such as Continent$\rightarrow$Country$\rightarrow$State, Country$\rightarrow$State$\rightarrow$County, and State$\rightarrow$County$\rightarrow$City [2510.00002].

The bit-level semantics are explicit. Each child under a common parent receives a zero-based child index $c_{id}$, and standard bitwise operators implement updates and membership tests:
$$
\text{Set: } B' = B \mid (1 \ll c_{id}), \qquad
\text{Clear: } B' = B \& \sim(1 \ll c_{id}),
$$
$$
\text{Toggle: } B' = B \oplus (1 \ll c_{id}), \qquad
\text{Selected}(c_{id}) = ((B \gg c_{id}) \& 1) = 1.
$$
For mask predicates, the scheme uses
$$
\text{Any among } S: (B \& M) \neq 0, \qquad
\text{All among } S: (B \& M) = M, \qquad
\text{Exactly } S: B = M.
$$
The implementation rule is to select the smallest bitmask type that fits the maximum child count: `INT (32 bits) if n ≤ 32`, `BIGINT (64 bits) if n ≤ 64`, and `VARCHAR(k) for n > 64`, with application-side `BigInteger` used for arbitrary-precision bitwise operations. The paper distinguishes the canonical structural triple from an optional “packed triple”; the latter is not part of the formalization and is not required.

The analytical properties are stated as bounded-hierarchy complexity theorems. Query and update complexity are
$$
T_{\mathrm{query}} =
\begin{cases}
O(1) & \text{if } n \le w\\
O(\lceil n/w\rceil) & \text{otherwise}
\end{cases}
\qquad
T_{\mathrm{update}} =
\begin{cases}
O(1) & \text{if } n \le w\\
O(\lceil n/w\rceil) & \text{otherwise}.
\end{cases}
$$
For parent-level scans with bounded branching factors $n_{\max}\le w$, batch traversal is
$$
T_{\mathrm{batch}} = O(P_{\mathrm{total}}),
$$
and storage reduction is expressed as
$$
\frac{S_{\mathrm{TLE}}}{S_{\mathrm{traditional}}}=\frac{\bar{C}}{\bar{c}\cdot k},
$$
where $\bar{C}$ is the average bitmask size, $\bar{c}$ is average children per parent, and $k$ is bits per relationship in the traditional foreign-key representation.

TLE is embedded in PBFD and related to PDFD. PBFD reuses a single pattern implementation across nodes that share a structural signature, such as identical bitmask layouts, and TLE supplies that signature together with constant-time data paths. PDFD uses bounded refinement ($R_{\max}$) and does not require TLE, although compact updates and predicate filters remain useful in its vertical validation phases. Both methodologies, and TLE’s operations, are modeled with unified state machines and verified in CSP/FDR.

The verification model defines states `S0 (Idle), S1 (Data Loaded), S2 (Hierarchy Resolved), S3 (Children Evaluated), S4 (Children Updated), S5 (Changes Committed), S6 (Workflow Finalized)` and events `TLE1–TLE11` over `LOAD`, `READ`, `WRITE`, `COMMIT`, and recurrence `S6→S0`. Reported FDR4 properties include deadlock freedom, divergence freedom, failures-divergences refinement to an abstract model, determinism, safe multi-unit composition, hostile-environment robustness, and recurrence/liveness. The paper states that these checks establish conformance of every `load→resolve→read→update→commit→finalize` cycle and of multiple concurrent units to the abstract specification.

Empirically, the reported enterprise deployment spans eight years with zero critical failures. The paper attributes to PBFD/TLE `7.64x faster median latency (P50), 8.54x faster tail latency (P95), and 7.44x faster average latency` than the aggregate of traditional controllers, together with `11.7× less reserved space, 85.7× smaller indexes, and 113.5× better page utilization`; it also reports reduction from `~4.7M (normalized)` rows to `~170K (core TLE tables)` and removal of all junction tables. The stated usage envelope is bounded fanout, repeated ancestor→descendant membership checks, frequent set/clear operations, and predictable $O(1)$ behavior. The stated limitations are schema rigidity, adoption challenges around bitmasking and scaffolding, speculative generalization to document stores, key-value stores, and graph databases, and the need for broader multi-context replication [2510.00002].

## 3. TLE as triple-layer liquid-liquid encapsulation

In multilayer droplet encapsulation, TLE denotes a single inner core wrapped by three distinct shell layers in radial order: `(1) the outer core liquid from the Y-junction (Lc,o), (2) an intermediate interfacial layer (Ls,1), and (3) an outer interfacial layer (Ls,2)`. The inner core itself is not counted as a shell layer, so `TLE = core + three shells`. The method couples a Y-junction compound droplet generator to impact-driven wrapping on stacked floating interfacial layers [2408.09026].

The reported geometry uses a distortion-free glass cuvette with inner dimensions `36 mm × 36 mm × 30 mm`, host bath `L_h` of deionized water, and a compound droplet in which ethylene glycol serves as the inner core `L_{c,i}` and “laser oil” as the outer core `L_{c,o}`. For TLE specifically, the outermost layer `L_{s,2}` is silicone oil dispensed first on water, and the intermediate layer `L_{s,1}` is mineral oil dispensed second on silicone. Each layer is allowed `~2 min` for uniform spreading before impact. The compound droplet then impinges through `L_{s,1}` and `L_{s,2}` in sequence. Example successful parameters are `D_c ≈ 3.46 mm`, `H ≈ 14 cm → v ≈ 1.66 m/s`, `Wei ≈ 356`, `V_{Ls,2} = 90 μL (h ≈ 69 μm)`, `V_{Ls,1} = 120 μL`, and ethylene-glycol inner volumes `3.15 μL` and `13.30 μL` [2408.09026].

The impact dynamics are resolved by high-speed imaging at `6400 fps`, with the wrapping event completing within tens of milliseconds. For double-layer encapsulation, the reported sequence is first contact, interface draw-down, interfacial penetration, necking, pinch-off, and settling. The underlying mechanism is stated as a competition among compound-drop inertia, viscous dissipation in the interfacial layer, and restorative interfacial forces at the `L_s–L_h` interface. Stable shell retention is governed by a spreading condition
$$
S = \gamma_{L_{c,o}-h} - \gamma_{L_{c,o}-L_s} - \gamma_{L_s-h} > 0.
$$
For `laser oil + canola oil + water`, the paper reports `S = 19.17 mN/m > 0`; for `laser oil + oil-based ferrofluid + water`, it reports `S = 11.8 mN/m > 0`.

The regime map is parameterized by the impact Weber number
$$
Wei = \frac{\rho_{\mathrm{eff}} v^2 D_c}{\sigma_{c,o}},
$$
the viscosity ratio `λ = μLs / μc,o`, and the thickness ratio `δ^* = h/D_c`. The reported boundary is qualitative in form:
$$
Wei > We_{\mathrm{crit}}(\lambda,\delta^*),
$$
with `Wecrit increasing monotonically with both λ and δ*`. Example thresholds include `Wei ≈ 165` for canola oil at `α = 0.062` and `V_{Ls} = 120 μL`, and `Wei ≈ 611` for PDMS at `α = 3.41` and `V_{Ls} = 120 μL`. For silicone oil, the paper states that higher `Wei` is needed than for canola at similar `δ^*`, and reports penetration at `Wei ≈ 356 (H = 14 cm)`.

Verification of shell morphology is performed by fluorescence microscopy and confocal Z-stacks. In double-layer encapsulation, red fluorophore in the shell and green dye in the ethylene-glycol core produce orthographic reconstructions that confirm concentric layers and core-shell morphology. In TLE runs, selective blue dye in mineral oil or silicone oil is used to distinguish shell order, with the order more apparent when the outermost `L_{s,2}` is dyed. The reported qualitative outcome is that the encapsulating layers effectively protect the water-soluble ethylene-glycol core inside the water bath, with no visible leakage or dissolution during the observation window.

The study’s design guidelines are explicit. If penetration fails, one may modestly increase `H`, reduce `V_{Ls}`, or switch to a lower-viscosity oil. If air is consistently entrapped, one may slightly reduce `H`, increase `V_{Ls}`, or increase `μ_{Ls}`. The stated limitations are volumetric layer-thickness control, the high `Wei` demanded by very viscous or thick layers, and the absence of quantified long-term diffusion or leakage measurements [2408.09026].

## 4. TLE as graphene/MoS2/graphene encapsulation in electron microscopy

In the microscopy literature, TLE denotes the “sandwich” configuration `graphene/MoS2/graphene`, in which a monolayer MoS2 specimen is enclosed between two single-layer graphene sheets. The comparison set is `bare/pristine single-layer MoS2` and `MoS2 atop single-layer graphene`. The central result is that full graphene encapsulation changes the radiation-damage behavior of monolayer MoS2 during STEM/TEM, allowing high-dose chemical imaging that is not sustained in the other two configurations [1310.4012].

Fabrication begins from mechanically exfoliated MoS2 and graphene. Bare MoS2 is wet-transferred directly to a `Quantifoil M TEM grid`. For the supported configuration, MoS2 is transferred onto single-layer graphene on `Si/SiO2`, then the two-layer stack is wet-transferred to the grid. For TLE, two sequential transfers are used: first MoS2 onto graphene, then another graphene layer on top, after which the substrate is removed and the three-layer stack is transferred to the TEM grid. Each transfer is followed by an acetone dip, and after final transfer the sample is dried in a critical point dryer. No thermal annealing is reported. Electron diffraction verifies that all three layers are single-layer, and in the TLE specimen the two graphene layers are rotated by `≈23°` relative to each other.

The imaging conditions are fixed across the study: `Nion UltraSTEM100`, `60 keV`, near-UHV `<5×10^-9 torr`, convergence semi-angle `30 mrad`, probe current `~100 pA`, and estimated probe size `~1.1 Å`. Under these conditions, the paper reports for bare MoS2 a sulfur vacancy within `~30 s` at `~2.5×10^8 e/Å^2` during HAADF imaging, dynamic expansion of the defect to a `~2 nm` hole over `~65 s`, and severe perforation during EELS spectrum imaging at a total dose of `~2.6×10^10 e/Å^2`. For MoS2 on graphene, the paper reports no defects in one HAADF image at `~5.1×10^8 e/Å^2`, first damage at a total accumulated dose of `~2.3×10^8 e/Å^2` during dynamic scanning, and EELS failure at `~9.2×10^10 e/Å^2`. For TLE, the reported values are no defects up to `~4.5×10^8 e/Å^2` during dynamic scanning and no damage during EELS spectrum imaging at `~1.7×10^11 e/Å^2`, with both Mo and S sublattices resolved throughout.

The paper interprets protection primarily in terms of ionization-damage mitigation rather than basal-plane knock-on sputtering. It states `E_d(S) ≈ 6.5 eV` for sulfur displacement in pristine monolayer MoS2, whereas the maximum transferable energy to `^{32}S` is `~4.3 eV at 60 keV`, `~6.5 eV at ~80 keV`, and `~7.4 eV at 100 keV`. Since `T_max < E_d` at `60 keV`, ionization dominates. The proposed protective mechanisms are graphene’s high electrical conductivity, which dissipates accumulated charge; its high thermal conductivity, which aids heat dissipation; environmental isolation due to graphene impermeability; mechanical confinement of displaced species; and possible interlayer interactions that may enhance electron transport.

The reported quantitative improvements are `~6.5× higher dose than bare and ~1.8× higher than single-side graphene support for defect-free chemical mapping`, together with `at least a ~2× increase in dose-to-first-defect` in dynamic scanning relative to single-side support. The study also reports side effects: a `~15% loss in signal-to-background` with encapsulation, minor loss of sharpness, and local damage in one graphene cap due to contaminant-mediated etching, likely associated with `Si` and `SiO2` clusters introduced during the double transfer. The MoS2 remains intact beneath such cap damage. The authors explicitly suggest broader applicability to other beam-sensitive specimens, especially under low-voltage, near-UHV, high-dose spectroscopy conditions [1310.4012].

## 5. TLE as an oil-encapsulated bubble-in-water compound

A separate fluid-mechanical usage defines TLE as the oil-encapsulated bubble-in-water compound formed when a rising oil droplet spreads over and fully engulfs an air bubble within water. The three phases are water as the outer phase, oil as the middle phase, and air as the inner phase. The thermodynamic control parameter is the oil spreading coefficient
$$
S_o \equiv \gamma_{aw} - (\gamma_{ao} + \gamma_{ow}),
$$
and total encapsulation requires `S_o > 0` [2505.18339].

The process is described as four stages: `collision/film drainage`, `encapsulation`, `reshaping`, and `compound rising`. In the base equal-size case `R_d = R_b ≈ 0.4 mm`, the paper reports `t_col ≈ 20 ms`, nearly instantaneous film drainage in simulation but `≈4.8 ms` in experiment, and encapsulation time `t_enc ≈ 6.35 ms`. During encapsulation, the droplet spreads over the bubble, capillary waves propagate along the three-phase lines, and the spreading fronts converge beneath the bubble. After full coverage, reshaping proceeds through exchange of kinetic and capillary energies, after which the compound rises toward terminal velocity.

The reported low-viscosity neck-growth laws for `Oh_s < 0.1` are
$$
\frac{r_n(t)}{R_0} = 1.23 \left(\frac{t}{T_{ci}}\right)^{0.44} \quad \text{for } D_a/D_b \ge 1,
$$
and
$$
\frac{r_n(t)}{R_0} = 1.30 \left(\frac{t}{T_{ci}}\right)^{0.50} \quad \text{for } D_a/D_b < 1.
$$
The paper states that the exponent depends on size ratio and is consistent with related bubble-crossing and coalescence literature.

Encapsulation-time scaling is split into spherical and deformed regimes. In the spherical regime `Bo < ~0.11`, the spreading speed scales as
$$
U_{sp} \approx U_{vc} \equiv \frac{S_o}{\mu_1 + \mu_2},
$$
and the encapsulation timescale is
$$
t_{vc} \approx \frac{(\mu_1 + \mu_2)R_b}{S_o}.
$$
In the deformed regime `0.11 \lesssim Bo \lesssim 2.2`, the reported capillary-gravitational scale is
$$
t_{enc} \approx t_{cg} \equiv \frac{t_{ic}}{1 + Bo_s}.
$$
The paper further reports an exponential increase of encapsulation time with viscosity ratio:
`t_enc(v*) ≈ 5.3849 exp(0.0192 v*) ms` for `v* ≡ ν2/ν1`, and
`t_enc(v**) ≈ 6.3996 exp(0.1032 v**) ms` for `v** ≡ ν1/ν2`.

The role of `S_o` is quantified directly. With other properties fixed, the study reports `S_o ≈ 0.29 mN/m → t_enc ≈ 7.30 ms` and `S_o ≈ 10.3 mN/m → t_enc ≈ 3.17 ms`. As `S_o → 0^+`, the spreading and reshaping stages overlap, bubble aspect ratio oscillates, and repeated saddle points appear in the velocity history. Size ratio also matters: for `R_a/R_b < 1`, smaller droplets can show faster early neck growth but longer overall `t_enc` because a larger bubble surface must be covered; for `R_a/R_b > 1`, `t_enc` approaches a plateau. Off-center impacts are reported to yield a non-monotonic `t_enc(B)` with a minimum around `B ≈ 0.5`, while `B ≳ 0.75` in deformed regimes may prevent contact.

The simulations use a `phase-field multiple-relaxation-time LBM (PF MRT-LBM), D2Q9`, with a triple-well free-energy functional and interface thickness `≈4 l.u.`. High-speed imaging at `2000–4000 fps` supports validation. The stated limitations are omission of explicit `DLVO/non-DLVO` forces in film drainage, effectively 2D or axisymmetric modeling, constant interfacial tensions without dynamic surfactant effects, and finite ranges in `Bo` and `Oh_s`. The paper presents the framework as relevant to `gas flotation` and `interfacial microfluidics` [2505.18339].

## 6. TLE as a three-level polarization-state construction in plasmonic photonics

In plasmonic photonics, TLE denotes neither shells nor layered matter, but a three-level quantum-state construction based on polarization channels of transmitted light. The physical platform is a cross-shaped nano-antenna or a biperiodic hole array that supports two orthogonal surface-wave channels. Their coherent superposition yields three distinguishable polarization classes associated with the Stokes axes: linear along structural axes, diagonal linear, and circular polarization. The paper uses these as qutrit labels `|0⟩ ≡ |S_1⟩`, `|1⟩ ≡ |S_2⟩`, and `|2⟩ ≡ |S_3⟩` [1806.00619].

The field model is
$$
E_{\text{tot}} = E_x e^{i\beta_x L_x} + E_y e^{i\beta_y L_y}, \qquad
\Delta\phi = \beta_x L_x - \beta_y L_y,
$$
with Jones operator
$$
J_{\text{dev}}=
\begin{pmatrix}
t_x e^{i\beta_x L_x} & 0\\
0 & t_y e^{i\beta_y L_y}
\end{pmatrix},
\qquad
J_{\text{out}} = J_{\text{dev}}J_{\text{in}}.
$$
The Stokes parameters are computed from the transmitted fields:
$$
S_0 = |E_x|^2 + |E_y|^2,\quad
S_1 = |E_x|^2 - |E_y|^2,\quad
S_2 = 2\,\mathrm{Re}(E_xE_y^*),\quad
S_3 = 2\,\mathrm{Im}(E_xE_y^*).
$$
Circular polarization arises when amplitudes are equal and the phase difference is `±\pi/2`. The rotating in-plane dipole is written as
$$
\mathbf{p}(t)=p_x\cos(\omega t)\,\hat{\mathbf{x}} + p_y\cos(\omega t+\Delta\phi)\,\hat{\mathbf{y}},
$$
with a circular trajectory for `p_x = p_y` and `\Delta\phi = \pm\pi/2`.

The encapsulated three-level state is parameterized as
$$
|\psi(\alpha,\lambda)\rangle
=
\alpha_0(\alpha,\lambda)|0\rangle
+
\alpha_1(\alpha,\lambda)|1\rangle
+
\alpha_2(\alpha,\lambda)|2\rangle,
$$
where
$$
\alpha_0 \equiv S_1,\quad
\alpha_1 \equiv S_2,\quad
\alpha_2 \equiv S_3,\quad
|\alpha_0|^2+|\alpha_1|^2+|\alpha_2|^2=1.
$$
The paper then “encapsulates” the discrete qutrit into a continuous orthonormal set indexed by input polarization angle and wavelength,
$$
\langle \psi(\alpha',\lambda')|\psi(\alpha,\lambda)\rangle
=
\delta_{\alpha',\alpha}\,\delta(\lambda'-\lambda),
$$
idealized for spectrally separated bands. A convenient parameterization is given as
$$
|\psi\rangle = e^{i\gamma}\big(\cos\theta\,|0\rangle + \sin\theta\cos\phi\,|1\rangle + \sin\theta\sin\phi\,|2\rangle\big).
$$

Device tuning is controlled by geometry, wavelength, and input polarization. For the biperiodic array, the surface-plasmon wave number is
$$
k_{\mathrm{spp}}=\frac{2\pi}{\lambda}\,\mathrm{Re}\left[\sqrt{\frac{\varepsilon_m\varepsilon_d}{\varepsilon_m+\varepsilon_d}}\right],
$$
and the phase condition is approximated by `\Delta\phi \approx k_{\mathrm{spp}}(P_x - P_y)`. Equal-amplitude launching is associated with an optimal input polarization `\alpha ≈ 46.5°` analytically, `~47°` numerically, and `~43°` experimentally in one design. Reported device examples include cross-arm lengths `L_x ≈ 80 nm` and `L_y ≈ 95 nm`, giving resonances at `λ ≈ 820 nm` and `λ ≈ 890 nm`, with near-circular radiation at `λ ≈ 850 nm`; and biperiodic arrays with `P_x ≈ 368–394 nm`, `P_y ≈ 407 nm`, target `λ = 700 nm`, and `S_3 ≈ 1` after correction of waveplate phase errors.

The paper extends the framework to two-photon entanglement by writing states of the form
$$
|\Psi\rangle = \sum_{i,j=0}^2 c_{ij}|i\rangle\otimes|j\rangle,
\qquad
\sum_{i,j}|c_{ij}|^2 = 1,
$$
with joint detection probabilities
$$
P_{k\ell} = \langle \Psi|\,\Pi_k^{(A)}\otimes\Pi_\ell^{(B)}\,|\Psi\rangle.
$$
It also makes a nonstandard claim that a single photon carries no spin and that circular polarization requires at least two photons. The same source explicitly notes that this departs from standard quantum electrodynamics, which assigns definite helicity `±1` to single photons. The status of that claim is therefore controversial within the paper’s own framing and should be read as the author’s inference rather than as established consensus [1806.00619].

## 7. Comparative perspective and terminological cautions

Across these literatures, the term TLE consistently marks a three-part organization, but the underlying objects differ categorically. In the software-engineering usage, the three levels are logical and schema-bound: table, column, and cell-bitmask. In liquid encapsulation, the three levels are radial shells around a core. In MoS2 microscopy, the three levels are material layers in a heterostructure. In bubble encapsulation, the three levels are outer, middle, and inner phases. In plasmonic photonics, the three levels are state labels associated with orthogonal Stokes channels [2510.00002] [2408.09026] [1310.4012] [2505.18339] [1806.00619].

This suggests a limited but useful cross-domain abstraction: TLE is repeatedly invoked when a problem is reformulated around a bounded ternary structure rather than an unbounded continuum of relations, layers, or states. Beyond that abstraction, the methods are not interchangeable. Bitmask operations, Weber-number regime maps, dose-tolerance metrics, spreading coefficients, and Stokes-parameter qutrit encodings belong to different mathematical and experimental frameworks.

A second caution concerns validation. The software-engineering TLE is supported by CSP/FDR verification, formal complexity theorems, and enterprise deployment. The droplet and bubble TLE variants are supported by regime maps, high-speed imaging, fluorescence, and lattice Boltzmann modeling. The graphene-encapsulation TLE is supported by atomic-resolution STEM/EELS measurements under controlled beam conditions. The photonic TLE is supported by superposition modeling, device simulations, and polarization measurements, but also contains explicitly nonstandard theoretical inferences. The name therefore carries no uniform evidentiary standard across fields.

The resulting encyclopedic conclusion is not that TLE denotes one mature doctrine, but that it is a recurrent label for technically precise three-level constructions whose meaning must be fixed by disciplinary context before any substantive interpretation is attempted.

Source: https://www.emergentmind.com/topics/three-level-encapsulation-tle