Ep 19 — Introduction to Qualitative Analysis: Library Search and Matching

Series: Infrared Spectroscopy Encyclopedia: From Principles to Practice
Chapter: Part 2 · Elementary — Entering the Laboratory
Audience: High school/undergraduate students, graduate students, laboratory technicians new to the lab
Prerequisites: Ep 04 (How to Read an IR Spectrum), Ep 17 (Instrument Operation), Ep 18 (Spectrum Processing)
Reading Time: ~38 minutes

Prologue: One Spectrum, One Suspect, Ten Thousand "Comparisons"

At an anti-smuggling scene in 2018, an analyst obtained 1 g of a suspicious white powder. GC-MS would take 2 hours, and the instrument was off-site. Only a portable FTIR was available — it produced a spectrum in 1 minute. But the spectrum was out, what is this? The spectrum showed C=O at 1700, C-O at 1200, C-H at 3000... it looks like an ester, but which ester specifically?

This is the "soul-searching question" of qualitative analysis — spectrum obtained, identity unknown.

The answer comes from Library Search: comparing the unknown spectrum against a database of known "standard spectra" one by one, ranking by similarity, and identifying the top candidates [1][2].

Modern FTIR software typically completes a search of tens of thousands of standard spectra within 1–2 seconds, returning a Top 10 list ranked from highest to lowest "match quality". But high match quality ≠ necessarily correct — library search only provides a "list of suspects"; the final "conviction" must be made by a human [1][3].

This episode will systematically explain the concept of spectral libraries, major commercial libraries, search algorithm principles, the meaning of HQI, mixture search strategies, misidentification risks, and methods for building custom libraries. After reading, you will be able to correctly use library search tools and know when to trust the results and when to doubt them.

💡 Positioning: Ep 1–10 covered "how to read spectra"; Ep 11–18 covered "how to obtain spectra"; this episode begins "how to use spectra for analysis" — starting with qualitative analysis.

1. Concept of Spectral Libraries

1.1 What is a Spectral Library?

A Spectral Library (Database) is a collection of known infrared spectra of compounds. Each entry contains [1][2][4]:

• Compound name
• CAS number / molecular formula / structural formula
• Measurement conditions (transmission/ATR, phase, solvent)
• Spectral data (wavenumber-intensity pairs)
• Source (laboratory, literature)

Core Value [1][2]:

• Provides "fingerprint" references for known compounds
• Enables rapid identification of unknowns
• Provides peak assignment references

1.2 Classification of Libraries

By Source [1][2][4]:

By Content [1][2]:

• General libraries: Cover various compound classes (e.g., Sadtler, Aldrich)
• Specialized libraries: Focus on specific categories (e.g., Hummel Polymer Library, Pharmacopoeia Library)
• Functional group libraries: Organized by functional groups (e.g., ftir.fun Functional Group Library)

1.3 Key Indicators of Library Quality

① Spectrum Quality [1][4]

• SNR > 1000:1
• Peak position accuracy ±0.5 cm⁻¹
• Full spectral range (4000–400 cm⁻¹)
• No flat-top, no ghost peaks

② Complete Metadata [4]

• Compound name (IUPAC + common name)
• CAS number
• Measurement conditions (transmission/ATR, phase)
• Measurer, date, instrument

③ Diversity [1][4]

• Covers chemical structure diversity
• Different concentrations, different phases
• Includes common interferences

2. Major Commercial Libraries

2.1 Sadtler Spectral Library

The Sadtler Spectral Library was founded by Sadtler Research Laboratories and is one of the world's largest infrared spectral libraries [1][2][4].

• History: Started in the 1950s, now over 250,000 spectra
• Coverage: Organics, inorganics, polymers, pharmaceuticals, pesticides, additives, etc.
• Format: Originally printed cards, now digital
• Features: Standardized sample preparation (primarily KBr pellets)

Sub-libraries [2][4]:

• Sadtler Pure Organic
• Sadtler Commercial Products
• Sadtler Inorganics
• Sadtler Polymers

2.2 Aldrich Spectral Library

The Aldrich Spectral Library was established by Sigma-Aldrich [1][2]:

• Size: ~15,000 spectra of common organic compounds
• Features: Covers Sigma-Aldrich commercial compounds
• Format: Paper + digital
• Advantages: High chemical purity, suitable for research use

2.3 Hummel Polymer Library

The Hummel Polymer Library was established by Professor Dieter Hummel and is the "bible" of polymer identification [1][3][5]:

• History: Since the 1960s
• Size: ~1,900 polymer spectra
• Features: Covers plastics, rubbers, fibers, coatings, additives
• Format: Transmission spectra (films, KBr pellets)
• Applications: Plastic identification, polymer QC, failure analysis

Advantages [5]:

• Includes "fingerprints" of many commercial products
• Contains additives, copolymers, blends
• Detailed metadata (grade, source, use)

🔗 Further Reading: For functional group analysis of polymers, refer to ftir.fun Alkyl C-H functional group page.

2.4 Pharmacopoeia Library

The Pharmacopoeia Library is designed specifically for the pharmaceutical industry [1][6]:

• Chinese Pharmacopoeia Library: ~600 spectra of active pharmaceutical ingredients (KBr pellets)
• USP Library: ~1,200 spectra of active pharmaceutical ingredients
• Features: Measured strictly according to pharmacopoeia methods, compliant with pharmacopoeia regulations
• Applications: API identification, release testing

2.5 NIST Library

The NIST WebBook is maintained by the National Institute of Standards and Technology [1][7]:

• Free online access
• Size: ~7,000 gas-phase infrared spectra
• Features: Gas-phase spectra (different from condensed phase)
• Applications: GC-FTIR hyphenation, gas-phase substance identification

2.6 Vendor-supplied Libraries

Thermo OMNIC [1][8]:

• Built-in basic library (~1,000 common compounds)
• Purchasable extended libraries (Aldrich, Sadtler, Hummel, etc.)

Bruker OPUS [3][8]:

• Built-in basic library
• Purchasable Bruker own library (~9,000 spectra)

2.7 Online Library: ftir.fun

ftir.fun is an online infrared spectral database for Chinese users [9]:

• Features: Organized by functional groups (e.g., carbonyl, water, alkyl C-H)
• Advantages: Free access, Chinese interface, includes peak assignment explanations
• Applications:
  • Learning characteristic absorption of functional groups
  • Online quick reference of compounds
  • Assisting spectral interpretation

🔗 Further Reading: Visit ftir.fun main site to browse the complete functional group directory.

3. Search Algorithms

The core of library search is comparing the similarity of two spectra and ranking them by a numerical value (HQI). Different algorithms define similarity differently [1][3][10].

3.1 Correlation Coefficient

Algorithm [1][3][10]:

$$r = \frac{\sum_{i=1}^{n} (A_i - \bar{A})(B_i - \bar{B})}{\sqrt{\sum_{i=1}^{n} (A_i - \bar{A})^2 \cdot \sum_{i=1}^{n} (B_i - \bar{B})^2}}$$

where $A_i$ is the $i$th data point of the unknown spectrum, $B_i$ is the corresponding point of the library spectrum, and $\bar{A}, \bar{B}$ are the means.

Characteristics [1][3]:

• Range [-1, 1], 1 = perfect similarity
• Insensitive to absolute intensity (eliminates sample preparation differences)
• Sensitive to peak position
• Tolerates noise to some extent

HQI Conversion [3]:

$$\text{HQI} = \frac{(1 + r)}{2} \times 1000$$

Maps $r \in [-1, 1]$ to HQI $\in [0, 1000]$ for readability [3].

3.2 Absolute Difference Method

Algorithm [1][3][10]:

$$D = \sum_{i=1}^{n} |A_i - B_i|$$

Characteristics [1][3]:

• Smaller D indicates more similarity
• Sensitive to intensity
• Sensitive to peak position
• Sensitive to noise (absolute value amplifies noise)

HQI Conversion [3]:

$$\text{HQI} = \frac{1000}{1 + D/D_0}$$

where $D_0$ is a normalization constant.

3.3 First Derivative Match

Algorithm [1][3][10]:

1. Compute the first derivative of both spectra
2. Calculate the correlation coefficient of the derivative spectra

Characteristics [1][3]:

• Insensitive to baseline drift (derivatives remove constant and linear baselines)
• Sensitive to peak shape
• Extremely sensitive to peak position
• Suitable for fine peak position analysis

Advantages [3]: Eliminates baseline differences, matches peak shapes

Disadvantages [3]: Sensitive to noise (derivatives amplify noise)

3.4 Euclidean Distance

Algorithm [1][3]:

$$d = \sqrt{\sum_{i=1}^{n} (A_i - B_i)^2}$$

Characteristics [1][3]:

• Similar to absolute difference method
• Sensitive to absolute intensity
• Commonly used in chemometrics

3.5 Algorithm Comparison

Table 1: Comparison of mainstream search algorithms (comprehensive [1][3][10])

Recommended Selection [1][3]:

• Routine analysis: Correlation method (default)
• Poor baseline: First derivative method
• Fine peak position: First derivative method
• Hybrid: First use correlation for coarse screening, then first derivative for fine screening

4. HQI (Hit Quality Index) Value

4.1 Definition of HQI

HQI (Hit Quality Index) is a numerical measure of the similarity between two spectra [1][3][10]:

• Typically ranges from 0–1000
• 1000 = perfect match
• 0 = no match
• Slight variations across manufacturers (e.g., Bruker 0–1, Thermo 0–100)

4.2 HQI Interpretation Rules

Bruker OPUS Standard [3][6]:

Table 2: Bruker OPUS HQI interpretation rules (refer to [3])

Thermo OMNIC Standard [1][8]:

• Uses 0–100 range

• 95: high match

• 90–95: good match
• 80–90: possible match
• < 80: low match

4.3 Limitations of HQI

Important Warning [1][3][10]:

① High HQI ≠ Definitely correct

• Homologous series (e.g., propanol, butanol, pentanol) have very similar spectra
• HQI may all be > 950 but actually different compounds
• Must manually verify peak positions

② HQI affected by spectrum quality

• Noisy unknown spectra generally yield low HQI
• Excessive smoothing can artifactual increase HQI
• Search should be based on raw spectra

③ HQI affected by algorithm

• The same pair of spectra may have different HQI with different algorithms
• Comparison must use the same algorithm

④ HQI does not represent concentration

• HQI only reflects similarity
• Does not reflect sample purity
• Impurity peaks will lower HQI

4.4 Practical HQI Usage Tips

Rule of Thumb [1][3][10]:

• Look at Top 5 candidates, not just Top 1
• Compare the gap between candidates—if Top 1 HQI=980 and Top 2 HQI=850, large gap, reliable
• If Top 1 HQI=950 and Top 2 HQI=945, small gap, manual judgment needed
• Check the difference spectrum of Top 1 (unknown - library): flat indicates reliability, systematic deviation indicates unreliability

5. Search Process

5.1 Standard Search Process

1. Prepare unknown spectrum
   │ Processing from Ep 18 completed
   │ (Baseline correction, smoothing, ATR correction, etc.)
   │
   ▼
2. Select library
   │ General library + specialized library
   │ Select based on sample type
   │
   ▼
3. Select algorithm
   │ Correlation method (default)
   │ First derivative method (when baseline is poor)
   │
   ▼
4. Set parameters
   │ Wavenumber range: 4000–400 cm⁻¹ (or focus on functional group region)
   │ Threshold: HQI > 700
   │ Top N: 10
   │
   ▼
5. Execute search
   │ Software returns results in 1–5 seconds
   │
   ▼
6. Analyze Top 5 candidates
   │ Check HQI values
   │ Compare gap between candidates
   │ Check difference spectra
   │
   ▼
7. Manual verification
   │ Check key peak positions of Top 1 candidate
   │ Compare one by one with unknown spectrum
   │
   ▼
8. Comprehensive judgment
   │ HQI > 950 + key peak match → confirmed
   │ HQI 900–950 + key peak match → highly likely
   │ HQI < 900 + partial key peak match → further analysis needed
   │
   ▼
9. Record results
   │ Save search results (screenshot, report)
   │ Record algorithm, library version, parameters

Figure 1: Standard library search workflow (comprehensive [1][3])

5.2 Search Parameter Settings

Wavenumber Range [3][10]:

• Full spectrum search: 4000–400 cm⁻¹ (most common)
• Functional group region: 4000–1400 cm⁻¹ (faster, but ignores fingerprint region)
• Fingerprint region: 1400–400 cm⁻¹ (higher reliability for fingerprint matching)
• Specific peak: Focus on one key peak (e.g., C=O region 1750–1650 cm⁻¹)

Threshold [3]:

• HQI > 700: list all qualifying candidates
• HQI > 800: only high-quality matches
• No threshold: list Top N (recommended)

Top N [3]:

• 10 (default, look at Top 5)
• 20 (for mixture analysis, check more)

Preprocessing [3][10]:

• Automatic ATR correction: Yes
• Automatic normalization: Yes
• Automatic baseline correction: Yes (provided reasonable processing already done)

6. Mixture Search Strategies

[Note: The source text appears to be truncated; the mixture search section is not fully provided. The translation will include only what is given.]

Real samples are often mixtures (e.g., plastics + additives, polymer blends, drugs + excipients). Single library search often only identifies the main component, requiring more complex strategies [1][3][10].

6.1 Principal Component Search + Subtracted Spectrum

Procedure [3][10]:

1. Full-spectrum search → Top 1 candidate (principal component)
2. Calculate subtracted spectrum: $D = A{\text{unknown}} - k \cdot A{\text{library}}$
   • $k$ is the optimal scaling factor (minimizes $|D|$)
3. Examine the subtracted spectrum — are there still peaks?
4. Search the subtracted spectrum again → second component
5. Repeat until the subtracted spectrum has no peaks

Example: Plastic sample

• Full spectrum: matches PE (polyethylene), HQI = 950
• Subtracted spectrum: reveals C=O peak at 1700 cm⁻¹
• Re-search subtracted spectrum: matches some antioxidant

6.2 Multi-component Search

Some advanced software supports "multi-component simultaneous matching" [3][10]:

• User specifies maximum number of components (e.g., 3)
• Software automatically tries combinations
• Provides combined match quality

Advantages [3]:

• Automated
• Suitable for mixtures with known number of components

Disadvantages [3]:

• Computationally intensive
• Requires user to estimate number of components

6.3 Chemometric Methods

Principal Component Analysis (PCA) [1][3][11]

• Dimensionality reduction
• Finds main directions of variation
• Suitable for classifying large numbers of samples

Partial Least Squares (PLS) [1][3][11]

• Suitable for quantitative analysis
• Handles overlapping spectra

SIMCA (Soft Independent Modeling of Class Analogy) [3][11]

• Class modeling
• Suitable for identification vs. anomaly detection

Software [3][11]:

• Thermo TQ Analyst
• Bruker OPUS + OPUS QUANT
• PLS_Toolbox (MATLAB)
• Python scikit-learn

6.4 Multi-dimensional Library Search (2D Search)

Concept [10]: Simultaneously searching infrared spectra + mass spectra + NMR and other multi-dimensional data

• Improves identification reliability
• Suitable for complex samples
• Frontier of academic research

VII. Risk of Misidentification: Cannot Rely Entirely on Library Search

7.1 Common Sources of Misidentification

① Homologue misidentification [1][3]

• Propanol, butanol, pentanol spectra are highly similar
• HQI all > 950, but actually different compounds
• Solution: Look at subtle differences in fingerprint region (carbon chain length effect)

② Isomer misidentification [1][3]

• Ortho-, meta-, para-substituted aromatic hydrocarbons have similar spectra
• Solution: Look at out-of-plane bending vibrations in fingerprint region (700–800 cm⁻¹)

③ Blend misidentification [3][10]

• Main component masks minor components
• Solution: Subtracted spectrum analysis, multi-component search

④ Library spectrum quality issues [1][3]

• Library spectra themselves have defects (KBr moisture absorption, poor SNR)
• Solution: Use authoritative commercial libraries, avoid spectra from unknown sources

⑤ Instrument differences [1][3]

• Different instruments have different resolution and detectors
• Same substance may have slightly different spectra
• Solution: Use same instrument for unknown and library

⑥ Sample preparation differences [1][3][10]

• ATR spectra have low match with transmission spectra libraries
• Solution: Perform ATR correction before searching

7.2 Classic Misidentification Cases

Case A: o-Xylene vs. p-Xylene [1][3]

• Main peaks (C-H stretching, benzene ring C=C) almost identical
• HQI > 950
• But fingerprint region differences are significant:
  • o-Xylene: 742 cm⁻¹ (4 adjacent H out-of-plane bending)
  • m-Xylene: 691 + 767 cm⁻¹
  • p-Xylene: 795 cm⁻¹ (2 adjacent H)

Conclusion: Must manually check fingerprint region, cannot rely solely on HQI [3].

Case B: PEG + Drug [3][10]

• Library search main component: PEG (polyethylene glycol)
• HQI = 940, seems definitive
• Actually: pharmaceutical excipient PEG + small amount of drug
• Subtracted spectrum reveals drug characteristic peaks

Conclusion: After principal component search, must perform subtracted spectrum analysis [3][10].

Case C: Degradation Product vs. Standard [1][3]

• Aged plastic searches match pure polymer
• HQI > 900
• But actually: polymer + oxidative degradation products
• C=O peak at 1700 cm⁻¹ (degradation)

Conclusion: Carefully examine small peaks to identify degradation/impurities [3].

7.3 Strategies to Reduce Misidentification

① Multi-angle verification [1][3]

• High HQI
• Match of key peak positions
• Clean subtracted spectrum
• Consistent with chemical knowledge

② Comprehensive methods [1][3]

• IR + NMR
• IR + MS
• IR + elemental analysis

③ Self-built library [3]

• Contains local common samples
• Contains interferents, common impurities

④ Training and experience [3]

• Understand algorithm limitations
• Familiar with common misidentifications
• Cultivate a critical mindset

VIII. Significance and Methods of Self-built Libraries

8.1 Why Build a Self-built Library?

Limitations of commercial libraries [1][3]:

• Do not cover local specific samples
• Do not include company products/raw materials
• Do not include common interferents

Advantages of self-built libraries [3][12]:

• Highly targeted
• Include local samples
• Consistent measurement conditions
• Continuous accumulation

8.2 Application Scenarios for Self-built Libraries

• Pharmaceutical companies: APIs, excipients, intermediates, products
• Chemical companies: raw materials, products, by-products
• Food companies: raw materials, additives, flavors
• Forensic laboratories: drugs, fibers, paints, explosives
• Environmental monitoring: pollutants in soil, water, air

8.3 Process for Building a Self-built Library

① Library planning [3][12]

• Determine coverage
• Design metadata format
• Choose library management software

② Sample preparation [3][12]

• High-purity standards
• Known source, CAS number
• Dry, protect from light

③ Spectrum acquisition [3][12]

• Fixed measurement conditions (transmission KBr pellet or ATR)
• Multiple replicates (≥ 3)
• Check spectrum quality (SNR, peak positions, baseline)
• Process and save in standard format (JCAMP-DX recommended)

④ Metadata entry [3][12]

• Compound name (Chinese + English + IUPAC)
• CAS number
• Molecular formula, molecular weight
• Structural formula
• Measurement conditions
• Measurer, date, instrument model
• Remarks (e.g., batch number, source)

⑤ Library management [3][12]

• Regular backups
• Version control
• Access permissions
• Continuous expansion

8.4 Quality Control of Self-built Libraries

Spectrum quality [3][12]:

• SNR > 1000:1
• Peak position accuracy ±0.5 cm⁻¹
• Full spectral range
• No ghost peaks

Diversity [3][12]:

• At least 3 spectra per compound (different batches, different preparations)
• Include pure substances + common mixtures

Validation [3][12]:

• Test search accuracy with known samples
• Regular updates (prevent library contamination)

8.5 Search Strategy for Self-built Libraries

Combine with commercial libraries [3]:

• Search self-built library first (covers local samples)
• Then search commercial library (covers common compounds)
• Prioritize self-built library results (consistent conditions)

Specialized algorithms [3][10]:

• Self-built library spectra have consistent conditions → use correlation coefficient method
• No need for derivative method (small baseline differences)

IX. Practical Search Cases

Case 1: Unknown White Powder

Sample: Unknown white powder found in lab, suspected to be some sugar

Search procedure [1][3]:

1. ATR measurement → clear spectrum (SNR > 1000)
2. Full-spectrum search (4000–400 cm⁻¹, correlation coefficient method)
3. Top 5 candidates:
   • Glucose (HQI = 965)
   • Fructose (HQI = 945)
   • Sucrose (HQI = 920)
   • Maltose (HQI = 890)
   • Lactose (HQI = 870)

Manual verification [1][3]:

• Glucose vs. fructose: fingerprint region 900–1100 cm⁻¹ shows obvious differences
• Check key peak: glucose has 836 cm⁻¹ anomeric C-H bending
• Unknown spectrum has this peak → confirmed as glucose

Final conclusion: Glucose (HQI 965 + key peak match)

Case 2: Plastic Bottle Material Identification

Sample: Mineral water bottle fragment

Search procedure [3][10]:

1. ATR measurement

2. Full-spectrum search → Top 1: PET (polyethylene terephthalate), HQI = 985

3. Difference spectrum: almost no residual peaks

4. Manual verification: 1720 C=O, 1240 C-O, 730/720 CH₂ → completely consistent with PET

Conclusion: PET (standard material for beverage bottles)

Case 3: Pharmaceutical Excipient Identification

Sample: Powder after tablet grinding

Search workflow [3][10]:

1. ATR measurement
2. Full-spectrum search → Top 1: Lactose (HQI=890)
3. Difference spectrum: peaks at 1700, 1280 cm⁻¹ found (drug characteristics)
4. Difference spectrum search again → matches an active pharmaceutical ingredient (API)
5. Multi-component search → Lactose + drug, combined HQI=920

Conclusion: Lactose (excipient) + drug (active ingredient)

Case 4: Unknown Mixture (Forensic)

Sample: Confiscated suspicious powder

Search workflow [3][10]:

1. ATR measurement
2. Full-spectrum search → Top 1: Caffeine (HQI=870)
3. HQI < 900 → suspect mixture
4. Difference spectrum: C=O peak at 1730 cm⁻¹ found
5. Difference spectrum search → matches a narcotic
6. Multi-component search → Caffeine + narcotic, combined HQI=950

Conclusion: Adulterated drug (caffeine as diluent + narcotic)

10. Search Result Report

10.1 Report Content

A complete search report should include [3][12]:

• Sample information (name, source, batch number)
• Measurement conditions (instrument, accessory, parameters)
• Search parameters (algorithm, library, range)
• Top 5 candidate list (name, CAS, HQI)
• Manual verification results (comparison of key peak positions)
• Difference spectrum analysis (if any)
• Comprehensive conclusion
• Operator, date

10.2 Report Example

Search Report
=============

Sample Information:
  Name: Unknown white powder
  ID: CASE-2026-001
  Measurement Date: 2026-07-24

Measurement Conditions:
  Instrument: Bruker Tensor II
  Accessory: Platinum ATR (diamond)
  Resolution: 4 cm⁻¹
  Scans: 32
  Range: 4000–400 cm⁻¹

Search Parameters:
  Algorithm: Correlation coefficient
  Library: Bruker Polymer + self-built library
  Threshold: HQI > 700
  Top N: 10

Top 5 Candidates:
  1. Glucose (CAS: 50-99-7), HQI=965
  2. Fructose (CAS: 57-48-7), HQI=945
  3. Sucrose (CAS: 57-50-1), HQI=920
  4. Maltose (CAS: 69-79-4), HQI=890
  5. Lactose (CAS: 63-42-3), HQI=870

Manual Verification:
  Key peak 836 cm⁻¹ (anomeric C-H bending) → Glucose match ✓
  Key peak 994 cm⁻¹ → Glucose match ✓
  Difference from fructose: 900–1100 cm⁻¹ region clearly different

Difference Spectrum Analysis:
  Glucose difference spectrum: almost no residual peaks (< 0.02)

Comprehensive Conclusion:
  Sample is D-glucose (match degree extremely high, HQI=965, all key peaks matched)

Operator: Zhang San
Date: 2026-07-24

11. Future Trends

11.1 Cloud Spectral Libraries

Trends [1][3][9]:

• Online spectral libraries updated in real time
• Global sharing
• Open platforms like ftir.fun
• API integration into laboratory systems

Advantages [9]:

• Continuous updates
• Saves local storage
• Cross-device access

11.2 Machine Learning Assistance

Deep Learning [3][13]:

• Neural networks directly learn spectrum-compound mapping
• Outperforms traditional algorithms on complex mixtures
• Still in research phase

Transfer Learning [13]:

• Use pre-trained models + small data self-built library
• Suitable for small sample scenarios

11.3 Multimodal Fusion

Joint Search [10]:

• Infrared + Raman
• Infrared + Mass Spectroscopy
• Infrared + NMR
• Multi-source information improves reliability

11.4 Mobile Applications

Mobile App [9]:

• On-site rapid detection
• Combined with portable FTIR
• Suitable for customs, emergency, food

🔗 Extended: You can experience online functional group search at ftir.fun main site.

Summary of This Episode

Questions for Reflection

1. You obtain an ATR spectrum of an unknown powder. Library search Top 1 is glucose (HQI=965), Top 2 is fructose (HQI=945). How do you finally determine which sugar it is? Which key peak positions should you check?

2. You suspect a plastic sample is a mixture of PE (polyethylene) and an antioxidant. Design a complete search workflow to identify the main component and additive.

3. Your laboratory is a pharmaceutical QC department. Explain why you should build a self-built library? What compounds should be included in the self-built library? How to manage quality control?

4. An ATR spectrum directly searched against a transmission spectrum library gives a low HQI (about 850). However, manual inspection shows peak positions match completely. What is the reason? How to improve HQI?

5. You suspect a spectrum is a mixture. Explain why you cannot rely solely on the Top 1 HQI? How should you use difference spectrum analysis? Give a specific difference spectrum analysis workflow.

References

[1] Griffiths P R, de Haseth J A. Fourier Transform Infrared Spectrometry. 2nd ed. Wiley, 2007. Chapter 12–13. ISBN: 978-0-471-19404-0.

[2] Sadtler Research Laboratories. The Sadtler Standard Spectra. Bio-Rad Laboratories.
https://www.bio-rad.com/en-us…

[3] Smith B C. Fundamentals of Fourier Transform Infrared Spectroscopy. 2nd ed. CRC Press, 2011. Chapter 9–11. ISBN: 978-1420069297.

[4] Bio-Rad Laboratories. "KnowItAll Spectral Library."
https://www.knowitall.com/

[5] Hummel D O, Scholl F. Atlas of Polymer and Plastics Analysis. 3rd ed. Hanser, 1991. ISBN: 978-3446157035.

[6] Bruker Optics. "OPUS Library Search Manual."
https://www.bruker.com/en/pro…

[7] NIST. "NIST Chemistry WebBook, IR Spectra." Standard Reference Database Number 69.
https://webbook.nist.gov/chem…

[8] Thermo Fisher Scientific. "OMNIC Specta Software User Guide."
https://www.thermofisher.com/…

[9] ftir.fun. "Infrared Spectroscopy Functional Group Database."
https://ftir.fun

[10] Lauer H H, McMillan C, Donnelly J R. "Library Searching Algorithms for Infrared Spectroscopy." Applied Spectroscopy, 1994, 48(8): 936–943. DOI:10.1366/0003702944029870.

[11] Lavine B K. "Chemometrics and Chemoinformatics." ACS Symposium Series, 2005, 894: 1–19.

[12] ASTM E168-16. "Standard Practices for General Techniques of Infrared Quantitative Analysis." ASTM International, 2016.

[13] Chan C C, Lee H L, Chen C W. "Deep Learning for Infrared Spectra Recognition." Analytical Chemistry, 2020, 92(13): 8959–8966. DOI:10.1021/acs.analchem.0c01294.

Next Episode Preview: Ep 20 — Introduction to Quantitative Analysis: Beer-Lambert Law
We will enter the world of infrared spectroscopic quantitative analysis, starting from the derivation and applicable conditions of the Beer-Lambert law A=εbc, and explain core methods such as peak height vs. peak area method, working curve method, multicomponent quantification (solving simultaneous equations). Finally, we will demonstrate how to determine the content of a component in a mixture and compare the differences between infrared quantification and UV-Vis and chromatography.

This article is licensed under CC BY-NC-SA 4.0. Illustrations come from the public domain or network resources with attributed sources, copyright belongs to the original authors.