Additional Research

Project Examples

(Ordered from recent to past, reflecting Ph.D. dissertations and related publications)

AI Meets MRI: Deep Learning for Enhanced Imaging and Diagnostics

Magnetic resonance imaging (MRI) plays a vital role in modern diagnostics, yet its high cost, lengthy scan times, and reliance on specialized infrastructure limit accessibility worldwide. Our research addresses these challenges through computational methods rooted in deep learning and informed by MRI physics. We develop algorithms that reconstruct high-quality images from limited or undersampled data, enabling faster and more efficient scans without compromising diagnostic utility. Our work includes techniques to improve reliability in accelerated imaging, transfer image quality between low- and high-field systems, and design reconstruction models that generalize across anatomies, protocols, and acquisition conditions. These methods are evaluated not only through quantitative benchmarks but also in clinically relevant settings, such as stroke detection. Rather than treating MRI as a purely data-driven problem, we integrate physical modeling, statistical reasoning, and rigorous validation to ensure our tools are robust, interpretable, and clinically applicable. The examples highlighted here reflect a selection of our published work; ongoing efforts continue to expand these capabilities in collaboration with clinicians, physicists, and data scientists.

Representative Publications
(# denotes supervised by X. Zhang; * denotes corresponding author: X. Zhang)
Few-shot deployment of pretrained MRI transformers in brain imaging tasks
M. Li#, G. Shen#, C.W. Farris, X. Zhang*
arXiv preprint arXiv:2508.05783
+Abstract
Machine learning using transformers has shown great potential in medical imaging, but its real-world applicability remains limited due to the scarcity of annotated data. In this study, we propose a practical framework for the few-shot deployment of pretrained MRI transformers in diverse brain imaging tasks. By utilizing the Masked Autoencoder (MAE) pretraining strategy on a large-scale, multi-cohort brain MRI dataset comprising over 31 million slices, we obtain highly transferable latent representations that generalize well across tasks and datasets. For high-level tasks such as classification, a frozen MAE encoder combined with a lightweight linear head achieves state-of-the-art accuracy in MRI sequence identification with minimal supervision. For low-level tasks such as segmentation, we propose MAE-FUnet, a hybrid architecture that fuses multiscale CNN features with pretrained MAE embeddings. This model consistently outperforms other strong baselines in both skull stripping and multi-class anatomical segmentation under data-limited conditions. With extensive quantitative and qualitative evaluations, our framework demonstrates efficiency, stability, and scalability, suggesting its suitability for low-resource clinical environments and broader neuroimaging applications.
Magnetic resonance image processing transformer for general accelerated image restoration
G. Shen#, M. Li#, S.W. Anderson, C.W. Farris, X. Zhang*
Scientific Reports — Nature, 2025, 15: 40064
+Abstract
Recent advancements in deep learning have enabled the development of generalizable models that achieve state-of-the-art performance across various imaging tasks. Vision Transformer (ViT)-based architectures, in particular, have demonstrated strong feature extraction capabilities when pre-trained on large-scale datasets. In this work, we introduce the Magnetic Resonance Image Processing Transformer (MR-IPT), a ViT-based image-domain framework designed to enhance the generalizability and robustness of accelerated MRI restoration. Unlike conventional deep learning models that require separate training for different acceleration factors, MR-IPT is pre-trained on a large-scale dataset encompassing multiple undersampling patterns and acceleration settings, enabling a unified framework. By leveraging a shared transformer backbone, MR-IPT effectively learns universal feature representations, allowing it to generalize across diverse restoration tasks. Extensive experiments demonstrate that MR-IPT outperforms both CNN-based and existing transformer-based methods, achieving superior quality across varying acceleration factors and sampling masks. Moreover, MR-IPT exhibits strong robustness, maintaining high performance even under unseen acquisition setups, highlighting its potential as a scalable and efficient solution for accelerated MRI. Our findings suggest that transformer-based general models can significantly advance MRI restoration, offering improved adaptability and stability compared to traditional deep learning approaches.
Regularization by neural style transfer for MRI field-transfer reconstruction with limited data
G. Shen#, Y. Zhu, M. Li#, R. McNaughton#, H. Jara, S.B. Andersson, C.W. Farris, S.W. Anderson, X. Zhang*
Frontiers in Artificial Intelligence, 2025, 8: 1579251
+Abstract
Recent advances in MRI reconstruction have demonstrated remarkable success through deep learning-based models. However, most existing methods rely heavily on large-scale, task-specific datasets, making reconstruction in data-limited settings a critical yet underexplored challenge. While regularization by denoising (RED) leverages denoisers as priors for reconstruction, we propose Regularization by Neural Style Transfer (RNST), a novel framework that integrates a neural style transfer (NST) engine with a denoiser to enable magnetic field-transfer reconstruction. RNST generates high-field-quality images from low-field inputs without requiring paired training data, leveraging style priors to address limited-data settings. Our experiment results demonstrate RNST’s ability to reconstruct high-quality images across diverse anatomical planes (axial, coronal, sagittal) and noise levels, achieving superior clarity, contrast, and structural fidelity compared to lower-field references. Crucially, RNST maintains robustness even when style and content images lack exact alignment, broadening its applicability in clinical environments where precise reference matches are unavailable. By combining the strengths of NST and denoising, RNST offers a scalable, data-efficient solution for MRI field-transfer reconstruction, demonstrating significant potential for resource-limited settings.
Learning to reconstruct accelerated MRI through K-space cold diffusion without noise
G. Shen#, M. Li#, C.W. Farris, S.W. Anderson, X. Zhang*
Scientific Reports — Nature, 2024, 14: 21877
+Abstract
Deep learning-based MRI reconstruction models have achieved superior performance these days. Most recently, diffusion models have shown remarkable performance in image generation, in-painting, super-resolution, image editing and more. As a generalized diffusion model, cold diffusion further broadens the scope and considers models built around arbitrary image transformations such as blurring, down-sampling, etc. In this paper, we propose a k-space cold diffusion model that performs image degradation and restoration in k-space without the need for Gaussian noise. We provide comparisons with multiple deep learning-based MRI reconstruction models and perform tests on a well-known large open-source MRI dataset. Our results show that this novel way of performing degradation can generate high-quality reconstruction images for accelerated MRI.
Attention hybrid variational net for accelerated MRI reconstruction
G. Shen#, B. Hao, M. Li#, C.W. Farris, I.C. Paschalidis, S.W. Anderson, X. Zhang*
APL Machine Learning, 2023, 1(4): 046116
+Abstract
The application of compressed sensing (CS)-enabled data reconstruction for accelerating magnetic resonance imaging (MRI) remains a challenging problem. This is due to the fact that the information lost in k-space from the acceleration mask makes it difficult to reconstruct an image similar to the quality of a fully sampled image. Multiple deep learning-based structures have been proposed for MRI reconstruction using CS, in both the k-space and image domains, and using unrolled optimization methods. However, the drawback of these structures is that they are not fully utilizing the information from both domains (k-space and image). Herein, we propose a deep learning-based attention hybrid variational network that performs learning in both the k-space and image domains. We evaluate our method on a well-known open-source MRI dataset (652 brain cases and 1172 knee cases) and a clinical MRI dataset of 243 patients diagnosed with strokes from our institution to demonstrate the performance of our network. Our model achieves an overall peak signal-to-noise ratio/structural similarity of 40.92 ± 0.29/0.9577 ± 0.0025 (fourfold) and 37.03 ± 0.25/0.9365 ± 0.0029 (eightfold) for the brain dataset, 31.09 ± 0.25/0.6901 ± 0.0094 (fourfold) and 29.49 ± 0.22/0.6197 ± 0.0106 (eightfold) for the knee dataset, and 36.32 ± 0.16/0.9199 ± 0.0029 (20-fold) and 33.70 ± 0.15/0.8882 ± 0.0035 (30-fold) for the stroke dataset. In addition to quantitative evaluation, we undertook a blinded comparison of image quality across networks performed by a subspecialty trained radiologist. Overall, we demonstrate that our network achieves a superior performance among others under multiple reconstruction tasks.
Bayesian reconstruction of magnetic resonance images using Gaussian processes
Y. Xu, C.W. Farris, S.W. Anderson, X. Zhang, K.A. Brown
Scientific Reports — Nature, 2023, 13: 12527
+Abstract
A central goal of modern magnetic resonance imaging (MRI) is to reduce the time required to produce high-quality images. Efforts have included hardware and software innovations such as parallel imaging, compressed sensing, and deep learning-based reconstruction. Here, we propose and demonstrate a Bayesian method to build statistical libraries of magnetic resonance (MR) images in k-space and use these libraries to identify optimal subsampling paths and reconstruction processes. Specifically, we compute a multivariate normal distribution based upon Gaussian processes using a publicly available library of T1-weighted images of healthy brains. We combine this library with physics-informed envelope functions to only retain meaningful correlations in k-space. This covariance function is then used to select a series of ring-shaped subsampling paths using Bayesian optimization such that they optimally explore space while remaining practically realizable in commercial MRI systems. Combining optimized subsampling paths found for a range of images, we compute a generalized sampling path that, when used for novel images, produces superlative structural similarity and error in comparison to previously reported reconstruction processes (i.e. 96.3% structural similarity and < 0.003 normalized mean squared error from sampling only 12.5% of the k-space data). Finally, we use this reconstruction process on pathological data without retraining to show that reconstructed images are clinically useful for stroke identification. Since the model trained on images of healthy brains could be directly used for predictions in pathological brains without retraining, it shows the inherent transferability of this approach and opens doors to its widespread use.
Distributionally robust image classifiers for stroke diagnosis in accelerated MRI
B. Hao, G. Shen#, R. Chen, C.W. Farris, S.W. Anderson, X. Zhang, I. Paschalidis
Medical Image Computing and Computer Assisted Intervention − MICCAI 2023, Springer, 2023, 14224: 768-777
+Abstract
Magnetic Resonance Imaging (MRI) acceleration techniques using k-space sub-sampling (KS) can greatly improve the efficiency of MRI-based stroke diagnosis. Although Deep Neural Networks (DNN) have shown great potential on stroke lesion recognition tasks when the MR images are reconstructed from the full k-space, they are vulnerable to the lower quality MR images generated by KS. In this paper, we propose a Distributionally Robust Learning (DRL) approach to improve the performance of stroke recognition DNN models when the MR images are reconstructed from the sub-sampled k-space. For Convolutional Neural Network (CNN) and Vision Transformer (ViT)-based models, our methods improve the stroke classification AUROC and AUPRC by up to 11.91% and 9.32% on the KS-perturbed brain MR images, respectively, compared against Empirical Risk Minimization (ERM) and other baseline defensive methods. We further show that DRL models can successfully recognize the stroke cases from highly perturbed MR images where clinicians may fail, which provides a solution for improved diagnosis in an accelerated MRI setting.
Bending, Deformation, and Thermomechanical Behavior

Project Example 1:
Multilayer cantilever structures are widely used in micro/nanosystems, yet their manufacturability, planarity, and reliability have long presented significant challenges. While much of our understanding of the thermomechanical behavior of layered systems has originated from microelectronics, it is important to recognize the substantial differences present in many MEMS applications. Fully understanding these distinctions is essential for optimizing the design of reliable micro/nanosystems. The overarching goal of this research is to uncover the deformation mechanisms inherent to MEMS thin-film materials, establish their connection to MEMS design and analysis, and apply these insights to improve device performance. This work advances our understanding of the interplay between the structure and properties of free-standing thin films, with potentially broad implications for both microelectronics and MEMS-based devices and systems.

Project Example 2:
Electroactive conducting polymers are known for their unique redox-state-dependent mechanical responses, which can be precisely controlled by various stimuli. When electrons are donated to or removed from the polymer chains, the conformations of the polymers undergo significant relaxation, accompanied by the insertion or extraction of compensating ionic species. These combined electronic and ionic changes lead to macroscopic volume variations in the polymer. Characterized by low activation voltage, large stress and strain, and the ability to operate in liquid electrolytes—including physiological saline solutions—conducting polymers are promising materials for artificial muscles. This research focuses on actuator applications of conducting polymers, where volume change is driven by ion and solvent migration. A general multilayer beam model for conducting polymer-based actuators has been developed. While electrostatic and piezoelectric materials have traditionally dominated the field of electromechanical actuation, conducting polymers offer attractive features that more closely mimic the behavior of natural muscles.

Representative Publications
(# denotes supervised by X. Zhang; * denotes corresponding author: X. Zhang)
Inelastic deformation of bilayer microcantilevers with nanoscale coating
I-K Lin#, X. Zhang*, Y. Zhang*
Sensors and Actuators A: Physical, 2011, 168(1): 1-9
+Abstract
The application and commercialization of microelectromechanical system (MEMS) devices suffer from reliability problems due to the structural inelastic deformation during device operation. Nanocoatings have been demonstrated to be promising solutions for suppressing creep and stress relaxation in bilayer MEMS devices. However, the micro/nano-mechanics within and/or between microcantilevers and coatings are not fully understood, especially when temperature, time, and geometric and material nonlinearities play significant roles in the thermomechanical responses. In this study, the thermomechanical behavior of alumina-coated/uncoated Au/SiNx bilayer microcantilevers was characterized by using thermal cycling and isothermal holding tests. Finite element analysis with power-law creep was used to simulate the mechanical behavior of microcantilevers during isothermal holding. To better understand the stress evolution and the mechanism of inelastic deformation, scanning electron microscopy and atomic force microscopy was employed to explore the grain growth and grain boundary grooving after isothermal holding at various temperatures of 100 °C, 150 °C and 200 °C. The methods and results presented in this paper are useful for the fundamental understanding of many similar bilayer microcantilever-based MEMS devices.
A multilayer bending model for conducting polymer actuators
P. Du#, X. Lin, X. Zhang*
Sensors and Actuators A: Physical, 2010, 163(1): 240-246
+Abstract
Electroactive conducting polymers (CPs) have been frequently used for fabricating bending actuators. To model this type of actuation, the traditional double-layer beam bending theory was implemented by neglecting the thickness of the thin intermediate metal layers for the sake of simplification. However, this common assumption has not been carefully validated and the associated errors have not been well acknowledged. In this work, a generic multilayer bending model was introduced to account for the actuators consisting of an arbitrary number of layers. Our model found the bending curvature, strain, stress, and in particular work density of the multilayer actuator as explicit functions of the thickness and modulus of each individual layer. The thickness of metals and conducting polymers were controlled in thermal evaporation and electrochemical synthesis, respectively. The modulus of polypyrrole (PPy), the conducting polymer used in this work, was determined within our model by the bending curvature measured using the charge-coupled device (CCD). This gave a modulus of our electrochemically synthesized PPy of 80 MPa, corresponding to an actuation strain of 2% in our model. It was concluded that neglecting the intermediate metal layers would lead to substantial errors. For instance, using a PPy/Au/Kapton trilayer actuator, a 5% error or below in strain can only be found if the Au layer is one thousand times thinner than Kapton. To enhance the actuation, a PPy/Pt/PVDF/Pt/PPy five-layer actuator has been often used. In this case, even if the Pt layer was reduced to 10 nm, our predicted error of neglecting the two metal layers would be 12.59%. Our results showed that the work density, chosen to measure the overall performance of the actuator, was highly sensitive to the modulus of the substrate polymer layer so that it was generally desirable of using a soft polymer substrate. With the multilayer bending model, we intend to provide an accurate and reliable tool for systematically analyzing the bending behavior and performance of the CP-based actuators.
Thermomechanical behavior and microstructural evolution of SiNx/Al bimaterial microcantilevers
I-K Lin#, X. Zhang*, Y. Zhang*
Journal of Micromechanics and Microengineering, 2009, 19(8): 085010
+Abstract
Bimaterial microcantilevers are used in numerous applications in microelectromechanical systems (MEMS) for thermal, mechanical, optical, tribological and biological functionalities. Unfortunately, the residual stress-induced curvature and combined effects of creep and stress relaxation in the thin film significantly compromises the performance of these structures. To fully understand the themomechanical deformation and microstructural evolution of such microcantilevers, SiNx/Al bilayer cantilever beams were studied in this work. These microcantilevers were heated and subsequently cooled for five cycles between room temperature and 250 °C, with the peak temperature in each successive cycle increased in increments of 25 °C using a custom-built micro-heating stage. The in situ curvature change was monitored using an interferometric microscope. The general behavior of the bimaterial microcantilever beams can be characterized by linear thermoelastic regimes with (dκ/dT)ave = 0.079 mm−1 °C−1 and inelastic regimes. After thermal cycling with a maximum temperature of 225 °C, upon returning to room temperature, the bimaterial microcantilever beams were flattened and the curvature decreased by 99%. The thermoelastic deformation during thermal cycling was well described by the Kirchhoff plate theory. Deformation of bimaterial microcantilevers during long-term isothermal holding was studied at temperatures of 100 °C, 125 °C and 150 °C with a holding period of 70 h. The curvature of bimaterial microcantilever beams decreased more for higher holding temperatures. Finite element analysis (FEA) with power-law creep in Al was used to simulate the creep and stress relaxation and thus the curvature change of the bimaterial microcantilever beams. The microstructure evolutions due to isothermal holding in SiNx/Al microcantilevers were studied using an atomic force microscope (AFM). The grain growth in both the vertical and lateral directions was present due to isothermal holding. As the isothermal holding temperature increased, the surface roughness of the film increased with more prominent grain structures.
The deformation of microcantilever-based infrared detectors during thermal cycling
I-K Lin#, Y. Zhang, X. Zhang*
Journal of Micromechanics and Microengineering, 2008, 18(7): 075012
+Abstract
Uncooled microcantilever-based infrared (IR) detectors have recently gained interest due to their low noise equivalent temperature difference (NETD), while concurrently maintaining low costs. These properties have made them available for a wider range of applications. However, the curvature induced by residual strain mismatch severely compromises the device’s performance. Therefore, to meet performance and reliability requirements, it is important to fully understand the deformation of IR detectors. In this study, bimaterial (SiNx/Al) microcantilever-based IR detectors were fabricated using surface micromachining with polyimide as a sacrificial layer. Thermo-mechanical deformation mechanisms were studied through the use of thermal cycling. A temperature chamber with accurate temperature control and an interferometer microscope were adopted in this study for thermal cycling and full-field curvature measurements. It was found that thermal cycling reduced the residual strain mismatch within the bimaterial structure and thus flattened the microcantilever-based IR detectors. Specifically, thermal cycling with a maximum temperature of 295 °C resulted in a 97% decrease in curvature of the microcantilever-based IR detectors upon return to room temperature. The thermoelastic deformation of the IR detectors was modeled using both finite element method (FEM) and analytical methods. A modified analytical solution based on plate theory was established to describe the thermoelastic mechanical responses by using a correction factor derived from FEM. Although in the current study Al and SiNx were chosen for the application of microcantilever-based IR detectors, the general experimental protocol and modeling approach can be applied to describe thermoelastic mechanical responses of bimaterial devices with different materials. Toward the end of this paper, we studied the correction factors in the modified analytical solution while varying parameters such as Young’s modulus ratio, thickness ratio and coefficient of thermal expansion (CTE) mismatch to investigate the influences of these parameters.
Development of double-cantilever infrared detectors: Fabrication, curvature control and demonstration of thermal detection
S. Huang#, H. Tao#, I-K Lin#, X. Zhang*
Sensors and Actuators A: Physical, 2008, 145-146: 231-240
+Abstract
his paper reports the recent progress on the development of double-cantilever infrared (IR) detectors, including the fabrication, the post-process curvature control, and also the first-time demonstration of thermal detection using capacitive-based IR focal plane arrays (FPAs). In this work, simplified double-cantilever IR FPAs based on bimaterial SiNx/Al and Al/SiNx cantilevers are fabricated using a surface micromachining module with polyimide as the sacrificial material. Thermal-cycling experiments of both 200 nm-thick Ebeam Al and 200 nm-thick PECVD SiNx films reveal that the residual stresses in IR materials can be significantly modified by thermal annealing. Therefore, an engineering approach to flattening IR FPAs is developed by using rapid thermal annealing (RTA). This article also demonstrates the thermal detection of cantilever IR FPAs using commercialized weak capacitance readout IC.
Study of gradient stress in bimaterial cantilever structures for infrared applications
S. Huang# and X. Zhang*
Journal of Micromechanics and Microengineering, 2007, 17(7): 1211-1219
+Abstract
Bimaterial SiNx/Al infrared cantilever structures are always initially curved because of the imbalanced residual stress in the two layers. Their performance and functionality are therefore significantly decreased. A thorough study of the residual stress (strain) has then become a key issue in the development of bimaterial SiNx/Al cantilever structures. In the curvature-based approach to the film stress, the residual strain is derived from the measured curvature based on certain assumptions on the distribution of the residual strain in the thickness direction. Previous models for a bimaterial cantilever structure, however, are not sufficient to characterize the residual strain in bimaterial SiNx/Al infrared structures. The main goal of this paper is to investigate gradient residual strain in bimaterial SiNx/Al infrared structures. To achieve this goal, the relationship between the residual strain and bending curvature is developed with the assumption that the residual strain in each layer is linearly distributed rather than uniform throughout the thickness. The profile of the gradient strain is then derived from the curvatures measured during the continuous etching of the top-most SiNx in the bimaterial cantilevers. The derived residual strain can then be inverted to predict curvature change further in the etching process. This paper demonstrates that a linear assumption of the residual strain yields a stronger agreement with the measured data in comparison to previously used models. In addition, several factors that may affect measurement accuracy are discussed at the end of the paper.
Gradient residual stress induced elastic deformation of multilayer MEMS structures
S. Huang# and X. Zhang*
Sensors and Actuators A: Physical, 2007, 134(1): 177-185
+Abstract
Multilayered structures are widely used as sensing or actuating components in MEMS devices. Since the thin films of multilayered structures are always subject to residual stresses, it is important to model the relation between these residual stresses and the resultant elastic deformation. The main purpose of this paper is to explore two different approaches to addressing this issue when the residual stress in each thin film is not necessarily uniform throughout the thickness. These two approaches are first briefly introduced and then used to arrive at identical solutions for a monolayer cantilever and a bilayer cantilever, both with arbitrary residual strain distributions throughout the thickness. The analytical formulas for a bilayer cantilever are further verified by the numerical simulation of a special case. After the discussion on the errors induced by assuming the gradient residual strains in the bilayer cantilevers are uniform, the relation between the bending plane and the neutral plane in bilayer cantilevers is also explored. Finally, we present an approach to characterizing residual stresses in thin films by using micromachined bilayer cantilevers in conjunction with the theory developed in this paper.
Elimination of stress-induced curvature in microcantilever infrared focal plane arrays
S. Huang#, B. Li, and X. Zhang*
Sensors and Actuators A: Physical, 2006, 130-131: 331-339
+Abstract
This paper reports an approach to eliminating stress-induced curvature in microcantilever-based infrared focal plane arrays (FPAs). Using a combination of argon ion beam machining and rapid thermal annealing (RTA), we successfully modified curvatures of free-standing SiNx/Al bimaterial FPAs. The SiNx/Al FPAs were fabricated using a surface micromachining technique with polyimide as a sacrificial material. The as-fabricated FPAs were concavely curved because of the imbalanced residual stresses in the two materials. To modify the FPAs curvature, first, Ar ions with energies of 500 eV was used, which sputter etched PECVD SiNx at a rate of 4 nm/min, and 20 min of ion beam machining reduced the FPAs curvature from –1.92 to –0.96 mm−1. Then based on the investigation on the thermomechanical behavior of both the e-beam Al and PECVD SiNx films during the thermal cycling, RTA was proposed to further modify the FPAs curvature. It is found that 5 min of RTA at 375 °C resulted in flat FPAs with acceptable curvatures (<0.10 mm−1).
Extension of the Stoney formula for film–substrate systems with gradient stress for MEMS applications
S. Huang# and X. Zhang*
Journal of Micromechanics and Microengineering, 2006, 16(2): 382-389
+Abstract
Using the Stoney formula and its modifications, curvature-based techniques are gaining increasingly widespread application in evaluating the stress in a film on a substrate. In principle, the formula applies only when the stress is uniform throughout the film thickness. The main purpose of this paper is to extend the Stoney formula when the residual strain in the film is no longer uniform, but dependent on the z position. To achieve this goal, a general theory was introduced for the elastic deformation of an arbitrary, multilayered system. By practicing this general theory, we used a polynomial function to describe the gradient stress in a film, and contributions by different elements of the polynomial to both the curvature and the bending strain were derived. A finite element simulation for a typical film–substrate structure was then carried out, leading to the verification of the theory developed in this paper. In the discussion section, we explored the relation between the surface curvature and the bending curvature as well as the difference between the stress in the constrained planar state and that in the relaxed state. In addition, the accuracy of the simplified formula, using thin film approximation, was evaluated. Finally, a SiNx-Al MEMS structure was studied by using the formula in this paper.
Ph.D. Dissertation
Mechanical and material characterization of bilayer microcantilevers for MEMS-based IR detector applications
I-Kuan Lin, Ph.D. Dissertation, Boston University
(Advisor: Xin Zhang; January 2011)
Development of double-cantilever infrared detectors
Shusen Huang, Ph.D. Dissertation, Boston University
(Advisor: Xin Zhang; May 2007)
Controlled Synthesis and Characterization of Micro/Nanoscale Materials

Project Example 1:
This research aims to synthesize copper oxide (CuO) nanowires (NWs) with controlled properties and to develop a novel sensing platform based on a single CuO NW coated with discrete nanoparticles for use in miniaturized gas and photonic sensors with superior performance. The project focuses on three main goals: (1) controlling the structure, morphology, and electronic properties of CuO NWs by tuning the oxidative environment during growth; (2) enabling the rapid growth of long CuO NWs from nanocrystalline copper produced via surface mechanical attrition treatment (SMAT) of standard copper substrates; and (3) fabricating, characterizing, and packaging single CuO NW sensors with enhanced gas and photon sensing capabilities through nanoparticle surface decoration. Copper oxide represents a state-of-the-art material in nanowire research, offering low cost, excellent processability, and ease of integration into micro-templates specifically designed for sensing applications.

Project Example 2:Research
Amorphous thin-film materials are increasingly used in emerging MEMS applications, yet their mechanical behavior remains less understood compared to their crystalline counterparts. The primary objective of this research is to advance scientific understanding of the underlying mechanisms that govern the mechanical responses of amorphous thin films. This deeper insight will enable the design and fabrication of more efficient MEMS structures and devices. The project systematically investigates the mechanical behavior of thin films across a range of thermal conditions, stress levels, and size scales, encompassing both elastic and plastic regimes. Distinct mechanical characteristics are thoroughly characterized, and the physical mechanisms driving these responses are analyzed in detail.

Representative Publications
(# denotes supervised by X. Zhang; * denotes corresponding author: X. Zhang)
Ballistic transport enhanced heat convection at nanoscale hotspots
S. Xu, Y. Xu, J. Zhang, J. Gao, X. Wang, X. Zhang, Y. Yue#
Journal of Applied Physics, 2024, 136(16): 164306
+Abstract
Along with device miniaturization, severe heat accumulation at unexpected nanoscale hotspots attracts wide attentions and urges efficient thermal management. Heat convection is one of the important heat dissipating paths at nanoscale hotspots but its mechanism is still unclear. Here shows the first experimental investigation of the convective heat transfer coefficient at size-controllable nanoscale hotspots. A specially designed structure of a single-layer graphene supported by gold-nanorod array is proposed, in which the gold nanorods generate hundreds of nanometers heating sources under laser irradiation and the graphene layer works as a temperature probe in Raman thermometry. The determined convective heat transfer coefficient (1928+155 −147 W m−2 K−1 for the 330 nm hotspot and 1793+157 −159 W m−2 K−1 for the 240 nm hotspot) is about three orders of magnitude higher than that of nature convection, when the simultaneous interfacial heat conduction and radiation are carefully evaluated. Heat convection, thus, accounts to more than half of the total energy transferred across the graphene/gold nanorods interface. Both the plasmon induced nanoscale hotspots and ballistic convection of air molecules contribute to the enhanced heat convection. This work reveals the importance of heat convection at nanoscale hotspots to the accurate thermal design of miniaturized electronics and further offers a new way to evaluate the convective heat transfer coefficient at nanoscale hotspots.
High thermal conductivity of free-standing skeleton in graphene foam
J. Gao, D. Xie, X. Wang, X. Zhang, Y. Yue#
Applied Physics Letters, 2020, 117(25): 251901
+Abstract
Due to the high-porosity structure, the low thermal transport property of graphene foam (GF) is expected. However, the interconnected skeleton can still act as excellent thermal conductor branches if phonon scattering is not severely affected in the structure of graphene flakes. Such a property has not been validated experimentally due to the difficulty in sample manipulation and the fragility of the structure. In this work, we report the characterization results of thermal properties of the free-standing skeleton in GF. Three individual skeleton samples from one GF piece are prepared under the same condition. The thermal diffusivity of GF skeletons is characterized in the range of 3.26–3.48 × 10−4 m2/s, and the thermal conductivity is determined to be 520–555 W/(m K), which is two orders of magnitude larger than the value of bulk GF. These high thermal conductivity values originate from the intrinsic thermal property of graphene, while the contact interfaces, wrinkled structures, and defects induced in the synthesis process do not affect the phonon transport property significantly, which proves that the three-dimensional hierarchical graphene structure can still be implemented in energy-intensive applications.
Graphene nanofluids as thermal management materials: Molecular dynamics study on orientation and temperature effects
J. Gao, H. Wu, A. Li#, Y. Yue#, D. Xie, X. Zhang*
ACS Applied Nano Materials, 2019, 2(11): 6828-6835
+Abstract
The dispersion of graphene nanoparticles in a fluid can enhance the thermophysical properties of the base liquid. The physics behind such a phenomenon has yet to be uncovered in the community. In this work, thermal transport in a graphene–water mixture is studied by classical molecular dynamics simulations. Several factors including orientation angle, curvature, thermal rectification, temperature, and van der Waals interaction are investigated, and special attention is paid to the effect on thermal conductance across graphene–water interfaces. It is found that thermal conductance increases from 13.92 to 26.70 MW/m2 K as the orientation angle is increased from 0° to 90°. When the curved graphene is introduced by altering the length to width ratio from 1.0 to 1.8, the thermal conductance is elevated. However, as the length to width ratio exceeds 1.8, such a trend does not continue due to the variation of the intrinsic thermal conductivity of graphene and the formation of the complex graphene–water interface. Even though the curved graphene introduces an asymmetric assembly, no thermal rectification effect is observed for diverse directions of heat flux. It is demonstrated that the enhancement of overall thermal conductance of nanofluids is ascribed to the interface thermal transport rather than the base liquid with increasing temperature. This correlation is suppressed in a hydrophilic interface due to the structural change of liquid layer adjacent to the interface.
Plasmonic heating induced by Au nanoparticles for quasi-ballistic thermal transport in multi-walled carbon nanotubes
Y. Xu+, X. Zhao#+, A. Li#, Y. Yue#, J. Jiang, X. Zhang*
Nanoscale, 2019, 11(16), 7572-7581
+Abstract
The plasmon resonances of nanostructures enable wide applications from highly sensitive sensing to high-resolution imaging, through the improvement of photogeneration rate stimulated by the local field enhancement. However, quantitative experimental studies on the localized heating and the thermal transport process in the vicinity of plasmonics are still lacking because of the diffraction limit in conventional optothermal methodologies. In this work, we demonstrate an approach based on Raman thermometry to probe the near-field heating caused by plasmonics. An array of Au nanoparticles (AuNPs) fabricated by the template-assisted method is used to generate the near field effect. Multi-walled carbon nanotubes (MWCNTs) dispersed on the AuNPs are employed to quantify the near-field heating from their Raman peak shifts. Results show that the temperature rise in MWCNTs on AuNPs is much higher than that in a control group under the same laser irradiation. Further analysis indicates that the enhanced photon absorption of MWCNTs attributed to plasmon resonances is partially responsible for the different heating effect. The nonuniform thermal hot spots at the nanoscale can result in the quasi-ballistic thermal transport of phonons in MWCNTs, which is another reason for the temperature rise. Our results can be used to understand plasmonic heating effects as well as to explore quasi-ballistic thermal transport in carbon-based low-dimensional materials by tailoring the geometry or size of plasmonic nanostructures.
+Abstract
Silicon oxynitride (SiON) is an important material to fabricate micro-electro-mechanical system (MEMS) devices due to its composition-dependent tunability in electronic and mechanical properties. In this work, the SiON film with 41.45% silicon, 32.77% oxygen and 25.78% nitrogen content was deposited by RF magnetron sputtering. Two types of optimized micro-structures including micro-cantilevers and micro-rotating-fingers were designed and fabricated using MEMS surface micromachining technology. The micro-cantilever bending tests were conducted using a nanoindenter to characterize the Young’s modulus of the SiON film. Owing to the elimination of the residual stress effect on the micro-cantilever structure, higher accuracy in the Young’s modulus was achieved from this technique. With the information of Young’s modulus of the film, the residual stresses were characterized from the deflection of the micro-rotating-fingers. This structure was able to locally measure a large range of tensile or compressive residual stresses in a thin film with sufficient sensitivities. The results showed that the Young’s modulus of the SiON film was 122 GPa and the residual stresses of the SiON film were 327 MPa in the crystallographic orientation of the wafer and 334 MPa in the direction perpendicular to the crystallographic orientation, both in compression. This work presents a comprehensive methodology to measure the Young’s modulus and residual stresses of a thin film with improved accuracy, which is promising for applications in mechanical characterization of MEMS devices.
Effects of composition and thermal annealing on the mechanical properties of silicon oxycarbide films
P. Du#, X. Wang#, I-K Lin#, X. Zhang*
Sensors and Actuators A: Physical, 2012, 176: 90-98
+Abstract
There is an increasing trend to incorporate silicon carbide (SiC) into silicon oxides to improve the mechanical properties, thermal stability, and chemical resistance. In this work the silicon oxycarbide (SiOC) films were deposited by RF magnetron co-sputtering from silicon dioxide and silicon carbide targets. Subsequently rapid thermal annealing was applied to the as-deposited films to tune the mechanical properties. Energy dispersive spectroscopy, scanning electron microscopy, Fourier transform infrared spectroscopy and ellipsometry were employed to characterize the compositions and microstructure of the films. The residual stress of the films was calculated from the film–substrate curvature measurement using Stoney’s equation. The film stress changed from compressive to tensile after annealing, and it generally increased with carbon contents. The Young’s modulus and hardness were investigated by the depth-sensing nanoindentation, which were found to increase with the carbon content and annealing temperature. A thorough microstructural analysis was conducted to investigate the effect of carbon content and annealing temperature on the mechanical properties of SiOC films.
Mechanical property characterization of sputtered and plasma enhanced chemical deposition (PECVD) silicon nitride films after rapid thermal annealing
P-H Wu, I-K Lin#, H-Y Yan, K-S Ou, K-S Chen, X. Zhang*
Sensors and Actuators A: Physical, 2011, 168(1): 117-126
+Abstract
In this paper, the mechanical and fracture properties of silicon nitride films subjected to rapid thermal annealing (RTA) have been systemically tested. The residual stress, Young’s modulus, hardness, fracture toughness, and interfacial strength of both sputtered and plasma-enhanced chemical vapor deposition (PECVD) silicon nitride films deposited on silicon wafers were measured and compared. The results indicated that the Young’s modulus and hardness of both types of silicon nitride films significantly increased when the RTA temperature increased. Furthermore, RTA processes could also alter the state of residual stress. The initial residual compressive stress of sputtered silicon nitride film was gradually relieved, and the film became tensile after the RTA process. For PECVD silicon nitride, the tensile residual stress reached its peak after a 600 °C RTA, then dropped after further increases in RTA temperature, due to stress relaxation. The tendency of the equivalent fracture toughness was to exhibit a strong correlation with that shown in the residual stress of silicon nitride. By considering the effect of residual stress, the real fracture toughness of both types of silicon nitride films were slightly enhanced by using RTA processes. Finally, experimental results indicated that the interfacial strength of PECVD silicon nitride could also be significantly improved by RTA processes at 600–800 °C. On the other hand, the initial interfacial strength of the sputtered silicon nitride was sufficiently strong, and the RTA processes only resulted in minor improvements. The characterization flow could be applied to other brittle films, and these specific test results should be useful for improving the structural integrity and process optimization of related MEMS and IC applications.
Mechanical properties of sputtered silicon oxynitride films by nanoindentation
Y. Liu#, I-K Lin#, X. Zhang*
Materials Science and Engineering A, 2008, 489(1-2): 294-301
+Abstract
Silicon oxynitride (SiON) has received a great deal of attention in micro-electro-mechanical system (MEMS) integration due to its composition-dependent tunability in optical, electronic and mechanical properties. In this work, silicon oxynitride films with different oxygen and nitrogen content were deposited by RF magnetron sputtering. Energy dispersive X-ray (EDX) spectroscopy and Fourier-transform infrared (FT-IR) spectroscopy were employed to characterize the SiON films with respect to stoichiometric composition and atomic bonding structure. Time-dependent plastic deformation (creep) of SiON films were investigated by depth-sensing nanoindentation at room temperature. Young’s modulus and indentation-hardness were found correlated with the nitrogen/oxygen ratio in SiON films. Results from nanoindentation creep indicated that plastic flow was less homogenous with increasing nitrogen content in film composition. Correspondingly, a deformation mechanism based on atomic bonding structure and shear transformation zone (STZ) plasticity theory was proposed to interpret creep behaviors of sputtered SiON films.
+Abstract
Plasma-enhanced chemical vapor deposited (PECVD) silicon oxide (SiOx) thin films have been widely used in Micro/Nano Electro Mechanical Systems to form electrical and mechanical components. In this paper, we explore the use of nanoindentation techniques as a method of measuring equivalent stress–strain curves of the PECVD SiOx thin films. Four indenter tips with different geometries were adopted in our experiments, enabling us to probe the elastic, elasto-plastic, and fully plastic deformation regimes of the PECVD SiOx thin films. The initial yielding point (σI) and stationary yielding point (σII) are separately identified for the as-deposited and annealed PECVD SiOx thin films, as well as a standard fused quartz sample. Based on the experimental results, a shear transformation zone based amorphous plasticity theory is applied to depict the plastic deformation mechanism in the PECVD SiOx.
Nanoindentation Creep of Plasma-Enhanced Chemical Vapor Deposited Silicon Oxide Thin Films
Z. Cao# and X. Zhang*
Scripta Materialia, 2007, 56(3): 249-252
+Abstract
The time-dependent plastic properties of both as-deposited and annealed plasma-enhanced chemical vapor deposited (PECVD) silicon oxide (SiOx) thin films were probed by nanoindentation creep tests at room temperature. Our experiments found a strong size effect in the creep responses of the as-deposited PECVD SiOx thin films, which was much reduced after annealing. Based on the experimental results, the deformation mechanism is depicted by the ‘shear transformation zone’ (STZ) based amorphous plasticity theories.
Size-dependent creep behaviour of plasma-enhanced chemical vapour deposited silicon oxide films
Z. Cao# and X. Zhang*
Journal of Physics D: Applied Physics, 2006, 39(23): 5054-5063
+Abstract
The time-dependent plastic deformation (creep) behaviours of both as-deposited and annealed plasma-enhanced chemical vapour deposited (PECVD) silicon oxide (SiOx) films were probed by nanoindentation load relaxation tests at room temperature. Our experiments found a strong size effect in the creep responses of the as-deposited PECVD SiOx thin films, which was much reduced after rapid thermal annealing. Based on the experimental results, the deformation mechanism is depicted by the ‘shear transformation zone’ (STZ)-based amorphous plasticity theories. The physical origin of the STZ is elucidated and linked with the shear banding dynamics. It is postulated that the high strain gradient at shallow indentation depths may be responsible for the reduction in the stress exponent n = ∂ log(strain rate)/∂ log(stress), characteristic of a more homogeneous flow behaviour.
+Abstract
Plasma-enhanced chemical vapor deposited (PECVD) silane-based silicon oxide (SiOx) films were chosen as an example to study the thermally-induced stress relaxation phenomena of amorphous dielectric films. Wafer curvature was measured optically both during and after various thermal conditions, including temperature cycling, constant peak temperature annealing, and varying peak temperature annealing experiments. From these measurements, we are able to obtain a complete stress evolutional history in the thin film, and further derive a series of mechanical/material properties such as viscoelasticity, density and viscosity, coefficient of thermal expansion (CTE), etc. We found that the stress in the amorphous silicon oxide films is sensitive to both viscosity and density changes. Stress relaxation was then theoretically investigated by a viscoelastic model, which was associated with a defects-based microstructural causal mechanism. Our theoretical model elucidates defects movement and consequent microstructure rearrangement as a source of damping, accompanied by viscous flow. This theory was applied to explain a series of experimental results, including stress hysteresis generation and reduction, stress relaxation, and CTE changes, etc.
Microbridge testing of plasma-enhanced chemical-vapor deposited silicon oxide films on silicon wafers
Z. Cao#, T-Y Zhang, X. Zhang*
Journal of Applied Physics, 2005, 97(10): 104909
+Abstract
Plasma-enhanced chemical-vapor deposited (PECVD) silane-based oxides (SiOx) have been widely used in both microelectronics and microelectromechanical systems (MEMS) to form electrical and/or mechanical components. In this paper, a nanoindentation-based microbridge testing method is developed to measure both the residual stresses and Young’s modulus of PECVD SiOx films on silicon wafers. Theoretically, we considered both the substrate deformation and residual stress in the thin film and derived a closed formula of deflection versus load. The formula fitted the experimental curves almost perfectly, from which the residual stresses and Young’s modulus of the film were determined. Experimentally, freestanding microbridges made of PECVD SiOx films were fabricated using the silicon undercut bulk micromachining technique. Some microbridges were subjected to rapid thermal annealing (RTA) at a temperature of 400 °C, 600 °C, or 800 °C to simulate the thermal process in the device fabrication. The results showed that the as-deposited PECVD SiOx films had a residual stress of –155 ± 17 MPa and a Young’s modulus of 74.8 ± 3.3 GPa. After the RTA, Young’s modulus remained relatively unchanged at around 75 GPa, however, significant residual stress hysteresis was found in all the films. A microstructure-based mechanism was then applied to explain the experimental results of the residual stress changes in the PECVD SiOx films after the thermal annealing.
+Abstract
The structural relaxation of plasma-enhanced chemical-vapor-deposited (PECVD) silane-based silicon oxide films during thermal cycling and annealing has been studied using wafer curvature measurements. These measurements, which determine stress in the amorphous silicon oxide films, are sensitive to both plastic deformation and density changes. A quantitative case study of such changes has been done based upon the experimental results. A microstructure-based mechanism elucidates seams as a source of density change and voids as a source of plastic deformation, accompanied by a viscous flow. This theory was then used to explain a series of experimental results that are related to thermal cycling as well as annealing of PECVD silicon oxide films including stress hysteresis generation and reduction and coefficient of thermal-expansion changes. In particular, the thickness effect was examined; PECVD silicon oxide films with a thickness varying from 1 to 40 µm were studied, as certain demanding applications in microelectromechanical systems require such thick films serving as heat∕electrical insulation layers.
Ph.D. Dissertation
Mechanical behaviors of PECVD dielectric films for MEMS applications
Zhiqiang Cao, Ph.D. Dissertation, Boston University
(Advisor: Xin Zhang; January 2007)
The growth and characterization of copper (II) oxide nanowires with single nanowire electrical, gas sensing, and photoconduction measurements
Benjamin Hansen, M.S. Thesis, Boston University
(Advisor: Xin Zhang; May 2009)
Diatom-Enabled Functional Materials

ResearchAs a major group of microalgae, diatoms have long served as both inspiration and source material across diverse research fields. The intricate morphology of their abundant nanostructured silicon dioxide exoskeletons—known as frustules—has attracted significant scientific attention. Diatom frustules have been employed in fluorescence immunoassays, where their nanopores provide large surface areas for biomolecular binding, and as anode materials in dye-sensitized solar cells due to their excellent light-scattering properties. Although the nanostructures on diatom frustules are remarkably uniform, they have rarely been leveraged for scalable nanopatterned manufacturing. This project advances our understanding of how to harness diatom frustule nanostructures at scale by investigating their interactions with surrounding gas, liquid, and solid environments. These insights enable the fabrication of large-area, uniform frustule assemblies, followed by two- and three-dimensional replication of their nanostructures. In particular, the photonic properties of these assemblies and their replicas are studied and tailored for targeted photonic applications.

Representative Publications
(# denotes supervised by X. Zhang; * denotes corresponding author: X. Zhang)
Diatom Cribellum-inspired hierarchical metamaterials: Unifying perfect absorption towards subwavelength color printing
X. Xie#, Y. Huang#, Z. Yang#, A. Li#, X. Zhang*
Advanced Materials, 2024, 36(33): 2403304
+Abstract
Diatom exoskeletons, known as frustules, exhibit a unique multilayer structure that has attracted considerable attention across interdisciplinary research fields as a source of biomorphic inspiration. These frustules possess a hierarchical porous structure, ranging from millimeter-scale foramen pores to nanometer-scale cribellum pores. In this study, this natural template for nanopattern design is leveraged to showcase metamaterials that integrates perfect absorption and subwavelength color printing. The cribellum-inspired hierarchical nanopatterns, organized in a hexagonal unit cell with a periodicity of 300 nm, are realized through a single-step electron beam lithography process. By employing numerical models, it is uncovered that an additional induced collective dipole mode is the key mechanism responsible for achieving outstanding performance in absorption, reaching up to 99%. Analysis of the hierarchical organization reveals that variations in nanoparticle diameter and inter-unit-cell distance lead to shifts and broadening of the resonance peaks. It is also demonstrated that the hierarchical nanopatterns are capable of color reproduction with high uniformity and fidelity, serving as hexagonal pixels for high-resolution color printing. These cribellum-inspired metamaterials offer a novel approach to multifunctional metamaterial design, presenting aesthetic potential applications in the development of robotics and wearable electronic devices, such as smart skin or surface coatings integrated with energy harvesting functionalities.
Diatom frustule-inspired metamaterial absorbers: The effect of hierarchical pattern arrays
A. Li#, X. Zhao#, G. Duan#, S.W. Anderson, X. Zhang*
Advanced Functional Materials, 2019, 29(22): 1809029
+Abstract
Diatoms are photosynthetic algae that exist ubiquitously throughout the planet in water environments. Over the preceding decades, the diatom exoskeletons, termed frustules, featuring abundant micro- and nanopores, have served as the source material and inspiration for myriad research efforts. In this work, it is demonstrated that frustule-inspired hierarchical nanostructure designs may be utilized in the fabrication of metamaterial absorbers, thereby realizing a broadband infrared (IR) absorber with excellent performance in terms of absorption. In an effort to investigate the origin of this absorption characteristic, numerical models are developed to study these structures, revealing that the hierarchical organization of the constituent nanoparticulate metamaterial unit cells introduce an additional resonance mode to the device, broadening the absorption spectrum. It is further demonstrated that the resonant peaks shift linearly as a function of inter-unit-cell spacing in the metamaterial, which is attributed to the induced collective dipole mode by the nanoparticles. Ultimately, the work herein represents an innovative perspective in terms of the design and fabrication of IR absorbers inspired by naturally occurring biomaterials, offering the potential to lead to advances in metamaterial absorber technology.
Silica nanowire growth on Coscinodiscus Species diatom frustules via vapor-liquid-solid process
A. Li#, X. Zhao#, S. Anderson, X. Zhang*
Small, 2018, 14(47): 1801822
+Abstract
Diatom frustules are a type of porous silicon dioxide microparticle that has long been used in applications ranging from biomedical sensors to dye-sensitized solar cells. The favorable material properties, enormous surface area, and enhanced light scattering capacity support the promise of diatom frustules as candidates for next generation biomedical devices and energy applications. In this study, the vapor–liquid–solid (VLS) method is employed to incorporate silica nanowires on the surface of diatom frustules. Compared to the original frustule structures, the frustule–nanowire composite material’s surface area increases over 3-fold, and the light scattering ability increases by 10%. By varying the gold catalyst thickness during the VLS process, tuning of the resultant nanowire length/density is achieved. Through material characterization, it is determined that both float growth and root growth processes jointly result in the growth of the silica nanowires. From a thermodynamics point of view, the preferential growth of the silica nanowires on frustules is found to have resulted from the enormous partial surface area of gold nanoparticles on the diatom frustules. The frustule–nanowire composite materials have potential applications in the development of novel biomedical sensing devices and may greatly enhance next generation solar cell performance.
Towards uniformly oriented diatom frustule monolayers: Experimental and theoretical analyses
A. Li#, W. Zhang#, R. Ghaffarivardavagh#, X. Wang#, S.W. Anderson, X. Zhang*
Microsystems & Nanoengineering — Nature, 2016, 2: 16064
+Abstract
Diatoms are unicellular, photosynthetic algae that are ubiquitous in aquatic environments. Their unique, three-dimensional (3D) structured silica exoskeletons, also known as frustules, have drawn attention from a variety of research fields due to their extraordinary mechanical properties, enormous surface area, and unique optical properties. Despite their promising use in a range of applications, without methods to uniformly control the frustules’ alignment/orientation, their full potential in technology development cannot be realized. In this paper, we realized and subsequently modeled a simple bubbling method for achieving large-area, uniformly oriented Coscinodiscus species diatom frustules. With the aid of bubble-induced agitations, close-packed frustule monolayers were achieved on the water–air interface with up to nearly 90% of frustules achieving uniform orientation. The interactions between bubble-induced agitations were modeled and analyzed, demonstrating frustule submersion and an adjustment of the orientation during the subsequent rise towards the water’s surface to be fundamental to the experimentally observed uniformity. The method described in this study holds great potential for frustules’ engineering applications in a variety of technologies, from sensors to energy-harvesting devices.
Biologically enabled micro- and nanostencil lithography using diatoms
J. Cai#, X. Wang#, A. Li#, S.W. Anderson, and X. Zhang*
Extreme Mechanics Letters, 2015, 4: 186-192
+Abstract
The development of a biologically enabled micro- and nanostencil lithography approach using diatoms is demonstrated. Diatom frustules are initially purified, sorted, and aligned into compact monolayers on underlying silicon substrates. Subsequently, the diatom monolayers are employed as shadow masks during the electron beam deposition of gold (Au) thin films, a process which enables the capacity to mirror the intricate micro- and nanoporous frustule architecture on the underlying silicon substrates. Following Au deposition and diatom frustule dissolution, both sub-micron and nanoscale gold patterns on silicon are realized using this approach. This unique method yields the highly structured patterning of gold and other materials on a variety of substrates, with feature sizes ranging from the sub-micron to the nanoscale, enabling a host of diverse applications.
Ph.D. Dissertation
Diatom enabled advanced functional materials
Aobo Li, Ph.D. Dissertation, Boston University
(Advisor: Xin Zhang; December 2019)
Elastic, Viscoelastic, and Thermal Characterization of Polymers

Project Example 1:Research
Polydimethylsiloxane (PDMS) is a key polymeric material widely used in biological and biomedical devices. However, PDMS is inherently viscoelastic—its elastic modulus varies with loading rate and over time. This work presents a comprehensive method for the viscoelastic characterization, modeling, and analysis of the bending behavior of PDMS micropillar arrays. Our approach provides a more physically accurate and in-depth conversion model for force measurement applications. The scientific insights gained from this study of PDMS viscoelasticity contribute to the analysis of many other soft polymer materials at the micro/nano scale, which are broadly used in biological and biomedical research.

Project Example 2:
Polymer-based materials have demonstrated significant potential across a wide range of applications, including biomedical devices, photovoltaic solar cells, supercapacitors, and LEDs. Thermal conductivity and thermal diffusivity are critical thermophysical parameters that govern both steady-state and transient heat transport in these materials. However, most polymers are inherently poor thermal conductors, which often necessitates the use of ultra-thin structures in electronic devices. Moreover, techniques available for their thermal characterization remain limited. In this work, we have developed a time-domain fluorescence spectroscopy technique for characterizing the thermophysical properties of polymers. Since many polymers exhibit strong fluorescence excitation, this method offers broad potential for evaluating emerging polymer materials.

Representative Publications
(# denotes supervised by X. Zhang; * denotes corresponding author: X. Zhang)
Time-domain transient fluorescence spectroscopy for thermal characterization of polymers
H. Wu, K. Cai, H. Zeng, W. Zhao, D. Xie, Y. Yue#, Y. Xiong, and X. Zhang
Applied Thermal Engineering, 2018, 138: 403-408
+Abstract
In this work, a time-domain fluorescence spectroscopy technique is developed to characterize thermophysical properties of polymers. The method is based on fluorescence thermometry of materials under periodic pulse heating. In the characterization, a continuous laser (405 nm) is modulated with adjustable periodic heating and fluorescence excitation. The temperature rise at sample surface due to laser heating is probed from simultaneous fluorescence spectrum. Thermal diffusivity can be determined from the relationship between normalized temperature rise and the duration of laser heating. To verify this technique, thermal diffusivity of a polymer material (PVC) is characterized as 1.031 × 10-7 m2/s, agreeing well with reference data. Meanwhile, thermal conductivity can be obtained by the hot plate method. Then, both steady and unsteady thermophysical properties are available. Quenching effect of fluorescence signal in our measurement can be ignored, as validated by longtime laser heating experiments. The uncertainty induced by uniformity of laser heating is negligible as analyzed through numerical simulations. This non-destructive fluorescence-based technique does not require exact value about laser absorption and calibration experiment for temperature coefficient of fluorescence signals. Considering that most polymers can excite sound fluorescence signal, this method can be well applied to thermal characterization of polymer-based film or bulk materials.
A role for matrix stiffness in the regulation of cardiac side population cell function
Y. Qiu#, A.F. Bayomy, M.V. Gomez, M. Bauer, P. Du#, Y. Yang, X. Zhang, R. Liao
AJP-Heart and Circulatory Physiology, 2015, 308(9): H990-H997
+Abstract
The mechanical properties of the local microenvironment may have important influence on the fate and function of adult tissue progenitor cells, altering the regenerative process. This is particularly critical following a myocardial infarction, in which the normal, compliant myocardial tissue is replaced with fibrotic, stiff scar tissue. In this study, we examined the effects of matrix stiffness on adult cardiac side population (CSP) progenitor cell behavior. Ovine and murine CSP cells were isolated and cultured on polydimethylsiloxane substrates, replicating the elastic moduli of normal and fibrotic myocardium. Proliferation capacity and cell cycling were increased in CSP cells cultured on the stiff substrate with an associated reduction in cardiomyogeneic differentiation and accelerated cell ageing. In addition, culture on stiff substrate stimulated upregulation of extracellular matrix and adhesion proteins gene expression in CSP cells. Collectively, we demonstrate that microenvironment properties, including matrix stiffness, play a critical role in regulating progenitor cell functions of endogenous resident CSP cells. Understanding the effects of the tissue microenvironment on resident cardiac progenitor cells is a critical step toward achieving functional cardiac regeneration.
Tunable electrical and mechanical responses of PDMS and polypyrrole nanowire composites
P. Du#, X. Lin, and X. Zhang*
Journal of Physics D: Applied Physics, 2013, 46(19): 195303
+Abstract
A generic experimental procedure is presented in this work to enhance the electrical responses of polydimethylsiloxane (PDMS) through incorporation of conducting polymer nanowires, while maintaining the desirable mechanical flexibility of PDMS. The conducting polypyrrole (PPy) nanowires are synthesized using a template method. The dielectric constants of the composites are characterized by impedance measurements, and the effect of nanowire concentration is investigated by the percolation theory. Using a continuous hyperbolic tangent function, critical volume fraction is estimated to be 9.6 vol%, at which an 85-fold enhancement in the dielectric constants is achieved. The viscoelastic properties of the composites are characterized by the stress relaxation nanoindentation tests, and the effect of nanowire concentration on the elastic modulus of composites is found to deviate significantly from the Wang–Pyrz model at the critical volume fraction. The tunable multifunctionality of PDMS composites that possess significantly enhanced electrical and moderate viscoelastic responses is desirable for many sensing and actuation applications.
Investigation of cellular contraction forces in the frequency domain using a PDMS micropillar-based force transducer
P. Du#, C. Chen, H. Lu, and X. Zhang*
Journal of Microelectromechanical Systems, 2013, 22(1): 44-53
+Abstract
Polydimethylsiloxane (PDMS) micropillar-based biotransducers are extensively used in cellular force measurements. The accuracy of these devices relies on the appropriate material characterization of PDMS and modeling to convert the micropillar deformations into the corresponding forces. Cellular contraction is often accompanied by oscillatory motion, the frequency of which ranges in several hertz. In this paper, we developed a methodology to calculate the cellular contraction forces in the frequency domain with improved accuracy. The contraction data were first expressed as a Fourier series. Subsequently, we measured the complex modulus of PDMS using a dynamic nanoindentation technique. An improved method for the measurement of complex modulus was developed with the use of a flat punch indenter. The instrument dynamics was characterized, and the full contact region was identified. By incorporating both the Fourier series of contraction data and the complex modulus function, the cellular contraction force was calculated by finite-element analysis (FEA). The difference between the Euler beam formula and the viscoelastic FEA was discussed. The methodology presented in this work is anticipated to benefit the material characterization of other soft polymers and complex biological behavior in the frequency domain.
Effect of loading rates on cellular force measurements by polymer micropillar based transducers
P. Du#, X. Zheng#, I-K Lin#, and X. Zhang*
Applied Physics Letters, 2011, 99(8): 083701
+Abstract
Polymeric deformable sensor arrays have been employed to measure cellular forces and offered insights into the study of cellular mechanics. Previous studies have been focused on using transducers in static domain and assumed elastic beam theory as the force conversion model. Neglecting the inherent viscoelastic behavior of polydimethylsiloxane and low aspect ratios of the sensor arrays compromised the accuracy of these devices. In this work, a more in-depth viscoelastic Timoshenko beam model was developed incorporating dynamic cellular forces. We studied chemically stimulated contractions of cardiac myocytes and found that the loading rate has a considerable influence on the sensitivity of the sensor arrays.
Extension of the beam theory for polymer bio-transducers with low aspect ratios and viscoelastic characteristics
P. Du#, I-K Lin#, H. Lu, and X. Zhang*
Journal of Micromechanics and Microengineering, 2010, 20(9): 095016
+Abstract
Polydimethylsiloxane (PDMS)-based micropillars (or microcantilevers) have been used as bio-transducers for measuring cellular forces on the order of pN to µN. The measurement accuracy of these sensitive devices depends on appropriate modeling to convert the micropillar deformations into the corresponding reaction forces. The traditional approach to calculating the reaction force is based on the Euler beam theory with consideration of a linear elastic slender beam for the micropillar. However, the low aspect ratio in geometry of PDMS micropillars does not satisfy the slender beam requirement. Consequently, the Timoshenko beam theory, appropriate for a beam with a low aspect ratio, should be used. In addition, the inherently time-dependent behavior in PDMS has to be considered for accurate force conversion. In this paper, the Timoshenko beam theory, along with the consideration of viscoelastic behavior of PDMS, was used to model the mechanical response of micropillars. The viscoelastic behavior of PDMS was characterized by stress relaxation nanoindentation using a circular flat punch. A correction procedure was developed to determine the load–displacement relationship with consideration of ramp loading. The relaxation function was extracted and described by a generalized Maxwell model. The bending of rectangular micropillars was performed by a wedge indenter tip. The viscoelastic Timoshenko beam formula was used to calculate the mechanical response of the micropillar, and the results were compared with measurement data. The calculated reaction forces agreed well with the experimental data at three different loading rates. A parametric study was conducted to evaluate the accuracy of the viscoelastic Timoshenko beam model by comparing the reaction forces calculated from the elastic Euler beam, elastic Timoshenko beam and viscoelastic Euler beam models at various aspect ratios and loading rates. The extension of modeling from the elastic Euler beam theory to the viscoelastic Timoshenko beam theory has improved the accuracy for the conversion of the PDMS micropillar deformations to forces, which will benefit the polymer-based micro bio-transducer applications.
Viscoelastic characterization and modeling of polymer transducers for biological applications
I-K Lin#, K-S Ou, Y-M Liao, Y. Liu#, K-S Chen, and X. Zhang*
Journal of Microelectromechanical Systems, 2009, 18(5): 1087-1099
+Abstract
Polydimethylsiloxane (PDMS) is an important polymeric material widely used in bio-MEMS devices such as micropillar arrays for cellular mechanical force measurements. The accuracy of such a measurement relies on choosing an appropriate material constitutive model for converting the measured structural deformations into corresponding reaction forces. However, although PDMS is a well-known viscoelastic material, many researchers in the past have treated it as a linear elastic material, which could result in errors of cellular traction force interpretation. In this paper, the mechanical properties of PDMS were characterized by using uniaxial compression, dynamic mechanical analysis, and nanoindentation tests, as well as finite element analysis (FEA). A generalized Maxwell model with the use of two exponential terms was used to emulate the mechanical behavior of PDMS at room temperature. After we found the viscoelastic constitutive law of PDMS, we used it to develop a more accurate model for converting deflection data to cellular traction forces. Moreover, in situ cellular traction force evolutions of cardiac myocytes were demonstrated by using this new conversion model. The results presented by this paper are believed to be useful for biologists who are interpreting similar physiological processes.
Viscoelastic mechanical behavior of soft microcantilever-based force sensors
I-K Lin#, Y-M Liao, Y. Liu#, K-S Ou, K-S Chen, and X. Zhang*
Applied Physics Letters, 2008, 93(25): 251907
+Abstract
Polydimethylsiloxane (PDMS) microcantilevers have been used as force sensors for studying cellular mechanics by converting their displacements to cellular mechanical forces. However, PDMS is an inherently viscoelastic material and its elastic modulus changes with loading rates and elapsed time. Therefore, the traditional approach to calculating cellular mechanical forces based on elastic mechanics can result in errors. This letter reports a more in-depth method for viscoelastic characterization, modeling, and analysis associated with the bending behavior of the PDMS microcantilevers. A viscoelastic force conversion model was developed and validated by proof-of-principle bending tests.
Ph.D. Dissertation
Viscoelastic characterization and modeling of PDMS micropillars for cellular force measurement applications
Ping Du, Ph.D. Dissertation, Boston University
(Advisor: Xin Zhang; January 2013)
Flexible Fabrication of 3D Multilayered Microstructures

Project Example 1:Research
The objective of this research is to develop a biocompatible, engineered platform for magnetic resonance imaging (MRI) and computed tomography (CT) contrast agents. These agents feature novel properties, including sensory capabilities, and hold broad potential for biomedical applications. The approach combines top-down and self-assembly methods, incorporating environmentally responsive materials to create a class of contrast agents designed to overcome the major limitations of current options. In collaboration with Dr. Stephan Anderson (Radiology), this project focuses on developing next-generation MRI and CT contrast agents. By employing micro- and nanofabrication techniques and integrating biocompatible magnetic materials, we aim to create geometrically precise particulate contrast agents with distinct spectral signatures and multiplexing capabilities. Furthermore, the inclusion of biocompatible hydrogels—tunable to respond to myriad environmental stimuli—endows these agents with in vivo biological sensing capabilities.

Project Example 2:Research
We have developed a method for fabricating embedded microchannels using a scanning laser direct-write technique. This approach enables the creation of embedded microchannels with variable heights within a single-layer polymer, offering exceptional flexibility and versatility. An apparatus has been designed to align multiple exposures, facilitating the large-scale production of complex microstructures and microdevices. Environmentally friendly, this technique avoids the use of toxic chemicals and does not produce harmful gases. It also significantly simplifies the fabrication process, potentially reducing the design-to-fabrication turnaround time to just a few hours. Using this laser direct-write method, a wide variety of 3D microstructures can be fabricated and applied across numerous fields. The next phase of our research focuses on elucidating the theory and methodology for determining the Young’s modulus of exposed polymers using a laser acoustic microscopy system. Through this laser-based ultrasonic technique, we analyze the Young’s moduli of polymer microbeams fabricated under varying exposure doses. This method provides a non-contact, non-destructive means for evaluating and characterizing materials.

Representative Publications
(# denotes supervised by X. Zhang; * denotes corresponding author: X. Zhang)
Biocompatible, micro- and nano-fabricated magnetic cylinders for potential use as contrast agents for magnetic resonance imaging
C. Wang#, X. Wang#, S.W. Anderson, and X. Zhang*
Sensors and Actuators B: Chemical, 2014, 196: 670-675
+Abstract
This paper reports novel MEMS and NEMS-based fabrication processes for biocompatible, hollow cylindrical ferromagnetic structures for potential use as contrast agents for magnetic resonance imaging (MRI). Compared to previous works on Ni-based cylindrical-nanoshells and Fe-based double-disk particles, biocompatibility and yield issues were strongly considered in this development of a simplified fabrication process incorporating iron oxide thin films. The novel, simplified fabrication processes developed herein yield robust, reproducible fabrication methodologies for the further development of this new class of MRI contrast agents. Specifically, both micron- and nano-scale hollow cylindrical agents were successfully fabricated, the size regimes of which enable a wide array of potential imaging applications. The use of top-down engineering approaches to MRI contrast agent design such as reported herein offers the capacity for multiplexed imaging which may dramatically potentiate the capabilities of MRI imaging.
Fabrication and characterization of composite hydrogel particles with x-ray attenuating payloads
C. Wang#, X. Wang#, S. Anderson, and X. Zhang*
Journal of Vacuum Science & Technology B, 2014, 32(3): 032001
+Abstract
This paper reports the fabrication and characterization of composite hydrogel particles composed of poly(ethylene glycol) diacrylate (PEG-DA)-based hydrogels and x-ray attenuating payloads. The top–down fabrication method employed herein is demonstrated to yield composite hydrogel particles of varying size and shape for use as computed tomography (CT) imaging contrast agents. Characterization of the materials properties of the PEG-DA hydrogels was undertaken, demonstrating tunable mechanical properties of composite hydrogels based on hydrogel composition and UV cross-linking time. Analyses of the leakage rates of a conventional iodine-based small molecular contrast agent as well as a nanoparticulate x-ray attenuating material from the PEG-DA hydrogels were undertaken. In contradistinction to clinically available iodinated CT contrast agents, as well as recently developed nanoparticulate CT contrast agents, the approach presented herein yields an engineering flexibility to the design of CT contrast agents which may be leveraged to optimize this class of agents to a wide array of specific imaging and sensing applications.
Microfabricated iron oxide particles for tunable, multispectral magnetic resonance imaging
X. Wang#, C. Wang#, S. Anderson, and X. Zhang*
Materials Letters, 2013, 110: 122-126
+Abstract
Recently, a novel class of magnetic resonance imaging (MRI) contrast agents developed using top-down microfabrication approaches has been reported. To realize the full capacity of this potentially paradigm-shifting approach to MRI contrast agent design, the integration of biocompatible materials with tunable magnetic properties was sought. To this end, deposition techniques yielding iron oxide thin films with a large range of readily tunable saturation magnetic polarization were developed using reactive sputtering under various conditions. Following the characterization of their chemical compositions and crystalline structures, the iron oxide thin films were subsequently utilized in the fabrication of size and shape specific magnetic double-disk microparticles, yielding the advantages of this new class of MRI contrast agents, including multiplexing capability, diffusion-driven signal amplification, and functional imaging capacity. The integration of iron oxides into this class of fabricated contrast agents offers several distinct advantages, including biocompatibility and the additional degree of freedom in the design of these agents achieved by the tunability of the iron oxide thin film magnetism, both of which are critical features in further optimizing these agents.
Fabrication of three-dimensional microstructures based on singled-layered SU-8 for lab-on-chip applications
H. Yu#, O. Balogun, B. Li, T.W. Murray, and X. Zhang*
Sensors and Actuators A: Physical, 2006, 127(2): 228-234
+Abstract
This paper introduces a novel 3D manufacturing approach to the rapid processing of microfluidic components such as embedded channels and microvalves, using a scanning laser system. Compared to existing manufacturing techniques, our direct UV laser writing method greatly simplifies fabrication processes, potentially reducing the design-to-fabrication time to a few hours, which is extremely beneficial during the product development stages. The initial process validation has been presented by using SU-8 material. With the fine-tuning of the laser processing parameters, the depth of SU-8 polymerization can be controlled. This paper also describes the underlying theory and method to determine the Young’s modulus of the exposed SU-8 material by using a laser acoustic microscopy system. The laser-based ultrasonic technique offers a non-contact, nondestructive means of evaluation and materials characterization. More importantly, it allows for local inspection of material properties. The results presented in this paper potentially could serve as the first crucial step towards the rapid manufacturing of microdevices for lab-on-chip applications.
Flexible fabrication of three-dimensional multi-layered microstructures using a scanning laser system
H. Yu#, B. Li, and X. Zhang*
Sensors and Actuators A: Physical, 2006, 125(2): 553-564
+Abstract
In this paper, we developed a scanning laser system, which allows rapid processing of freeform multi-layered microstructures. More importantly it enables rapid prototyping of three-dimensional (3D) microdevices at low cost. The capabilities of three-dimensional manufacturing, inclined patterning, and multi-layered manufacturing have been demonstrated. Specifically, both in-plane and out-of-plane processing is feasible using spot-by-spot controllable laser pulsing. The laser processing perpendicular to the specimen surface is realized by fine tuning the focus level and laser intensity. A large number of microfluidic components such as cantilever beams, embedded channels and other shapes requiring gaps between layers are demonstrated in a single layer. Compared to the existing manufacturing techniques, our direct laser writing method greatly simplifies fabrication processes, potentially reducing the design-to-fabrication cycle to a few hours.
Building embedded microchannels using a single layered SU-8, and determining Young’s modulus using a laser acoustic technique
H. Yu#, O. Balogun, B. Li, T.W. Murray, and X. Zhang*
Journal of Micromechanics and Microengineering, 2004, 14(11): 1576-1584
+Abstract
In this paper, an innovative method to create embedded microchannels is presented. The presented technology is based on a direct-write technique using a scanning laser system to pattern a single layered SU-8. The enormous flexibility of the scanning laser system can be seen in two key features: the laser pulsing can be controlled spot-by-spot with variable exposure doses, and the laser intensity penetrating into samples can be adjusted by varying the laser focus level. The UV laser direct-write method greatly simplifies the fabrication processes. Moreover, it can be set up in a conventional manufacturing environment without the need for clean room facilities. The second part of this paper describes the underlying theory and method to determine Young’s modulus of exposed SU-8 by using a laser acoustic microscopy system. The laser-based ultrasonic technique offers a non-contact, non-destructive means of evaluation and material characterization. This paper will determine Young’s modulus of UV exposed SU-8 generated with different exposure doses. Measurements show that Young’s modulus is highly dependent on exposure dose. Young’s modulus ranges from 3.8 to 5.4 GPa when the thickness of a fully cross-linked SU-8 microbeam varies from 100 to 205 µm with a gradually increased UV exposure dose.
Rapid three-dimensional manufacturing of microfluidic structures using a scanning laser system
B. Li, H. Yu#, A. Sharon, and X. Zhang
Applied Physics Letters, 2004, 85(12): 2426-2428
+Abstract
This letter introduces a three-dimensional manufacturing approach to the rapid processing of microfluidic structures using a scanning laser system. This technique takes advantage of the nonuniform distribution of laser power along its incident axis. The laser processing perpendicular to the specimen surface is realized by fine-tuning focus levels and laser intensity. A large number of microfluidic components such as cantilevered valves, embedded channels, and other shapes requiring gaps between layers are demonstrated in a single layer. With this process, a class of microstructures with designed-in functionalities can be developed.
Ph.D. Dissertation
Size and shape specific particles toward biomedical imaging: Design, fabrication, and characterization
Xiaoning Wang, Ph.D. Dissertation, Boston University
(Advisor: Xin Zhang; September 2014)
Flexible fabrication of three-dimensional multi-layered microstructures using a scanning laser system
Hu Yu, Ph.D. Dissertation, Boston University
(Advisor: Xin Zhang; May 2006)