A machine learning–based microorganism classification and a novel synthesis of stabilized tetragonal and cubic ZrO₂ nanoparticles: A future carrier for antimicrobial drug delivery

Technical Summary: Machine Learning-Based Microorganism Classification and Synthesis of Stabilized ZrO₂ Nanoparticles

Problem Statement
The research addresses two distinct but converging challenges in biomedical and materials science. First, there is a critical need for accurate, automated identification of microorganisms (including bacteria, protozoa, and fungi) for applications in medical diagnostics, food safety, and environmental monitoring. Second, there is a demand for the development of biocompatible, non-toxic nanomaterials that can serve as both potent antimicrobial agents and effective carriers for drug delivery. Specifically, the synthesis of stabilized tetragonal (t) and cubic (c) phases of Zirconia (ZrO₂) nanoparticles at moderate temperatures remains a technical hurdle, as these high-energy polymorphs are naturally metastable and prone to transforming into the monoclinic (m) phase at lower temperatures.

Methodology
The study employs a dual-pronged approach combining computational biology and advanced materials synthesis:

1. Microorganism Classification:

   • Dataset: A dataset of 757 microscopic images comprising eight representative microorganism classes: Amoeba, Euglena, Hydra, Paramecium, rod-shaped bacteria, spherical bacteria, spiral bacteria, and yeast.
   • Model: A supervised deep learning model based on the pre-trained Xception architecture was utilized.
   • Process: Images were preprocessed (scaled and normalized) and split into training and testing sets. The model was tuned for multi-class classification, with performance evaluated using precision, recall, F1-scores, and accuracy.

2. Synthesis of ZrO₂ Nanoparticles:

   • Precursor Formation: A novel chemical method involving Al-metal surface hydrolysis was used. Aluminum metal plates with nascent surfaces were reacted in an aqueous solution of zirconium oxychloride (ZrOCl₂·8H₂O) at room temperature. This produced a fully transparent, colorless Al³⁺ modified ZrO(OH)₂·xH₂O gel via polycondensation over 24–48 hours.
   • Thermal Decomposition: The amorphous gel precursor was subjected to reconstructive thermal decomposition in air at 500°C for 2 hours.
   • Variable Control: The phase of the resulting nanoparticles was controlled by varying the Al₂O₃ content (doping level) from 2 mol% to 15 mol%.
   • Characterization: Structural and phase analysis were conducted using X-ray Diffraction (XRD) with CoKα radiation, Scanning Electron Microscopy (SEM), and Energy-Dispersive X-ray (EDX) spectroscopy. Crystallite sizes were calculated using the Debye–Scherrer relation.

Key Results

• Classification Performance: The machine learning model achieved an overall accuracy of 98%, with macro and weighted F1-scores ranging from 0.98 to 0.99.

  • Hydra, Paramecium, and yeast were classified with perfect precision, recall, and F1-scores (1.00).
  • Spherical bacteria showed slightly lower recall (0.94) and F1-score (0.96).
  • The confusion matrix indicated near-perfect classification with minimal misclassification, even among morphologically similar bacterial groups.

• Nanoparticle Synthesis and Phase Stabilization:

  • Gel Structure: The precursor gel exhibited an amorphous, liquid-like structure with interconnected fibrils or layers, confirmed by XRD (broad halos) and SEM.
  • Phase Control:
    • 2–5 mol% Al₂O₃: Resulted in a single-phase metastable tetragonal (t-ZrO₂) structure, with minor monoclinic (m-ZrO₂) impurities (5–10%).
    • ≥10 mol% Al₂O₃: Induced a phase transformation to a single-phase cubic (c-ZrO₂) structure.
  • Crystallite Size: The average crystallite size (D) ranged between 7 and 9 nm, depending on the Al₂O₃ concentration.
  • Solid Solution: No independent crystallized Al₂O₃ was detected; the material formed a perfect Al₂O₃:ZrO₂ solid solution.
  • Stabilization Mechanism: The study proposes that smaller Al³⁺ ions substitute Zr⁴⁺ ions, occupying 6-fold oxygen coordination sites. This induces neighboring Zr⁴⁺ cations into distorted 8-fold sites containing oxygen vacancies, creating a covalently oriented structure that stabilizes the metastable t- and c-phases at low temperatures.

Significance and Claims
The authors claim this work offers a "new perspective" by integrating computational microorganism recognition with the development of a novel nanosystem.

1. Computational Reliability: The study demonstrates that supervised machine learning (specifically Xception) is a robust tool for accurate microbial recognition, achieving high reliability (98% accuracy) across diverse biological classes.
2. Novel Synthesis Route: The paper introduces a unique, room-temperature surface hydrolysis method to create transparent Al-modified zirconia gels, which can be converted into stabilized nanoparticles at a moderate temperature (500°C). This avoids the high-temperature requirements often associated with phase stabilization.
3. Biomedical Potential: The resulting Al₂O₃-stabilized ZrO₂ nanoparticles are described as nontoxic and biocompatible. The authors posit that these nanoparticles possess intrinsic antimicrobial capabilities (via ion release, membrane rupture, and ROS formation) and serve as a promising nanocarrier for antimicrobial drug delivery.
4. Future Outlook: The authors state that while the current work establishes the synthesis and classification capabilities, future research will focus on detailed antimicrobial efficacy studies, drug targeting mechanisms, and broader applications in biosensing and biomedical engineering.