Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Pengfei Zhu has a total of 65 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    An operator-theoretic interpretation of exceptional spectral dynamics in Maxwell–Cattaneo heat transport

    by Pengfei Zhu

    Heat transport exhibits distinct regimes ranging from ballistic propagation to diffusive relaxation, traditionally described by disparate theoretical frameworks. Here, we introduce a unified first-order operator formulation in which temperature and heat flux are treated as a coupled state vector, yielding a physically constrained first-order operator representation of heat transport. The resulting generator is intrinsically non-Hermitian and gives rise to a spectral structure governed by an exceptional point that separates overdamped diffusion from underdamped wave-like propagation. In this framework, Fourier’s law emerges as a singular limit of a hyperbolic dissipative system, while the Cattaneo equation arises naturally as the minimal representative hydrodynamic closure of kinetic theory. We show that the exceptional point induces nonanalytic spectral transitions, nonmodal transient dynamics, and a breakdown of conventional modal decomposition. The theory further generalizes to anisotropic media, where direction-dependent exceptional surfaces enable intrinsic steering of heat flow. Rather than introducing a new transport equation, the present work provides an operator-theoretic reformulation of the classical Maxwell–Cattaneo dynamics, showing that its non-normal generator naturally admits an exceptional-point interpretation.

  • Open Access

    Geometry adaptive photothermal coherence tomography and its application in additively manufactured metal structures

    by Rongbang Wang, Pengfei Zhu, Hai Zhang, Andreas Mandelis

    Infrared thermography has been widely applied in aerospace and other technological fields. However, constrained by perspective projection effects and the diffusive, non-wave nature of heat propagation, conventional methods are limited to generating 2D projections that reflect depth-integrated information, primarily surface temperature. Consequently, they fail to provide intuitive inspection results for complex structures and make it difficult to quantitatively evaluate internal defects. Here, we propose a geometry adaptive photothermal coherence tomography (GA-PCT) technique to enable 3D morphology reconstruction and quantitative defect depth analysis in additively manufactured metallic components. GA-PCT performs line-structured light scanning via galvanometer-controlled laser deflection, while an infrared camera synchronously captures the thermal fringe for rapid reconstruction of surface morphology. By incorporating beam homogenization and delayed-sampling strategies into the photothermal tomography technique, accurate defect-depth analysis is achieved. Unlike existing photothermal tomography methods that form a volumetric impression by assembling depth-related 2D feature maps, GA-PCT directly outputs 3D coordinates encompassing both surface and internal features. Compared with optical 3D imaging and optical coherence tomography, which are limited to reconstructing only surface or near-subsurface features, as well as point-by-point scanning techniques such as X-ray CT and photoacoustic tomography, GA-PCT offers remarkable advantages in information completeness (with an effective penetration depth of 1.4 mm in stainless steel) and detection efficiency (≥10 cm 2 ·s −1 ), showing great potential for applications in non-destructive testing and reverse engineering.


  • Making neural networks understand internal heat transfer using Fourier-transformed thermal diffusion wave fields

    by Pengfei Zhu, Hai Zhang, Clemente Ibarra-Castanedo, Xavier Maldague, Andreas Mandelis

    Heat propagation, governed by phonon interactions, is described by partial differential equations linking thermal transport to material properties. Conventional thermography relies on surface emissions, limiting subsurface resolution, while existing tomographic methods capture only single-layer frames. Physics-informed neural networks (PINNs) integrate data with physics but mainly fit external temperatures, lacking direct access to internal fields. Here, we propose a Helmholtz-informed neural network (HINN) to predict internal temperature distributions without interior measurements. By converting the time-domain diffusion equation into a pseudo-Helmholtz form, HINN incorporates both real and imaginary components of the thermal field, followed by inverse Fourier transform to reconstruct 3D thermal maps with defects. A truncated operation and conjugate symmetry repair further improve efficiency and accuracy. Results show that HINN surpasses PINNs and inverse solvers, enabling non-invasive thermography for materials, biomedicine, and nondestructive evaluation.


  • Enhanced low-power ultrasonic-induced thermography for evaluation of low-velocity impact damage in aerospace fiber metal laminate structures

    by Yubin Zhang, Pengfei Zhu, Changhang Xu, Clemente Ibarra-Castanedo, Yunfei Deng, Zhiyuan Zhang, Rui Liu, Linglin Chen, Xavier Maldague

  • Three-Dimensional Wide-Bandwidth Quantum Energy Truncation Terahertz Coherence Tomography

    by Pengfei Zhu, Hai Zhang, Stefano Sfarra, Fabrizio Sarasini, Xavier Maldague, Andreas Mandelis

    We report a quantum energy truncation terahertz coherence tomography technique that enables highly localized extraction of molecular-species-specific THz responses. The method exploits resonant photon loss when the emitter THz energy matches molecular vibration quanta. Reflected THz radiation is recorded and mapped into depth-coded image sequences, from which subsurface structural information and molecular fingerprints are retrieved to construct three-dimensional tomograms. This approach resolves major limitations of conventional continuous-wave (CW) THz imaging-namely, its lack of spectroscopic specificity and dependence on angular scanning-by directly correlating molecular excitation spectra with depth information. As a result, quantum energy truncation terahertz coherence tomography enables rapid, three-dimensional THz imaging with selective sensitivity to discrete molecular quantum energy transitions, providing new opportunities for nondestructive characterization of material refractive indices, molecular composition, and subsurface structures.


  • Open Access

    Field-material coupled neural network: A novel zero-shot physics-informed learning inverse problem solver for extracting complex dielectric constant in terahertz band

    by Pengfei Zhu, Stefano Sfarra, Elena Pivarciova, Carlo Santulli, Xavier Maldague

    Accurate extraction of the complex dielectric constant in the terahertz (THz) band is essential for material characterization and non-destructive evaluation yet remains challenging due to the ill-posed nature of electromagnetic inverse problems and the limited availability of reliable reference data. In this work, a field-material couple neural network (FMCNN) is proposed to retrieve the complex dielectric constant directly from THz measurements. The FMCNN consists of a field neural network and a material neural network that are strongly coupled through the frequency-domain Maxwell equations in the form of a Helmholtz equation, with the governing physics enforced by partial differential equation and boundary condition constraints. This formulation enables zero-shot physics-informed learning inversion, requiring only measured test data as input. The extracted dielectric constants are validated by comparison with results from a one-dimensional normal-incidence model and the Drude–Lorentz model, showing good agreement over a broad frequency range, particularly above 0.2 THz. These results demonstrate that the FMCNN provides a physics-consistent and data-efficient approach for material parameter extraction in the THz band, offering an alternative to conventional model-based methods.


  • Thermal Diffusivity Measurement Based on Thermal/Cooling Excitation: Theory and Experiments

    by Pengfei Zhu, Hai Zhang, Stefano Sfarra, Fabrizio Sarasini, Clemente Ibarra-Castanedo, Xavier Maldague, Andreas Mandelis

    Photothermal methods for measuring thermal diffusivity inherently involve an ill-posed inverse problem influenced by multiple factors, including sample thickness, heating or cooling duration, and excitation energy. Measurement accuracy becomes particularly challenging under non-impulsive pulsed excitation when the observation timescale is comparable to the pulse duration. This difficulty arises from poorly defined pulse profiles, broadened thermal responses, and the lack of clear boundary condition criteria - especially under significant interfacial temperature gradients where natural convection dominates (Grashof number Gr >> 2,000). Although the classic Parker solution is widely used, it is physically unrealistic because it assumes adiabatic heat flux and shallow-region heat absorption. In this study, we demonstrate that Parker’s assumption is mathematically equivalent to a Dirac pulse boundary condition. We then derive comprehensive analytical solutions for both thermal and cooling responses under Dirac and rectangular pulse excitations. By comparing these with eigenfunction-based solutions under well-posed boundary conditions, we reveal that Parker’s model is valid only before the thermal peak occurs. Through dimensionless analysis, we further show that Parker’s solution represents the limiting case of the rectangular pulse model as the heating duration approaches zero. Finally, we propose a novel excitation approach—evaporative cryocooling—for thermal diffusivity measurements. This technique provides a compact, low-cost, and easily implemented alternative to conventional excitation schemes. The theoretical model is further validated through comparison with experimental results.

  • Physics-Guided Deep Neural Networks for Mineral Identification Using Terahertz Time-Domain Spectroscopy and Micro-X-Ray Fluorescence

    by Pengfei Zhu, Silum Teaghes Ghebreyesus, Yubin Zhang, Mariam Coulibaly, Jonathan Côté, Marc Constantin, Xavier Maldague

    Terahertz time-domain spectroscopy techniques (THz-TDS), as a relatively new spectroscopic technique, exhibit great potential in geoscience fields. In this work, THz-TDS was applied for mineral identification. Three rock core samples of lithium-bearing with unknown elemental distributions were investigated. Due to the large data volume of THz-TDS measurements, the acquired signals contain both effective and redundant information, making feature extraction a crucial step. A comprehensive evaluation of different feature extraction approaches was performed, including the original time-domain signals, spectroscopic knowledge-based representations (amplitude, phase, refractive index, and absorption coefficient maps), and a conventional statistical algorithm – principal component analysis (PCA). Among these methods, PCA achieves the most effective feature separation; however, many principal components were redundant. To overcome this limitation, a novel unsupervised deep learning framework, termed terahertz physics-guided neural network (THz-PGNN) was developed. This network incorporates physical knowledge from THz wave propagation into an unsupervised autoencoder structure, enabling the extraction of more sensitive and physically meaningful features. Comparative experiments between the conventional autoencoder and THz-PGNN show that the proposed model achieves higher sensitivity and captures deeper hidden representations of the THz data. Finally, the normalized cross-correlation algorithm from four reference pre-characterized samples and micro-X-ray fluorescence were employed to further validate the accuracy and robustness of the proposed methods.


  • Thermal Diffusivity Characterization of Impacted Composites Using Evaporative Cryocooling Excitation and Inverse Physics-Informed Neural Networks

    by Pengfei Zhu, Hai Zhang, Stefano Sfarra, Fabrizio Sarasini, Rubén Usamentiaga, Gunther Steenackers, Clemente Ibarra-Castanedo, Xavier Maldague

    The thermal diffusivity measurement of impacted composites using pulsed methods presents an ill-posed inverse problem influenced by multiple factors such as sample thickness, cooling duration, and excitation energy. In this study, a novel excitation method—evaporative cryocooling—was introduced for measuring the thermal diffusivity of tested samples. Compared to conventional excitation modalities, evaporative cryocooling excitation is compact, portable, and low cost. However, evaporative cryocooling cannot be considered a pulsed method due to its prolonged excitation duration. In general, it is difficult to measure thermal diffusivity based on non-impulsive pulsed excitation at times commensurate with the pulse duration, often due to ill-defined pulse shape and width and the subsequent potentially complicated thermal response which may be subject to diffusive broadening. To address this challenge, inverse physics-informed neural networks (IPINNs) were introduced in this work and integrated with an evaporative cryocooling method. The Parker method combined with a photothermal method was employed as a reference. To improve the accuracy of both IPINNs and Parker methods, terahertz time-domain spectroscopy (THz-TDS) was employed for measuring the thickness of impacted composites. Simulations and experimental results demonstrated the feasibility and accuracy of the IPINN-based approach.


  • THz-PINNs: Time-Domain Forward Modeling of Terahertz Spectroscopy With Physics-Informed Neural Networks

    by Pengfei Zhu, Xavier Maldague

    Terahertz time-domain spectroscopy (THz-TDS) is a powerful tool for extracting optical and electrical parameters, as well as probing internal structures and surface morphology of materials. However, conventional simulation techniques, such as the finite-element method and finite-difference time-domain, face challenges in accurately modeling THz wave propagation. These difficulties arise from the broad operating bandwidth of THz-TDS (∼0.1–10 THz), which demands extremely high temporal resolution to capture pulse dynamics and fine spatial resolution to resolve microstructures and interface effects. To address these limitations, we introduce physics-informed neural networks (PINNs) into THz-TDS modeling for the first time. Through a comprehensive analysis of forward problems in the time domain, we demonstrate the feasibility and potential of PINNs as a powerful framework for advancing THz wave simulation and analysis.