Unveiling The True Shape Of Conductive Filaments: Adaptive Scalpel Spm Redefines 3d Volumetric Sensing

19 Aug 2025
Volume 0

Poster Winner at NanoScientific Symposium Americas 2024
Md Ashiqur Rahman Laskar and Team, Arizona State University
Adapted from Presentation, Edited by NanoScientific

In the nanoscopic realm of advanced materials and electronic devices, precision is everything. The need to understand, control, and characterize structures at the nanoscale has driven the evolution of microscopy into a high-art of engineering and physics. At the 2024 NanoScientific Symposium, Researcher Md Ashiqur Rahman Laskar and his interdisciplinary team from Arizona State University, INRiM (Italy), and Neno-Vision (Czech Republic) presented an award-winning breakthrough in this domain—Adaptive Scalpel Scanning Probe Microscopy (SPM), a transformative approach to three-dimensional tomographic analysis. This work addresses a longstanding issue in conductive atomic force microscopy (C-AFM): the parasitic effects caused by tip-sample junction leakage and inhomogeneous electric fields, which distort lateral and vertical resolution in 3D reconstructions. Laskar’s innovation offers a path to sub-1000 nm³ tomographic precision, especially for characterizing resistive switching memories—ushering in a new era of volumetric sensing in nanoelectronics.

The Problem: Artifacts in Nanoscale Tomography
At the heart of modern nanoelectronics lies a growing class of devices that rely on resistive switching—such as valence change memory (VCM) systems based on transition metal oxides like NbOx. These systems depend on the reversible formation and rupture of nanoscale conductive filaments (CFs), which define the ON and OFF states of a memory cell. Understanding these filaments, including their 3D morphology and evolution, is critical for optimizing device performance and longevity.

Yet a key barrier to true 3D insight has been tip-induced artifacts in conventional tomographic atomic force microscopy. These artifacts fall into two primary categories:

In-plane leakage: Current spreads laterally due to parasitic conduction pathways through the surrounding oxide matrix. As a result, the detected current footprint significantly overestimates the actual lateral dimensions of the conductive feature (i.e., the CF).
Vertical overbiasing: As material is removed layer by layer during the tomographic process, the oxide thickness reduces, causing an unintentional increase in electric field strength. This non-uniformity leads to variations in contrast and feature depth, skewing vertical resolution.

The result? A bloated and blurred picture of what should be nanosharp and sharply localized.

The Innovation: Adaptive Scalpel SPM
To combat these dual distortions, Laskar’s team devised Adaptive Scalpel SPM, an approach that dynamically adjusts the tip-sample bias throughout the tomographic process. Instead of applying a constant voltage, which leads to uncontrolled increases in electric field as oxide layers thin, the adaptive method calibrates the bias to maintain a steady internal electric field—thus preserving contrast integrity as depth increases.

This innovation was tested on VCM cells with gold (Au) top electrodes and NbO_x-based switching layers. After removing the Au layer (see Fig. 1), a sharp conductive probe was used to iteratively erode the NbOx material while mapping current flow. Two acquisition modes were compared:

• Constant bias mode: Fixed voltage during the entire scalpel scan.

• Adaptive bias mode: Voltage tuned according to oxide thickness to stabilize the internal electric field.

Fig. 1. VCM Device and Scalpel C-AFM: A) Scanning electron microscopy image. B) Schematic of the Au dots forming the top electrodes of the VCM under study. C) Detailed view with SEM and conductive atomic force microscopy (C-AFM) of the AU top electrode of the VCM cell. D) Schematic of the tip-induced filament formation and top-electrode removal process.

Uncovering the True Filament Geometry
The difference between these two approaches is vividly captured in 3D volumetric reconstructions of the conductive filament. Under constant bias, the filament appears larger and less defined, a result of leakage spreading and overbiasing. In contrast, adaptive bias yields a filament image that is smaller, sharper, and more confined—closely matching theoretical expectations and physical models.

This is reinforced in the cross-sectional comparisons of x-y and x-z slices, where the adaptive bias captures a sharply bounded filament located precisely at the core of the active device area. Such precision is crucial when designing or troubleshooting high-density memory arrays or emerging logic components and improved resolution.

Quantitative Insights: Resolution and Leakage Control
The experimental data quantifies the improvement.

Lateral resolution improves due to reduced in-plane leakage, confirmed by varying the bias during scanning and observing a clear decrease in current footprint size.
Vertical sensitivity is optimized through adaptive voltage control, leading to significant decrease in vertical leakage and charging artifacts.

Introduction. • Controlling nanoscale tip-induced material removal is crucial for achieving atomic-level precision in tomographic sensing with atomic force microscopy (AFM). • This work addresses in-plane and vertical tip-sample junction leakage as sources of parasitic contrast in tomographic AFM, hindering real-space 3D reconstructions.
• An adaptive tip-sample biasing scheme is demonstrated for reduction
in overestimation of the lateral dimensions and the mitigation of a class of artefacts induced by the high electric field inside the thin oxide when volumetrically reduced.

Fig. 3. Vertical Depth Sensitivity In Volumetric Reconstruction. Illustration of the second physical effect impacting vertical depth sensitivity in volumetric C-AFM reconstruction. A) This idealized case depicts a conductive path (light blue) embedded within a thin oxide layer. B) The biased probe is used to remove material, thus eroding
the oxide progressively. C) As the oxide thickness is reduced, the internal electric field will rise during the measurement. D) The concept of adaptive scalpel SPM is to maintain a constant electric field during the analysis by adjusting the tip-sample bias as the oxide is made thinner. E) Evolution of the tip-sample current as a function of depth for a CF (red) and pristine oxide (blue). In the inset, we show the schematic
of the acquisition area. F) The evolution of the tip-sample current for a pristine oxide region compares the leakage in the case of constant bias (blue) and adaptive bias (red).

Fig. 2 provides an instructive visualization of in-plane leakage effects, while Fig. 3 explains how thickness-dependent field increases can distort vertical profiling. The adaptive bias strategy not only restores fidelity but allows for reliable current-depth analysis, further substantiating its effectiveness.

Fig. 4. 3D Observation Of The Conductive Filament. 3D observation of the conductive filament in NbO_x cell. The tomographic reconstruction is visualized by volume rendering comparing A) constant bias scalpel SPM with B) adaptive scalpel SPM. In the inset we compare the 2D observation of the CF section plane (xz) and four section x-y planes for comparison of the resolution obtainable with the two modes of acquisitions (scalebar 5nm).

Fig. 5. 2D Slice Views. Top view of selected 2D slices throughout the NbO_x device (scan size 100 × 100 nm) for A) constant voltage analysis, and B) for adaptive voltage. The CF appears in the middle of the active area after top electrode removal. The highly conductive features on the top and bottom corners are the exposed parts of the Nb electrode, which is completely exposed during the removal of NbO_x.

Beyond Resistive Memories: Broader Impact
Although this study focuses on NbOx-based VCM cells, the implications of Adaptive Scalpel SPM stretch far beyond memory technology. In photovoltaics, 3D conductive networks and ion migration paths are critical to device stability and efficiency. In ferroelectrics and piezoelectrics, understanding charge domain structures and their evolution during operation demands similar volumetric precision. In 2D and layered devices, heterostructures such as transition metal dichalcogenides or perovskite-based architectures require reliable tomography to decode buried interface physics. By enabling 3D reconstructions with sub-1000 nm³ resolution, Adaptive Scalpel SPM positions itself as a universal tool for high-precision electrical nanocharacterization.

Fig. 6. Electrical I-V readout from tip-induced resistive switching. (Note: here the current flowing in the filament is limited by the high series resistance of the tip-sample junction acting as a form of current compliance for the tip-sample device system).

Fig. 7. Schematic of the current collection scheme using the tip-Au-NbO_x-stack for switching. Note, the associated equivalent resistive network. Here, the emphasis should be placed on the high resistance of the tip-sample junction once the Au is removed.

Practical Implementation: Tip-Device Interface
A detailed understanding of the tip-sample interaction was also key to the study. As the schematic in Fig. 7 shows, the team modeled the tip-Au-NbOx stack as an equivalent resistive network. Once the Au top electrode is removed, the tip-sample junction becomes the primary resistance limiter. This naturally introduces a form of current compliance, a fortunate artifact that protects the device from damage during scanning.

Moreover, I-V curves demonstrate stable switching characteristics, and consistent current evolution under adaptive bias indicates improved reproducibility. Notably, Fig. 8 reveals that the filament area measured at the deepest slices of the NbOx device was significantly smaller in adaptive mode, confirming better spatial confinement.

Fig. 8. *Filament area comparison under constant and adaptive bias for the last four slices at the very bottom of the NbO_x layer.*

Collaboration and Outlook
This project is a testament to international scientific collaboration. Co-authors from ASU, INRiM, and NenoVision brought together expertise in materials science, metrology, AFM technology, and device engineering. The equipment and infrastructure support of ASU’s Eyring Materials Center and funding from DOE and NSF were critical to its success. Looking ahead, Laskar and his colleagues envision integrating adaptive scalpel techniques into broader AFM platforms, potentially in combination with machine learning algorithms for automated reconstruction and classification of 3D features. The goal is clear: bring unprecedented clarity and confidence to nanoscopic tomography.

Conclusion
The work of Md Ashiqur Rahman Laskar and his team exemplifies the spirit of NanoScientific: combining creativity, technical mastery, and scientific rigor to break new ground in nanoscale exploration. Their Adaptive Scalpel SPM approach not only solves a real and pressing problem in AFM tomography, it redefines the resolution standards for the entire field. As nanoscale systems become ever more complex, tools like Adaptive Scalpel SPM will be essential in decoding their structure-function relationships, helping engineers build the next generation of smart, efficient, and responsive devices.

**About Md Ashiqur Rahman Laskar**
Md Ashiqur Rahman Laskar is a PhD candidate in Electrical Engineering
at Arizona State University (ASU), specializing in semiconductor
metrology and failure analysis. He is doing research with Dr. Umberto
Celano at ASU. The ASU team aims to design experiments and develop
advanced metrology techniques for the emerging devices and
materials. The research goal is to understand and observe the fundamental processes that govern the operation of emerging devices,
through nanoscale analytical imaging and correlative metrology. This
generates fundamental understanding to engineer improved materials
for nanoelectronics. The ASU team behind the research: PhD students A. Laskar, S. Chakrabarti; Prof. U. Celano, the principal investigator of Nanoelectronics Metrology & Failure Analysis Lab at ASU.
For more information, https://labs.engineering.asu.edu/celano/

*From L-R: Md Ashiqur Laskar, Srijan Chakrabarti, Dr. Umberto Celano*

References
• M. A. R. Laskar, G. Leonetti, G. Milano, O. Novotný, J. Neuman, S. Tongay, U.
Celano, Adaptive Scalpel Scanning Probe Microscopy for Enhanced Volumetric
Sensing in Tomographic Analysis. Adv. Mater. Interfaces 2024, 11, 2400187.
• Song, J., Zhou, Y., Padture, N.P. et al. Anomalous 3D nanoscale photoconduction in hybrid perovskite semiconductors revealed by tomographic atomic force microscopy. Nat Commun 11, 3308 (2020).

Cookie categories	Purpose of cookie
Essential Cookies	To provide you with the Service you have requested. Without these cookies, Services that you have asked for cannot be provided. For example, essential cookies enable us to remember the items you place in the shopping basket so that you can order products.
Performance Cookies (Audience measurement)	Advertising CoTo analyze the use, performance and design of our Services, detect errors and improve user experience. For example, this type of cookie enables us to understand how often you are using our Services, to recognize that you have visited our websites before and to identify which parts of our Services and which sections of our website are most popular, for example by allowing us to see which pages visitors access most frequently and how much time visitors spend on each page.okies (Behavioral advertising)
Functional Cookies (Content customization)	To deliver a better user experience, and build up customer usage patterns and profiles. For example, this type of cookie ensures that the information displayed when you use our Services (e.g. on your next visit to our website) will match up with your user preferences (such as language preferences, font size or auto-fill options).
Advertising Cookies (Behavioral advertising)	To personalize our website, deliver customized advertisements to you, or contact you directly where you have separately consented to such communications, in a way which is relevant to you and which matches your interests by, for example, using information about products you have browsed or ordered on our website.