スキル一覧に戻る

interpret-ir-spectrum

pjt222
更新日 Yesterday
2 閲覧
17
2
17
GitHubで表示
その他general

について

このスキルは、診断領域と指紋領域を分析することで官能基を体系的に同定するIRスペクトル解釈機能です。水素結合の影響を検出し、信頼度評価付きの官能基インベントリを作成します。化合物の初期スクリーニング、反応モニタリング、官能基の有無確認にご利用いただけます。

クイックインストール

Claude Code

推奨
メイン
npx skills add pjt222/agent-almanac -a claude-code
プラグインコマンド代替
/plugin add https://github.com/pjt222/agent-almanac
Git クローン代替
git clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/interpret-ir-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 IR Spectrum

Analyze IR absorption → id functional groups, assess H-bonding, inventory structural features.

Use When

  • ID functional groups in unknown (first screen)
  • Confirm presence/absence (e.g., rxn converted OH → ketone?)
  • Monitor rxn progress → appear/disappear of absorptions
  • Distinguish similar compounds by functional group
  • Complement NMR + MS w/ vibrational info

In

  • Req: IR data (abs freq cm-1 + intensities, %T or Abs)
  • Req: Prep method (KBr, ATR, Nujol, thin film, soln cell)
  • Opt: Mol formula / expected class
  • Opt: Known frags from other spectra
  • Opt: Instrument params (res, range, detector)

Do

Step 1: Spectrum Quality + Format

Verify suitability before peak analysis:

  1. y-axis format: %T (peaks down) / Abs (peaks up). Keep consistent.
  2. Wavenumber range: ≥ 4000-400 cm-1 for mid-IR. Note truncation.
  3. Baseline: Flat + near 100%T (or 0 Abs) in no-abs regions. Slopes/noise → reduce reliability.
  4. Resolution: Adjacent peaks < instrumental res → can't distinguish. Typical FTIR: 4 cm-1.
  5. Prep artifacts: KBr → broad OH ~3400 cm-1 (moisture). Nujol obscures CH stretch. ATR distorts low wavenumbers. Note.

→ Spectrum suitable; format, range, artifacts documented.

If err: Severe baseline probs, saturation (flat-bottom peaks → too-conc sample), prep artifacts obscuring critical regions → note limitation + flag regions unreliable.

Step 2: Diagnostic Region (4000-1500 cm-1)

High-freq region → most functional groups:

  1. O-H (3200-3600 cm-1): Broad abs. Sharp ~3600 → free OH; broad 3200-3400 → H-bonded OH (alcohols, acids, water).
  2. N-H (3300-3500 cm-1): Primary amines → 2 peaks (sym+asym); secondary → 1. Sharper than OH.
  3. C-H (2800-3300 cm-1):
Frequency (cm-1)Assignment
3300sp C-H (alkyne, sharp)
3000--3100sp2 C-H (aromatic, vinyl)
2850--3000sp3 C-H (alkyl, multiple peaks)
2700--2850Aldehyde C-H (two peaks from Fermi resonance)
  1. Triple-bond (2000-2300 cm-1):
Frequency (cm-1)AssignmentNotes
2100--2260C triple-bond CWeak or absent if symmetric
2200--2260C triple-bond NMedium to strong
~2350CO2Atmospheric artifact, disregard
  1. Carbonyl (1650-1800 cm-1) — most diagnostic single region:
Frequency (cm-1)Assignment
1800--1830, 1740--1770Acid anhydride (two C=O stretches)
1770--1780Acid chloride
1735--1750Ester
1700--1725Carboxylic acid
1705--1720Aldehyde
1705--1720Ketone
1680--1700Conjugated ketone / alpha-beta unsaturated
1630--1690Amide (amide I band)
  1. C=C + C=N (1600-1680 cm-1): Alkene C=C → 1620-1680 (weak-med). Aromatic C=C → multiple 1450-1600. C=N (imine) → 1620-1660.

→ All abs in diagnostic ID'd w/ func group + confidence (strong/tentative/absent).

If err: Carbonyl obscured (water in KBr, atm CO2) → note gap. Expected group absent → confirm w/ 2nd prep before concluding absent.

Step 3: Fingerprint (1500-400 cm-1)

Low-freq region → confirmation + structural detail:

  1. C-O (1000-1300 cm-1): Ethers, esters, alcohols, acids → strong C-O. Esters → characteristic strong band 1000-1100 + carbonyl.
  2. C-N (1000-1250 cm-1): Amines + amides; overlap C-O → tentative w/o other evidence.
  3. C-F, C-Cl, C-Br:
Frequency (cm-1)Assignment
1000--1400C-F (strong)
600--800C-Cl
500--680C-Br
  1. Aromatic subst pattern (700-900 cm-1): OOP C-H bending → substitution:
Frequency (cm-1)Pattern
730--770Mono-substituted (+ 690--710)
735--770Ortho-disubstituted
750--810, 860--900Meta-disubstituted
790--840Para-disubstituted
  1. Fingerprint comparison: Region unique per compound. Ref spectrum avail → overlay + compare → identity confirm.

→ Confirmatory assignments for Step 2 groups + structural detail (subst patterns, C-O/C-N).

If err: Fingerprint inherently complex + overlapping. Ambiguous → flag tentative + rely on diagnostic + other spectra.

Step 4: H-bonding + Intermolecular Effects

Evaluate sample state + interactions:

  1. H-bonding broadening: Compare width+pos of OH, NH. Free OH sharp ~3600; H-bonded broad + shifted 3200-3400. Acid dimers → very broad OH 2500-3300.
  2. Conc + state effects: Soln spectra at diff conc → distinguish intramolecular (conc-indep) from intermolecular (conc-dep) H-bonds.
  3. Fermi resonance: 2 overlapping bands → doublet. Classic: aldehyde C-H pair ~2720 + 2820. Recognize → avoid mis-assign as separate groups.
  4. Solid-state effects: KBr + Nujol → solid packing → broadens bands + shifts 10-20 cm-1 vs soln. ATR closest to neat liquid.

→ H-bonding characterized, prep artifacts accounted, anomalous band shapes explained.

If err: H-bonding unresolved (overlap OH + NH) → note ambiguity. D2O exchange / var-temp → helps, requires add'l data.

Step 5: Compile Func Group Inventory

Assemble findings → structured report:

  1. Confirmed groups: Strong unambiguous abs in diagnostic (e.g., sharp C=O at 1715 = ketone/aldehyde).
  2. Tentative: Weaker evidence / overlap → >1 possible group.
  3. Absent: Characteristic strong abs clearly missing (no broad OH → no free alcohol/acid).
  4. Discrepancies: Abs not fitting proposed groups, or expected abs missing.
  5. Cross-ref: Compare IR inventory vs NMR, MS, UV-Vis if avail.

→ Complete inventory by confidence, specific freqs + intensities cited as evidence.

If err: Inventory incomplete/contradictory → ID which add'l exps (ATR vs KBr, var conc, D2O exchange) resolve ambiguity.

Check

  • Quality assessed (baseline, res, artifacts, y-axis)
  • Solvent, prep, atm artifacts ID'd + excluded
  • All abs in diagnostic (4000-1500) assigned / flagged
  • Carbonyl region → sub-type assignment where possible
  • Fingerprint examined for confirmation
  • H-bonding evaluated + peak shape/pos impact documented
  • Inventory compiled w/ confidence
  • Absent groups explicit (neg evidence informative)
  • Cross-ref vs other spectra

Traps

  • Ignore prep artifacts: KBr moisture (broad 3400), Nujol C-H (2850-2950), ATR distortion at low wavenumbers → mimic/obscure real. Always consider prep.
  • Over-interpret fingerprint: Region < 1500 complex+overlapping. Use for confirm not primary ID. Don't assign every peak.
  • Confuse atm CO2 w/ sample: Sharp doublet ~2350 = atm CO2 usually, not sample. BG subtraction removes, verify.
  • Neglect intensity+width: Strong broad ≠ weak sharp at same freq. Report intensity (str/med/weak) + shape (sharp/broad) + freq.
  • Single-peak assignment: Never ID func group from single abs. Carbonyls → supported by additional bands (C-O for esters, N-H for amides, C-H for aldehydes).
  • Absence from weak abs: Some groups → inherently weak IR (sym C=C, triple bonds sym alkynes). Absence ≠ always absence of group.

  • interpret-nmr-spectrum — detailed connectivity + H environments
  • interpret-mass-spectrum — mol formula + fragmentation
  • interpret-uv-vis-spectrum — chromophores complementing IR
  • interpret-raman-spectrum — complementary vibrational → IR-inactive modes
  • plan-spectroscopic-analysis — select + sequence techniques pre-acquisition

GitHub リポジトリ

pjt222/agent-almanac
パス: i18n/caveman-ultra/skills/interpret-ir-spectrum
0
agentsagentskillsai-assisted-developmentclaude-codeskillsteams

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。

スキルを見る

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。

スキルを見る

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。

スキルを見る

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。

スキルを見る