Peifu Zhang

Nanorobots, with miniature size, wireless operation, and flexible mobility, can access narrow spaces and are widely used in biomedicine. High-precision transport is vital for targeted drug delivery, disease diagnosis, and treatment, yet current research has limitations like insufficient single-imaging diagnosis and active tumor targeting. In this research, a magnetic nanorobot platform is proposed, combining magnetic ferric oxide substrate with rare earth nanoparticles (RENPs). This magnetic nanorobot features multi-modal imaging including NIR II fluorescence, MRI, and PA with high moving target effect (<5 s). After conjugated with antibodies against MET and VEGF, this platform enables multi-molecular targeting. Under an external magnetic field, the dual-targeting (active and physical magnetic) allows in vitro and in vivo precise tumor cell targeting with tumor imaging. Moreover, its NIR II luminescence enables enhanced real-time temperature detection (24.6–41.5°C). Especially, the deep learning allows NIR II imaging to simultaneously perform increased fluorescence imaging for surgical boundary definition and temperature sensing display (R² = 0.973, RMSE = 1.50°C). With high penetration and spatial resolution, this proposed magnetic/rare earth hybrid nanorobots holds great promise for precision biomedicine.

@article{li2026deep,
  title     = {Deep learning enabled magnetic/rare earth hybrid nanorobots for multi-modal bioimaging and temperature sensing with surgical boundary determination},
  author    = {Li, Wenjing and Lin, Bi and Li, Bixiao and Zhang, Peifu and Ju, Ziyue and Ansari, Anees A. and Lv, Ruichan},
  journal   = {Sensors and Actuators B: Chemical},
  volume    = {455},
  pages     = {139673},
  year      = {2026},
  doi       = {10.1016/j.snb.2026.139673}
}

High-performance electronic equipment integrated manufacturing (IM) is essential for intelligent manufacturing, yet the stability and reliability of IM monitoring traditionally rely on offline detection after manufacturing, which leads to late defect discovery and potential losses from production stoppages. Existing multi-modal fault diagnosis algorithms lack physical-information fusion and contextual modeling, making it difficult to fully exploit the spatial and nonlinear dynamic features of signals. To address these limitations, we propose a multi-modal physical information fusion network (MPIF-Net) for online fault monitoring in IM-SLM (selective laser melting) processes. Multi-sensor deployment collects multi-modal acoustic emission data, which are transformed into fused images preserving spatial and nonlinear dynamic features while retaining the original temporal modality. A dual-attention convolution module and Mamba block capture both local and global information from the fused images; a BiLSTM architecture extracts temporal dependencies from the raw acoustic emission signals; and a physics-informed neural network (PINN) generates high-fidelity thermal modality signals to improve monitoring accuracy. Finally, a channel-concatenation fusion strategy integrates physical information with data-driven features. Experimental results show MPIF-Net significantly outperforms existing methods, achieving a 1.48% improvement in F1 over conventional multi-modal algorithms — offering a generalizable solution for intelligent monitoring in integrated manufacturing.

@article{lv2025physical,
  title     = {Physical Information Fusion of Multi-modal Data for Online Fault Monitoring of Selective Laser Melting Integrated Manufacturing},
  author    = {L{\"u}, Ruichan and Xu, Han and Zi, Bin and Zhang, Peifu and Li, Yuji and Huang, Jin},
  journal   = {Journal of Mechanical Engineering},
  volume    = {61},
  number    = {11},
  year      = {2025}
}

Scintillator-based X-ray imaging is widely applied in medical diagnostics, security screening, and nondestructive evaluation. However, continuous irradiation inevitably generates scattered photons that degrade image quality, and it remains a major challenge to design intrinsically noise-resistant imaging platforms. Here, we demonstrate that X-ray-excited persistent luminescence (XEPL) from scalable lanthanide-doped SrFCl phosphors enables low-dose, noise-free delayed X-ray imaging. Trace Ce³⁺ incorporation creates an efficient trap–Ce–Tb energy transfer cascade, while Na⁺ co-doping increases trap density, yielding a 13.4-fold enhancement in Tb³⁺ XEPL. The system exhibits remarkable stability under repeated irradiation and after 80 days of ambient storage, outperforming the commercial SrAl₂O₄:Eu/Dy. Bright afterglow is achieved with only 5 s of X-ray exposure, delivering a spatial resolution of 21.4 lp/mm. Delayed image acquisition permits extended camera integration without further irradiation, thereby eliminating scattering noise and enabling background-free 3D reconstruction. Furthermore, we show that delayed imaging can be combined with machine learning for intelligent detection. This work addresses a fundamental limitation of real-time X-ray detection and offers a practical route toward high-resolution, low-dose, and intrinsically noise-resistant imaging.

@article{zhao2025dual,
  title     = {Dual-Heterovalent Codoping Enables Bright and Rapid X-Ray Excited Persistent Luminescence for Noise-Free 3D Imaging and Intelligent Detection},
  author    = {Zhao, Jian and Li, Deyang and Zhang, Peifu and Lv, Ruichan and Wang, Yubin and Zhou, Su and Deng, Degang and Xu, Shiqing and Lei, Lei},
  journal   = {Laser \& Photonics Reviews},
  year      = {2025},
  pages     = {e02875},
  doi       = {10.1002/lpor.202502875}
}