Non-Enzymatic Glucose Sensors

• Non-enzymatic glucose sensors are devices that detect glucose using physical, chemical, or electronic transduction methods without relying on enzymes, offering improved stability and shelf-life.
• They employ mechanisms such as boronic acid binding, electrocatalytic oxidation, and optical/dielectric property shifts to achieve sensitive and selective glucose quantification.
• Advances in materials and device integration have enabled wearable and non-invasive formats, making these sensors promising for continuous and reliable glucose monitoring.

Non-enzymatic glucose sensors are analytical devices that detect glucose without relying on enzyme-catalyzed reactions, instead utilizing physical, chemical, or electronic transduction based on glucose's intrinsic chemical or physical properties. These sensors exploit mechanisms such as selective chemical binding, direct electrochemical oxidation, optical absorption, resonance frequency shifting due to dielectric changes, or charge-transfer phenomena at functionalized electrode interfaces. The technology spans a broad range of platforms, including nanostructured electrodes, field-effect transistors, fiber-optic and photonic elements, and microwave or millimeter-wave dielectric resonators. Non-enzymatic approaches are pursued to circumvent the limitations of enzymatic sensors—primarily enzyme instability, drift, cofactor dependence, and poor shelf-life—enabling robust, potentially lifelong continuous glucose monitoring (CGM) in both invasive and non-invasive form factors.

1. Principles of Non-Enzymatic Glucose Recognition and Transduction

Non-enzymatic recognition mechanisms can be divided into three primary categories: (1) specific molecular interactions (e.g., boronic acid–diol binding), (2) direct electrocatalytic oxidation on metal or metal-oxide surfaces, and (3) measurement of perturbations in physical properties (optical absorption, refractive index, permittivity) as a function of glucose concentration.

Selective Binding via Boronic Acid Chemistry

Phenylboronic acids covalently and reversibly bind the cis-diol groups of D-glucose, forming cyclic boronate esters. This chemistry is exploited both in electronic sensors (e.g., CNT-FETs functionalized with pyrene-1-boronic acid) and optical transduction (e.g., 4-MPBA on plasmonic nanostructures) (Liu et al., 24 Apr 2025, Lerner et al., 2013). The equilibrium constant typically is $K_b \approx 10^2$–$10^3$ M$^{-1}$, and binding events shift local electronic or refractive environments, modifying the sensor's electrical or optical readout.

Direct Electrocatalytic Oxidation

Transition metal oxides (e.g., NiO, Fe$_2$O$_3$, CuO), noble metals (Pt, Au), and emerging 2D heterostructures (e.g., BP/g-CN) act as electrocatalysts for glucose oxidation without enzymes (Ozkahraman et al., 2024, Choudhary et al., 24 Jan 2026, Sakr et al., 2018). In alkaline medium, glucose is oxidized to gluconolactone, releasing electrons detected as a current. Charge transfer is strongly modulated by electrode geometry, surface area, and interfacial chemistry.

Physical Property Transduction

Glucose alters refractive index, dielectric permittivity, and absorption spectra in optical or electromagnetic fields. Surface plasmon resonance (SPR) sensors (Liu et al., 24 Apr 2025, Akafzade et al., 2020), fiber-optic evanescent wave sensors (Pattanayak et al., 2024), microwave/millimeter-wave resonators (Omer et al., 2019, Tay et al., 2 Mar 2025), and NIR spectroscopy devices (Jain et al., 2019, Gani, 12 Sep 2025) exploit these intrinsic property changes to quantify glucose without need for chemical recognition.

2. Materials Platforms and Device Architectures

A diverse set of materials and device concepts underpin modern non-enzymatic glucose sensors. Representative examples include:

Platform	Mechanism	Sensitivity/Range
BP/g-CN heterostructure patch	Direct oxidation, P–N coupling	1.1 $\mu$A mM$^{-1}$ cm$^{-2}$ (0.1–1.0 mM) (Ozkahraman et al., 2024)
Boronic acid–CNT FET	Diol binding, field effect	LOD ~0.3 $\mu$M, dynamic 1 $\mu$M–100 mM (Lerner et al., 2013)
Ag/SiNW plasmonic watch (4-MPBA)	SPR, SERS + boronic acid	LOD 0.02 mM (SPR), 0.12 mM (SERS) (Liu et al., 24 Apr 2025)
Graphene–Schottky junction	Direct H$_2$O$_2$ detection	LOD 0.5 mM, range 0–15 mM (Serry et al., 2015)
NiO nanoflower on Ex-G electrode	Ni(OH)$_2$/NiOOH redox	304.12 $\mu$A mM$^{-1}$ cm$^{-2}$, LOD 0.1 mM (Choudhary et al., 24 Jan 2026)
Co-doped hematite FOEW fiber	Evanescent wave absorption	LoD $\approx$ 4 mM, S = –0.0674 $\mu$A/mM (Pattanayak et al., 2024)
NIR dual-wavelength spectroscopy	Beer–Lambert absorption	Mean absolute error 4.6%, 70–450 mg/dL (Jain et al., 2019)
Microwave triple-pole CSRR	Dielectric shift (S$_{21}$)	S$_L$ up to $9.16 \times 10^{-3}$ dB/(mg/dL) (Omer et al., 2019)
WGM disc mm-wave resonator	Permittivity shift (S$_{21}$)	0.025–0.077 dB/(mg/dL), LOD 0.13 mg/dL (Omer et al., 2019)

Material selection dictates core properties: electrocatalysts like NiO, BP/g-CN, and PtO/graphene enable direct, enzyme-free glucose oxidation; chemical functionalization (boronic acid, 4-MPBA) imparts diol specificity; plasmonic or high-Q dielectric structures provide field enhancement and sensitivity to subtle permittivity or refractive changes.

3. Transduction Mechanisms and Performance Metrics

Electrochemical Approaches

Amperometric and cyclic voltammetric responses quantify the electron flux arising from glucose oxidation on the sensor surface. Sensitivity ($S$) is typically reported as $\mu$A mM$^{-1}$ cm$^{-2}$, with state-of-the-art systems exceeding 300 $\mu$A mM$^{-1}$ cm$^{-2}$ for nanostructured NiO/Ex-G (Choudhary et al., 24 Jan 2026). Limits-of-detection (LOD) range from sub-micromolar for FETs (Lerner et al., 2013) to $\sim$0.1 mM for planar electrodes (Ozkahraman et al., 2024). Selectivity is mediated by surface chemistry; antifouling properties arise either from intrinsic charge (BP/g-CN) or redox potential (NiO, PtO).

Optical and Plasmonic Sensors

SPR and SERS provide highly sensitive, label-free optical signatures. For example, in the Ag/SiNW dual-mode platform, resonance shifts and SERS peaks correspond quantitatively to glucose–boronic acid adduct formation, giving linear responses from 10 $\mu$M to 100 mM and LODs as low as 0.02 mM (SPR) and 0.12 mM (SERS, portable) (Liu et al., 24 Apr 2025). Functionalized plasmonic constructs (Ag/Si$_3$N$_4$/Au) enhance evanescent field intensity and sensing depth, with resolvable glucose concentration changes below 0.1% w/v (Akafzade et al., 2020).

Microwave and mm-Wave Sensors

Dielectric and permittivity changes induced by glucose are transduced via S$_{11}$ (reflection) or S$_{21}$ (transmission) coefficients. High-Q split-ring resonators and WGM dielectric discs yield sensitivity up to 0.077 dB/(mg/dL) with LOD below 1 mg/dL (Omer et al., 2019, Omer et al., 2019). Field enhancement in split-ring or triple-pole geometries boosts detection limits and sharpens resonance dips, critical for wearable or non-contact implementations.

Optical/NIR Spectroscopy

Glucose exhibits weak yet specific vibrational overtone and combination absorptions in the 940–1300 nm NIR region. Beer–Lambert-based quantification using multiple wavelengths and machine learning models achieves mean absolute errors of 4.6% and validation in the 70–450 mg/dL range (Jain et al., 2019). Advanced approaches, such as dual-modal SWIR imaging and CNN regression, further improve non-contact accuracy (MAPE $<$5%, 100% Clarke Error Grid Zone A) (Belfarsi et al., 15 Jun 2025). Robust physics-informed regression outperforms unconstrained DNNs in ultra-realistic, noisy scenarios (RMSE 13.6 mg/dL, Clarke A 95.8%) (Gani, 12 Sep 2025).

4. Device Integration: Wearable and Non-Invasive Platforms

Non-enzymatic glucose sensors have achieved high degrees of miniaturization, wearability, and integration.

Wearable Optical Devices

The Ag/SiNW–boronic acid plasmonic platform is implemented as a fully integrated smart watch: a 638 nm diode laser interrogates the functionalized nanowires, with a photodiode spectrometer, BLE transmission, and real-time app visualization; total power is ~20 mW and LOD is 0.052 mM in wearable use. Validation confirms stable, accurate, real-time monitoring in sweat, with <5% drift and high correlation to blood glucose (R$^2>$0.95) (Liu et al., 24 Apr 2025).

NFC-Powered Patches

The BP/g-CN heterostructure is integrated into a flexible microfluidic sweat patch, with an NFC chip enabling battery-free operation and smartphone readout. Microchannels ensure consistent sweat delivery, and on-body treadmill trials demonstrate response times $<$30 s, baseline drift $<$5%/2 h, and robust selectivity against sweat interferents (Ozkahraman et al., 2024).

Non-Contact and Photonic Sensing

Center-illumination-area-detection (CIAD) geometry enables truly non-contact reflectance glucose assays over meter-scale distances, linear up to $\sim$18 mg/dL of glucose in scattering media. This geometry maximizes the collection of deeply scattered photons, critical for high SNR in turbid tissues (Piao et al., 2020).

Point-of-Care and Optical Fiber Solutions

Co-doped hematite on fiber-optic evanescent wave probes enables compact, low-cost, and biocompatible solutions with LoD $\approx$4 mM and response times $\sim$60 s (Pattanayak et al., 2024). TMO-based (NiO) nanostructured electrodes fabricated on exfoliated graphite extend utility to disposable strip formats (Choudhary et al., 24 Jan 2026).

5. Analytical Trade-offs and Comparative Performance

Non-enzymatic sensors surpass enzyme-based platforms in several aspects:

• Stability and Longevity: Absence of protein catalysts eliminates denaturation, drift, and shelf-life constraints (Serry et al., 2015, Lerner et al., 2013).
• Response Time: Many architectures exhibit equilibrium times $<$1 s (CNT/BP/G heterostructures) or short incubation (15 min, Ag/SiNW–SPR) (Liu et al., 24 Apr 2025, Ozkahraman et al., 2024).
• Selectivity: Molecular recognition via boronic acids or redox-potential-based discrimination (e.g., BP/g-CN, NiO) imparts high specificity; however, direct electrochemical sensors can possess reduced selectivity relative to enzyme-linked systems (Serry et al., 2015, Sakr et al., 2018).
• Sensitivity and Limit of Detection: Platforms span sub-micromolar (boronic acid–CNT) to sub-millimolar LOD (NiO nanoflowers, BP/g-CN, SPR/SERS). High specific surface area, interfacial engineering (P–N coupling, metal oxide doping), and field enhancement (plasmonic, microwave) are key determinants of analytical performance (Ozkahraman et al., 2024, Choudhary et al., 24 Jan 2026, Liu et al., 24 Apr 2025).
• Calibration and Interference: Physics-informed and regression models can exploit intrinsic multivariate contrasts to compensate for confounders, outperforming uninformed DNNs in realistic scenarios (Gani, 12 Sep 2025).

6. Representative Methodologies and Mathematical Frameworks

Mathematical expressions and calibration models are central to device operation and performance prediction:

• Langmuir and Hill–Langmuir Isotherms for boronic acid–glucose binding:

$$\theta = \frac{K_a[\mathrm{Glucose}]}{1 + K_a[\mathrm{Glucose}]}$$

or

$$\frac{\Delta I}{I} = A\frac{(c/K_a)^n}{1 + (c/K_a)^n} + Z$$

(Lerner et al., 2013).

• Direct Electrochemical Sensing:
  • Ni(OH)$_2$/NiOOH redox cycling and DPV/chronoamperometry for NiO-based electrodes:

  $$\mathrm{NiOOH} + \mathrm{Glucose} \rightarrow \mathrm{Ni(OH)_2} + \mathrm{Gluconolactone}$$

  Sensitivity:

  $S = \frac{\Delta I}{\Delta C \times A}$

  (Choudhary et al., 24 Jan 2026).

• SPR Optical Transduction:

$$k_{\mathrm{sp}} = \frac{\omega}{c}\sqrt{\frac{\varepsilon_m\varepsilon_d}{\varepsilon_m+\varepsilon_d}}$$

$\Delta\lambda_{\mathrm{res}} \approx m\Delta n_s$

(Liu et al., 24 Apr 2025, Akafzade et al., 2020).

• Microwave Resonator Sensitivity:

$$S = \frac{\partial |S_{21}|}{\partial C} = \frac{\partial |S_{21}|}{\partial \omega_r}\frac{\partial \omega_r}{\partial \epsilon}\frac{\partial \epsilon}{\partial C}$$

(Omer et al., 2019).

• Optical/NIR Spectroscopy:

Beer–Lambert Law:

$$A(\lambda) = \log_{10}\Bigl(\frac{I_0(\lambda)}{I(\lambda)}\Bigr) = \varepsilon(\lambda)c\ell$$

(Jain et al., 2019, Belfarsi et al., 15 Jun 2025, Gani, 12 Sep 2025).

7. Future Directions and Open Challenges

Non-enzymatic glucose sensors present several research challenges:

• In Vivo and Real-World Validation: Robustness against physiological variables (hydration, temperature, skin melanin, perfusion) and environmental factors (humidity, ambient light, contact pressure) is critical. Simulated frameworks (e.g., ρ$_{\text{glucose–NIR}}$≈0.21) guide statistical and algorithmic calibration (Gani, 12 Sep 2025).
• Wearable Integration: Advances in NFC, BLE connectivity, microfluidics, and flexible substrates have been demonstrated, but require broader clinical deployment for validation (Ozkahraman et al., 2024, Liu et al., 24 Apr 2025).
• Machine-Learning-Driven Calibration: Physics-informed regression or hybrid analytical–AI models outperform unconstrained DNNs for embedded platforms, balancing interpretability, computational efficiency, and resilience to domain shift (Belfarsi et al., 15 Jun 2025, Gani, 12 Sep 2025).
• Multiplexed and Multi-Modal Sensing: Combination of optical, electrical, and electromagnetic signatures, and sensor fusion architectures, offer avenues to enhance specificity, minimize confounders, and broaden physiological compatibility (Belfarsi et al., 15 Jun 2025).
• Materials Discovery: Rational design of heterostructures (BP/g-CN), doped nanomaterials (Co–Fe$_2$O$_3$), and surface modification remains central to optimizing surface affinity, charge transfer, and stability (Ozkahraman et al., 2024, Pattanayak et al., 2024).

Non-enzymatic glucose sensing, leveraging interdisciplinary advances in nanofabrication, materials chemistry, photonics, and data science, now encompasses platforms with performance metrics (sensitivity, selectivity, integration) suitable for prospective, continuous, and non-invasive glucose monitoring in real-world biomedical contexts (Liu et al., 24 Apr 2025, Ozkahraman et al., 2024, Belfarsi et al., 15 Jun 2025, Gani, 12 Sep 2025, Jain et al., 2019, Choudhary et al., 24 Jan 2026).