Portable Raman leaf-clip sensor for rapid detection of plant stress

The study addresses the need for rapid, specific, and early diagnosis of plant stress phenotypes to prevent yield loss in the context of global food insecurity and climate change. Conventional optical phenotyping methods (imaging and reflectance spectroscopy) are non-destructive and allow repeated measurements, but their spectral changes are often non-specific across different abiotic stresses and may manifest too late for early intervention. Raman spectroscopy provides molecularly specific fingerprints of plant metabolites and has shown promise for early, in vivo diagnostics of plant health, development, and stress. Prior approaches typically required excised samples or positioning whole plants near a stationary instrument. This work aims to develop and validate a portable leaf-clip Raman sensor enabling rapid, reproducible, in vivo measurements on growing plants to specifically diagnose stresses such as nitrogen deficiency before visible symptoms appear, thereby advancing precision agriculture.

Optical techniques for agricultural phenotyping include imaging for morphology and structure, and reflectance spectroscopy for rapid, non-destructive assessment of leaf biochemicals like chlorophyll and carotenoids. However, reflectance changes can be similar across multiple abiotic stresses and general stress responses, limiting specificity and early detection. Handheld Raman spectrometers have been explored for monitoring plant health, early disease diagnosis, and detection of abiotic/biotic stresses. Reported plant Raman studies often relied on detached leaf samples or placing intact leaves on holders near bench instruments. Leaf-clip sensors have historically measured transmission or fluorescence for chlorophyll estimation but are susceptible to leaf anatomical variability (cuticle reflectance, veins, flatness), requiring species-specific calibration. Raman spectroscopy, by measuring inelastic scattering linked to molecular vibrations, offers metabolite-specific markers (e.g., carotenoids, nitrates), which can enable earlier and more specific diagnosis. This study builds on these findings by integrating Raman capabilities into a leaf-clip format to mitigate prior limitations and enhance field usability.

Instrumentation for portable Raman leaf-clip: A fiber-based Raman probe (InPhotonics; anodized aluminum body 4.2" x 1.5" x 0.5" with 1.5" stainless steel tip) coupled to a portable instrument using an 830 nm laser was integrated by TechnoSpex Pte Ltd. Excitation is delivered via a 105 µm core fiber; thin-film filters remove ASE; a focusing lens/window provides a 7.5 mm working distance; back-scattered Raman is collected through the same optics with excitation blocking filters. The instrument spans 100–2000 cm⁻¹ with 10 cm⁻¹ resolution (Avantes HSC spectrometer with TE-cooled 1024 x 58 back-thinned CCD). Laser power tests (75–405 mW at 830 nm) showed no significant spectral differences; typical operation used 130 mW at the sample. Leaf-clip interface: A stereolithography 3D-printed methacrylate photopolymer clip holds the fiber probe (set screws) and leaf (embedded opposing magnets) at fixed distance, suppresses beam displacement, blocks ambient light and transmitted laser for eye safety, and provides a handhold. The clip’s tilted surface suppresses back-reflection into the probe. Reference benchtop instrument: Kymera 328i spectrograph (Andor) with 600 g mm⁻¹ grating and 830 nm excitation. Leaf sections (3 mm diameter; placed on 100 µm fused silica slide) were measured at two locations per leaf (either side of midvein). Prior work validated nitrate Raman measurements against chemical analysis and mutant lines, and demonstrated time-series diagnosis of nitrogen deficiency. Spectral acquisition and processing: For each sample spot, five spectra were collected with 10 s integration. Cosmic rays were detected and removed. Spectra were smoothed (Savitzky-Golay filter) and averaged to obtain a representative spectrum. Residual fluorescence was removed by positive-residual polynomial subtraction. Frequency calibration used the instrument’s attenuated 830 nm line and a polystyrene standard. Peaks corresponding to carotenoids were identified at 1520, 1150, and 1004 cm⁻¹ from chemical standards; nitrate was identified at 1046 cm⁻¹ using NH4NO3 and KNO3 standards. Plant materials and growth: Arabidopsis thaliana WT (Col-0), Pak Choi (Brassica rapa chinensis), Choy Sum (Brassica rapa var. parachinensis) were germinated on agar MS medium and grown at 22 °C, 60% RH, long-day (16 h light/8 h dark), white light 100 µmol m⁻² s⁻¹. Nutrient treatments used modified Hoagland’s solution: +N (with 2 mM Ca(NO3)2, 3 mM KNO3, pH 5.8) and −N (replacing nitrates with 2 mM CaCl2 and 3 mM KCl, pH 5.8). Additional soil-based Choy Sum plants were subjected to drought (no water for 3 days) and heat stress (25 °C vs control 20 °C) after growth to 3 weeks. Experimental design: Arabidopsis: five biological replicates per condition (+N, −N), two positions per leaf, five spectra per position. Comparisons were made between in vivo leaf-clip portable measurements and benchtop measurements on leaf sections. Vegetables (Pak Choi and Choy Sum): three biological replicates, two locations per leaf of same age; five spectra per location (plant spectrum is mean of 10 spectra). The nitrate peak at ~1045–1046 cm⁻¹ was quantified; the adjacent 1067 cm⁻¹ peak was used as an internal reference to compute the 1045/1067 ratio. Signal-to-noise ratio (SNR) was calculated from 30 spectra (10 s each) at the same location from two biological replicates for both systems. PCA was used to distinguish control vs drought and temperature stressed plants in Choy Sum.

• Developed a portable 830 nm Raman leaf-clip sensor enabling rapid, reproducible in vivo measurements under full light conditions, with spectra comparable to benchtop Raman measurements of leaf sections.
• Specific detection of nitrogen status via the nitrate Raman peak at ~1045–1046 cm⁻¹: clear separation between +N and −N plants in Arabidopsis, Pak Choi, and Choy Sum.
• Internal referencing using the adjacent 1067 cm⁻¹ peak reduced variability across biological replicates and leaf locations, improving classification between +N and −N conditions.
• SNR results for Choy Sum nitrate peak (1045 cm⁻¹): benchtop SNR = 5.15 and 5.31; portable SNR = 2.51 and 3.31. For the ratio 1045/1067: benchtop SNR = 6.28 and 6.38; portable SNR = 5.05 and 5.06, indicating comparable performance for peak ratios despite smaller light collection in the portable system.
• Robust acquisition across diverse leaf morphologies (Arabidopsis, Lettuce, Pak Choi, Choy Sum, Spinach, Kailan) with consistent identification of common plant Raman bands, including carotenoids (1520, 1150, 1004 cm⁻¹), lignin/cellulose, phenylalanine (1003 cm⁻¹), pectin (747 cm⁻¹), and nitrate (~1046 cm⁻¹).
• Early stress diagnostics beyond nitrogen: within 3 days, drought and elevated temperature stresses in Choy Sum produced detectable spectral changes in secondary metabolites; PCA clearly separated control, drought, and heat-stressed plants.
• Operational parameters: 10 s integration per spectrum; five spectra per spot; typical laser power 130 mW at 830 nm; spectral span 100–2000 cm⁻¹; resolution 10 cm⁻¹; laser power variations (75–405 mW) did not significantly change spectra.

The portable leaf-clip Raman sensor directly addresses the need for early, specific detection of plant stresses by measuring molecular fingerprints in vivo. The nitrate peak at ~1045–1046 cm⁻¹ serves as a specific marker of nitrogen status, overcoming limitations of reflectance-based methods that conflate multiple stresses. Internal referencing to the 1067 cm⁻¹ peak compensates for spatial heterogeneity (veins, surface orientation, reflectance), improving reproducibility and classification between nutrient-sufficient and deficient plants. Comparisons show that in vivo portable measurements yield results consistent with benchtop systems, enabling field-deployable diagnostics without sacrificing specificity. The device captured early metabolic changes under drought and heat stress, and carotenoid/anthocyanin signatures suggest broader applicability to diverse stress phenotypes (heat, cold, saline, light). These findings are significant for precision agriculture, enabling real-time monitoring and management of nutrient application and stress mitigation to prevent yield loss and environmental impacts from over-fertilization.

The study demonstrates a portable leaf-clip Raman sensor capable of rapid, reproducible in vivo detection of plant stress, with a focus on early diagnosis of nitrogen deficiency in Arabidopsis thaliana, Pak Choi, and Choy Sum. Spectra obtained under growth conditions are comparable to benchtop measurements. Using the nitrate peak (~1045–1046 cm⁻¹) with an internal reference (1067 cm⁻¹) enables robust discrimination between nitrogen-sufficient and deficient plants. Preliminary results show applicability to drought and temperature stress, and observable carotenoid and other metabolite peaks indicate potential for diagnosing a wider range of stresses. The simplicity and speed of the probe make it suitable for field use and rapid cultivar screening. Future work could expand validated spectral markers for additional stresses (e.g., salinity, light extremes), refine calibration across species and leaf architectures, and integrate automated analytics for on-site decision support.

Intensity variability arises from leaf heterogeneity (veins, surface orientation, reflectance), leading to scatter in raw peak amplitudes; internal referencing mitigates but does not eliminate this. The portable system has lower absolute SNR than benchtop instruments due to smaller light collection and miniature components, though peak ratios are comparable. Demonstrations of drought and heat stress classification are preliminary with limited replicates and short duration (3 days). Prior leaf-clip optical sensors are sensitive to leaf anatomy; while Raman internal referencing reduces sensitivity, species-specific factors may still affect measurements and require validation. The study focuses primarily on nitrogen deficiency, with broader stress diagnostics indicated but not fully validated across diverse conditions and species.