RDRS SpectraLib

About RDRS SpectraLib

A comprehensive, instrument-grade web application designed for the visualization, advanced processing, and high-speed mathematical matching of Raman and infrared spectra. Built to process proprietary laboratory data alongside massive open-source reference libraries.

Data Privacy Guarantee: All spectrum parsing and processing happens completely locally in your browser. Your raw data files and spectra are never uploaded, saved, or collected by RDRS SpectraLib server.

Basic Workflow (Multiple Spectra)

RDRS SpectraLib is designed to handle dozens of spectra simultaneously. When you open any processing tool (like Smooth or Baseline), you will see a list of all your loaded spectra.

1. Check the boxes next to the spectra you want to process. (Visible spectra in the chart are checked by default).
2. Select your settings and click Apply.
3. By default, RDRS SpectraLib keeps your original data and creates a brand new, processed copy with a tag (like [SG] or [Cut]) added to the name.

Data Protection & Export Restrictions

RDRS SpectraLib operates under strict licensing agreements with external providers (such as RRUFF™ and ROD). To protect the intellectual property of these databases, certain restrictions are enforced:

• Export Block: Reference spectra loaded from the Database or generated via the Match tool cannot be exported as raw data files (CSV, TXT, or JCAMP-DX).
• Derived Protection: Any spectrum created by processing a reference file (e.g., Smoothing, Baseline Correction, or Averaging a reference) inherits this protection and remains locked.
• Project Saves: When saving a Project (.json), protected reference spectra are automatically stripped from the file to prevent unauthorized redistribution of raw database arrays.

Reference Database

Load pure, verified reference spectra directly into your chart for visual comparison.

How to use: Type a mineral name (e.g., "Quartz") into the Global Search, or use the dropdowns to browse by Technique and Class. Check the boxes of the minerals you want, and click Add to Chart. Hovering over a result will show a live preview overlay.

Match Spectrum to Database

Identify an unknown sample by comparing it against our 11,400+ reference library using traditional mathematical algorithms or advanced Artificial Intelligence.

Available Algorithms:

• Machine Learning: Features models trained specifically on pure Raman, Infrared (FT-IR), and X-Ray Diffraction (XRD) datasets. Achieves high validation accuracy by intelligently ignoring noise and background humps. Best for identifying the single dominant phase in a sample.
• AI: Multi-Phase Mixture (Experimental): A multi-label neural network designed to untangle and identify multiple overlapping minerals in a single mixed rock. (Available for standard Raman and XRD). Use the Sensitivity Slider to adjust detection limits: lower it to find hidden trace minerals, or raise it to prevent false positives.
• 1st Derivative: The traditional gold-standard. Mathematically strips away sloped baselines to match pure peak inflection points.
• Pearson / Cosine: General curve shape and intensity correlation matchers.
• Shift-Invariant (X-Axis Drift): Uses mathematical cross-correlation to perfectly match spectra even if your spectrometer's laser suffered from thermal drift, completely avoiding the need to manually align data first!

How to use:

1. Select exactly one unknown spectrum from the list.
2. Select the correct Database Category to route your data to the appropriate AI brain. (Ensure your X-axis units match the required format for that technique!)
3. Select your algorithm. (If using the Multi-Phase Mixture, adjust your Sensitivity threshold).
4. Click Find Best Matches.

Note: Results are paginated. You can load up to the top 50 closest matches by clicking "Show Next 10 Results" at the bottom of the list.

Spectra Galaxy

Visualize the entire reference database as an interactive 2D map using UMAP (Uniform Manifold Approximation and Projection). Minerals with similar molecular structures and spectral fingerprints naturally cluster together into physical "galaxies".

Available Analytics & Color Maps:

• Chemical Superclass (Dana): Categorizes minerals into standard families (Silicates, Carbonates, Sulfates, Oxides, etc.) based on their elemental composition.
• Primary Peak Position: Applies a continuous thermal heatmap based on the location of the absolute highest intensity peak in the spectrum.
• Spectral Entropy: Maps the mathematical "messiness" or complexity of the signal. Clean, sharp spectra contrast heavily against highly fluorescent data or complex mixtures.
• Center of Mass: Maps the weighted centroid of the spectral curve, highlighting regions where the bulk of the scattering or diffraction occurs.

How to use:

1. Select one unknown spectrum from your workspace to automatically project it onto the map as a Yellow Star (⭐). (Or select none to freely browse the database).
2. Select the correct Instrument Category (Raman, FT-IR, or XRD) from the Galaxy Controls to load the corresponding universe.
3. Use the Color Map By dropdown to analyze the clusters, and drag/scroll to pan and zoom around the map.
4. Click any dot in the cloud to instantly compare its database reference against your target in the bottom-right 1D Spectrum Comparison panel.

Note: The points physically closest to your Target Star on the 2D map are mathematically your most likely phase matches. If you lose your place while panning, click "Locate My Target Spectrum" to snap the camera back.

Plot Digitizer

A massive quantity of historical and technical data is only available as chart images in published PDFs. The Plot Digitizer uses a combination of computer vision techniques and manual inputs to recover raw data arrays from these images.

Phase 2: Data Extraction

• Manual Mode: Click Add Points Mode and manually click along the curve to extract data points. Use the Undo or Clear buttons to adjust your output.
• Auto (AI) Mode: Uses computer vision to automatically trace colored lines.
  • 1. Mask: Use the highlighter to paint over the specific curve you want. This limits the AI's search to your specified region, ignoring intersecting lines and text.
  • 2. Color Target: Click the eye-dropper and select the exact color of the curve from the image.
  • 3. Run: Click Run Auto-Extraction. The AI will place points along the center of the line.

Phase 3: Generate Spectrum

Once you are satisfied with the extracted points, click the green Generate Spectrum button in the top right. RDRS SpectraLib will automatically convert your pixels into physical units using your calibration data and push the new spectrum directly into your main workspace for analysis!

Advanced Plot Configuration

RDRS SpectraLib gives you deep, publication-level control over how your data is rendered on the screen. These settings are cached in your browser so your preferred lab aesthetic loads automatically every time.

• Global - Rendering Engine: SVG is the default engine (crisp lines, perfect for PDF exports). WebGL offloads rendering to your GPU for 60fps panning when analyzing massive datasets (50+ spectra), though filled areas may show minor artifacts.
• Global - Mirror Axes: Draws a solid bounding box around the top and right sides of the chart, replicating your tick styles. This is the standard aesthetic for most scientific journals.
• Axis Scales (Linear vs Logarithmic): Logarithmic scales are incredibly useful when a massive fluorescence hump dwarfs a tiny Raman signal. Note: Log scales require strictly positive data.
• Ticks & Grids: Independently toggle gridlines, axis lines, and minor ticks for the X and Y axes. You can also change the direction of major ticks (Inside, Outside, Cross, or Hidden) to match specific publication formatting rules.

Workspace & Sidebar Management

Manage large datasets efficiently using the built-in file explorer tools.

• Folders: Click the New button at the bottom of the sidebar to create virtual folders. Grouping spectra allows you to instantly toggle visibility () or delete an entire dataset with a single click.
• ☰ Drag & Drop: Reordering is permanently unlocked! Click and drag the handle next to any spectrum to arrange your list, or drop it onto a folder header to organize it. The chart's layering updates dynamically to match your list.
• Live Filter: When 2 or more spectra are loaded, a search bar appears at the top of the sidebar. Type any part of a name to instantly filter your list. Folders containing matching spectra will automatically expand to show you the results.
• Renaming: Click the pen icon next to any spectrum or folder to rename it inline. Press Enter to save.
• Batch Rename: Click the signature icon on any folder header to rename all spectra inside it simultaneously. Use dynamic variables like <Name> and <Inc 0Nr> to instantly generate organized naming patterns for your clusters.

Display Settings

Visual toggles to customize your workspace appearance. These do not mathematically alter raw data arrays.

• Flip X: Reverses the X-axis direction.
• Legend: Cycles the chart legend between Outside, Floating, and Hidden.
• Grid: Toggles background layout gridlines.
• Reset View: Instantly restores default zoom, offsets, tools, and display states.
• Lines: Cycles the rendering engine between continuous Lines, independent Points, or Both. Use Width to adjust thickness.
• Labels: Adjusts the font size of detected peak annotations.
• Fill: Shades the area under the curve. Reveals a dynamic UV-Vis rainbow gradient if your X-axis is set to [nm].

Interactive Tools

Tools that modify the chart's physical layout and active tracking.

• Stack: Vertically separates overlapping spectra by a standard offset (Δ).
• Auto-Norm: Temporarily min-max scales all spectra to a 0-1 range for visual comparison without changing the underlying raw detector counts.
• Drag Y: Activate this to physically click and drag spectra vertically across the chart for rapid visual alignment.
• Cursor: Click anywhere on the chart to drop vertical tracking lines. Useful for identifying peak boundaries before integration.
• Ridgeline: Creates a pseudo-3D overlapping "mountain" effect. Spectra are automatically normalized to [0,1] and filled to their baseline. Perfect for visualizing peak shifts across massive datasets. Adjust the Δ (Stack Offset) input to control the exact amount of overlap!
• Heatmap: Instantly transforms all visible spectra into a dense 2D color contour plot. Perfect for visualizing temperature series or spatial mapping.
• Freehand Edit: Danger! Manually "paint" over massive artifacts to force the data down to zero. Note: This permanently alters the raw data points.
• Manual Baseline: Click to physically drop and draw a custom baseline under your peaks. Turn on the Snap magnet to lock your clicks directly to the true data line.

Cut Range & Scale Y

• Cut Range: Deletes all data outside of the X-bounds you provide. Excellent for removing noisy detector edges.
• Scale: Multiplies the intensity by a specific factor. Note: You can also visually scale spectra without altering the raw data by changing the ×Scale box in the left sidebar next to the spectrum name!

Data Interpolation (Resampling)

Increases or decreases the data density of your spectrum and forces it onto a strict, uniform X-axis grid. Note: Interpolation does not increase the true optical resolution of your spectrometer; it is primarily used to align disparate datasets so they can be processed by matrix-based chemometrics (like PCA or 2D-COS).

Available Algorithms:

• Akima Spline: The gold standard for spectroscopy. A specialized monotone spline that resists overshooting. It perfectly smooths data while strictly preserving the local shape, preventing fake negative dips around sharp Raman peaks.
• Natural Cubic Spline: Creates an extremely smooth polynomial curve. Warning: Susceptible to "Runge's Phenomenon" (Ringing). If you have a sharp peak next to a flat baseline, a cubic spline will mathematically overshoot and create an artificial negative crater at the base of the peak.
• Linear: Simply draws straight lines between existing points. Mathematically safe with zero artifacts, but leaves low-density peaks looking like jagged triangles.

How to use: Enter a new Step Size (e.g., 1.0 for a point every 1 cm⁻¹). Leave the X Start and X End bounds blank to automatically use the spectrum's current limits.

Zap Peak (Excise)

Removes bad data points (like stubborn artifacts or saturated solvent peaks) and heals the gap with a clean, interpolated line.

How to use: Input the exact X-axis center of the bad peak, and how wide the cut should be. RDRS SpectraLib will delete the data in that window and draw a straight line connecting the two broken ends.

Cosmic Ray Removal (Despiking)

Automatically detects and removes artificially sharp spikes caused by high-energy particles hitting the camera sensor.

• Derivative (Gradient) algorithm: The industry standard. Identifies cosmic rays by hunting for near-vertical instantaneous slopes (which true Raman peaks do not have). Excellent for preserving sharp diamond or silicon peaks.
• Z-Score (Median) algorithm: Best for broader artifacts or saturated pixels.

Tip: Keep Live Preview checked! The red dashed line will show you exactly what the cleaned spectrum looks like before you apply it.

Spectral Alignment (Cross-Correlation)

Fixes instrument drift. Minor temperature changes in a laboratory can cause a spectrometer's laser to drift, shifting the X-axis slightly over time. If you run PCA or LCF on shifted peaks, the math will model the instrument drift instead of the actual chemical variance.

The Math: RDRS SpectraLib uses a 1st Derivative Global Cross-Correlation algorithm. Instead of dynamically stretching or warping the X-axis (which destroys quantitative peak areas), this algorithm slides the entire spectrum left and right to find the perfect overlap with your anchor reference. This strictly preserves your physical peak shapes, FWHM, and Integration Areas.

Smoothing

Reduces high-frequency detector noise. Selecting the right algorithm and window size is critical to reducing noise without artificially broadening your actual chemical peaks.

Available Algorithms:

• Savitzky-Golay: The gold standard for vibrational spectroscopy. It fits a local polynomial across the sliding window. (Preserves sharp peak heights and widths much better than standard averaging. Lower polynomial order = smoother curve).
• Moving Average: A traditional sliding filter. Rectangular weights all points in the window equally, while Triangular gives more weight to the center point (better for preserving peak tips).
• Percentile Filter: A non-linear ranking filter. Setting it to <50% creates a standard Median filter. Setting it lower (<50%) is excellent for mathematically flattening out sharp cosmic rays or artificial spikes that ride on top of the true baseline.

Normalization & Preprocessing

Mathematical scaling and scattering corrections to prepare your data for advanced chemometrics.

• 0 → 1 / 0 → 100% (Min-Max): Scales the Y-axis so the lowest point is 0 and the highest is 1 (or 100). Essential before running Hierarchical Cluster Analysis (HCA) so weak and strong acquisitions of the same mineral group together properly.
• Divide by Max: Scales the highest peak to 1, but leaves the baseline wherever it naturally falls.
• SNV & MSC (Scatter Correction): Standard Normal Variate and Multiplicative Scatter Correction eliminate physical baseline shifts and multiplicative scattering effects caused by sample surface roughness (e.g., in powders). Highly recommended before PCA Clustering.
• Mean-Centering: Subtracts the average of the dataset from every spectrum. Mandatory prerequisite before running Train & Predict (PLS-R/SVM) models.

Automatic Baseline Correction

Removes thermal noise, fluorescence, or sloped backgrounds from your data.

• ALS (Asymmetric): The gold standard for massive fluorescent humps.
• Rolling Ball: Acts as a morphological filter, perfectly tracing highly irregular or "wavy" backgrounds (like glass slides).
• Adaptive (SNIP): Highly effective iterative clipping, excellent for standard Raman.
• Scattering (Poly): Fits a polynomial to handle broad, wavy backgrounds.

Shift X-Axis

Corrects instrumental calibration errors by manually shifting the data horizontally.

How to use: Enter a positive number to shift the peaks to the right (higher wavenumbers), or a negative number to shift them to the left. The Y-intensities remain completely unchanged.

Unit & Axis Conversions

A unified tool to mathematically transform your data arrays or fix imported metadata.

• Vibrational Math: Converts physical X units for optical spectroscopy (e.g., nm ↔ cm⁻¹, eV, THz). Note: Formulas like nm to eV mathematically reverse the array, which RDRS SpectraLib handles and re-sorts automatically.
• Diffraction Math (XRD): Uses Bragg's Law (nλ = 2d sinθ) to convert between 2-Theta angles and physical d-spacing.
• XRD Target Conversion: Crucial for AI Matching! The RDRS AI models are trained exclusively on Copper (Cu Kα1) radiation. If your diffractometer used a different anode (like Cobalt or Molybdenum), you must use this tool to mathematically convert your 2-Theta angles to the Copper standard before running a match.
• Y-Axis Math: Converts intensity scales (e.g., Transmittance % ↔ Absorbance).
• Rename Labels: Overwrites the global axis titles for your PDF reports without altering the underlying raw data points.

Batch Processing Pipeline (Macros)

The ultimate power-user tool. Apply an entire sequence of mathematical operations to dozens of spectra with a single click.

How to use: Check the boxes for the operations you want to apply (Cut, Smooth, Baseline, Normalize), adjust the specific parameters, and click Run Pipeline. RDRS SpectraLib will sequentially chain the math together in the background.

Find Peaks

Automatically detect and label peaks (maxima) or valleys (minima) based on height thresholds.

When peaks are found, RDRS SpectraLib automatically calculates the FWHM (Full Width at Half Maximum) for crystallographic analysis. The results are automatically logged to the Console and exported in your PDF reports.

Integrate Peak Area

Calculates the area under a curve, which is proportional to chemical concentration.

How to use: Click the ✛ Cursor tool in the left panel to find the X-axis start and end points of your peak. Enter those numbers here. The Console Log will open and display the Total Area (down to zero) and the Net Area (above the local background).

Colorimetric Analysis (CIE 1931)

Extracts the true mathematical color (Hex code and x,y coordinates) from an emission or photoluminescence spectrum.

How to use: Your X-axis must be set to Wavelength [nm]. Input the X bounds to isolate the material's emission peak (carefully excluding the excitation laser/LED). The application integrates the area using standard CIE 1931 Color Matching Functions and outputs the precise color swatch directly to the Console.

Peak Curve Fitting (Deconvolution)

Resolves heavily overlapping peak clusters into individual mathematical components using Levenberg-Marquardt optimization. You can now model true physical and instrumental states by selecting specific Profile Shapes.

How to use:

1. (Highly Recommended) Click Visually Pick Centers. The modal will become transparent. Click directly on the chart to drop vertical markers where you suspect hidden peak shoulders are located. (Right-click to undo, or use the trash icon to start over).
2. Auto-Bounding: As you drop centers, RDRS SpectraLib will automatically calculate the optimal X Start and X End bounds to isolate the cluster for you! (You can still type these manually if needed).
3. Click Finish Picking.
4. Select your Profile Shape. (See the guide below).
5. If you choose to skip visual picking, manually enter your X-bounds and select the Number of Peaks (up to 10) to force the solver to blindly guess.
6. Click Run Optimization.

Advanced Solver Parameters: If your data is extremely noisy and triggers an 'Optimization failed to converge' error, expand the Advanced Parameters drawer. Increase the Max Iterations (e.g., 20000) or lower the Tolerance (e.g., 1e-6) to force the math to solve it.

Derivative Spectroscopy

Calculates the 1st or 2nd derivative to identify hidden shoulders or overlapping bands. Because derivatives amplify signal noise exponentially, always apply a Savitzky-Golay Smooth prior to using this tool.

Spectral Stripping (Subtract)

Used to mathematically remove a pure reference phase from a mixed sample spectrum.

How to use: Select your mixed sample (Minuend) and the pure database reference (Subtrahend). Slowly drag the multiplier slider. Watch the chart update live until the reference peaks visually disappear from your mixed sample.

Merge Spectra (Stitch)

Combines multiple spectral segments (e.g., from a dual-grating spectrometer) into one continuous trace.

How to use: Select the spectra you want to weave together. RDRS SpectraLib will automatically sort them by X-axis, interpolate any empty gaps, and average any overlapping regions to create a single, seamless line.

Add & Divide Spectra

Perform direct mathematical operations between multiple target spectra and a single reference spectrum.

• ＋ Add Spectra: Adds the Y-values of a reference spectrum to your selected targets.
• ÷ Divide Spectra: Divides your target spectra by a reference spectrum. Essential for calculating Transmission/Reflectance ratios.

Auto-Alignment Magic: RDRS SpectraLib will automatically interpolate the reference data to perfectly align with each target's X-axis before calculating the results.

Spectrum Averaging & Variance

Calculates the Mean (average) of multiple spectra. Highly recommended for heterogeneous samples where multiple acquisitions were taken. If you check Draw ±1 Std. Dev., it will draw a semi-transparent shaded area behind the main line representing the sample variance!

Spectral Density Analysis

Converts a massive, unreadable overlap of spectra (a "hairball") into a 2D statistical density heatmap.

How to use: Select your spectra and adjust the Y-Axis Resolution (bins). The resulting visualization will show bright/hot regions where the majority of your spectra share a path, making it easy to see the dominant trend while ignoring rare outliers or isolated noise.

Exploratory Cluster Analysis

Advanced unsupervised Machine Learning techniques that reduce the dimensionality of complex spectra, compressing thousands of data points into a 2D scatter plot to group chemically similar spectra together.

Available Algorithms:

• PCA (Linear): Best for understanding global variance and identifying the exact physical peaks that drive the separation between samples.
• t-SNE (Non-Linear): Highly effective at untangling complex, overlapping data into tight, distinct local clusters.
• UMAP (Hybrid): The modern successor to t-SNE. Creates distinct local clusters while mathematically preserving the global distances between groups.

How to use: Select at least 3 spectra, choose your algorithm, and click Compute. Depending on your choice, the dashboard will display:

• Scree Plot (PCA Only): Shows how much physical variance is explained by each Principal Component (PC). The curve flattens out when the remaining components are just noise.
• Loadings (PCA Only): Shows the physical Raman shifts (or wavelengths) that are driving the separation in the clusters. Peaks here indicate why spectra are grouping together.
• Scores Matrix (All Algos): The main scatter plots. Chemically identical spectra will cluster together. The diagonal plots (KDE) show the statistical distribution (bell curves) of your data. (Expands to full screen for t-SNE and UMAP).

Hierarchical Cluster Analysis (HCA)

HCA is an advanced unsupervised chemometric technique that calculates the mathematical distance (variance) between spectra and groups them into a "family tree" called a Dendrogram. It is the gold standard for classifying unknown samples or proving that different acquisitions belong to the exact same mineral phase.

How to use:

1. Select at least 3 spectra (the more, the better).
2. (Optional) Check Include Distance Matrix Heatmap if you want to visualize the exact Euclidean distance between every single pair of spectra as a 2D color grid.
3. Click Build Tree.

Train & Predict (Chemometrics)

Build custom Machine Learning models (PLS-R or SVM) directly in your browser to predict continuous concentrations or classify unknown samples.

• PLS Regression (PLS-R): Predicts continuous numbers (e.g., % Concentration, Temperature, Hardness). Requires strictly numerical training values.
• SVM Classification: Predicts categorical groups (e.g., "Cancer" vs "Healthy", or "Quartz" vs "Calcite").
• LDA Discriminant Analysis: A supervised classification method that projects spectra into a 2D scatter space, maximizing the mathematical separation between your defined categories.

The Proper Workflow:

1. Pre-process: Your data must be mathematically clean to prevent the algorithm from memorizing background noise. Apply Auto Baseline → Cut Range (to isolate the active peaks) → SNV → Mean-Centering.
2. Assign Training Data: Open the Train & Predict tool. Type your known values (e.g., "85") into the boxes next to your reference spectra.
3. Assign Unknowns: Leave the boxes completely blank for the spectra you want the AI to predict.
4. Click Train Model & Predict. The results will be automatically logged to the Console for easy copying to Excel.

Univariate Calibration (Beer's Law)

Builds a standard linear calibration curve linking peak intensity (or area/FWHM) to physical concentration.

How to use: Select a target peak, and input the known concentrations for your standard spectra. The system will calculate the linear regression (R²) and predict the concentrations of any blank "Unknown" spectra. A visual calibration curve and detailed metrics are logged to the Console.

Sequential Kinetic Curve Fitting

Analyzes time-series reactions by tracking a target peak's intensity or area over time to calculate physical rate constants and half-lives.

How to use: Visually pick a target peak. The algorithm will automatically extract the time intervals from your filenames (e.g., "10s", "15min") or use a fixed interval. Select the Reaction Order (0, 1st, or 2nd) and compute. The resulting decay/growth curve and mathematical constants (k, t½, R²) will be displayed in a dedicated dashboard.

Peak Trend Analysis (SPLOM)

Generates a Scatterplot Matrix (SPLOM) to track how a specific peak's Position (X), Intensity (Y), FWHM, and Area evolve across a sequence of files. Essential for kinetics, thermal degradation, and phase transition studies.

How to use:

1. Run the Find Peaks tool on your loaded dataset to extract the mathematical parameters.
2. Open the Trend / SPLOM tool.
3. Click Visually Pick Target Peak and click the specific peak you want to track on the main chart.
4. Click Generate Correlogram. The algorithm will hunt for that peak in every file (within your specified ± Tolerance window) and map the correlations.

2D Correlation Spectroscopy (2T2D)

Two-Trace Two-Dimensional (2T2D) Correlation Spectroscopy mathematically analyzes how spectral bands change across a dynamic dataset. It spreads highly overlapping 1D peaks across a 2D plane to resolve hidden components and track physical changes.

2. How to read the Contour Maps:

• Synchronous Map (In-Phase): Highlights simultaneous changes.
  • Auto-peaks (The Diagonal): The dense spots running bottom-left to top-right. These show the regions experiencing the greatest intensity changes across your dataset.
  • Cross-peaks (Off-Diagonal): If Red (Positive), the two intersecting peaks are growing or shrinking together. If Blue (Negative), one peak grows while the other shrinks.
• Asynchronous Map (Out-of-Phase): The ultimate tool for resolving hidden phases.
  • The Diagonal: Always mathematically zero (white/flat).
  • Cross-peaks ("Butterflies"): Red and Blue off-diagonal clusters prove that the peaks are changing independently of each other. This is definitive proof of a peak shifting (due to chemical substitution/stress) or the presence of a completely different, hidden mineral phase.

3. The interactive slicer (numerical purification)

The Slicer isn't just a visualization tool; it acts as a numerical curve-resolution tool. By mathematical definition, a horizontal slice of an asynchronous 2T2D map taken at a specific characteristic peak is entirely devoid of the spectral contribution from that peak's species.

Workflow: Ensure Show 1D Margins is checked. Hover over the center of an Asynchronous "butterfly" cross-peak. Click it. The 1D slice on the right has now mathematically stripped away the interference from the peak pointing to the Y-axis. Click Save Extracted Slice to push this purified trace back into your workspace for further analysis!

Linear Combination Fitting (LCF)

Quantifies the percentage composition of a mixed sample using Non-Negative Least Squares (NNLS).

How to use: Baseline correct your mixed sample. Load the pure reference spectra from the database. Open the LCF tool, select the mixture as the Target, check the references, and calculate. The algorithm builds a synthetic "Fit" curve and logs the exact mathematical percentages to the Console.

MCR Blind Unmixing

Multivariate Curve Resolution (MCR-ALS) mathematically extracts the pure "hidden" components from a mixed dataset without needing a reference database.

How to use:

1. Select your mixed spectra (e.g., a time-series of a chemical reaction or a spatial map).
2. Input the number of pure components you suspect are in the mixture.
3. Keep Use SIMPLISMA initial guess checked (highly recommended) to seed the algorithm with real physical peaks rather than blind algebraic guesses.
4. Click Run Extraction. The algorithm will mathematically deduce the pure spectral signatures and add them directly to your chart.

Synthetic Mixture Builder

Dynamically mix pure references to visually match an unknown target spectrum, or build a hypothetical mixture from scratch.

How to use: Select a Target (or check "Blank Target"), then select up to 5 pure components. Use the sliders to adjust the proportional weight of each component. The cyan dotted line previews the synthetic mix in real-time on your chart.

Evaluate & Prune (SNR)

A dual-purpose dataset cleanup tool. Evaluate SNR calculates signal-to-noise ratios to flag noisy acquisitions, while Smart Prune automatically isolates the most pristine and representative spectra from massive datasets.

UV-VIS Acquisition (Live Hardware)

Connect USB spectrometers, CMOS sensors, or Virtual Cameras (via OBS/SharpCap) directly to the browser for real-time photon digitization.

• Region of Interest (ROI): Use Extract Y and Slit Height to isolate the diffraction streak. Adjust Start/End X to crop dead sensor pixels; this significantly reduces CPU overhead on 4K sensors.
• Sensor Extraction Channel: Select standard Luminance (Grayscale) or isolate specific Bayer RGB channels (Red, Green, Blue). Isolating the physical color channel that matches your optical diffraction peak heavily reduces noise on standard webcams.
• Live Smooth (IIR): An Infinite Impulse Response temporal filter. Higher values integrate light over more frames, effectively suppressing high-frequency sensor noise (Dark Current) at the cost of responsiveness.
• Dark Frame Subtraction (D): Turn off your light source and click Set Dark to capture the sensor's thermal noise floor. Toggle Apply Dark to mathematically subtract this from every frame.
• Flat-Field Correction & Analysis Mode (W): Capture a "White" reference (e.g., a halogen lamp). Enabling Apply Flat unlocks the Analysis Mode dropdown. Select Trans % to calculate Transmittance (Sample - Dark) / (White - Dark), or select Abs (A) to calculate true quantitative Absorbance (-log10(T)) according to Beer's Law.
• Kinetics Series: Automated time-lapse acquisition. Define an Interval and Total Captures to monitor chemical reactions. Results can be automatically piped into a 2D Heatmap or 3D Waterfall view upon completion.

Wavelength Calibration [nm]

To convert raw pixels into physical units, use a known emission source (like a fluorescent lamp or laser pointer):

1. Drop a Cursor on a known peak (e.g., Mercury 546.1 nm).
2. Click Set P1 in the calibration panel. Repeat for P2 (and P3 for non-linear gratings).
3. Click Apply Math. The system will calculate a Regression Fit (Linear or Quadratic) and project a dynamic UV-Vis rainbow beneath your data.

Privacy Policy & GDPR Compliance

RDRS SpectraLib is committed to protecting your personal data and your scientific intellectual property in strict accordance with the General Data Protection Regulation (GDPR).

1. What data is collected?

• Account Data: When an account is requested, your Name, Institutional Email, Affiliation, and stated Intended Use are collected. This information is stored securely in the database to verify academic/researcher status.
• Usage Telemetry: To monitor server health and improve the application, basic interaction events are logged (e.g., "User ran LCF Unmixing"). Actual spectral data arrays are NEVER logged, uploaded, or viewed.
• Connection Data: IP addresses are temporarily logged during sessions strictly to prevent abuse and access rate-limit violations.

2. How is your data used?

Your email is used exclusively to send initial login credentials, password resets, and critical operational updates regarding the RDRS SpectraLib platform (such as major algorithmic changes or server migrations). Your data will never be sold or used for third-party marketing.

3. Scientific Data Privacy Guarantee

4. Your Rights (Data Deletion)

You have the "Right to be Forgotten." To permanently delete an RDRS SpectraLib account and wipe all associated telemetry/logs from the servers, please send an email to andrei.apopei@uaic.ro from the registered institutional email address.