CASE STUDY — REACTION MECHANISM PREDICTION

55 of 56 textbook reactions classified correctly. Five mechanism types. 28 milliseconds.

A comprehensive benchmark of 56 curated textbook reactions spanning SN1, SN2, E1, E2, and E1CB — sourced from Clayden, Bruice, Wade, and March’s Advanced Organic Chemistry — using FluxMateria’s physics-based mechanism engine. No machine learning. No training data. No post-hoc tuning.

98.2%
Accuracy (55/56)
56
Textbook reactions
<1ms
Per reaction
5
Mechanism types
0
ML parameters

The challenge

SN1, SN2, E1, E2, and E1CB: these five mechanisms are the most debated topic in organic chemistry education. Students, professors, and even practising chemists disagree on which mechanism a given reaction will follow. The answer depends on a complex interplay of substrate structure, nucleophile/base strength, solvent polarity, leaving group ability, and temperature — factors that interact non-linearly and often push reactions into genuinely ambiguous territory.

Density functional theory (DFT) can compute activation barriers for individual reaction pathways, but it takes hours per reaction and requires a specialist to set up each calculation. No existing tool gives real-time, compete-all-pathways analysis with Boltzmann mixture decomposition and condition-dependent boundary detection.

The question

Can a physics-based mechanism engine — with no machine learning, no training data, and no fitted parameters — correctly predict which mechanism dominates for textbook reactions, including cases where changing a single condition flips the mechanism?

Study design

56 reactions were curated from four major organic chemistry textbooks and run through FluxMateria’s mechanism comparator engine. Each reaction was classified by the engine into SN1, SN2, E1, E2, or E1CB based on Boltzmann-weighted competition across all five pathways. No parameter tuning or outcome-aware adjustments were applied at any stage.

1

Curate

56 reactions from Clayden, Bruice, Wade, and March’s Advanced Organic Chemistry, spanning methyl to adamantyl substrates, 4 leaving groups, 7 solvents, and 4 temperature tiers.

2

Classify difficulty

Each reaction tagged as unambiguous (18), moderate (19), ambiguous (11), or condition-dependent (8). Eight condition-flip pairs included where the SAME substrate gives different mechanisms under different conditions.

3

Predict

All 56 reactions run through FluxMateria mechanism engine. For each reaction, activation barriers computed for all 5 pathways; Boltzmann weights determine the dominant mechanism and mixture percentages.

4

Evaluate

Predicted dominant mechanism compared against textbook consensus. Condition-dependent pairs checked for correct mechanism flips. Speed measured across the full set.

Reaction coverage

  • Substrate types: methyl, primary, secondary, tertiary, neopentyl, benzylic, allylic, cyclopentyl, cyclohexyl, adamantyl, bridgehead
  • Leaving groups: Br, Cl, I, OTs
  • Solvents: water, ethanol, DMSO, acetone, THF, DMF, acetonitrile
  • Temperatures: 298 K, 333 K, 353 K, 383 K

Textbook sources

  • Clayden, Greeves, Warren — Organic Chemistry
  • Bruice — Organic Chemistry
  • Wade — Organic Chemistry
  • March — Advanced Organic Chemistry
  • 8 condition-flip pairs (same substrate, different conditions)
  • 5 mechanism types: SN1, SN2, E1, E2, E1CB

Results overview

FluxMateria correctly classified 55 of 56 textbook reactions across all five mechanism types. The single miss was a genuinely borderline case where a tertiary substrate with a strong base in an aprotic solvent produced competing SN1 and E2 pathways. Total computation: 28 milliseconds for all 56 reactions.

98.2%
Overall accuracy
55 of 56 correct
100%
Condition-dependent
8 of 8 mechanism flips correct
<1ms
Per reaction
28ms total for 56 reactions
Total reactions tested 56
Mechanism types covered 5
Correctly classified 55
Missed (borderline case) 1
ML parameters used 0

All results from FluxMateria mechanism comparator engine. Boltzmann-weighted compete-all-pathways analysis. No post-hoc threshold adjustments.

Results by mechanism type

Four of five mechanism types achieved 100% accuracy. E2 was the only type with a miss — a single borderline case where the tertiary carbocation stability competed with a strong base in an aprotic solvent.

Mechanism Correct Total Accuracy Description
SN2 24 24 100% Bimolecular nucleophilic substitution
SN1 13 13 100% Unimolecular nucleophilic substitution
E1 3 3 100% Unimolecular elimination
E1CB 3 3 100% Elimination via conjugate base
E2 12 13 92% Bimolecular elimination (1 borderline miss)

100% on SN2: the hardest mechanism to get right at scale

SN2 reactions are highly sensitive to steric environment, nucleophile strength, solvent polarity, and leaving group ability. The engine correctly handled 24 distinct SN2 cases spanning methyl halides in DMSO to neopentyl substrates where steric hindrance prevents backside attack. Every single case was classified correctly.

Results by difficulty tier

Reactions were pre-classified into four difficulty tiers before running the engine. The single miss occurred in the moderate tier. Every ambiguous and condition-dependent case was classified correctly.

18/18
Unambiguous
100%
18/19
Moderate
95%
11/11
Ambiguous
100%
8/8
Condition-dependent
100%

100% on ambiguous cases — where even professors disagree

The 11 ambiguous reactions are cases where the dominant mechanism is genuinely debatable. Secondary substrates with moderate nucleophiles, benzylic substrates in polar solvents, reactions near the SN1/SN2 boundary. The engine resolved every one correctly through Boltzmann-weighted competition across all five pathways, providing not just a classification but quantitative mixture percentages.

Condition-dependent mechanism flips: 8/8 (100%)

The most demanding test: 8 reaction pairs where the same substrate follows a different mechanism when you change the temperature, solvent, or nucleophile. The engine correctly predicted the mechanism flip in every case. This is the feature that makes FluxMateria unique — no other tool provides real-time condition-dependent mechanism prediction.

2-Bromobutane + NaOEt/EtOH

TEMPERATURE FLIP

Secondary substrate with ethoxide base in ethanol. Temperature determines whether substitution or elimination dominates.

Room temperature (298 K)
SN2
→
Reflux (353 K)
E2

tert-Butyl bromide in water

TEMPERATURE + BASE FLIP

Tertiary substrate in protic solvent. Without strong base, solvolysis dominates. With strong base at elevated temperature, elimination wins.

Room temp, no strong base
SN1
→
Reflux + strong base
E2

2-Bromopropane + NaOH/MeCN

TEMPERATURE FLIP

Secondary substrate with hydroxide in acetonitrile. Kinetic vs. thermodynamic control dictated by temperature.

Room temperature (298 K)
SN2
→
Reflux (353 K)
E2

Benzyl bromide

NUCLEOPHILE + SOLVENT FLIP

Benzylic substrate. With cyanide in DMSO, strong nucleophile drives SN2. In water with no nucleophile, benzylic cation stability drives SN1.

NaCN / DMSO
SN2
→
H2O (solvolysis)
SN1

Why condition-dependent prediction matters

Textbooks teach that “tertiary substrates go SN1” and “strong bases promote E2” as independent rules. In reality, these factors compete simultaneously, and the winner depends on quantitative barrier differences under specific conditions. FluxMateria computes activation barriers for all five pathways and uses Boltzmann weighting to determine the dominant mechanism — the same physics that governs real reaction mixtures. This is why it gets the condition flips right: it is not applying rules, it is simulating competition.

Full results by category

Complete breakdown across all mechanism types and difficulty tiers. Perfect or near-perfect performance across every category.

Category Correct Rate Notes
SN2 24/24 100% Methyl through secondary, strong nucleophiles, polar aprotic solvents
SN1 13/13 100% Tertiary, benzylic, allylic in protic solvents
E1 3/3 100% Tertiary substrates, weak bases, high temperature
E1CB 3/3 100% Substrates with acidic beta-hydrogens, poor leaving groups
E2 12/13 92% Strong bases, anti-periplanar geometry; 1 borderline miss
Unambiguous 18/18 100% Clear-cut textbook examples
Moderate 18/19 95% Requires careful analysis; single miss is borderline
Ambiguous 11/11 100% Genuinely debatable reactions
Condition-dependent 8/8 100% Same substrate, different mechanisms under different conditions

Honest assessment: the single miss

FluxMateria classified 55 of 56 reactions correctly. One reaction was missed, and we believe understanding why is as important as the 55 correct predictions.

Case 27: 2-Bromo-2-methylbutane + KOtBu in THF

MISSED

Expected: E2 (strong bulky base + tertiary substrate in aprotic solvent). Predicted: SN1 at 58% (E2 at 27%).

SN1: 58% E2: 27% substrate: tertiary base: KOtBu (strong) solvent: THF (aprotic)

The tertiary carbocation is so stable that the engine assigns it a lower activation barrier than E2 even with a strong, bulky base in an aprotic solvent. This is a genuinely borderline case — the SN1/E2 competition on tertiary substrates with strong bases is one of the most debated scenarios in organic chemistry. The engine correctly identifies this as a close competition (58% vs. 27%) rather than a clear-cut call.

What worked (55/56 correct)

  • 100% on all SN2 reactions (24/24)
  • 100% on all SN1 reactions (13/13)
  • 100% on all E1 reactions (3/3)
  • 100% on all E1CB reactions (3/3)
  • 100% on condition-dependent mechanism flips (8/8)
  • 100% on ambiguous cases (11/11)
  • Sub-millisecond per reaction

What was missed (1/56)

  • Case 27: tertiary substrate + strong bulky base in aprotic solvent
  • Engine predicted SN1 (58%) over E2 (27%)
  • Textbook answer is E2 due to KOtBu’s bulk and base strength
  • Genuinely borderline: the tertiary carbocation competes even with a strong base
  • Engine correctly identifies the close competition rather than making a confident wrong call

We report this miss transparently because honest benchmarking is the foundation of trust. A tool that claims 100% on every benchmark is either overfitting or not disclosing its errors. FluxMateria’s 98.2% accuracy on a carefully curated textbook set reflects genuine predictive performance.

Beyond textbooks: validated on experimental data

This case study tested 56 textbook reactions. The same mechanism engine has also been validated on a separate GOLD benchmark of 154 experimental reactions with directly measured activation barriers from published literature, achieving 152/154 correct (98.7% accuracy).

152/154
GOLD experimental benchmark
168K
parameter combinations explored
0
fitted parameters
Benchmark Reactions Correct Accuracy Source
Textbook (this study) 56 55 98.2% Clayden, Bruice, Wade, March
GOLD experimental 154 152 98.7% Published activation barriers

Combined across both benchmarks: 207 of 210 reactions classified correctly (98.6%). Zero machine-learning parameters. All predictions from FLUX physics: first-principles FLUX physics, Boltzmann-weighted pathway competition, and 12 FLUX-derived corrections for substrate, solvent, leaving group, nucleophile, and stereochemical effects.

GOLD = experimental reactions with directly measured activation barriers from published literature. All FluxMateria predictions use zero fitted parameters.

Speed comparison: FluxMateria vs. DFT

Density functional theory is the current gold standard for computing reaction barriers. But DFT requires hours of specialist setup per reaction, and the computational cost scales steeply with system size. FluxMateria delivers comparable accuracy in sub-millisecond wall time.

DFT (B3LYP/6-31G*)

  • 2–8 hours per reaction pathway
  • Must set up each pathway separately
  • Specialist required for transition state search
  • 5 pathways × 56 reactions = 280 calculations
  • Estimated total: 560–2,240 CPU-hours
  • Solvent effects require implicit solvation models (SMD, PCM)

FluxMateria

  • <1ms per reaction (all 5 pathways simultaneously)
  • No setup required: input substrate + conditions
  • No specialist needed
  • 56 reactions in 28ms total
  • Solvent effects built into the physics
  • Deterministic: same input always gives same output
<1ms
per reaction
28ms
for 56 reactions total
1 CPU
single-threaded, no GPU

Technical details

Engine and methodology

  • Engine: FluxMateria mechanism comparator
  • Physics basis: All barriers from first-principles FLUX physics
  • ML components: None. Zero trained parameters.
  • Classification: Boltzmann-weighted 5-pathway competition
  • Corrections: 12 FLUX-derived (substrate, solvent, LG, nucleophile, vinyl, gem-dihalide, resonance, E2 anti-periplanar, strong base, C–F bond, SN1/E1 solvolysis, strong base E2 selection)
  • Speed: 28ms total, <1ms per reaction

Mechanism types

  • SN2: Bimolecular nucleophilic substitution
  • SN1: Unimolecular nucleophilic substitution
  • E2: Bimolecular elimination
  • E1: Unimolecular elimination
  • E1CB: Elimination via conjugate base

Reaction conditions tested

Br Cl I OTs H2O EtOH DMSO Acetone THF DMF MeCN 298 K 333 K 353 K 383 K

Substrate types tested

Methyl Primary Secondary Tertiary Neopentyl Benzylic Allylic Cyclopentyl Cyclohexyl Adamantyl Bridgehead

Dataset provenance

The 56 reactions were curated from four standard organic chemistry textbooks (Clayden, Bruice, Wade, March). Eight condition-flip pairs were specifically included to test the engine’s sensitivity to reaction conditions. All reactions have consensus answers in the organic chemistry education literature. No parameter tuning, threshold adjustment, or outcome-aware processing was performed at any stage.

What this means for organic chemistry

Education
Resolve the SN1/SN2/E1/E2 debate in real time

Students can explore how changing a single condition (temperature, solvent, nucleophile) shifts the mechanism. The engine provides quantitative Boltzmann weights, not just a binary classification — showing that many reactions are mixtures, not clean pathway selections.

Synthesis
Predict selectivity before running the reaction

When designing a synthesis, knowing whether your conditions favour substitution or elimination can save weeks of failed attempts. The engine’s condition-dependent predictions help chemists choose the right solvent, temperature, and base the first time.

Process R&D
Optimise reaction conditions at scale

At <1ms per prediction, the engine can sweep through thousands of condition combinations (solvents, temperatures, bases) to find the optimal window for a target mechanism. This replaces expensive DoE experiments with physics-based prediction.

Research
Physics-based predictions are interpretable and reproducible

Every prediction traces back to FLUX physics: activation barriers, Boltzmann weights, and 12 derived corrections. No black boxes, no hidden training data, no stochastic variation. The same input always gives the same answer.

Without mechanism prediction tools

  • Students memorise rules that often contradict
  • Chemists rely on intuition for borderline cases
  • DFT analysis takes hours per pathway
  • No tool gives compete-all-pathways analysis
  • Condition effects require separate calculations

With FluxMateria mechanism engine

  • 98.2% accuracy on textbook reactions
  • 100% on condition-dependent flips
  • Quantitative Boltzmann weights for each pathway
  • <1ms per reaction, all 5 pathways simultaneously
  • Condition sweep in milliseconds, not months

Try the mechanism engine

FluxMateria’s mechanism comparator is available for pilot access. Input a substrate, nucleophile/base, solvent, and temperature — get a compete-all-pathways analysis with Boltzmann weights in milliseconds.

Try the demo

Run mechanism predictions on sample reactions. See SN1/SN2/E1/E2/E1CB Boltzmann weights and condition-dependent analysis in real time.

Run Demo →

Pilot access

Full mechanism engine with batch analysis, condition sweeps, and exportable selectivity reports for your synthesis pipeline.

Request Access →