Based on the provenance, I can now give a complete answer. The batch 48 record merges values from three source types:

1. LBN 48.pdf (scanned handwritten ELN, no text layer) — aim, conclusion, materials[0–3], process_steps[0–2], critical_parameters[0–1], samples[0], outcome
2. LBN 49.pdf (scanned handwritten ELN, no text layer) — materials[4–7], process_steps[3–5], critical_parameters[2–3], procedure_summary
3. 28112025 9076 Expt Table 2025.xls (digital spreadsheet) — observations, record_type, process_steps[6–9]
4. APP DATA xlsx (digital) — samples[1]
5. NMR PDFs (digital, with text layer) — samples[2], date, scientist

The LBN 48 and LBN 49 PDFs are the only handwritten sources (confirmed: no embedded text layer, only PaddleOCR output exists, and they are absent from the ocr-accuracy.md table of digital PDFs).

Answer:

The values in batch 48 that came from handwriting are all the fields sourced from LBN 48.pdf and LBN 49.pdf (both scanned handwritten ELN pages with no text layer):

From LBN 48.pdf [pages 1–2]:

• aim, conclusion, conditions.atmosphere
• materials[0–3]: Vicaspin LUB 1010 178 gms, P2O5 9.357 gms, Epomine SP-018 30 gms, 2-ethyl hexyl acrylate 70 gms
• process_steps[0–2]: AV-52.52, Mono-24.44, Di-51.76, Free acid-0.24, AV-0.45, IV-26.1, AV-44.36, HV-46.58, SAP-62.14, Amine-36.25, IV-5.8
• critical_parameters[0–1]
• samples[0], outcome

From LBN 49.pdf [pages 1–2]:

• materials[4–7]: Vicaspin LUB 1010 208 gms, P2O5 10.93 gms, Epomine SP-018 30 gms, 2-ethyl hexyl acrylate 70 gms
• process_steps[3–5]: AV-51.87, Mono-22.56, Di-48.93, Free acid-1.096, AV-0.45, IV-26.1, AV-46.5, HV-49.4, SAP-61.98, Amine-20.2, IV-2.58
• critical_parameters[2–3], procedure_summary

Not from handwriting (digital sources): observations, record_type, process_steps[6–9] ← Expt Table .xls; samples[1] ← APP DATA xlsx; samples[2], date, scientist ← NMR PDFs (digital).