Geng Academic Integrity Report: Luminescent Properties of High-Power White LEDs Based on Phosphor Films (DOI: 10.11901/1005.3093.2015.029)

Academic fraud report · Geng Detector

Summary

This report assesses a 2016 paper in the Chinese Journal of Materials Research (材料研究学报) on phosphor-film-based high-power white LEDs. Overall verdict: highly questionable, driven by one confirmed internal contradiction. The key confirmed finding is a data mismatch between the body text and the abstract/conclusions: the device with color rendering index (CRI) 91.7 is described in the body (Figure 5 discussion) as having a correlated color temperature of ~2700 K and luminous efficacy of ~52.6 lm/W, while the abstract and conclusions attribute 4460 K and CRI 91.7 to the same device; 4460/4450 K in the body actually belongs to a different device (Sr₃SiO₅:Eu²⁺ orange phosphor film) with CRI 73.7 and 76.5 lm/W. This is likely a serious transcription error rather than fabrication, but requires correction. Three further findings remain unresolved with plausible benign explanations: near-identical 525 nm yellow emission peaks across five film thicknesses (Figure 3), unusually stable luminous efficacy (~77 lm/W, Table 1) despite large CIE shifts, and uniform copy-move forensics signals at offset (0,32) across all five figures, likely attributable to regular chart elements or a shared export pipeline but unverifiable without original images. Confidence in the confirmed discrepancy is high; fraud characterization would require journal-level investigation and raw data.

Verdict

Highly questionable (🟠). One confirmed, title-level internal inconsistency regarding the correlated color temperature of the flagship device. Other findings carry plausible benign explanations and remain unresolved pending original data. The automated Bayesian synthesis (99.9% posterior) should be treated cautiously, as it rests mainly on copy-move signals consistent with regular chart graphics.

Key findings

Confirmed (Finding 1): CRI 91.7 device inconsistent between sections. Body text (Figure 5 discussion, p. 391): ~2700 K, 52.6 lm/W, CRI 91.7. Chinese abstract, English abstract ("correlated color temperature Tc of 4460K, and the color rendering index of 91.7") and conclusions: 4460 K with CRI 91.7. In the body, 4450 K belongs to a different device (Sr₃SiO₅:Eu²⁺ orange film: ~4450 K, 76.5 lm/W, CRI 73.7). Three values (CCT, efficacy, CRI) are cross-mismatched and cannot be self-consistent. Most likely a serious writing/proofreading error rather than fabrication, but a correction is required.
Finding 2 (unresolved, 🟡): Figure 3 electroluminescence spectra (IF = 300 mA) show near-perfect overlap of the 525 nm yellow peak across 0.88–1.52 mm thicknesses, while 455 nm blue intensity declines with thickness. The paper explains this via blue-light absorption saturation above ~1 mm thickness; needs raw spectral data to confirm.
Finding 3 (unresolved, 🟡): Table 1 luminous efficacy values (69.89, 76.35, 76.96, 76.28, 77.79 lm/W) vary <2% for four of five thickness points, while CIE coordinates shift substantially (x: 0.3277→0.3695; y: 0.4066→0.4864). Paper attributes the plateau to constant 525 nm emission; raw test records needed.
Finding 4 (unresolved, 🟡): Copy-move detection flags block matches at the identical offset (0,32) in all 5 embedded images (42–83 pairs each), plus inconsistent noise variance signals. Benign explanations (regular axes/grid elements, uniform export pipeline) are speculative and unsupported by the paper, but the signals cannot be cleared without original high-resolution images. PRNU weak similarity (NCC 0.016–0.165) is negligible and non-diagnostic.

Evidence highlights

Verified quotations from the original text: body Figure 5 discussion (2700 K / 52.6 lm/W / CRI 91.7); Figure 4 device (4450 K / 76.5 lm/W / CRI 73.7); abstract/conclusion claims (4460 K / CRI 91.7).
Copy-move detections: img-000 (83 pairs), img-001 (65 pairs), img-002 (61 pairs), img-003 (42 pairs), img-004 (57 pairs), all at offset (0,32).
Automated Bayesian synthesis: posterior 99.9%, 95% CI [98.8%, 100%], BF = 1.64e+4 — generated under a fixed 5% prior and conditional-independence assumptions; treat as heuristic only, not a finding of misconduct.

Notes

DOI: 10.11901/1005.3093.2015.029; Journal of Materials Research (材料研究学报), Vol. 30, No. 5, May 2016; authors from Wuyi University. Source file: ChinaXiv 202303.10757v1 (uploaded March 2023).
Recommended actions: request author clarification of the Figure 5 device CCT (2700 K vs 4460 K) and abstract/conclusion correction (confirmed error); request raw spectra (Fig. 3) and Table 1 test records; submit a correction request to the journal; hold off on public posting (e.g., PubPeer) until image forensics on originals is complete.
AI-assisted report; false positives and negatives are possible. Final misconduct determination requires institutional investigation and the official record takes precedence.