update

Author: yjw0510

WRITING_POLICY.md

@@ -0,0 +1,161 @@
+# Yu Lab Website Writing Policy
+
+This policy applies to every reader-facing sentence on the site, including metadata,
+captions, accessibility labels, news, recruitment copy, and Korean localization.
+
+## 1. Reader and author
+
+- Write for prospective graduate students and public-sector readers who do not yet
+  know computational chemistry.
+- Write as a computational chemist. Lead with the question and the modeling choice;
+  use an application system as evidence, not as a claim of domain authority.
+- English is the source text. Korean is a fresh rendering of the same meaning, not a
+  sentence-by-sentence translation.
+
+## 2. House voice
+
+The reference style is compact, factual, and result-led. A research paragraph should
+normally move through three functions:
+
+1. Name the material behavior or scientific problem.
+2. State the computational approach in plain language, defining the one term the
+   reader needs.
+3. Give the finding and why it changes what can be predicted or designed.
+
+Use concrete verbs such as `compute`, `compare`, `trace`, `resolve`, and `predict`.
+Prefer one claim per sentence. End when the finding lands; do not add a generic promise
+about future impact.
+
+State the claim directly. Avoid rhetorical antithesis such as `not X but Y`, `rather
+than`, `instead of`, `A가 아니라 B`, `뿐 아니라`, and concessive `~지만` framing.
+Scientific scope belongs in the subject or domain of the claim:
+
+- Write “For the salts and models examined, the results follow one comparison curve.”
+- Write “Training coverage defines the chemistry represented by the MLFF.”
+- Write “The trajectories measure reduced mobility; the model attributes it to
+  confinement resistance.”
+
+Keep functional negation when it conveys an actual state or instruction, such as an
+eligibility rule, empty result, loading error, or genuinely unoccupied orbital.
+
+Avoid mission-language and AI boilerplate: `delve`, `robust`, `comprehensive`,
+`innovative`, `transformative`, `leverage`, `foster`, `enhance`, `pivotal`,
+`multifaceted`, `crucial`, `vital`, `furthermore`, `moreover`, and `it is worth noting`.
+Delete claims such as “X is important” unless the sentence states what X changes.
+
+## 3. Depth and length by section
+
+### Home
+
+- Purpose: identify the lab, its method-centered identity, and the class of questions it
+  answers.
+- Hero description: 18–30 English words.
+- Section subtitle: no more than 24 English words.
+- Specimen captions may identify the modeling scale, system, and one size cue. Do not
+  list software, ensembles, temperatures, runtimes, electronic-structure settings, or
+  other protocol details.
+
+### Multiscale
+
+- Purpose: teach why several representations are needed and what each can answer.
+- Interactive `question`: 12–28 English words.
+- Interactive `concept`: 45–80 English words.
+- Interactive `takeaway`: 12–30 English words.
+- Method-page body: 220–320 English words, excluding publication citations.
+- Explain the tradeoff between resolution and reach. Do not teach basis sets, levels of
+  theory, thermostat or barostat choices, integration schemes, software packages, or
+  parameter recipes.
+
+### Research Topics
+
+- Purpose: show what the lab has actually studied, why, how, and what the papers found.
+- Use the paper corpus as the evidence boundary.
+- Keep each description near 45–55% of its August 2026 pre-edit length:
+
+| Topic | English target |
+| --- | ---: |
+| Polymer-directed self-assembly | 215–255 words |
+| Aqueous chemistry with ML force fields | 205–240 words |
+| Glass dynamics and microrheology | 150–185 words |
+| Colloidal structure, transport, and optics | 220–260 words |
+| Multiscale modeling of hydrogels | 130–165 words |
+| Collaborative projects | 65–90 words |
+| Research directions under development | 125–160 words |
+
+- Two paragraphs are usually enough for a multi-paper topic. Use a short list only when
+  the reader is genuinely choosing among independent projects.
+- One informative number may remain when it changes the interpretation. A chain of
+  temperatures, exponents, sample counts, rates, or dimensions does not belong here.
+- Define the evidence boundary in positive terms. Name the systems, models, observables,
+  and training domain covered by the claim. A matching exponent supports a proposed
+  mechanism; a direct mechanism test requires its own evidence.
+
+### Publications, People, News, Funding, and Contact
+
+- Publication summary: 25–45 English words; method plus principal finding.
+- PI biography, if shown: 110–150 English words; research identity before chronology.
+- News body: 80–140 English words unless the announcement needs instructions.
+- Recruitment instructions: direct, welcoming, and specific. State explicitly when
+  beginners are welcome.
+- Official paper titles, grant names, affiliations, dates, and bibliographic metadata are
+  records, not prose; preserve them exactly.
+
+## 4. Terms for non-specialists
+
+Do not use a specialist term before explaining it. Define it in the sentence where it
+first matters, then use the short form:
+
+- `density functional theory (DFT)`: a quantum-mechanical calculation of how electrons
+  are distributed.
+- `machine-learning force field (MLFF)`: a model trained on quantum calculations to
+  predict atomic energies and forces.
+- `molecular dynamics (MD)`: a simulation that follows molecular motion over time.
+- `coarse-graining`: grouping several atoms into one interaction site to reach larger
+  systems and longer collective motion.
+- `active microrheology`: reading local mechanics from the force on a small driven probe.
+
+Definitions should serve the immediate claim. Do not turn them into textbook asides.
+
+## 5. Korean localization
+
+- Rebuild the sentence around natural Korean information order. Do not mirror English
+  subjects, passives, or article structure.
+- Keep standard acronyms and names such as DFT and MLFF in English. On first use, pair an
+  English technical term with a short Korean explanation when the acronym alone would
+  exclude the reader.
+- Prefer direct verbs over nominalized chains. Remove `이를 통해`, `나아가`, `종합적으로`,
+  `~라고 할 수 있다`, and `~에 기여할 수 있다` unless they carry real logical content.
+- Research and explanatory copy uses concise `-다` style. Recruitment may address the
+  reader directly and politely.
+- Preserve chemical formulas, proper names, paper titles, and protected technical tokens.
+
+## 6. Evidence and review gates
+
+Every substantive sentence must pass all of these checks:
+
+1. **Purpose:** Does it belong in this section at this depth?
+2. **Actor or claim:** Does someone do something, or could the claim be tested and found
+   wrong?
+3. **Evidence:** Is it supported by the provided paper, an official record, or the actual
+   visualization source?
+4. **Definition:** Can a non-specialist understand every necessary term from context?
+5. **Deletion:** If the sentence disappears, does the paragraph lose information? If not,
+   delete it.
+6. **Balance:** Does the sentence make its claim and evidence boundary directly, without
+   promotional overreach or defensive retreat?
+7. **Specificity:** Is it concrete without drifting into protocol trivia or number lists?
+8. **Parallel meaning:** Do English and Korean make the same claim without sounding like
+   translations of one another?
+
+## 7. Final polish workflow
+
+1. Finish and fact-check the English source copy.
+2. Write the Korean version from the settled meaning.
+3. Run the English copy through the site’s human-writing review rules.
+4. Run only the Korean prose through the pinned `im-not-ai` v2.3.0 workflow at commit
+   `82137e858763dadb99561f194c5c00465735017b`. The tool is Korean-only; never feed it the
+   English source or live TypeScript/JSON files.
+5. Protect acronyms, formulas, names, and headings in the staging document. Review the
+   generated diff by hand and re-integrate only changes that preserve scientific meaning.
+6. Re-run word budgets, terminology checks, type checks, tests, build, and bilingual
+   visual review.

content/multiscale/allatom.mdx

@@ -1,27 +1,28 @@
 ---
 title: "All-Atom Molecular Dynamics"
 titleKo: "전원자 분자동역학"
-shortDescription: "A single arrangement of atoms defines neither diffusion nor a hydrogen-bond lifetime. Fixing the chemistry buys the millions of steps those averages are made of."
-shortDescriptionKo: "확산 계수도 수소결합 수명도 수백만 스텝을 평균해서 얻는 값이다. 화학을 고정해 힘 계산을 싸게 만들어야 그 스텝이 나온다."
+shortDescription: "All-atom molecular dynamics follows every atom with fixed interaction rules, reaching the trajectories needed to study structure, transport, and relaxation."
+shortDescriptionKo: "전원자 분자동역학은 정해 둔 상호작용 규칙으로 모든 원자의 움직임을 따라가며 구조와 이동, 이완을 통계로 읽는다."
 icon: "Atom"
 color: "#06b6d4"
 order: 3
 scale: "1-100 nm"
 contentKo: |
-  ## 개요
+  ## 분자의 움직임을 원자 단위로 본다
 
-  전원자 분자동역학은 시스템 내 모든 원자(물 분자, 이온, 결합 모두)를 추적하며 물리적 힘, 즉 결합 상호작용, 전기적 인력과 척력, 약한 반데르발스 인력 아래에서 원자들이 어떻게 이동하는지 계산한다. 원자를 단순화된 비드로 묶는 조대화 시뮬레이션과 달리, 전원자 모형은 과냉각 액체의 국소 구조 비균일성, 고분자 질서화의 동역학, 물 속 이온 용해의 단계별 메커니즘 등을 분해하는 데 필요한 완전한 분자 세부 구조를 보존한다.
+  전원자 분자동역학(MD)은 모든 원자의 움직임을 시간에 따라 계산한다. 결합한 원자와 서로 가까운 원자의 상호작용을 미리 정한 수학식인 고전 역장으로 각 스텝의 힘을 구한다. 원자의 종류와 국소 분자 구조, 용매, 이온을 표현한 긴 궤적이 쌓인다.
 
-  ## 현재 연구
+  ## 궤적 전체에서 물성을 구한다
 
-  - **유리의 능동 미시유변학**: 모형 금속 유리를 관통하는 단일 나노 규모 탐침의 마찰력 측정. 이 마찰력이 유리 전이를 반영하며 벌크 측정으로는 볼 수 없는 물질 내부의 국소 역학적 차이를 드러냄
-  - **블록 공중합체 질서화 경로**: 무질서 고분자 용융체가 육방 원통 배열로 조직되는 과정 추적. 원통이 형성되었으나 정렬되지 않은 일시적 중간 상태 포함
-  - **이온 용해 메커니즘**: Al³⁺ 같은 금속 이온의 배위 껍질에 물 분자가 협동적으로 진입하는 과정 추적. 접근 순서와 기하가 용해 경로를 결정함을 규명
-  - **조대화 모형의 전원자 검증**: 전원자 궤적에서 얻은 구조·동역학 참조 측정량이 단순화된 조대화 모형의 교정 기준 제공
+  궤적의 여러 시점에서 같은 관측량을 계산하면 확산 계수와 이완 시간, 값의 분포를 얻는다. 전원자 MD는 액체 구조, 분자 이동, 고분자 질서화, 주변의 변형 반응을 연구하는 데 적합하다. 결합 관계가 고정된 고전 역장은 구조와 수송을 다룬다. 결합이 바뀌는 반응은 반응형 역장이나 전자 계산으로 연구한다.
 
-  ## 방법론
+  ## 연구실에서 쓰는 방식
 
-  멀티스케일 인터랙티브 경험은 물에 녹인 카페인의 명시적 용매 궤적으로 전원자 MD의 핵심 개념을 예시한다. 원자 표현 방식, 역장 항, 국소 분자 구조의 차이, 주기 경계 조건을 차례로 보여 준다. 이 궤적은 하나의 학습용 예시이며, 핵심 개념은 액체와 용액의 다양한 전원자 시뮬레이션에 적용된다.
+  모형 유리에서는 작은 탐침을 끌 때 받는 힘으로 주변의 역학을 읽는다. 이를 능동 미시유변학이라 한다. 힘의 변화는 원자가 이웃이 만든 임시 우리 안에서 움직이는 빠른 과정과 그 우리를 벗어나는 느린 구조 이완을 나눠 보여 준다. 블록 공중합체는 서로 다른 화학 성분을 이어 만든 고분자다. 시뮬레이션에서 드물게 나타나는 구조까지 찾도록 설계한 방법을 이용하면 원통들이 정렬되는 도중 방향이 어긋난 원통망이 잠시 머무는 과정을 볼 수 있다. 두 연구는 궤적에서 구조가 생기는 순서를 해석한다.
+
+  ## 더 단순한 모형의 기준
+
+  전원자 결과는 여러 원자를 한 단위로 묶는 조대화 모형이 맞춰야 할 기준도 제공한다. 구조 분포와 이완 거동, 각 값의 불확실성을 함께 넘기면 단순한 모형이 재현할 목표와 작동 범위가 정해진다.
 
   ## 관련 논문
 
@@ -31,20 +32,21 @@ contentKo: |
   - Kim et al., *J. Phys. Chem. Lett.* (2024), Al³⁺ 용해
 ---
 
-## Overview
+## Following Molecular Motion
+
+All-atom molecular dynamics (MD) follows every atom over time. A classical force field supplies fixed mathematical rules for interactions between bonded atoms and nearby nonbonded atoms. This representation preserves chemical identity and local molecular geometry. Each inexpensive step retains solvent and ions, allowing long trajectories to accumulate.
+
+## Properties from a Full Trajectory
 
-All-atom molecular dynamics tracks every atom in a system, including every water molecule, every ion, and every bond, and calculates how they move under physical forces: bonding interactions, electrical attractions and repulsions, and weak van der Waals attractions. Unlike coarse-grained simulations that group atoms into simplified beads, all-atom models preserve the full molecular detail needed to resolve phenomena such as local structural heterogeneity in supercooled liquids, the kinetics of polymer ordering, and the step-by-step mechanism of ion dissolution in water.
+Measurements across many frames yield diffusion rates, structural relaxation, averages, and variation. All-atom MD is suited to liquid structure, molecular transport, polymer ordering, and local mechanical response. Its fixed bonding pattern covers structure and transport; bond-changing reactions use reactive force fields or electronic methods.
 
-## Current Focus
+## How We Use It
 
-- **Active microrheology of glasses**: Pulling a single nanoscale probe through a model metallic glass and measuring the friction force it encounters. This friction encodes the glass transition and reveals local mechanical differences within the material that bulk measurements cannot see
-- **Block copolymer ordering pathways**: Following how a disordered polymer melt organizes into hexagonally packed cylinders, including transient intermediate states where cylinders have formed but not yet aligned
-- **Ion dissolution mechanisms**: Tracking how water molecules cooperatively enter the coordination shell of dissolving metal ions such as Al³⁺, revealing that the sequence and geometry of approach determine the dissolution pathway
-- **Atomistic validation of coarse-grained models**: Providing reference structural and dynamical measurements from all-atom trajectories as calibration targets for the lab's simplified coarse-grained models
+In model glasses, we pull a small probe and read the surrounding material's response from the resisting force, an approach called active microrheology. The force separates motion inside a temporary cage of neighboring atoms from slower structural rearrangement. In block copolymers, polymers made from linked segments with different chemistry, simulations designed to cross rarely visited structures reveal an intermediate network of misaligned cylinders on the way to ordered packing. Across both systems, complete trajectories resolve the sequence of structural change.
 
-## Methods
+## A Reference for Simpler Models
 
-The Multiscale interactive experience uses an explicit-solvent trajectory of caffeine dissolved in water to illustrate core all-atom MD concepts: atomistic representation, force-field terms, local molecular structure, and periodic boundary conditions. The trajectory is a teaching example; the concepts apply across all-atom simulations of liquids and solutions.
+All-atom results also provide targets for coarse-grained models, which group several atoms into one unit. Structural distributions, relaxation behavior, and their uncertainty define what the simpler model should reproduce and set its working range.
 
 ## Related Publications