MK2: Multifaceted Contexts in Research

• MK2 is a context-dependent label used across fields such as thermoelectrics, patent-based idea generation, nanophotonics, and single-pion production, with its meaning derived from local disciplinary definitions.
• In thermoelectrics, MK2 represents the power factor (S²/ρ) and is associated with metal-like transport behavior influenced by defect-mediated scattering and specific Seebeck and resistivity values.
• Across advanced applications, MK2 denotes tailored system and design parameters—optimizing prompt-centric ideation, nanolens focusing capabilities, and pion production modeling via precise engineering metrics.

MK2 is a context-dependent label rather than a single cross-disciplinary technical concept. In recent arXiv literature, it appears in at least four distinct senses: as shorthand for the thermoelectric power factor in a Ru-based Heusler alloy; as the name of a prompt-centric system for patent-based idea generation; as the identifier of a second silicon-rich nitride metalens design; and as part of the lineage of a single-pion production model in lepton–nucleon scattering (Mondal et al., 2022, Xu et al., 11 Jul 2025, Khalilian et al., 2024, Kabirnezhad, 2022). This diversity of usage means that the term acquires its meaning entirely from local disciplinary context.

1. Context-dependent uses of MK2

Across the cited works, MK2 serves different semantic roles: a unit-bearing shorthand, a system name, a device-design label, and a model-designation marker. The term therefore does not denote a uniform parameter, architecture, or physical observable across fields (Mondal et al., 2022, Xu et al., 11 Jul 2025, Khalilian et al., 2024, Kabirnezhad, 2022).

Domain	Meaning of MK2	Representative technical markers
Thermoelectrics	Shorthand for power factor $PF$	$PF = S^2/\rho$; $\sim 1$ mW m$^{-1}$ K$^{-2}$ at 300 K; $\sim 2$ mW m$^{-1}$ K$^{-2}$ at 390 K
Patent-based idea generation	Prompt-centric pipeline name	Gemini 2.5 prompt engineer; GPT-4.1 generator; Qwen3-8B Elo judge
Nanophotonics	Second SRN metalens design	$NA = 0.99$; efficiency $42\%$ or $PF = S^2/\rho$0; $PF = S^2/\rho$1
Single-pion production	Earlier model lineage within MK / MK2 framework	Low-$PF = S^2/\rho$2, low-$PF = S^2/\rho$3 resonance-region model later extended

A plausible implication is that technical interpretation of MK2 requires immediate specification of the target literature, since even adjacent fields use the label for fundamentally different entities.

2. MK2 as thermoelectric power factor in Ru$PF = S^2/\rho$4TiGe

In the Ru$PF = S^2/\rho$5TiGe study, “MK2” is explicitly not a separate material parameter; it is shorthand for the thermoelectric power factor, reported in the usual units of mW m$PF = S^2/\rho$6 K$PF = S^2/\rho$7, sometimes written as mW/mK$PF = S^2/\rho$8 in the text and figure caption (Mondal et al., 2022). The defining relation is

$PF = S^2/\rho$9

with the figure of merit given by

$\sim 1$0

In this usage, MK2 is therefore the numerator of $\sim 1$1 without the thermal-conductivity term.

The reported thermoelectric values are specific. The Seebeck coefficient is positive over the whole range $\sim 1$2–$\sim 1$3 K, with $\sim 1$4 and $\sim 1$5; $\sim 1$6 exhibits a hump-like maximum at about $\sim 1$7 K and becomes roughly linear above $\sim 1$8 K. The electrical resistivity is $\sim 1$9 and $^{-1}$0, yielding a residual resistivity ratio

$^{-1}$1

The transport is described as metal-like, with the low $^{-1}$2 indicating behavior more characteristic of a bad metal / semimetal than a conventional semiconductor.

Within this framework, the power factor increases monotonically with temperature, reaching $^{-1}$3 mW m$^{-1}$4 K$^{-1}$5 at $^{-1}$6 K and $^{-1}$7 mW m$^{-1}$8 K$^{-1}$9 at $^{-2}$0 K. Thermal conductivity shows a sharp peak near $^{-2}$1 K and falls to $^{-2}$2 W m$^{-2}$3 K$^{-2}$4 at $^{-2}$5 K; the lattice thermal conductivity is dominant, with $^{-2}$6 W m$^{-2}$7 K$^{-2}$8. The corresponding $^{-2}$9 values are $\sim 2$0 at $\sim 2$1 K and $\sim 2$2 at $\sim 2$3 K.

The interpretation given in the paper links these results to an unusual combination of metal-like resistivity, large positive thermopower, and relatively low thermal conductivity. The positive $\sim 2$4 is taken to indicate hole-dominated transport, while the comparatively low $\sim 2$5 is associated with heavy Ru and Ge atoms that suppress phonon transport. Magnetic measurements further show superparamagnetically interacting clusters below $\sim 2$6 K, attributed to antisite disorder and small structural deviations; these defects are also invoked to explain the low-temperature resistivity anomaly and possible Kondo-like scattering. In this domain, then, MK2 is a transport-performance quantity embedded in a broader defect-mediated thermoelectric picture.

3. MK2 as a prompt-centric PBIG system

In the Patent-Based Idea Generation shared task, MK2 designates a deliberately lightweight, prompt-centric system rather than a physical parameter or device (Xu et al., 11 Jul 2025). The task setting is tightly specified: given $\sim 2$7 real patents, $\sim 2$8 each from NLP, Computer Science, and Materials Chemistry, the system must generate exactly one product idea per patent that plausibly could reach the market within three years. Each submission is a JSON object with four short fields—title, product_description, implementation, and differentiation—and evaluation is pairwise via an Elo-style ranking by both automated LLM judges and human experts on six criteria: technical validity, innovativeness, specificity, need validity, market size, and competitive advantage.

MK2 treats this as a prompt-engineering problem rather than a training problem. Gemini 2.5 functions as the prompt engineer and iterative editor; GPT-4.1 is the final generation model; and Qwen3-8B is the inexpensive judge used in an internal Elo loop to select the best prompt. The pipeline begins with prompt drafts from the official PBIG guidelines by multiple models. Gemini 2.5 then inspects outputs from weaker prompts, identifies useful parts, and grafts those fragments into the strongest prompt. GPT-4.1 uses the resulting prompt to generate the actual ideas, and the internal leaderboard compares candidate prompts pairwise across all six criteria in one step rather than separately per criterion. To reduce position bias, the two outputs in each comparison are swapped in $\sim 2$9 of cases, and GPT-4.1-mini serves as a fast preliminary gatekeeper before outputs are sent to the leaderboard.

The prompt is treated as the main artifact. It instructs the model to behave like an expert business strategist and product-innovation analyst, to identify the single most unique and non-obvious mechanism in the patent, to treat that mechanism as indispensable, and to build exactly one commercially viable concept around it. The design also emphasizes direct optimization for the competition criteria, strict output format, and character limits. A notable engineering observation is that long system prompts made it harder for models to obey those limits; post-editing often harmed score quality, while simple truncation was less damaging. The final prompt therefore restated the character limit again at the end of the user prompt, close to the generation point.

The reported results are strong in the competition’s automatic evaluation. MK2 topped the automatic leaderboard overall and won $^{-1}$0 of $^{-1}$1 tests. In the detailed description, it achieved the top automatic scores in all three domains, and human judges favored it in NLP and CS. In NLP, MK2 obtained the highest scores in five of six criteria under both automatic and human evaluation, with market size as the main weaker spot. In CS, it took the top automatic score in all six criteria and remained strong under human evaluation, though specificity and market size were less robust. Materials Chemistry is the main limitation: automatic judging remained favorable, but human judges did not place it first on any criterion. The authors interpret this mismatch as evidence that automatic metrics can miss the scientific rigor or plausibility demanded by domain experts, suggesting stronger domain grounding and hybrid evaluation as future directions.

4. Mk2 as an SRN polarization-based metalens

In nanophotonics, Mk2 is the second silicon-rich nitride polarization-based metalens design in a study of high-numerical-aperture flat optics (Khalilian et al., 2024). It was created as the higher-NA demonstration relative to Mk1, specifically to push numerical aperture toward the limit in air and to show how reduced pitch size can improve efficiency while retaining a compact, CMOS-compatible platform. The device operates at $^{-1}$2 nm under P-polarized illumination and was simulated using the finite-difference time-domain method with PML boundaries.

The paper attributes to Mk2 a numerical aperture of $^{-1}$3, focusing efficiency of $^{-1}$4, with the comparison table reporting $^{-1}$5, and $^{-1}$6. Its focal length is $^{-1}$7 at $^{-1}$8 nm. The comparison lens, Mk1, has $^{-1}$9, efficiency $^{-2}$0, $^{-2}$1, and focal length $^{-2}$2 at the same wavelength. Mk2 therefore represents the more extreme focusing case: tighter focal spot and higher angular capability at the cost of reduced efficiency.

The phase synthesis follows the standard hyperbolic phase profile

$^{-2}$3

where $^{-2}$4 is the focal length. The shorter $^{-2}$5 used in Mk2 is consistent with its stronger focusing curvature and higher $^{-2}$6. The paper defines focusing efficiency by integrating the Poynting vector over the incident area and over a focal region spanning three times the $^{-2}$7 of the intensity distribution, so the reported $^{-2}$8 or $^{-2}$9 measures the fraction of incident power concentrated near the focus.

Pitch size is the central design variable. The authors argue that reducing pitch decreases scattering area and phase-profile discretization error, so the metasurface better approximates an ideal continuous lens. This is formalized with a Rayleigh-Gans-style criterion,

$NA = 0.99$0

and with the optical-path-difference expression

$NA = 0.99$1

The paper further includes

$NA = 0.99$2

and reports a pitch size limit of $NA = 0.99$3 nm, together with a fill ratio constrained between $NA = 0.99$4 and $NA = 0.99$5 to achieve the required $NA = 0.99$6 delay.

The enabling material is silicon-rich nitride. At $NA = 0.99$7 nm, the measured refractive index is $NA = 0.99$8. The SRN layer was deposited by PECVD with $NA = 0.99$9 SiH$42\%$0, $42\%$1 NH$42\%$2, $42\%$3, $42\%$4 torr, and $42\%$5 nm thickness. The paper states that the $42\%$6 nm thickness is $42\%$7 times the minimum height needed for effective phase-delay control across the full $42\%$8 range. In this context, Mk2 is a device label that encapsulates a specific high-NA design point and an associated pitch-optimization argument.

5. MK2 in single-pion production model lineage

In lepton–nucleon scattering, MK2 refers to an earlier single-pion production model that is cited and extended in later work on electron–proton interactions (Kabirnezhad, 2022). The extension targets the transition region between resonance and DIS, retaining the MK structure of resonant production, non-resonant background, and their interference, but replacing low-$42\%$9 descriptions with parameterizations that satisfy pQCD asymptotics, are guided by vector meson dominance, and are constrained by a broader set of exclusive electron–proton data.

The paper states that the original MK/MK2 model was designed mainly for the resonance region, with vector form factors tuned to electron-scattering data at relatively modest kinematics, roughly $PF = S^2/\rho$00 and $PF = S^2/\rho$01 concentrated below the second resonance region. The updated formulation extends the model to higher $PF = S^2/\rho$02 using VMD form factors consistent with QCD via quark-hadron duality, and to higher $PF = S^2/\rho$03 using Regge phenomenology for the non-resonant background. The hadronic vector-current helicity amplitude is written as

$PF = S^2/\rho$04

and the electroproduction cross section is expressed through the usual decomposition into $PF = S^2/\rho$05, $PF = S^2/\rho$06, $PF = S^2/\rho$07, $PF = S^2/\rho$08, and $PF = S^2/\rho$09, multiplied by the virtual-photon flux $PF = S^2/\rho$10.

The resonant-sector update introduces VMD-based transition form factors such as

$PF = S^2/\rho$11

and

$PF = S^2/\rho$12

with logarithmic renormalization factors enforcing the asymptotic pQCD behavior. Superconvergence constraints such as

$PF = S^2/\rho$13

ensure the required high-$PF = S^2/\rho$14 suppression. For the non-resonant background, dipole vector form factors are replaced by VMD-based nucleon form factors, and low-energy $PF = S^2/\rho$15-channel propagators are Reggeized:

$PF = S^2/\rho$16

$PF = S^2/\rho$17

with a smooth interpolation

$PF = S^2/\rho$18

The parameter extraction uses all available CLAS exclusive data in the ranges

$PF = S^2/\rho$19

for the channels $PF = S^2/\rho$20 and $PF = S^2/\rho$21. The fit involves $PF = S^2/\rho$22 free parameters, optimized with MINUIT2 MIGRAD, and reports

$PF = S^2/\rho$23

Uncertainties are propagated by randomly throwing all fitted parameters within their correlated post-fit covariance matrix and taking the envelope containing the central $PF = S^2/\rho$24 of predictions. In this literature, then, MK2 does not label the new extension itself; it labels the predecessor whose limited low-$PF = S^2/\rho$25, low-$PF = S^2/\rho$26 scope motivates the extension.

6. Nomenclature, misconceptions, and nearby acronyms

A recurrent source of confusion is that MK2 can appear to be a generic technical shorthand, whereas the cited works show that its meaning is entirely local. In the Ru$PF = S^2/\rho$27TiGe paper, the point is explicit: “MK2” is not a separate material parameter, but shorthand for the thermoelectric power factor expressed in mW m$PF = S^2/\rho$28 K$PF = S^2/\rho$29 (Mondal et al., 2022). In the PBIG paper, it is a system name. In the metalens paper, it is a design identifier. In the pion-production paper, it is part of a model lineage.

Another nearby source of confusion is orthographic similarity to MKID. The CCAT-prime instrumentation paper concerns microwave kinetic inductance detectors and a 280 GHz instrument module, including more than $PF = S^2/\rho$30 feedhorn-coupled, polarization-sensitive MKIDs across $PF = S^2/\rho$31 networks, operated within cryogenic stages at $PF = S^2/\rho$32 K, $PF = S^2/\rho$33 K, $PF = S^2/\rho$34 K, and $PF = S^2/\rho$35 mK (Vavagiakis et al., 2022). That terminology is unrelated to any of the MK2 usages above. This suggests that acronym-level similarity is not a reliable guide to conceptual relatedness.

Taken together, these usages show that MK2 functions as a contextual label attached to different epistemic objects: a thermoelectric performance measure, a prompt-optimization pipeline, a metasurface design instance, and a hadronic-interaction model family. The commonality lies only in nomenclature. The technical content resides in the surrounding domain-specific definitions, equations, and evaluation protocols.