interpret-raman-spectrum
について
このスキルは、ラマンスペクトルを解析して分極率選択則を用いて分子振動を同定し、赤外分光法を補完します。特に、赤外分光法が限界を示す水性試料、対称振動、および中心対称分子の分析に有効です。主な機能には、ラマン・赤外モードの比較、対称性解析のための偏光解消度の計算、蛍光の影響を抑えながら参照スペクトルとの照合などが含まれます。
クイックインストール
Claude Code
推奨npx skills add pjt222/agent-almanac -a claude-code/plugin add https://github.com/pjt222/agent-almanacgit clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/interpret-raman-spectrumこのコマンドをClaude Codeにコピー&ペーストしてスキルをインストールします
ドキュメント
Interpret Raman Spectrum
Analyze Raman scattering → id mol vibrations, apply selection rules complementary to IR, integrate Raman + IR → comprehensive vibrational.
Use When
- Samples difficult for IR (aqueous, sealed, remote sensing)
- ID symmetric vibrations weak/inactive in IR
- Complement IR via mutual exclusion (centrosymmetric mol)
- Characterize C materials (graphene, CNT, diamond) via Raman bands
- Inorganic, minerals, crystalline phases (Raman often > informative than IR)
- Non-destructive in situ (no sample prep for many Raman)
In
- Req: Raman data (Raman shift cm-1 vs int)
- Req: Excitation laser λ (e.g., 532, 633, 785, 1064 nm)
- Opt: IR of same sample → complementary
- Opt: Polarization data (parallel + perpendicular → depolarization ratios)
- Opt: Mol formula / compound class
- Opt: Physical state (solid, liquid, soln, gas, thin film)
Do
Step 1: Quality + Artifacts
Evaluate reliability before peak analysis:
- Laser λ + fluorescence: Fluorescence = most common interference. Broad intense BG obscures Raman peaks. Shorter λ (532) → more fluorescence; longer (785, 1064) → less, weaker Raman (int scales λ^-4).
- S/N: Peaks distinguishable from noise? Weak scatterers → longer acquisition / higher power.
- Cosmic ray spikes: Sharp narrow spikes random pos = cosmic artifacts, not Raman. Appear in one of time-avg set; remove w/ spike filters.
- Baseline correction: Slope/curve (fluorescence / thermal) → subtract before measuring.
- Photodegradation: High power → damage/transform sample. Check spectral changes between successive acquisitions at same spot. Reduce power if degradation.
- Range: Standard 100-4000 cm-1. Low-freq cutoff depends on edge/notch filter blocking Rayleigh. Note truncation.
→ Quality assessed, fluorescence documented, artifacts (cosmic, baseline) ID'd / corrected, usable range confirmed.
If err: Fluorescence dominates (broad BG >> peaks) → recommend re-measure w/ longer λ (785 / 1064) or SERS. Sample degrades → reduce power / rotating stage.
Step 2: Raman-Active Modes + Selection Rules
Determine Raman-active + how complement IR:
- Raman rule: Vibration Raman-active if changes polarizability. Symmetric stretches (change mol vol) → typically strong Raman.
- IR rule (compare): Vibration IR-active if changes dipole moment. Asymmetric stretches → typically strong IR.
- Mutual exclusion: Mol w/ center of inversion (centrosymmetric) → no vibration both Raman-active + IR-active. Band in both → no center of symmetry.
- General complementarity: Even non-centrosymmetric → Raman-strong tend IR-weak + vv. Combined dataset > either alone.
- Raman-favored modes: Sym stretches (C-C, C=C, S-S, N=N), breathing modes of rings, stretches of homonuclear bonds (no dipole change → IR-inactive) → typically strong Raman.
→ Selection rules applied, Raman-active vs IR-active distinguished, mutual exclusion tested if centrosymmetric.
If err: Mol symmetry unknown → use combined Raman + IR to infer. Band in both w/ comparable int → not centrosymmetric.
Step 3: Raman Shift Positions
Assign bands → vibrational modes via characteristic freqs:
- C-H stretch (2800-3100 cm-1): Similar IR but Raman int differ. Aromatic + olefinic C-H (3000-3100) often > Raman than aliphatic.
- Triple bonds (2100-2260 cm-1): C≡C sym stretch strong Raman, often weak/absent IR. C≡N active in both.
- Double bond stretches:
| Shift (cm-1) | Assignment | Raman Intensity |
|---|---|---|
| 1600--1680 | C=C stretch | Strong |
| 1650--1800 | C=O stretch | Medium (weaker than IR) |
| 1500--1600 | Aromatic C=C | Medium to strong |
- Aromatic ring modes:
| Shift (cm-1) | Assignment | Notes |
|---|---|---|
| 990--1010 | Ring breathing (monosubstituted) | Very strong, diagnostic |
| 1000 | Ring breathing (sym. trisubstituted) | Strong |
| 1580--1600 | Ring stretch | Medium |
| 3050--3070 | Aromatic C-H stretch | Medium |
- Other characteristic:
| Shift (cm-1) | Assignment |
|---|---|
| 430--550 | S-S stretch (disulfide) |
| 570--705 | C-S stretch |
| 800--1100 | C-C skeletal stretch |
| 630--770 | C-Cl stretch |
| 500--680 | C-Br stretch |
| 200--400 | Metal-ligand stretch |
- C materials: G band (~1580, graphitic sp2) + D band (~1350, defect/disorder) → diagnostic for C allotropes. 2D (~2700) → graphene layer count. Diamond sharp peak 1332.
→ All significant bands assigned to vibrational modes w/ ref to freq ranges.
If err: Band unassignable from tables → consult DBs (RRUFF minerals, SDBS organics). Unassigned → combination modes, overtones, lattice vibrations in crystalline.
Step 4: Compare Raman vs IR
Integrate two complementary techniques:
- Tabulate corresponding bands: Per mode → Raman shift, IR freq, rel int each technique.
- Modes in one only: Raman but not IR (or vv) → symmetry info. Sym stretches of non-polar bonds (S-S, C=C sym env) → Raman only.
- Resolve ambiguities: IR tentative (e.g., overlapping C-O + C-N fingerprint) → Raman may be clearer (diff rel int).
- Functional group confirm: Confirm IR-ID'd groups via Raman counterparts. Ester → C=O IR (~1735) + C-O-C Raman. Acid → broad OH IR + C=O both.
- Assess consistency: Raman + IR mutually consistent. Contradictions (band assigned sym stretch strong both for centrosymmetric) → err in assignment / symmetry.
→ Unified vibrational analysis table combining Raman+IR, func groups confirmed / refined by complementary.
If err: No IR → Raman alone useful but reduced certainty. Note assignments benefiting from IR confirm.
Step 5: Polarization + Document
Depolarization ratios → symmetry + compile final:
- Depolarization ratio (ρ): ρ = I⊥ / I∥, from polarized Raman.
- ρ = 0-0.75 → polarized band. Totally symmetric vibrations (A-type) polarized.
- ρ = 0.75 → depolarized. Non-totally-sym vibrations → ρ = 0.75.
- Symmetry assignment: Polarized bands → totally sym irrep of point group. Helps distinguish modes of diff sym at similar freqs.
- Compile: Complete table per observed:
- Raman shift (cm-1)
- Rel int (strong/medium/weak)
- Depolarization ratio (if measured)
- Assignment
- Corresponding IR band (if observed)
- Compare ref spectra: Known compound → compare vs published (RRUFF, SDBS, NIST). Agreement ±3 cm-1 + matching rel int → identity.
- Report uncertainties: Flag tentative assignments, note which add'l exps (temp-dep Raman, resonance Raman, SERS) resolve ambiguity.
→ Complete analysis, all bands assigned, polarization → symmetry, results integrated w/ IR + other.
If err: No polarization → symmetry relies on freq+int alone. Note limitation + recommend polarized measurements if symmetry critical.
Check
- Quality assessed (fluorescence, cosmic, baseline, photodegradation)
- Selection rules applied, Raman-active modes ID'd
- Mutual exclusion tested if centrosymmetric
- All significant bands assigned
- Raman vs IR compared + integrated where avail
- Depolarization ratios → symmetry (if polarization avail)
- Assignments consistent w/ known / proposed structure
- Results compared w/ ref spectra where poss
Traps
- Fluorescence overwhelm signal: Most common prob. Switch longer λ / time-gated detection. Don't interpret broad fluorescent humps as Raman bands.
- Confuse cosmic spikes w/ real peaks: Random pos sharp intense spikes → present in single, absent in averaged. Always check reproducibility.
- Neglect polarizability rule: Strong IR modes (asym polar) → weak/absent Raman + vv. Don't expect same int pattern as IR.
- Ignore degradation: High power → char, polymerize, phase-transform. Spectrum changes between measurements at same spot = degradation.
- Assume all Raman = fundamentals: Overtones (2× fundamental) + combination bands appear. Weaker than fundamentals but cause confusion.
- Overlook low-freq: Lattice vibrations, torsional, metal-ligand below 400 cm-1. Many setups don't access. Verify notch/edge filter allows low-freq if these modes relevant.
→
interpret-ir-spectrum— complementary vibrational → dipole-activeinterpret-nmr-spectrum— connectivity → complete structureinterpret-mass-spectrum— formula + fraginterpret-uv-vis-spectrum— electronic transitions + chromophoresplan-spectroscopic-analysis— select + sequence techniques pre-acquisition
GitHub リポジトリ
関連スキル
llamaguard
その他LlamaGuardは、暴力やヘイトスピーチなど6つの安全性カテゴリーにおいて、LLMの入力と出力をモデレートするMetaの70-80億パラメータモデルです。94〜95%の精度を提供し、vLLM、Hugging Face、Amazon SageMakerを使用してデプロイ可能です。このスキルを使用して、AIアプリケーションにコンテンツフィルタリングと安全策を簡単に統合できます。
cost-optimization
その他このClaudeスキルは、リソースの適正サイジング、タグ付け戦略、支出分析を通じて、開発者がクラウドコストを最適化することを支援します。AWS、Azure、GCPにわたるクラウド支出の削減とコストガバナンスの実施のためのフレームワークを提供します。インフラコストの分析、リソースの適正サイジング、または予算制約への対応が必要な際にご利用ください。
quantizing-models-bitsandbytes
その他このスキルは、bitsandbytesを使用してLLMを8ビットまたは4ビット精度に量子化し、精度の低下を最小限に抑えつつ50〜75%のメモリ削減を実現します。限られたGPUメモリでより大規模なモデルを実行したり、推論を高速化するのに理想的で、INT8、NF4、FP4などのフォーマットをサポートしています。HuggingFace Transformersと統合され、QLoRAトレーニングや8ビットオプティマイザーを可能にします。
dispatching-parallel-agents
その他このClaudeスキルは、複数のエージェントを配備し、3つ以上の独立した問題を並行して調査・修正します。共有状態や依存関係がなく解決可能な、無関係な障害が発生するシナリオ向けに設計されています。中核となる機能は並列問題解決であり、効率を最大化するために独立した問題領域ごとに1つのエージェントを割り当てます。
